Comparative standards for the evaluation of clinical and epidemiological data in oncology: methodical development and verification on population data

Transcription

1 Dissertation thesis in oncology Comparative standards for the evaluation of clinical and epidemiological data in oncology: methodical development and verification on population data Jan Mužík 2010 Masaryk University, Faculty of Medicine, Institute of Biostatistics and Analyses, Brno, Czech Republic Supervisor: doc. RNDr. Ladislav Dušek, Ph.D. Co-supervisors: prof. MUDr. Jan Žaloudík, CSc. prof. MUDr. Rostislav Vyzula, CSc.

2

3 Bibliographic identification Author: Title of dissertation: Title of dissertation (in Czech): Doctoral study program: Supervisor: Co-supervisors: Jan Mužík Comparative standards for the evaluation of clinical and epidemiological data in oncology: methodical development and verification on population data Srovnávací standardy pro hodnocení klinických a epidemiologických dat v sou asné onkologii: vývoj metodiky a její verifikace na popula ních datech. Oncology doc. RNDr. Ladislav Dušek, Ph.D. prof. MUDr. Jan Žaloudík, CSc. Year of defense: 2010 prof. MUDr. Rostislav Vyzula, CSc. Keywords: Keywords (in Czech): cancer, epidemiology, data analysis novotvary, epidemiologie, analýza dat

4

5 Content Content... 5 Abstract... 7 Abstrakt (Abstract in Czech)... 9 Aims of the dissertation thesis Acknowledgements A. Data and Information Background of Czech Cancer Care current status Introduction overview of the main sources of information The main information sources of Czech oncology available on-line Demographic data and the Czech National Cancer Registry Demographic data and the Death Records Database The Czech National Cancer Registry The system of maintenance and check of records in CNCR Ensuring quality of data on the epidemiology of malignant tumours in the Czech Republic The Czech National Cancer Registry On-line Czech clinical registries monitoring the use of monoclonal antibodies in cancer therapy 27 5 Use of population-based data for cancer care assessment in the Czech Republic References B. Author s contribution to the development of data standards for the evaluation of Czech cancer care C. Author s contribution to the development of software tools for automated analysis and reporting of cancer data Glossary of specific terms Enclosures Paper I. Cancer Epidemiology in the Czech Republic: Current Load and the Time Trends. Paper II. Historical data of the Czech National Cancer Registry: information value and risk of bias Paper III. Epidemiology of chronic myeloid leukemia. Paper IV. Data registries form indispensable information base of current oncology. Paper V. The Czech National Cancer Registry and reference standards for health care assessment. Paper VI. Colorectal cancer screening in Europe. Paper VII. Multivariate analysis of risk factors for testicular cancer: a hospital-based casecontrol study in the Czech Republic Paper VIII. Analysis of population cancer risk factors in national information system SVOD. Paper IX. Epidemiology of Malignant Tumours in the Czech Republic [online] Paper X. What is the information availability to the cancer epidemiology data? Paper XI. UROWEB - web portal for analysis and visualization of epidemiology, diagnosis and treatment of urologic malignancies [online]. XII. Appendix to paper I. Summary of cancer epidemiology in the Czech Republic XIII. Curriculum Vitae - 5 -

6

7 Abstract Malignant diseases are a serious health problem in the Czech population, particularly with regard to annually increasing numbers of newly diagnosed patients. In international comparisons, the Czech Republic is ranked in the top positions in the incidence and mortality of cancer. Causes of development of many types of cancer are not clearly known, despite a still more extensive knowledge of processes of tumor growth. One of the most important sources of information in understanding this issue still remains cancer epidemiology, which allows us to assess the population burden, to identify the risk populations, to evaluate the success of diagnosis and treatment and to predict the number of patients for the planning of cancer care, etc. An essential prerequisite for the use of cancer epidemiology is availability of sufficiently large and complete population data and representativeness and completeness of such information. The Czech Republic has such data available thanks to the population-based National Cancer Registry (NCR), which is consistently managed since 1976 until now. NCR contains data on all diagnosed neoplasms in the Czech population, including relevant data on the tumour diagnosis and primary treatment of cancer. The database of NCR currently contains over 1.6 million records. Even a few years ago this extensive data was not fully accessible for professionals; also its level of quality, completeness and informational value was unknown. Availability of data on the epidemiology of neoplasms in the Czech Republic was previously limited only to publications Cancer Incidence in the Czech Republic issued annually by Institute for Health Information and Statistics (IHIS). Due to the limited capabilities of these publications and the growing need to use the extensive data collected in the NCR in 2001 was this database made first time available (with agreement of IHIS) for the cancer care professionals (the steering committee of the Czech Society for Oncology). Institute of Biostatistics and Analyses, Masaryk University (IBA) was commissioned to process these data and the author of this dissertation started to develop standardized data management over full NCR data. In addition to the definition of data standards, new measures for data quality assessment and scientifically based analytic reporting had to be developed. This dissertation thesis summarizes original articles and contributions of the author to the development of procedures for processing of large epidemiology data sets, algorithms completing information on the disease stage according to the recorded values (TNM, ICD and ICD-O classifications with respect to their changes over time) and finally algorithms for stratification of patients according to their information value and completeness of the data. The results initiated definition of the reference data set of patients as the reference pool for comparative evaluation of health care outcomes in oncology at the population level. The comprehensive summary of the current epidemiologic situation in the Czech Republic was published and basic standards for the comparative assessment of the population burden by cancer were defined. The standards include quantification of trends in incidence, mortality and prevalence, procedures to prepare data for international comparisons, calculation of lifetime cumulative cancer risk, assessment of regional distributions and trends. With regard to the lack of availability of these highly valuable data for professionals and for the lay public as well, the author also contributed to the development of several software analytic tools, which made the NCR data available to the users in user friendly environment. The first generation of this software was system SVOD (System for Visualisation of Oncologic Data), which provided a complete set of epidemiological data and data on the diagnosis and treatment recorded in NCR. This software was designed mainly to professionals. In the second generation a new Portal of cancer epidemiology of the Czech - 7 -

8 Republic (available from was developed based on web technology. In the last work the team headed by the author developed a specialized portal for analysis of data on urology malignancies ( This novel tools makes accessible epidemiological data for the public, including presentations oriented abroad

9 Abstrakt (Abstract in Czech) Nádorová onemocn ní p edstavují závažný zdravotní problém sou asné eské populace, zvlášt s ohledem na stále rostoucí po ty nov diagnostikovaných pacient ro n. eská republika zaujímá v mezinárodním m ítku p edním místa v incidenci a mortalit t chto onemocn ní. P í iny vzniku ady nádorových onemocn ní nejsou známy, a to i p es stále rozsáhlejší znalosti proces nádorového zvratu a r stu. Jedním z velmi významných zdroj informací p i porozum ní této problematice tak stále z stává epidemiologie nádorových onemocn ní, která umož uje hodnotit popula ní zát ž a identifikovat její trendy, poukazovat na rizikové skupiny v populaci, hodnotit úsp šnost diagnostiky a lé by, predikovat po ty lé ených p i plánování pé e o nemocné atp. Zásadním p edpokladem využití nádorové epidemiologie je však dostupnost dostate n rozsáhlých a úplných popula ních dat, jejich reprezentativnost a informa ní úplnost. Díky celopopula nímu Národnímu onkologickému registru (NOR) standardizovan vedenému od roku 1976 až do sou asnosti má eská republika taková data k dispozici. Národní onkologický registr obsahuje údaje o všech zachycených novotvarech v eské populaci v etn relevantních údaj o diagnostice a primární protinádorové lé b onkologických pacient a v sou asné dob obsahuje p es 1,6 miliónu záznam. Ješt v nedávné dob však tato rozsáhlá data nebyla b žn dostupná odborné ve ejnosti a nebyla známa úrove jejich kvality, úplnosti a informa ní hodnoty. Dostupnost údaj o epidemiologii novotvar v eské republice byla d íve omezena pouze na publikace Novotvary každoro n vydávané Ústavem zdravotnických informací a statistiky R (ÚZIS R). Vzhledem k omezeným možnostem t chto publikací a k rostoucí pot eb využívat rozsáhlé údaje shromážd né v NOR byla v roce 2001 tato data se souhlasem ÚZIS poprvé zp ístupn na odborné onkologické ve ejnosti zastupované vedením eské onkologické spole nosti LS JEP ( OS) pro možnosti vlastního zpracování a využití. Zpracováním t chto dat byl pov en Institut biostatistiky a analýz Masarykovy univerzity (IBA) a v rámci IBA byl hlavním analytikem dat NOR a dat o epidemiologii nádor jmenován autor diserta ní práce RNDr. Jan Mužík. Disertace shrnuje originální práce autora a jeho p ísp vek k vývoji zcela nového, mezinárodn konkurenceschopného managementu popula ních onkologických dat R. Byly vytvo eny postupy pro zpracování t chto objemových dat, byly vyvinuty algoritmy kompletující informaci o stadiu onemocn ní dle zaznamenaných hodnot (TNM, MKN a MKN-O klasifikace i s ohledem na jejich zm ny v ase) a následn algoritmy stratifikující pacienty do skupin podle jejich informa ní hodnoty a úplnosti dat. Výsledkem je definice referen ního souboru pacient jako komparativního standardu pro hodnocení výsledk lé ebné pé e v onkologii na popula ní úrovni. Dále bylo provedeno kompletní vyhodnocení aktuální situace epidemiologie novotvar v eské republice a byly nastaveny základní komparativní standardy pro hodnocení popula ní onkologické zát že (standard obsahuje trendy incidence, mortality a prevalence, mezinárodní srovnání, kumulativní riziko, v kovou strukturu, regionální zát ž a trendy stadií). S ohledem na nedostate nou dostupnost t chto vysoce hodnotných dat pro odbornou i laickou ve ejnost byly dále vyvinuty softwarové analytické nástroje zp ístup ující data NOR v uživatelsky p ijatelné podob pomocí interaktivních analytických nástroj. V první generaci byl vyvinut software SVOD (Systém pro Vizualizaci Onkologických Dat) zp ístup ující kompletní sadu epidemiologických data, dat o diagnostice a lé b zaznamenané v NOR. Tento software byl ur ený p edevším odborné ve ejnosti. Ve druhé generaci byl následn vytvo en Portál epidemiologie nádor v eské republice ( který v rozsahu základních komparativních standard pro hodnocení popula ní onkologické zát že zp ístupnil tato data nejširší ve ejnosti v etn prezentace do zahrani í. Nejaktuáln jším výsledkem práce - 9 -

10 týmu vedeného autorem diserta ní práce je nová generace webových analytických nástroj, publikovaná se zam ením na urologické malignity (

11 Aims of the dissertation thesis The general objective of the dissertation was to develop procedures for the use of available sources of population data on the cancer epidemiology with focus on data of the National Cancer Registry of the Czech Republic (NCR) and setting the reference standards and comparative standards for assessing the cancer burden and for evaluating the results of medical care. Aims of this work were as follows: - development of standardized procedures for the processing of population-based data - evaluation of completeness and information usability of these population data - development of algorithms for stratification of records according to their completeness and information value - setting standards for the comparative assessment of the population cancer burden - definition of a reference group of patients and the comparative standard for assessing the outcomes of medical care on the population level (mainly for evaluation of survival) - publication of the results in articles in scientific journals and in electronic form via the software tools

12 Acknowledgements First and foremost I would like to thank to my supervisor doc. RNDr. Ladislav Dušek, Ph.D. for the opportunity to work on this interesting and useful topic, for superior working conditions, for his motivation, professional guidance and assistance. I would like to thank also the steering committee of the Czech Society for Oncology, namely to prof. MUDr. Jan Žaloudík, CSc., prof. MUDr. Rostislav Vyzula, CSc. and to prof. MUDr. Jitka Abrahámová, DrSc. for valuable comments and suggestions to my work and for the help with orientation in this area. I would like to thank also to the management of the Institute for Health Information and Statistics and the board of Czech Society for Oncology for making the National Cancer Registry accessible for such extensive processing. Last, but not least, I would like to thank my colleagues from the Institute of Biostatistics and Analyses for creating an inspirational, creative and productive work environment and to my family and my friends for their support and patience

13 A. Data and Information Background of Czech Cancer Care current status Published in Dušek L. et al. Czech Cancer Care in Numbers Praha: Grada Publishing, a.s., Shortened, updated

14

15 1 Introduction overview of the main sources of information A comprehensive quantitative assessment of health care must cover its inputs, the processes and the outputs. The health care inputs are represented primarily by patient numbers and their health condition, i.e. the health care load. The health care processes involve the distribution and availability of health care, and the standardization of diagnostic and treatment approaches, as well as their control. And finally, the health care outputs are generally perceived as health care quality and the results. Of course, all these components of health care are strongly linked to costs. The costs, in turn, primarily concern the extent of care needed (the health care load); well-adjusted processes should ensure correct and timely utilization of the financial resources, and health care quality assessment is essential to evaluate whether these resources have been utilized reasonably and cost-effectively. It is generally known that any meaningful assessment cannot be done without specific and representative data, and this particularly applies to health care. It is, therefore, inherent that a solid information platform be developed in the long term if we are to properly assess different areas of health care (hence the need for background information). The Czech health care system and its medical professionals are certainly able to assess the results at the level of individual patients, while there are often serious problems with such assessments at regional or even national levels. The analyses of population-based data on diagnostic and therapeutic processes and the results are not available, and the same applies to standardized export of data from the hospital information systems. With regards to the facts mentioned above, the Czech Society for Oncology has based its information strategy on the combination of population-based and clinical registries, and makes every effort to develop it further in cooperation with the health care payers. This would not be possible without the very helpful participation of doctors and their colleagues who collect data beyond their call of duty. It is only thanks to such professionals that the Czech Society for Oncology has data available for all strategic components of health care assessment. The system is based on real data which is collected directly at clinical practice. There is no point to analyze international data, since the contained information there is not helpful and/or compatible for assessing the actual results achieved in the Czech healthcare system. The following paragraphs provide an overview of the main data sources: The Czech Statistical Office processes data on the demographic structure of the Czech population within the monitoring of demographic trends. This data describes the main demographic characteristics noted for the Czech population, namely the total number of inhabitants, age structure, life expectancy, etc. The Death Records Database serves as a source of ascertaining population mortality from malignant neoplasms in the Czech Republic; this database is also maintained by the Czech Statistical Office in compliance with the international methodology and based on the data from the Death Certificate (ICD-10 classification). The Czech National Cancer Registry (CNCR) is the principal source of data on the epidemiology of malignant tumours. CNCR has become an indispensable part of the complex cancer care, containing more than 1.6 million records over the period and covering the entire Czech population. The registration of malignant neoplasms is stipulated by the legislation and is obligatory. Three screening programmes of secondary prevention are currently in operation, focusing on breast cancer, colorectal cancer and cervical cancer. All these screening programmes

16 benefit from a solid background comprising population data, making it possible to assess their nationwide impact. Moreover, specific information systems for cancer screening programmes have been developed, focusing on the collection of data on performed examinations and their results. This so-called data audit of screening programmes serves for regular assessment of screening performance. Clinical registries of the Czech Society for Oncology represent as the main source of clinical data. These registries provide a reliable description of the clinical practice, focusing on the collection and analysis of data on diagnostic and therapeutic procedures, or monitoring the results of a specific type of treatment. These clinical databases in combination with the Czech National Cancer Registry constitute a complex source of data, making it possible to assess not only the healthcare results, but also population-based indicators of quality, including the nationwide and regional availability of health care. Some information cannot be obtained from commonly available data sources; for example, the time trends in the probability of relapse/progression of primary malignant disease, or the probability of application of higher-line treatments in differently affected cancer patients. In these cases, predictions and assessments are made by an expert panel of the Czech Society for Oncology. Each year, the expert panel processes relevant data from the above-mentioned sources in order to answer clearly defined questions. In 2005, the concept and structure of electronic documentation of cancer patient was developed by another expert panel of the Czech Society for Oncology. Recently, this information system has been progressively implemented in everyday practice of the Czech health facilities; in this way, data previously reported to health care payers only will become available for analyses by the Czech Society for Oncology. The combination of these reports with diagnostic records within CNCR creates a brand new information system, making very detailed clinical data readily available. The project called Fusion of CNCR and health care payers data was successfully accomplished in two major hospitals and led to the analysis of more than 110,000 complete and comprehensive records on the treatment of cancer patients. In 2008, the Czech Society for Oncology commenced cooperation with the Czech National Reference Centre (CNRC) of health care payers; the CNRC is able to centralize data from health care payers and to monitor diagnostic and treatment procedures nationwide. Population-based data on diagnostic examinations within the cancer screening programmes are currently being analyzed in order to provide support for the assessment of Czech cancer screening programmes. The above-mentioned overview documents the fact that the extent of background information at the Czech Society for Oncology covers all needed levels with regard to the collection and assessment of data ranging from individual hospitals to the population level (Figure 1)

17 3rd LEVEL INVESTIGATION Epidemiological reports Registration records for research purposes Clinical trials 2nd LEVEL ANALYSIS Diagnosisspecific records Continuous clinical records Records on complications Quality of life 1st LEVEL REGISTRATION Collection of clinical records Electronic case report forms Reports on provided health care Figure 1 The structure of background information as developed by the Czech Society for Oncology 2 The main information sources of Czech oncology available on-line The Czech Society for Oncology makes every effort to ensure unrestricted availability of important data sources on the internet: information and news for professionals and the lay public is made available here. The presented pages serve as a quick overview of the most important outputs. The NOP On-line project ( and the main communication portal of the Czech Society for Oncology ( These portals communicate, among others, the targets of the National Cancer Control Programmes and the way these targets are being accomplished. The portal contains a section for health professionals and a separate section for patients, their families and friends. The portal has been designed to provide an objective presentation of all cancer centres (CCs) and to provide information services to these centres in the respective regions (communication between CCs and their partners, regional data collection...). The individual CCs can use the portal to present their targets, the housed equipment and the quality of provided health care. The national web portal focused on the epidemiology of malignant tumours ( This web portal graphically presents more than 1.6 million records from the Czech National Cancer Registry. The portal is primarily focused on epidemiological data and the related time trends, inclusive of the regional differences and population risks. Users can

18 find here a number of interactive tools and automatic reporting systems, facilitating orientation in the complex database even for the lay users, who can assimilate the analyses and outputs according to their own preferences. The portal also analyzes the demographic data of the Czech population (this function has been made available thanks to the cooperation with the Czech Statistical Office) and the extensive databases on the condition of the environment. Cancer incidence and mortality in annual outputs of Institute of health information and statistics of the Czech Republic (IHIS) IHIS annually publish tabular overview of incidence and mortality of cancer in publications Cancer Incidence in the Czech Republic. They are available also in electronic form here 00&kind=1&mnu_id=5300 Web portals of the national cancer screening programmes ( These web portals deal with organized population screening programmes in the Czech Republic. Apart from organizational and educational materials, detailed reports on data audits and their results are regularly published on these websites. The DIOS project (Dose Intensity as Oncology Standard, This web portal deals with the parametric assessment and adherence to anticancer chemotherapy guidelines. Apart from several useful tools, which can be also used for educational purposes, the portal provides analytical and software support for its users, as well as the realization of multicentre projects. The most valuable tools involve the Central Library of Chemotherapy Regimens, and a set of automated software tools, such as Therapy Organizer, Dose Intensity Calculator, and reporting tools. The portal contributes to the standardization of diagnostic and therapeutic procedures. Methodical and data background for clinical and drug registries: This methodical portal provides the background and maintenance data for the drug and clinical registries in the Czech Republic. The portal informs both the professionals and lay persons about the development of methods for registering clinical data, and describes the standards for on-line information systems and on-line available registries. This platform also led to the establishment of eight drug registries developed by the Czech Society for Oncology, which are focused on expensive anticancer pharmacotherapy (for instance, subportals of respective projects such as or on the assessment of data collected within individual hospitals or data describing selected diagnostic groups of malignant tumours (such as A specialized information portal oriented on urological malignancies ( This portal is focused on the assessment of the epidemiological load, the diagnostic and therapeutic procedures, and the treatment results in urological malignancies in the Czech Republic. All population-based data from this domain is made available in an interactive manner, and reference survival rates are also provided. The project is the outcome of the cooperation between the Czech Society for Oncology and the Czech Urological Society

19 3 Demographic data and the Czech National Cancer Registry 3.1 Demographic data and the Death Records Database As a standard part of population monitoring, the Czech Statistical Office also administrates data on the demographic structure of Czech Republic and makes it available on its website ( This data describes the main demographic characteristics of the Czech population, such as the total population, the age structure, life expectancy, as well as predictions of the time trends up to The Czech legislation requires all deaths which occur in the Czech Republic to be registered in the Death Records Database; for this purpose, standardized Death Certificates have been designed to collect precise data on the cause of death in each individual. The causes of death are classified according to the International Classification of Diseases (ICD), which can be found on the WHO website ( at the time of writing this publication, the 10 th revision of ICD was valid in the Czech Republic and worldwide. Moreover, the primary cause of death is further specified according to standard guidelines. Apart from detailed information on the cause of death, the data structure of the Death Records Database also contains information on birth certificate number, the sex and address of the deceased, as well as the dates of birth and death. The Death Records Database is administered by the Czech Statistical Office and outputs are available on-line in aggregate form at or The Czech National Cancer Registry The Czech National Cancer Registry (CNCR) was established in 1976 in order to collect population-based and clinical data on newly diagnosed malignant tumours in the Czech Republic (or in Czechoslovakia, historically). More specifically, a database of individuals diagnosed with malignant tumours was set up, and its records also served for prospective monitoring during the follow-up period. Nowadays, the CNCR is an indispensable part of the Czech National Cancer Control Programme, containing more than 1.4 million records since 1976 and covering the entire Czech population. The CNCR is enshrined in the Czech legislation as part of the National Health Information System (NHIS) and is administered by the Institute of Health Information and Statistics of the Czech Republic (IHIS, The Czech National Cancer Registry (CNCR) contains personal data on patients, data describing malignant tumours and diagnostic details, data on patients' treatment, as well as data on post-treatment follow-up. The parametric structure of the CNCR can be divided into several groups (according to the clinical significance of respective parameters) as follows: Identification of record on cancer case Identification and demographic characteristics of the patient Data on the diagnosis of malignant tumour (such as the date of diagnosis, clinical and histological characteristics) All treatment modalities of primary anticancer therapy Actual condition of the patient and his/her disease, as monitored with follow-up reports Additional parameters

20 3.3 The system of maintenance and check of records in CNCR The registration of malignant tumours is enshrined in the Czech legislation and is obligatory. The Czech National Cancer Registry (CNCR) is administered by the Institute of Health Information and Statistics of the Czech Republic (IHIS), which is responsible for cohesion of the registry as regards its methodology and contents. The IHIS regularly checks the correctness of submitted data, distributes the methodology, processes, and provides and publishes statistical outputs, and defines access rights to authorized users. The Coordination Centre for Departmental Medical Information Systems (CCDMIS) processes data from the CNCR on nationwide level. The CCDMIS is responsible for smooth operation of the registry, the database status, technical support and data security. It also provides information technology (HW, SW, and communication), authentication and authorization. The CNCR Council is an advisory body and expert guarantor to CNCR. Members of the CNCR Council most typically include representatives of IHIS, the regional CNCR centres, the Czech Ministry of Health, and from the Czech Society for Oncology. Any health care facility in which a malignant tumour has been diagnosed, treated and followed-up, is legally obliged to report that tumour. Data on patients and their tumours are sent to CNCR by specifically instructed health care professionals. The health care facility fills in the so-called Report on Malignant Neoplasm, which forms an integral part of the obligatory medical documentation. In compliance with the legislation in force, this Report is submitted either in written or in electronic form, in a specified extent within one month from the date of diagnosis, to a respective regional CNCR centre. The reporting health care facilities are obliged to submit all data required to compile the complete Report on Malignant Neoplasm. The disease is then followed-up using the so-called Follow-up Report on Malignant Neoplasm, which must be filled in by the follow-up health care facility in the specified extent and within the defined time intervals; the Follow-up Report is then submitted to the regional CNCR centre for further processing. The regional CNCR centres are obliged to send this data to the national centre electronically, regularly (at least once a year), and before the deadline as stipulated by the CNCR administrator. The IHIS then ensures data check, and CNCR records are completed with data from the Death Records Database. 3.4 Ensuring quality of data on the epidemiology of malignant tumours in the Czech Republic Epidemiological data contained in the Czech National Cancer Registry (CNCR) present an essential source of information, which is employed for the predictive assessment of cancer load, among others. For this reason, the CNCR database undergoes regular quality audits, which include the assessment of data quality and credibility. The following part introduces the main principles and results of these audits. The Czech National Cancer Registry (CNCR) is based on the following fundamental principles, which clearly define its position and informational value: 1. CNCR is a nationwide registry focused on epidemiological data, i.e. it is not a collection of specific diagnostic and therapeutic records, which would only reflect development within individual diagnoses. 2. CNCR clearly defines the regional affiliation of each new record, assigning it unequivocally to a specific health care facility which is responsible for the patient's treatment, or later to a follow-up health care facility. These measures have made the registry ready for possible conversions should any changes in the administrative division occur, and have made it possible to sort the records by the type of health care facility

21 3. CNCR puts accent on the correctness and completeness of the diagnostic records. In most solid tumours, this involves particularly the identification of diagnosis and the location of primary tumour, as well as TNM and stage determination. 4. Records within CNCR are regularly checked and updated, particularly with data on patient's death. Each year, data within CNCR undergoes quality and completeness checks. Chronological order is checked as the minimum attribute of data consistency, and records are further checked with regard to their correctness and completeness. In the stage of input audit, incomplete or clearly incorrect records are marked as irrelevant for further processing, and are subsequently checked and updated. If a record cannot be completed retrospectively, it is marked by a code denoting that this record has various weights in various situations: summary code, defining the overall credibility and usability of the record, for example: a complete record, fully usable incomplete record for objective reasons, otherwise consistent record with serious problems and of compromised quality, incorrect and unusable record code denoting missing values in primary record code denoting record in which some values have been omitted due to their inconsistency code denoting the main reasons of record incompleteness (objective reasons / error) As explained above, the data audit detects all situations and reasons that might lead to incomplete diagnostic identification of records. However, many reasons are objective, and they often bring valuable information if detected properly. Here is a list of possible reasons for incomplete records: 1. In some diagnostic groups, the TNM classification has no sense, i.e. the incompleteness of those records is actually correct (in the Czech health system, this situation has been historically caused by the universal CNCR reporting form, which thus excluded the diagnostic identification of leukaemias, for example). 2. In other diagnostic groups, the respective TNM classification was not yet defined at the time of diagnosis. 3. Incomplete diagnostic identification due to diagnosis based on post-mortem examination (DCO records) or autopsy 4. Incomplete diagnostic identification due to early death of patients (shortly after diagnosis), or patients without the start of anticancer therapy. 5. Incomplete diagnostic identification due to unexpected development of a situation, moving the patient abroad, patient s refusal to treatment, etc. 6. Actually wrongly incomplete or otherwise problematic records which do not belong to any of the preceding categories. The above-mentioned coding system allows data administrators to sort the records by various degrees of incompleteness, without being obliged to remove the original records from registry: Records containing at least the diagnosis of malignant tumour can be included into basic epidemiological summarizations. Records containing the clinical stage only or TNM only can be retrospectively supplied with the missing item, using the rules which were valid at the time of diagnosis, and based on expert opinions. Records missing the clinical stage and TNM, whether this is wrongly incomplete or for objective reasons might be classified on the probability basis

22 stage or TNM available TNM and stage available Stage available only In some diagnoses, TNM only or stage only is available due to the status of TNM classification at that time (e.g., malignant tumours of the urogenital tract in ). In other cases, the record is incomplete and can be updated retrospectively, based on TNM version at that time and expert opinions. All patients TNM available only stage and TNM not available TNM and stage not available because classification was not introduced Diagnosis based on autopsy / DCO Early death of patient Patient was not treated Incomplete record Stage and TNM is determined by the TNM classification valid at the time of diagnosis. The incompleteness of records (with respect to TNM and stage) in some diagnoses is due to the absence of rules of TNM and stage determination. Other objective reasons for incompleteness of records are related to the patient's condition at the time of diagnosis, or the patient's refusal to be treated, etc. All these objective criteria having been taken into account, there are also some truly incomplete records, where TNM classification and stage are missing. Figure 2 Basic classification of record completeness in the Czech National Cancer Registry Figure 2 presents the basic scheme of stratification of CNCR records with regard to their completeness in the diagnostic section. This scheme has been designed according to the above-described and justified categories in an effort to differentiate partly incomplete records from records incomplete for objective reasons, and from records which must be marked as incorrect. Figure 3 presents the assessment of all CNCR records. There is a clear improvement in the quality of CNCR data over time, which cannot be justified by the development in TNM classification only, but also by the development of the CNCR itself. Nevertheless, the development in TNM classification is a very important factor: recent data contains 6.3% of incomplete records due to missing TNM classification. One must bear in mind, however, that even nowadays, there are specific groups of neoplasms for which the TNM system has not been defined. These diagnoses are objectively incompatible with the majority standard for the registration of solid tumours, and so should be assessed separately (haemato-oncological malignancies, malignant tumours of the brain and the central nervous system, in situ tumours, rare tumours, unspecified sites). Provided that the clinical stage can be supplied in records containing TNM only, then particularly data from the period are fully comparable with the highest-quality international data, as regards completeness of records. With respect to the incidence rates of DCO records and the findings during autopsy, such recent data is also acceptable from the international point of view. Figure 3 also documents the decreasing proportion of untreated patients and patients who died soon after diagnosis (early deaths). "Early death" in our analysis corresponds to death within one month of diagnosis, which corresponds to the criteria used in international analyses

23 Overall, CNCR records from the period contain only 5.5% records which are unfoundedly missing both the TNM classification and the clinical stage (Figure 3). A very positive message is that the completeness of data increases in time and that the most recent period (which is the most relevant one for clinical analyses) provides high-quality data. In the period , only 4.9% of records were detected to be unfoundedly incomplete (Table 1). Furthermore, it is obvious that since 1995, the problem of incomplete diagnostic classification applies much less to the most frequent diagnoses, such as the breast cancer, colorectal cancer and lung cancer. Each year, the audit focusing on the correctness of CNCR data results in a dataset of credible records with properly defined diagnosis, and a dataset of incomplete records with clearly specified objective causes of the problems. Similar filtration procedures are employed by most population-based studies which monitor the survival of cancer patient with regard to their diagnosis or treatment. Even records lacking information on the clinical stage can be employed for routine assessments of incidence and mortality rates, because the diagnosis of malignant tumours is specified. On the other hand, these records can be omitted for the purpose of health care assessment, without making any systemic distortion or reducing the information value of outputs. Table 1 provides an overview of the results of CNCR data audit for the recent period

24 All CNCR records Stage and/or TNM available TNM and stage available Stage available only TNM available only TNM and stage not available because of missing classification % 6.4% 3.0% 7.1% 5.5% 3.4% 10.4% 8.4% 6.5% 10.4% 40.8% 27.5% 22.0% 33.7% 52.0% 70.7% Stage and TNM not available Diagnosis based on autopsy / DCO Early death of patient Patient was not treated % 5.2% 4.5% 4.8% 7.4% 4.3% 2.9% 4.0% 13.4% 6.5% 3.5% 5.8% Stage and TNM not known % 6.4% 5.5% 5.5% 100% TNM 2nd edition Validity of TNM classifications in CNCR data TNM 3rd TNM 4th edition edition TNM 5th edition TNM 6th edition 80% 60% 40% 20% 0% Figure 3 Classification of records on malignant neoplasms (C00-C97) in CNCR with respect to completeness of data on the stage of disease. Year

25 Table 1 Analysis of CNCR records from the period with respect to the availability of clinical stage and TNM Stage or TNM available Stage and TNM available Diagnostic group of malignant tumours N Stage and TNM available Stage available TNM available TNM classification not introduced Diagnosis based on autopsy/dco Early death of patient Patient was not treated Stage and TNM not available unfoundedly Oral cavity and pharynx (C00-C14) 8, % 4.0% 1.0% 0.5% 2.3% 3.1% 2.7% 7.1% Oesophagus (C15) 3, % 4.2% 1.7% 5.8% 7.2% 10.0% 6.2% Stomach (C16) 11, % 3.1% 1.6% 5.4% 7.7% 9.1% 4.7% Colon and rectum (C18-C21) 55, % 2.5% 1.2% 3.1% 3.6% 3.4% 4.1% Liver and intrahepatic bile ducts (C22) 5, % 3.6% 3.6% 18.9% 15.1% 13.5% 4.8% Gallbladder and biliary tract (C23,C24) 6, % 4.5% 3.1% 9.8% 12.0% 10.2% 6.4% Pancreas (C25) 12, % 3.5% 3.6% 10.0% 11.3% 11.2% 4.4% Larynx (C32) 3, % 2.6% 1.0% 2.3% 2.0% 2.0% 4.6% Bronchus and lung (C33,C34) 43, % 2.9% 2.1% 0.1% 7.0% 7.2% 5.1% 3.9% Melanoma of the skin (C43) 12, % 2.3% 1.4% 0.5% 0.7% 0.5% 5.0% Other malignant neoplasm of the skin (C44) 108, % 0.6% 3.5% 0.1% 1.5% Connective and soft tissues (C47,C49) 1, % 5.2% 8.0% 22.8% 2.3% 3.5% 3.1% 1.9% Breast women (C50) 40, % 2.6% 0.6% 1.3% 1.7% 1.4% 4.0% Cervix uteri (C53) 7, % 8.8% 1.1% 1.1% 1.2% 2.7% 6.9% Uterus (C54,C55) 12, % 5.8% 0.8% 0.8% 1.5% 1.5% 3.4% 24.8% Ovary (C56) 8, % 6.6% 1.6% 3.4% 3.4% 2.7% 7.2% Prostate (C61) 29, % 5.4% 1.3% 2.8% 2.1% 7.3% 9.6% Testis (C62) 3, % 3.6% 0.6% 0.5% 0.3% 0.3% 6.0% Kidney (C64) 18, % 3.2% 2.3% 6.2% 3.5% 3.8% 5.6% Bladder (C67) 16, % 5.8% 1.0% 1.5% 2.0% 2.5% 11.2% Brain and spinal cord (C70-C72) 5, % Thyroid gland (C73) 4, % 3.3% 1.1% 3.1% 1.1% 1.1% 11.1% Hodgkin s lymphoma (C81) 1, % Non-Hodgkin s lymphoma (C82-C85,C96) 8, % Multiple myeloma (C90) 3, % Leukaemia (C91-C95) 8, % Other malignant tumours 19, % 22.1% 2.3% 7.7% 7.4% 12.0% 8.5% 6.3% Malignant tumours in total 461, % 3.5% 2.0% 6.3% 3.0% 3.3% 3.3% 4.9% Source: Czech National Cancer Registry (CNCR)

26 3.5 The Czech National Cancer Registry On-line Data from the Czech National Cancer Registry can be viewed and analyzed by anyone thanks to the project SVOD (System for Visualization of Oncology Data), which is available on-line at The objective of this comprehensive website is to provide a set of useful tools for performing analyses of several data sources which are automatically aggregated according to the user's choice (cancer epidemiology, demographic data, data on the risk status of the Czech population, etc.). All analytical tools provide clear outputs in the form of graphs or tables that are easy to understand; based on the results, the user can further specify or modify his/her analysis by setting a number of parameters to obtain the desired output. The software tools on this portal are primarily aimed for health care managers as well as professionals working in the field of human and ecological risk assessment. All graphical outputs (e.g. trends in incidence and mortality rates) have been prepared in a very safe way to be widely accessible to the general public. Automated analyses represent the core functions of the entire system. These analyses can be very straightforward, the results being displayed after just several mouse clicks, or they can be rather elaborate, depending on the users' experience and professional knowledge. Thanks to these software tools, anyone can analyze the epidemiological trends over the last three decades, stratify and filter cohorts of patients, and assess population risks in absolute or ageadjusted numbers. Some of the tools even offer stratification of cancer cases according to the clinical stage or TNM. Major epidemiological trends can be readily compared with the international data (GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide). The portal is equipped with the following automated analytical tools: Incidence and mortality: overall time trends in incidence rates, mortality rates, and mortality/incidence ratio. The user can choose from the absolute numbers of cases, crude rates (number of cases per 100,000 population) or age-standardized rates (ASR): European ASR (ASR-E) or World ASR (ASR-W) is available. Time trends: changes in time trends in the incidence and mortality rates over time. The user can choose from the growth indices relevant to a specific year, or year-on-year changes. Both parameters can be viewed as absolute numbers or as proportional changes (expressed in percents). Regional overviews: comparison of the incidence and mortality rates in individual regions of the Czech Republic. The user can choose from the crude rates and age-standardized rates, and the graphical output can be displayed either as a map or as a bar chart. Age of patients: age structure of living and deceased patients diagnosed with a specific type of cancer. The user can choose from the absolute numbers of cases, recalculation per 100,000 population (age-specific rate), and percentage of cases according to age categories (age structure). Clinical stages: time trends in the proportion of cancer patients diagnosed at different clinical stages. The user can choose from the absolute numbers, percents, and crude rates (recalculation per 100,000 persons); the graphical output can be displayed as bar chart, line graph or pie chart over a selected time period. Comparison with foreign countries: comparison of incidence and mortality rates in the Czech Republic with international data. All these analyses are based on data obtained from the IARC database GLOBOCAN Comparative analyses: time trends in the incidence or mortality rates in selected region(s) in comparison with the overall epidemiological situation in the Czech Republic. Summary presentation: a comprehensive overview of the basic analyses for individual diagnostic groups

27 4 Czech clinical registries monitoring the use of monoclonal antibodies in cancer therapy Apart from population-based data, the Czech Society for Oncology has been using a set of clinical registries to collect real clinical data on cancer patients in the form of medical documentation. The principal objectives of these projects involve monitoring and retrospective evaluation of the treatment results and safety in areas where treatment effectiveness and health care improvements are of particular importance. Special attention is paid to the modern treatment approaches based on monoclonal antibodies, which are specifically targeted at cancer cells. These clinical registries have been developed as noninterventional studies within the Czech National Cancer Control Programme ( and have been running in compliance with the Czech legislation in force, meeting all requirements with respect to security of the collected data (for more details, see The primary and secondary objectives of clinical registries are listed below. Primary objectives of clinical registries Monitoring the number of patients treated with monoclonal antibodies Assessing the treatment safety in terms of standardized toxicity scoring Assessing the treatment effectiveness in terms of treatment response and survival Secondary objectives of clinical registries Analysis of patients' survival in relation to the monitored clinical parameters Analysis of data from patients treated with monoclonal antibodies in relation to the reference population data The network of clinical registries is well organized and fully functional, covering all main Czech health care facilities dealing with cancer patients. Data collection is representative enough, bringing significant added value to both the doctors and patients: Clinical registries inform the public about the incidence rates of a given cancer type (or disease in general) and provide informational support to public health care facilities and physicians. Oncologists get a better overview of their patients and receive regular reports on the number of patients involved in the entire project. Due to the overall assessment of nationwide data, health care providers obtain statistically relevant information on the treatment results. Moreover, a clinical registry can promote more effective exchange of information and experience among doctors. The clinical registries are based on electronic databases which meet high security standards. The security of individual records within the registry is guaranteed via de-identified data collection. Each patient's identity is replaced with a number (ID) which does not allow any retrograde identification of that person. Data security is further ensured via user accounts and encrypted communication channels. Authorized users can only access the system after entering a valid username and password. Users can be assigned various levels of authorization so that they have access to selected functions or parts of the system. An encryption protocol is used for data transfer between the user and central database to prevent tapping into the communication between the client and server. For this reason, any communication between

28 the client and server is realized via the secure protocol HTTPS, using the SSL (Secure Socket Layer) encryption. The Czech Society for Oncology is currently running several clinical registries (Table 2 gives an overview of the most important projects). Each project is equipped with its own database which has a specific parametric structure. The system is rather flexible, making it possible to create subprojects or collect data focused on specific protocols. As regards monitoring of monoclonal antibodies in cancer therapy, the expert teams within the Czech Society for Oncology have more than 7,000 records at their disposal; all of them being stored, checked and validated in a total of seven databases (see Table 2.). Such amount of data, combined with an effective information platform, makes it possible to perform representative analyses of the treatment results, stratified survival analyses, as well as research analyses of the prognostic factors. Table 2 The principal Czech clinical registries focused on cancer therapy with monoclonal antibodies (status in June 2010) Project name Description No. of records Project website Avastin Clinical registry of patients with advanced colorectal carcinoma, breast carcinoma or lung 1,960 carcinoma treated with Avastin. Erbitux Clinical registry monitoring the treatment effectiveness of Erbitux, a targeted biological therapy used in the treatment of colorectal cancer and locally advanced head and neck cancer. GIST Clinical registry for the collection of epidemiological and clinical data on patients with gastrointestinal stromal tumours (GIST). Herceptin Clinical registry of breast cancer patients treated with Herceptin. 1,947 interb International clinical registry of ErbB2-positive breast cancer patients treated with lapatinib RenIS Clinical project for the collection and analysis of data on patients with malignant neoplasm of the kidney. Tarceva Clinical registry of patients with advanced nonsmall cell lung carcinoma or pancreatic 1, carcinoma treated with Tarceva. 1 patients from the Czech Republic only; 2 patients from the Czech Republic only, results from

29 5 Use of population-based data for cancer care assessment in the Czech Republic The Czech National Cancer Registry is a database maintained over a long period which contains a number of details that may be used for nationwide assessment of cancer care. These outputs may be used to set population-based reference standards, which can be compared with the partial values from regions or hospitals. In particular, CNCR provides the following information: Epidemiological overviews: Cancer incidence according to the tumour site (diagnosis: topographical code) Cancer incidence according to histological type (diagnosis: morphological code) Cancer prevalence Mortality from specific cancer types International epidemiological comparisons based on age-standardised incidence or mortality rates (in particular, standardisation to world population, ASR-W) Data related to individual records: Diagnosis of malignant tumour, described by standard classification systems Extent of the disease at the time of diagnosis (TNM, ptnm, clinical stage, risk group ) Diagnostic method of cancer detection Basic dates: Date of first visit to physician Date of diagnosis (Date of the first treatment procedure) Date of the last follow-up or date of death Patient s social status (education, current or past employment) Patient s age at the time of diagnosis, patient s sex Data on follow-up health care facility Other diagnoses including other malignant tumours Patient s clinical condition according to WHO criteria Specific analyses which can be employed in health care assessment: Comparison of the changes in epidemiological parameters of tumours according to diagnostic codes (topographical and morphological), according to regions or other classification criteria Monitoring the impact of changes in the society and medicine via changes in the extent of disease. For example, higher proportion of less advanced stages in individual diagnoses reflect improved diagnostic methods, better physician vigilance and education related to cancer, more effective prevention, etc. Comparison of all monitored parameters values sorted by individual health care facilities or by types of health care facilities Statistical (indirect) monitoring of potential delays or problems in diagnosis of specific cancer type (as follows from information on the first visit to physician and the date of diagnosis) Comparison of changes in mortality from individual cancer types and subsequent assessment of the overall effectiveness of the diagnostic and treatment process Monitoring of overall absolute or relative survival rates Region-specific analyses which can indirectly imply other associations, e.g. changes in the environmental parameters, etc

30 Specific analyses which can be employed in health care planning: Using the incidence and mortality curves to estimate the development of epidemiological load in future and long-term trends. Specific data on clinical stage is also available in these analyses. Population metrics to monitor the performance and effectiveness of implemented screening programmes Reference values of the overall mortality, as well as survival analyses, rank among the most valuable outputs of CNCR. Epidemiological results give evidence of the stabilised cancerrelated mortality in the long-term for the Czech Republic. Some of the malignancies have even seen a downward trend in mortality despite the continuously increasing incidence, which has confirmed the long-term improvements in cancer care delivered in the Czech Republic. The nationwide reference standards of the short- and long-term survival of cancer patients in the Czech Republic were evaluated. Table 3 shows an example of the relative survival assessment in a form which is available for all cancer diagnoses in the Czech Republic. This example shows that the achieved probability of survival is high, particularly in less advanced clinical stages. Moreover, it is evident that cancer care results have significantly improved over time. However, the prospect of long-term survival is rather poor in clinical stage IV. Unfortunately, Czech oncology still has to deal with a very high incidence of primarily diagnosed metastatic cancer, which significantly decreases the overall survival results if all clinical stages are assessed together. Table 3 Example of the reference values for the assessment of cancer patients survival in the Czech Republic 1 5-year survival rates of treated patients in different periods (95% confidence intervals) Diagnosis / clinical stage (CS) Comparison of two precedent periods 2 Comparison of two recent periods from which data is available Breast cancer CS I 92.0 ( ) 96.6 ( ) 96.5 ( ) 98.6 ( ) CS II 80.6 ( ) 84.5 ( ) 85.4 ( ) 87.0 ( ) CS III 53.8 ( ) 57.4 ( ) 59.3 ( ) 64.6 ( ) CS IV 21.1 ( ) 22.8 ( ) 24.3 ( ) 27.7 ( ) All CS 70.1 ( ) 77.6 ( ) 79.6 ( ) 83.5 ( ) Colorectal cancer CS I 66.1 ( ) 78.1 ( ) 81.2 ( ) 86.0 ( ) CS II 49.8 ( ) 64.5 ( ) 67.2 ( ) 71.0 ( ) CS III 41.6 ( ) 42.1 ( ) 45.7 ( ) 50.0 ( ) CS IV 11.8 ( ) 10.9 ( ) 11.7 ( ) 12.5 ( ) All CS 49.0 ( ) 53.3 ( ) 54.8 ( ) 57.0 ( ) 1 Reference values are illustrated on the example of two most common cancer diagnoses. Assessment based on CNCR records: treated cancer patients with verified cancer diagnosis. 2 Cohort analysis of patients diagnosed in a specific time period. 3 Period analysis; information on patients survival diagnosed recently has been included in the calculation

31 References Binding instructions of the National Health Information System (NHIS): Czech National Cancer Registry instruction for the contents of data structure, version /2, Institute of Health Information and Statistics of the Czech Republic (IHIS), Prague On-line version: section IHIS, part Binding instructions Cancer Incidence in the Czech Republic. Institute of health information and statistics of the Czech Republic. Available from: d=1&mnu_id=5300 Capocaccia D., Gatta G., Roazzi P., Carrari E., Santaquilani M., de Angelis R., Tavilla A., Eurocare Working Group: The EUROCARE-3 database: methodology of data collection, standardization, quality control and statistical analysis. Annals of Oncology, Supplement 5: v14-v27, Clegg L.X., Feuer E.J., Midthune D.N., Fay M.P.: Impact of reporting delay and reporting error on cancer incidence rates and trends. J. Nat. Cancer Inst., 94 (20), , Coordination Centre for Departmental Medical Information Systems (CCDMIS), Czech National Cancer Registry (CNCR), [ ], available on Demographic data of the Czech Republic and Death Records Database of the Czech Republic, Czech Statistical Office Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Výzula R. Epidemiology of malignant tumours in the Czech Republic [online]. Masaryk University, [2005], [cit ]. On-line available: Version 7.0 [2007], ISSN Dusek L., Muzik J., Kubasek M., Koptiková J., Snajdrova L., Ondrusova M. (2007) National portal of epidemiology of malignant tumours in the Slovak Republic [online]. Masaryk University, Brno, Czech Republic [2007]. ISBN Dušek L. et al. Czech cancer care in numbers Praha: Grada Publishing, a.s., 1st ed ISBN European health for all database (HFA-DB). World Health Organisation [cit ] Ferlay J., Bray F., Pisani P. and Parkin D.M. GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004, ICD-O-3: International Classification of Diseases for Oncology, 3rd Edition, World Health Organization, 2000; MKN-O-3: Mezinárodní klasifikace nemocí pro onkologii, T etí vydání, eská verze, Ústav zdravotnických informací a statistiky R, Praha 2004, ISBN Institute of Health Information and Statistics of the Czech Republic (IHIS), National Health Information System (NHIS) - legislation, [ ], available on Institute of Health Information and Statistics of the Czech Republic (IHIS), National Health Information System (NHIS), Czech National Cancer Registry, [ ], available on International Statistical Classification of Diseases and Health Related Problems (The) ICD-10 Second Edition, World Health Organization, 2005, ISBN International Union Against Cancer (UICC): TNM Classification of malignant tumours. 3rd ed. M.H. Harmer, ed. Geneva, Enlarged and revised TNM klasifikace zhoubných novotvar, 3. vydání, eská verze, Ústav zdravotnických informací a statistiky R, Praha International Union Against Cancer (UICC): TNM Classification of malignant tumours. 4th ed. P. Hermanek, L.H. Sobin, eds. Berlin, Heidelberg, New York: Springer Verlag, Revised 1992.; TNM klasifikace zhoubných novotvar, 4. vydání, 2. revize 1992, eská verze, Ústav zdravotnických informací a statistiky R, Praha 1994 International Union Against Cancer (UICC): TNM Classification of malignant tumours. 5th ed. Sobin LH, Wittekind Ch (ed.), New York, Wiley-Liss. 1997; TNM klasifikace zhoubných novotvar,

32 vydání 1997, eská verze, Ústav zdravotnických informací a statistiky R, Praha 2000, ISBN International Union Against Cancer (UICC): TNM Classification of malignant tumours. 6th ed. Sobin LH, Wittekind Ch (ed.), New York, Wiley-Liss. 2002; TNM klasifikace zhoubných novotvar, 6. vydání 2002, eská verze, Ústav zdravotnických informací a statistiky R, Praha 2004, ISBN Jemal A., Clegg L.X., Ward E. a kol.: Annual report to the nation on the status of cancer, , with a special feature regarding survival. Cancer, 101(1), 4 27, Micheli A., Baili P., Quinn M.: EUROCARE Working Group. Life expectancy and cancer survival in the EUROCARE-3 cancer registry areas. Ann Oncol, 14 (Suppl 5): v28-v40,

33 B. Author s contribution to the development of data standards for the evaluation of Czech cancer care

34

35 All newly diagnosed neoplasms have been collected in the National Cancer Registry of the Czech Republic (NCR) since In 2004 they represented over 1.2 million completed records of cases diagnosed in period Vast majority of outputs available from this large data set was focused only on basic description of epidemiological situation in the Czech Republic and its regions without any further information contributing to the assessment of quality of cancer care. This situation led to a closer cooperation between the Institute of Health Information and Statistics of the Czech Republic (IHIS) and the Czech Society for Oncology (CSO) and in 2004 the complete data set was provided to the CSO. At this time the close cooperation started between the Institute of Biostatistics and Analyses of Masaryk University (IBA) and CSO. IBA was commissioned to perform data processing, detailed analyses and expert utilization. Jan Mužík was responsible for most of the work in this field and under the supervision of Ladislav Dušek, Ph.D., Director of IBA and members of Steering Committee of CSO he contributed to the development of data standards for evaluation of Czech cancer care in these areas: - standardization of information sources related to cancer care - evaluation of cancer epidemiology and population burden - assessment of cancer management and therapeutic outcomes Standardization of information sources related to cancer care 1. Dušek L., Mužík J., Koptíková J., Žaloudík J., Klimeš D., Bourek A., Indrák K., Mihál V., Hajdúch M., Št rba J., Vyzula R., Abrahámová J. Data registries form indispensable information base of current oncology. Klinická onkologie 20, Supplement 1/2007, 53-62, ISSN X (original article in Czech) Summary: This paper comments methodical principles of data registration in oncology. Cancer registries should be accepted as an indispensable source of valuable information for evaluation of anticancer therapy. Fully functional population-based registration, however, requires control of all determining factors, namely data model, technological background and data quality assurance. The parametric structure of general population-based registries should be minimized, primarily focused on epidemiology with only limited number of clinical entries. Such population-based registry should be operated automatically, through direct exporting of data from hospital information systems. On the other hand, reasonable registration of cancer data for clinical conclusions must include following key components, that cannot be obtained universally for all diagnoses: (1) risk typology of newly diagnosed cases, (2) list of diagnostic and therapeutic procedures and reached therapeutic response, (3) time plan of follow-up, (4) survival monitoring, (5) time and cause of death. None of these items could be omitted if the registry is targeted for clinical interpretation. Such information standard can be guaranteed only in specialized clinical registries supervised by expert societies. 2. Mužík J., Koptíková J., Dušek L., Žaloudík J., Vyzula R., Abrahámová J. Historical data of Czech National Cancer Registry: information value and risk of bias; Klinická onkologie 20, Supplement 1/2007, 63-76, ISSN X (original article in Czech) Summary: The paper is focused on quality assessment of population-based cancer registries. The analyses are methodically based on the database of the Czech National Cancer Registry (NCR) that offers valuable model with epidemiological data of all diagnostic groups of cancer, all collected continuously since Quality control is necessary for applications of NCR in health care evaluation and for international

36 presentation of Czech cancer epidemiology. The audit resulted in a set of diagnostically completed and verified records and in a set of records with incomplete diagnostics (namely TNM) due to well distinguished objective reasons (DCO cases, cases diagnosed at autopsy, early death after diagnosis, etc.). All objective reasons were recognized and the database was enriched by codes that identify the status of the record and its reason. In addition, we found 5.1 % of cases with wrong and/or incomplete diagnostics and without any acceptable explanation. These cases typologically correspond to the whole NCR database and they can be therefore excluded from serious analyses with a minimum risk of bias. Increased probability of bias due to erroneously incomplete records exists only in region-specific comparisons. Contribution of author in this area [1, 2] included development of standardized management of the large NCR data set, identification of parameters crucial for stratification of cases and evaluation of cancer care results. In the next step author developed algorithms for completing information on disease stage from available TNM components and algorithms stratifying NCR records into groups according to their information value for cancer care assessment. Very important development covered mastering of classification rules used in cancer registration (TNM classification, ICD and ICD-O classification) including the changes between different editions of classification. Finally, reference data sets for further analyses were defined. Evaluation of cancer epidemiology and population burden 3. Mužík J., Dušek L., Koptíková J., Fínek J. Cancer Epidemiology in the Czech Republic: Current Load and the Time Trends. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN Summary: From the point of view of malignant neoplasms, the Czech Republic ranks among the most afflicted countries in Europe and worldwide. All neoplasms, with the exception of non-melanoma skin cancer (C44), have a total incidence rate of almost 500 cases per 100,000 population per year, and the annual mortality rate exceeds 270 deaths per 100,000 population. Although the mortality rate has stabilized since 1995 and the trend has even shown a downward course in some diagnoses, the prevalence of malignant neoplasms remains consistently high due to increasing incidence rate. In 2005, the prevalence rate of malignant diseases (apart from C44) exceeded 270,000 persons in the Czech Republic. The most frequent diagnoses include colorectal cancer, lung cancer, breast cancer in women and prostate cancer in men. The chapter provides an overview of the epidemiological trends, as well as an analysis of the data from the Czech National Cancer Registry, documenting the ever higher quality of records within that database. The epidemiological characteristics of malignant tumours in the Czech population are available on-line at This chapter in internationally distributed book comprehensively summarizes current situation in cancer epidemiology of the Czech Republic. The position of the Czech Republic is described in international context, incidence, mortality and prevalence trends, cumulative incidence risk, age distribution of the patients, regional distribution of the patients and trends of proportion of disease stage in newly diagnosed cases. This set of epidemiological information represents basic comparative standard for evaluation of cancer disease burden in population. Many other outputs focused on evaluation of epidemiology of given cancer diagnoses in the Czech Republic were subsequently published in Czech scientific journals, but are not mentioned in this thesis

37 4. Zavoral M., Suchánek Š., Závada F., Dušek L., Mužík J., Seifert B., Fri P. Colorectal cancer screening in Europe. World J Gastroenterol, Beijing, China, the WJG Press and Baishideng. vol. 15, no. 47, s Summary: Colorectal cancer (CRC) is the second most frequent malignant disease in Europe. Every year, people are diagnosed with this condition, and patients die of it. In 2003, recommendations for screening programs were issued by the Council of the European Union (EU), and these currently serve as the basis for the preparation of European guidelines for CRC screening. The manner in which CRC screening is carried out varies significantly from country to country within the EU, both in terms of organization and the screening test chosen. A screening program of one sort or another has been implemented in 19 of 27 EU countries. The most frequently applied method is testing stool for occult bleeding (fecal occult blood test, FOBT). In recent years, a screening colonoscopy has been introduced, either as the only method (Poland) or the method of choice (Germany, Czech Republic). In this work author contributed with analysis of epidemiology of colorectal cancer in the European countries. Techniques and standard procedures developed for evaluation of cancer epidemiology from the NCR data of the Czech Republic were used and were applied on available international data of colorectal cancer. 5. Mužík J., Faber E. Epidemiology of chronic myeloid leukemia. In Chronická myeloidní leukémie. 1 st edition. Praha : Galén, ISBN , s (original article in Czech) Summary: Not many epidemiological studies have been published on chronic myeloid leukemia (CML). In statistical studies and reviews dedicated to the occurrence of malignant diseases in general, patients with CML are often included in the category of myeloproliferative disorders, myeloid or chronic leukaemia. Even more often, it is usually not specified whether the diagnosis has been confirmed by cytogenetic or molecular tests of Ph chromosome or BCR-ABL gene presence or merely by blood and bone marrow tests, and Ph-positive CML is reported together with Ph- and BCR-ABL negative cases. CML is usually reported to constitute 15-20% of all leukaemia cases. Total incidence figures differ widely and range from 0.6 to 2.0 new cases per persons per year. Unlike lymphoproliferative conditions, whose numbers have increased greatly in the last few years, CML incidence has not according to some sources significantly changed in the last few decades. This work represents example of evaluation of population burden by one of the cancer types, for which there are not available adequately reliable data sources. CML suffers from lack of population data and therefore any easy accessible comparative standard for evaluation of cancer disease burden in population could not be used. Reliability of the data sources is closely related to the recorded information about specific diagnostic definition of the disease and to the epidemiological representativeness of the data set. Author newly aggregated numerous data sources and used NCR data, data of clinical registries and international epidemiological data for estimation of CML incidence in the Czech Republic

38 Assessment of cancer management and therapeutic outcomes 6. Dušek L., Pavlík T., Koptíková J., Mužík J., Gelnarová E., Žaloudík J., Vyzula R., Hajdúch M., Abrahámová J. Czech National Cancer Registry and reference standards for health care assessment. Klinická onkologie 20, Supplement 1/2007, 77-95, ISSN X (original article in Czech) Summary: The paper examines value of the Czech National Cancer Registry (CNCR) for the health care assessment. More than 1.3 millions of records collected since 1977 were audited from the viewpoint of comprehensiveness and correctness. The audit proved steadily increasing quality of the CNCR in time, diagnostic error rate (including incorrectly unstaged cases) decreased below 6 % after The most recent CNCR records are therefore fully usable for the evaluation of health care results. For that purpose, reference data set with > valid records was defined as a source of population standards for overall survival modelling. The reference data set covers a recent period ( ) and is clinically relevant (it contains only fully diagnosed and then treated patients). So called complete analysis of 5-year survival was proposed as optimal for benchmarking of cancer centres against population-based reference. If the benchmarking is supposed to reflect survival experienced by recently diagnosed patients, the period analysis was proposed for such more up-to-date assessment (Brenner and Gefeller, 1996). All calculated reference standards of overall survival are recommended for self-benchmarking of cancer centres. The values should not be applied for mutual comparisons of hospitals. Main author s contribution in this work was identification of crucial parameters that are important for assessment of the quality and completeness of NCR data and development of algorithms for stratification of the NCR records according to the information value for evaluation of cancer care. Identification of patients with single and multiple tumours was also important, because the multiplicity of malignances could influence results of evaluation of cancer care efficacy if focused on one diagnosis only. Final output of this process was definition of the reference data set as a comparative standard for assessment of therapeutic outcomes in cancer care on a population level. In this work the author also made first attempts to stratify the patients according to the type of facility performing medical treatment. Usage of such type of stratification was strongly limited by the completeness of information about facilities of medical treatment and by the fact that only information about primary cancer care is available in the NCR data. This stratification process was followed by overall survival modelling and led to a final output population benchmarking standards of 5-year survival of cancer patients in the Czech Republic. 7. Dusek L., Abrahamova J., Lakomy R., Vyzula R., Koptikova J., Pavlik T., Muzik J., Klimes D. Multivariate analysis of risk factors for testicular cancer: a hospital-based case-control study in the Czech Republic. Neoplasma Vol.55, No.4, p , 2008 Summary: Growing incidence of testicular cancer around the world stimulates research attempting to explain the trends. This study quantified the contribution of different types of potential risk factors for testicular germ-cell cancer (TGCC) with discrimination between seminoma and non-seminoma. A standardized questionnaire containing demographic data, pre- and perinatal factors, social, lifestyle and occupational parameters was prepared. The data file consists of n = 356 TGCSO (seminoma: n = 195; non-seminoma: n = 161) and n = 317 controls, frequency matched on age to cases. The following factors were significantly associated with the risk of TGCSO in univariate analyses (ORs): atrophic testis (5.3),

39 smoking over 12 pack-yr (4.9), cryptorchidism (2.9), testicular trauma (2.0), birth weight under 3,000 g (1.6), low degree of education (3.0) in correlation with manual occupation (2.3) and finally, overall familial cancer history (1.5) and familial history of breast (1.8) and prostate cancer (3.9). On the other hand, maternal age over 20 yr (OR < 0.4) and moderate recreational sport activity (OR = 0.5) significantly reduced the risk of TGCSO. A significant risk was associated with cryptorchidism (OR = 2.9; 95% CI = ) where orchidopexy was delayed after5 yr of age (OR = 5.2; 95% CI = ). Delayed orchidopexy was associated namely with the risk of seminomas (OR =7.5; 95% CI = ). Only some of the variables were retained in multivariate model for TGCSO as well as for histological subtypes (multivariate adjusted OR for all TGCSO): atrophic testis (5.9), family history of prostate cancer (4.8), cryptorchidism (3.8) and interaction term low degree of education & manual occupation (3.0). Familial history of breast cancer elevated risk of TGCSO and of seminomas (OR: ). Birth weight under 3,000 g was retained in a multivariate model for TGCSO with a borderline significance (OR = 1.67). We could not rule out any type of risk factors, as each one was significantly represented in the final multivariate models. Familial cancer history remained to be a significant risk factor, together with some lifestyle and occupational parameters. This suggests that both environmental exposures and genetic inheritance can play role in the moderation of the risk TGCC. This work serves as an example of usage of specific clinical data for evaluation of risk factors affecting development of cancers. Population-based cancer registries cannot contain such detailed information about anamnesis, risk factors and cancer aetiology, so this type of clinical research is indispensable for cancer risk evaluation. To reach the credibility of the results of this type of clinical research it is important to follow representativeness of such type of data. Contribution of author in this work included validation of representativeness of the group of patients in comparison with the population-based data from NCR and analytical support with specific outputs from NCR data set. Algorithms for evaluation of completeness of the diagnostic data were also used

40 - 40 -

41 C. Author s contribution to the development of software tools for automated analysis and reporting of cancer data

42

43 Development of comparative standards for the evaluation of Czech cancer care based on population data was accompanied by the development of tools, which have made such large and valuable data accessible to the whole community of cancer care specialists in an appropriate and user friendly form. The development resulted in the creation of software named SVOD Software for Visualisation of Oncological Data. The process was performed in two steps: - development of architecture and technological standards of SW tools - implementation of web-based reporting systems Development of architecture and technological standards of SW tools 8. Mužík J., Dušek L., Pavliš P., Koptíková J., Žaloudík J., Vyzula R. Analysis of population cancer risk factors in national information system SVOD. Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J. Summary: Human risk assessment requires analysis of multiple data sources to make correct interpretations of the results. One of these sources is population epidemiologic databases. In this work we introduce System for Visualization of Oncological Data (SVOD, version 6.0) built up over the database of the National Cancer Registry of Czech Republic. This system currently includes accessible data of all malignant diagnoses (C00-C97) from time period (over 1.2 million cases). Such a database represents a great pool of information that can be used in retrospective evaluation of population risk. The first software that we made was System for Visualization of Oncological Data SVOD (last version 6.0). Software SVOD was designed for installation on Windows operation systems (98/ME and above) on personal computer. User could choose one of three levels of analytical tools: - presentation predefined comprehensive set of analytical outputs - data browser enables analysis of individual parameters of the database - expert tools analytical tools fully controlled by user and focused on specific area Most valuable part of the software was expert tools which were focused in three main areas: - epidemiology (incidence and mortality trends, temporal fluctuation of trends, age structure of patients, age specific trends in time, temporal trends of stages, regional data - districts) - comparative analyses (time trends of incidence in regions, age structure in regions, comparison of ASR in regions) - health management (time trends of application of diagnostic and therapeutic methods, application of diagnostic and therapeutic methods according to age, used combinations of diagnostic and treatment methods, survival analysis - Kaplan-Meier analysis) In these expert tools user can select a specific group of patients for the analysis according to diagnosis, sex, age, time period, stage, detailed TNM, region and other specific parameters. The results are in graphical and tabular form and user can choose the units of the output and type of graphs in some cases. Author contributed to the development of this software in the area of database background design and data preparation and in the area of design of analytical tools. Main programmer of this software was Petr Pavliš, Ph.D

44 Over the time demand for the software SVOD dramatically increased, but only basic epidemiological outputs were mostly requested. This led to the decision to make such analytical tool with basic epidemiological outputs accessible on-line via internet network. With agreement of IHIS and steering committee of CSO we started to develop system, which would be freely available to the public. Implementation of web-based reporting systems The epidemiological part of the former software SVOD was transformed into a new technology. An analytical tool with basic epidemiological outputs was developed, which is accessible on-line via internet network. Author contributed to the development of this on-line analytical system by database structure design, preparation of the data of NCR and of the international data (GLOBOCAN 2002) and in the area of design and validation of analytical tools. Main programmer of this software was Miroslav Kubásek, Ph.D. 9. Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Vyzula R. Epidemiology of Malignant Tumours in the Czech Republic [online]. Masaryk University, Czech Republic, [2005], Version 7.0 [2007], Approved by National Technical Library, ISSN Published as Dušek L., Mužík J., Koptíková J., Brabec P., Žaloudík J., Vyzula R., Kubásek M. The National Web Portal for Cancer Epidemiology in the Czech Republic. Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J. Summary: The aim of this article is to inform about newly issued automated web portal focused on population risk analyses related to cancer epidemiology. Portal was built up on very representative database of National Cancer Register (Czech Republic, Ministry of Health Care, standardized collection of cancer data from ) that provides fully representative long-term trends. Nowadays the database consists of more than cases stratified according to main risk factors and diagnostic descriptors including TNM classification of tumours. The automated system of on-line analyses offers unique access to cancer registries, demographic and environmental databases. Portal is principally developed as tool increasing the information potential of risk assessment studies. Portal is available at This portal was launched in September 2005 and is available on the address Data have been updated annually; nowadays a complete database of the NCR for the period (over 1.6 million records) is available. Contribution of author is preparation of new data updates of the system, analytical validation of updated outputs and development and design of new analytical functions. 10. Ondrušová M., Ondruš D., Dušek L., Mužík J.: What is the information availability to the cancer epidemiology data? Iranian Red Crescent Medical Journal 2008, 10(3): Summary: National Cancer Registry of the Slovak Republic, National Health Information Centre, would like to respond to many requests for easy and comprehensible access to the

45 national and international data on cancer epidemiology. The working group created a new analytic web-page called "National portal on cancer epidemiology". All the data are valid, adapted for publications and quotation and the access to the web-page is free for the wide professional public. After the very successful launch of the portal of cancer epidemiology in the Czech Republic we were contacted by representatives of the Slovak National Cancer Registry (NCR) whether it is possible to develop analogous analytical tool for the Slovak NCR as well. The cooperation was successful and the portal was launched in Contribution of the author was the same as in the case of Czech version of the on-line software. Unfortunately, cooperation with the Slovak NCR was stopped in 2010, database of the on-line software was not updated and since 2010 the portal is not accessible. 11. Mužík J., Dušek L., Babjuk M., Kubásek M., Fínek J., Petruželka L. UROWEB - web portal for analysis and visualization of epidemiology, diagnosis and treatment of urologic malignancies [online]. Masarykova univerzita, Brno, Version 1.6d. Reviewed and guaranteed by the Czech Urological Society and the Czech Society for Oncology. Approving by National Technical Library in process. (original article in Czech) Summary: Project Uroweb.cz ( is an information web portal dedicated to all who are interested in malignancies of urogenital system. The primary objective of the project is to build an information portal guaranteed by the Czech Urological Society (CUS) and the Czech Society for Oncology (CSO), which will provide important information about importance of urology in the treatment of urological malignancies in the Czech Republic (CR). Another goal is presentation of existing diagnostic and therapeutic procedures and regional presentations of urological departments. Dominant part of the portal is interactive analytical software that provides detailed epidemiological information for each diagnosis. For kidney, bladder, prostate and testicles thematic interactive epidemiological analyses in following areas were prepared: Epidemiology and population data of CR; International epidemiological data; Regional reports. After success with the portal of cancer epidemiology in the Czech Republic we started to develop new version of on-line analytical tools, which are focused on specific diagnostic group and offers more detailed information compared to original portal. The first result was project Uroweb. Contribution of the author to the development of the Uroweb portal was as follows: - preparation of the data of NCR - for monitoring long-term trends of disease stages it was necessary to derive disease stage according to recorded T, N and M in period of validity of the 3 rd edition of the TNM classification; morphological codes were stratified into appropriate groups - preparation of international data new international data were acquired from the publication Cancer Incidence in Five Continents, Vol IX and were processed for usage in the analytical software - structure and design of all analytical tools on the Uroweb portal - validation of analytical outputs - processing of regular updates of the source database

46 - 46 -

47 Glossary of specific terms Age-specific rate Age-specific rate is calculated as crude rate in given age group. Age structure Age structure represents distribution of age groups in given population. ASR (age-standardised rate) An age-standardised rate (ASR) is a summary measure of the rate that a population would have if it had a standard age structure. Standardization is necessary when comparing several populations that differ with respect to age because age has a powerful influence on the risk of cancer. The ASR is a weighted mean of the age-specific rates; the weights are taken from population distribution of the standard population. The most frequently used standard population is the World Standard Population. The calculated incidence or mortality rate is then called age-standardised incidence or mortality rate (world). It is also expressed per 100,000. Age standards Age World Standard Population 12,000 10,000 9,000 9,000 8,000 8,000 6,000 6,000 6,000 6,000 European Standard Population 8,000 7,000 7,000 7,000 7,000 7,000 7,000 7,000 7,000 7,000 Age Total World Standard Population 5,000 4,000 4,000 3,000 2,000 1, ,000 European Standard Population 7,000 6,000 5,000 4,000 3,000 2,000 1,000 1, ,000 Crude rate Data on incidence or mortality are often presented as rates. For a specific tumour and population, a crude rate is calculated simply by dividing the number of new cancers or cancer deaths observed during a given time period by the corresponding number of person years in the population at risk. For cancer, the result is usually expressed as an annual rate per 100,000 persons at risk. Cumulative risk Cumulative incidence/mortality is the probability or risk of individuals getting/dying from the disease during a specified period. For cancer, it is expressed as the number of new born children (out of 100, or 1000) who would be expected to develop/die from a particular cancer before the given age (mostly 75 years) if they had the rates of cancer observed in the period in the absence of competing causes. Incidence Incidence is the number of new cases arising in a given period in a specified population. This information is collected routinely by cancer registries. It can be expressed as an absolute number of cases per year or as a rate per 100,000 persons per year (see Crude rate and ASR). The rate provides an approximation of the average risk of developing a cancer

48 Mortality Mortality is the number of deaths occurring in a given period in a specified population. It can be expressed as an absolute number of deaths per year or as a rate per 100,000 persons per year. Mortality data are provided by national statistical offices. Prevalence The prevalence of a particular cancer can be defined as the number of persons in a defined population who have been diagnosed with that type of cancer, and who are still alive at the end of a given year, the survivors. Complete prevalence represents the number of persons alive on a certain day who previously had a diagnosis of the disease, regardless of how long ago the diagnosis was, or if the patient is still under treatment or is considered cured. Partial prevalence, which limits the number of patients to those diagnosed during a fixed time in the past, is a particularly useful measure of cancer burden. Survival It is defined as the probability of survival, expressed as time elapsed since diagnosis (1,3 5- year survival). This observed survival probability is influenced by mortality both from cancer of interest and from other causes. For this reason, relative survival is usually calculated. It is defined as the ratio of the observed survival in the group of patients to the survival expected in a group of people in the general population, who are similar to the patients with respect to all possible factors affecting survival at the beginning of the follow-up period, except for the disease of interest. References: Isabel dos Santos Silva. Cancer Epidemiology: Principles and Methods. IARC Press, Lyon, France, No. of pages: ix ISBN CANCERMondial - Glossary of Statistical Terms. International Agency for Research on Cancer, Lyon,

49 Enclosures Part B of the thesis Paper I. Mužík J., Dušek L., Koptíková J., Fínek J. Cancer Epidemiology in the Czech Republic: Current Load and the Time Trends. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN Paper II. Mužík J., Koptíková J., Dušek L., Žaloudík J., Vyzula R., Abrahámová J. Historical data of the Czech National Cancer Registry: information value and risk of bias; Klinická onkologie 20, Supplement 1/2007, 63-76, ISSN X (original article in Czech) Paper III. Mužík J., Faber E. Epidemiology of chronic myeloid leukemia. In Chronická myeloidní leukémie. 1st edition. Praha : Galén, ISBN , s (original article in Czech) Paper IV. Dušek L., Mužík J., Koptíková J., Žaloudík J., Klimeš D., Bourek A., Indrák K., Mihál V., Hajdúch M., Št rba J., Vyzula R., Abrahámová J. Data registries form indispensable information base of current oncology. Klinická onkologie 20, Supplement 1/2007, 53-62, ISSN X (original article in Czech) Paper V. Dušek L., Pavlík T., Koptíková J., Mužík J., Gelnarová E., Žaloudík J., Vyzula R., Hajdúch M., Abrahámová J. Czech National Cancer Registry and reference standards for health care assessment. Klinická onkologie 20, Supplement 1/2007, 77-95, ISSN X (original article in Czech) Paper VI. Zavoral M., Suchánek Š., Závada F., Dušek L., Mužík J., Seifert B., Fri P. Colorectal cancer screening in Europe. World J Gastroenterol, Beijing, China, the WJG Press and Baishideng. vol. 15, no. 47, s Paper VII. Dusek L., Abrahamova J., Lakomy R., Vyzula R., Koptikova J., Pavlik T., Muzik J., Klimes D. Multivariate analysis of risk factors for testicular cancer: a hospital-based casecontrol study in the Czech Republic. Neoplasma Vol.55, No.4, p ,

50 Part C of the thesis Paper VIII. Mužík J., Dušek L., Pavliš P., Koptíková J., Žaloudík J., Vyzula R. Analysis of population cancer risk factors in national information system SVOD. Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J. Paper IX. Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Vyzula R. Epidemiology of Malignant Tumours in the Czech Republic [online]. Masaryk University, Czech Republic, [2005], Version 7.0 [2007], Approved by National Technical Library, ISSN Published as Dušek L., Mužík J., Koptíková J., Brabec P., Žaloudík J., Vyzula R., Kubásek M. The National Web Portal for Cancer Epidemiology in the Czech Republic. Proceedings of the 19 th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J. Paper X. Ondrušová M., Ondruš D., Dušek L., Mužík J.: What is the information availability to the cancer epidemiology data? Iranian Red Crescent Medical Journal 2008, 10(3): Paper XI. Mužík J., Dušek L., Babjuk M., Kubásek M., Fínek J., Petruželka L. UROWEB - web portal for analysis and visualization of epidemiology, diagnosis and treatment of urologic malignancies [online]. Masarykova univerzita, Brno, Version 1.6d. Reviewed and guaranteed by the Czech Urological Society and the Czech Society for Oncology. Approving by National Technical Library in process. (original article in Czech) XII. Appendix to paper I. Mužík J. Summary of cancer epidemiology in the Czech Republic. Published In Czech Cancer Care in Numbers Praha: Grada Publishing, a.s., ISBN (only selected examples of the most prevalent diagnoses) XIII. Curriculum Vitae

51 Paper I. Mužík J., Dušek L., Koptíková J., Fínek J. Cancer Epidemiology in the Czech Republic: Current Load and the Time Trends. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN

52 Cancer Epidemiology in the Czech Republic: Current Load and the Time Trends J. Mužík, L. Dušek, J. Koptíková, J. Fínek From the point of view of malignant neoplasms, the Czech Republic ranks among the most afflicted countries in Europe and worldwide. All neoplasms, with the exception of non-melanoma skin cancer (C44), have a total incidence rate of almost 500 cases per 100,000 population per year, and the annual mortality rate exceeds 270 deaths per 100,000 population. Although the mortality rate has stabilised since 1995 and the trend has even shown a downward course in some diagnoses, the prevalence of malignant neoplasms remains consistently high due to increasing incidence rate. In 2005, the prevalence rate of malignant diseases (apart from C44) exceeded 270,000 persons in the Czech Republic. The most frequent diagnoses include colorectal cancer, lung cancer, breast cancer in women and prostate cancer in men. This chapter provides an overview of the epidemiological trends, as well as an analysis of the data from the Czech National Cancer Registry, documenting the ever higher quality of records within that database. The epidemiological characteristics of malignant tumours in the Czech population are available on-line at 1 Introduction Cancer epidemiology is of an ever growing importance, particularly due to the high incidence rates of malignant tumours. The Czech Republic is no exception in this respect; quite the opposite, the Czech population ranks among the most afflicted countries worldwide. Tens of thousands of new cancer patients are diagnosed in the Czech Republic every year, and there are hundreds of thousands of cancer patients who were diagnosed, treated and followed-up in previous years. Under these circumstances, the analysis of information on cancer patients is of strategic importance, and cancer epidemiology plays a key role in both retrospective and prospective analyses. The Czech oncology is well equipped with population-based data on malignant tumours. A consolidated collection of demographic data is available in the Czech Republic, and the Czech National Cancer Registry (CNCR) has been maintained over a long period, covering 100% of all cancer diagnoses and the entire Czech population. The existence of data itself, albeit in high quality, is not enough. This data needs to be processed, analysed and adequately used. This seemingly banal statement requires an innovative solution, with respect to the volume and complex structure of the cancer data. The CNCR database contains more than 1.5 million records since The only way to make this comprehensive data available is to design an automated analysis which would provide outputs in the final form. Such a solution has been developed and put into operation in the Czech Republic: the Czech National Cancer Registry is equipped with an information system which, among others, provides an analytical web portal accessible to anyone (on-line at ISSN ).

53 Table 1 Overview of number of records in CNCR. Number of records in CNCR Diagnostic group Period Total ( ) Oral cavity and pharynx (C00 C14) 6,957 26,448 Oesophagus (C15) 2,645 9,654 Stomach (C16) 10,394 66,036 Colon and rectum (C18 C21) 47, ,859 Liver and intrahepatic bile ducts (C22) 4,798 19,766 Gallbladder and biliary tract (C23,C24) 6,081 28,956 Pancreas (C25) 10,089 40,279 Larynx (C32) 3,160 14,611 Trachea, bronchus and lung (C33,C34) 36, ,795 Melanoma of the skin (C43) 9,681 29,355 Other malignant neoplasm of the skin (C44) 85, ,975 Connective and soft tissues (C47,C49) 1,520 6,728 Breast (in women) (C50) 32, ,852 Cervix uteri (C53) 6,253 32,033 Uterus (C54,C55) 10,395 42,872 Ovary (C56) 7,420 30,261 Prostate (C61) 22,501 63,214 Testis (C62) 2,455 8,568 Kidney (C64) 15,362 48,705 Bladder (C67) 13,254 44,122 Brain and spinal cord (C70 C72) 4,661 16,999 Thyroid gland (C73) 3,782 11,188 Hodgkin s lymphoma (C81) 1,570 8,725 Non-Hodgkin s lymphoma (C82 C85,C96) 6,826 24,484 Multiple myeloma (C90) 2,611 10,489 Leukaemia (C91 C95) 6,880 29,664 Other malignant neoplasms 16,356 70,305 Neoplasms in situ (D00 D09) 18,640 42,526 Neoplasms of uncertain or unknown behaviour (D37 D48) 5,746 13,888 Non-malignant neoplasms (D10 D36) - - Neoplasms in total 402,506 1,464,357 With the epidemiological data accessible, the cancer load can be assessed throughout the population and for individual regions, the proportion of clinical stages as well as the success rate of early detection can be analysed, and the time trends can then be drawn from all assembled data. In addition, national web portals dedicated to various topics provide essential information background for compliance with the recommendation of the European Council dated 02 November 2003 (2003/878/EC) on cancer screening. Therefore, cancer data does not only serve for retrospective research; it is particularly essential for optimizing projects aimed at the fight against cancer from the very early stage.

54 Table 2 Major changes in CNCR history and in the classification of malignant neoplasms. Implemented in CNCR since Description Methodology of cancer reporting and of data collection establishment of obligatory record keeping on malignant diseases establishment of the Czech National Cancer Registry, start of population-based data collection (IHIS*, Methodology 6a/1976) modifications in cancer reporting (IHIS*, Methodology 24/86) modifications in cancer reporting (NHIS**, Methodology 56) modifications in cancer reporting (NHIS**, Methodology 61) establishment of automated check for the correctness and completeness of diagnostic data and logic relations of parameters modifications in cancer reporting (NHIS**, Methodology 61/2006d1) International Classification of Diseases (ICD) coding of basic diagnosis International Classification of Diseases 8 th revision International Classification of Diseases 9 th revision International Classification of Diseases 10 th revision International Classification of Diseases for Oncology (ICD-O) tumour morphology and topography ICD-O 1976 (translated into Czech in 1981) ICD-O (translated into Czech in 1994) ICD-O (translated into Czech in 2004) TNM classification of malignant tumours (TNM) extent and stage of disease TNM classification, 2 nd edition (1974) TNM classification, 3 rd edition (1978) TNM classification, 4 th edition, 2 nd revision (1992) (issued in Czech in 1994) TNM classification, 5 th edition (1997) (issued in Czech in 2000) TNM classification, 6 th edition (2002) (issued in Czech in 2004) * IHIS Institute of Health Information and Statistics of the Czech Republic ** NHIS National Health Information System Table 3 Sources of data on cancer mortality rates in the Czech Republic. Mortality rates of all malignant tumours except non-melanoma skin cancer (C00 C97 excluding C44) [number of cancer-related deaths per 100,000 persons] Source CNCR * CSO ** CNCR The Czech National Cancer Registry contains data on the cause of death according to the Death Certificate (dg1a, dg1c and dg2). Death of a given person from malignant tumour is then determined depending on accord with the cancer diagnose and the immediate cause of death (dg1a) or the primary cause of death (dg1c). CSO Data on the deceased person is recorded in the Death Certificate; this data is then passed on to CSO for subsequent processing of population-based statistics. In accordance with the international methodology, a single diagnosis is established as the primary cause of death of an individual; this diagnosis is subsequently used in statistical assessments. Slight differences in mortality rates according to CNCR and CSO might be caused by inaccuracies in the determination of a single primary cause of death in CSO data, as well as by possible incompleteness in reporting the causes of death in CNCR. * CNCR Czech National Cancer Registry ** CSO Czech Statistical Office

55 Number of newly diagnosed malignant tumours per 100,000 population Year men women Figure 1 Trend in the incidence rates of malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) in the Czech Republic. Source: Czech National Cancer Registry. Number of newly diagnosed malignant tumours per 100,000 persons in a given age group Patient'sage at the time of diagnosis Period: Figure 2 Age-specific incidence rate of malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) in the Czech Republic comparison among different time periods. Source: Czech National Cancer Registry.

56 Number of newly diagnosed malignant tumours per 100,000 persons in a given age group Year Patient's age at the time of diagnosis: yrs yrs yrs yrs yrs 80 yrs and above Figure 3 Trend in the incidence rates of malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) in the Czech Republic comparison among different age groups. Source: Czech National Cancer Registry. Figure 4 Newly diagnosed malignant tumours: proportion of diagnostic groups according to patient s age (Czech Republic, ). Source: Czech National Cancer Registry.

57 Figure 5 Proportion of diagnostic groups among patients who died of cancer, according to patient s age at the time of death (Czech Republic, ). Source: Czech National Cancer Registry. 2 The Czech National Cancer Registry and records on malignant neoplasms 2.1 The contents of the Czech National Cancer Registry database The Czech National Cancer Registry (CNCR) is a fully functional database that is maintained and updated according to long-term standards, containing records of all malignancies diagnosed in the Czech Republic since 1977, cover ing the entire Czech population. CNCR has been dealt with in detail in Chapter 5; this section is limited to the basic methodical aspects. CNCR has been maintained in standard regime since 1977; nevertheless, during more than three decades, both the CNCR database and the international classification of malignant tumours have undergone major changes. These changes have been recorded and are being considered in all analyses which include long-term trends. Table 6.1 provides a summary of the contents of CNCR database, while the overview of changes in CNCR history is given in Table 6.2.

58 Table 4 Trends in incidence rates of malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) in the Czech Republic. Number of neoplasms Men Women Whole population No. per 100,000 persons Number of neoplasms No. per 100,000 persons Number of neoplasms No. per 100,000 persons , , , , , , , , , , , , Growth index % 21.3% 17.8% Growth index % 38.6% 41.3% Table 5 Causes of death in the Czech Republic in the period Neoplasms (C00 D48) Diseases of the circulatory system (I00 I99) Injury, poisoning and certain other consequences of external causes (S00 T98) Diseases of the respiratory system (J00 J99) Diseases of the digestive system (K00 K93) Diseases of the genitourinary system (N00 N99) Diseases of the nervous system (G00 G99) Endocrine, nutritional and metabolic diseases (E00 E90) Other diseases, disorders and conditions Neoplasms (C00 D48) number of deaths per year 0 14 yrs yrs yrs yrs 65+ yrs Total N = 3,085 N = 6,927 N = 28,019 N = 96,398 N = 407,972 N = 542, % 10.0% 27.8% 41.5% 23.4% 26.6% 2.5% 6.1% 20.2% 33.1% 60.0% 52.1% 18.4% 69.7% 29.4% 7.6% 3.3% 6.3% 4.9% 2.7% 3.2% 4.0% 5.0% 4.7% 1.1% 2.0% 10.5% 7.8% 3.0% 4.2% 0.3% 0.4% 0.9% 1.2% 1.5% 1.4% 10.0% 4.3% 3.5% 1.8% 1.5% 1.8% 1.5% 0.7% 0.9% 1.2% 1.4% 1.3% 53.6% 4.1% 3.6% 1.9% 1.0% 1.6% 0 14 yrs yrs yrs yrs 65+ yrs Total ,560 8,004 19,103 28,854

59 Number of cases per 100,000 men Number of cases per 100,000 women Men trends in incidence and mortality rates incidence rate; 551 new cases per 100,000 men in 2005 mortality rate; 297 deaths per 100,000 men in Year Women trends in incidence and mortality rates incidence rate; new cases per 100,000 women in 2005 mortality rate; deaths per 100,000 women in Year Number of cancer survivors per 100,000 men Number of cancer survivors per 100,000 women Men trend in prevalence rate 2052 cancer survivors per 100,000 men in Women trend in prevalence rate Year 2900 cancer survivors per 100,000 women in 2005 Figure 6 Trends in incidence, mortality and prevalence rates of malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) in the Czech Republic. Source: Czech National Cancer Registry. Year

60 Table 6 The main epidemiological characteristics of malignant tumours in the Czech Republic Oral cavity and pharynx (C00 C14) Oesophagus (C15) Stomach (C16) Men Women Men Women Men Women Number of neoplasms per year ( ) Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) 4,263 2, ,420 2,057 Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 31% 44% 25% 75% 23% 27% Mortality growth index ( ) 23% 6% 18% 4% 26% 37% Age at diagnosis median (25% 75% percentile) 57 (51 65) 64 (54 76) 62 (54 70) 70 (58 80) 70 (60 77) 74 (63 80) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Colon and rectum (C18 C21) Liver and intrahepatic bile ducts (C22) Gallbladder and biliary tract (C23, C24) Men Women Men Women Men Women Number of neoplasms per year ( ) 4,670 3, Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) 2,539 1, Number of deaths per 100,000 persons , ( ) Prevalence number of persons (year 2005) 22,366 18, Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 29% 9% 8% 15% 5% 3%

61 Mortality growth index ( ) Age at diagnosis median (25% 75% percentile) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Colon and rectum (C18 C21) Liver and intrahepatic bile ducts (C22) Gallbladder and biliary tract (C23, C24) Men Women Men Women Men Women 5% 7% 10% 10% 8% 13% 68 (60 75) 72 (62 79) 67 (59 75) 73 (64 80) 71 (62 77) 74 (65 80) Pancreas (C25) Larynx (C32) Trachea, bronchus and lung (C33, C34) Men Women Men Women Men Women Number of neoplasms per year ( ) ,639 1,501 Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) ,268 1,332 Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) , ,629 2,433 Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 27% 28% 1% 23% 2% 45% Mortality growth index ( ) 25% 17% 7% 61% 11% 27% Age at diagnosis median (25% 75% percentile) 67 (58 75) 73 (63 79) 59 (54 67) 60 (54 70) 66 (59 73) 68 (58 76) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002)

62 Melanoma of skin (C43) Other malignant neoplasm of skin (C44) Connective and soft tissues (C47, C49) Men Women Men Women Men Women Number of neoplasms per year ( ) ,541 7, Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) 6,702 8,931 52,825 60,342 1,125 1,112 Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 59% 37% 70% 62% 11% 20% Mortality growth index ( ) 4% 12% 11% 10% 20% 47% Age at diagnosis median (25% 75% percentile) 63 (52 73) 59 (48 72) 71 (62 78) 72 (61 79) 59 (46 70) 59 (45 73) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Breast (women) (C50) Cervix uteri (C53) Uterus (C54,C55) Ovary (C56) Prostate (C61) Testis (C62) Women Women Women Women Men Men Number of neoplasms per year ( ) 5,584 1,038 1,745 1,232 3, Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) 1, , Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) 49,599 15,373 20,741 8,318 20,477 6,377 Prevalence per 100,000 persons (year 2005) Mortality / incidence index , Incidence growth index ( ) 24% 13% 14% 4% 115% 40%

63 Mortality growth index ( ) Age at diagnosis median (25% 75% percentile) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Breast (women) (C50) Cervix uteri (C53) Uterus (C54,C55) Ovary (C56) Prostate (C61) Testis (C62) Women Women Women Women Men Men 6% 22% 15% 1% 19% 26% 62 (53 73) 51 (40 63) 65 (57 73) 62 (52 73) 72 (65 77) 34 (27 43) Kidney (C64) Bladder (C67) Brain and spinal cord (C70 C72) Men Women Men Women Men Women Number of neoplasms per year ( ) 1, , Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) 9,032 6,363 10,534 4,009 1,420 1,278 Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 32% 11% 42% 43% 10% 16% Mortality growth index ( ) 2% 7% 0% 19% 18% 17% Age at diagnosis median (25% 75% percentile) 64 (56 72) 69 (59 76) 69 (61 76) 71 (62 77) 57 (45 68) 61 (49 72) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002)

64 Thyroid gland (C73) Hodgkin s lymphoma (C81) Non-Hodgkin s lymphoma (C82 C85,C96) Men Women Men Women Men Women Number of neoplasms per year ( ) Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) 1,219 5,678 2,148 2,123 3,560 3,400 Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 39% 100% 25% 31% 42% 39% Mortality growth index ( ) 0% -14% 57% 46% 16% 16% Age at diagnosis median (25% 75% percentile) 57 (44 69) 55 (43 65) 37 (25 56) 33 (24 57) 64 (52 74) 69 (56 77) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Multiple myeloma (C90) Leukaemia (C91 C95) Neoplasms in situ (D00 D09) Men Women Men Women Men Women Number of neoplasms per year ( ) ,574 Number of neoplasms per 100,000 persons ( ) Number of deaths per year ( ) Number of deaths per 100,000 persons ( ) Prevalence number of persons (year 2005) ,954 2,487 4,424 31,388 Prevalence per 100,000 persons (year 2005) Mortality / incidence index Incidence growth index ( ) 5% 18% 1% 0% 268% 104%

65 Mortality growth index ( ) Age at diagnosis median (25% 75% percentile) Cumulative risk (age 0 74 years) Incidence ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Mortality ranking of the Czech Rep. in international comparison (ASR, estimate from 2002) Multiple myeloma (C90) Leukaemia (C91 C95) Neoplasms in situ (D00 D09) Men Women Men Women Men Women 9% 13% 3% 4% 87% 24% 69 (58 75) 71 (62 77) 67 (56 75) 70 (57 77) 70 (60 77) 41 (31 59) Table 7 Classification of diagnostic groups of malignant tumours in the Czech Republic according to trends in incidence rates, and according to the cancer stages at the time of diagnosis. Classification according to trends in incidence rates Diagnoses of malignant tumours with incidence rates rising in the long-term C61 (GI: %) C73 (GI: +84.9%) C44 (GI: +66.0%) C43 (GI: +46.6%) C67 (GI: +42.2%) C62 (GI: +40.8%) C82 C85, C96 (GI: +40.8%) C00 C14 (GI: +34.5%) C15 (GI: +32.5%) C25 (GI: +27.4%) C50 (GI: +23.4%) C64 (GI: +23.3%) C18 C21 (GI: +20.0%) C47, C49 (GI: +15.6%) C54, C55 (GI: +13.6%) C70 C72 (GI: +12.9%) C90 (GI: +10.8%) C22 (GI: +10.2%) Diagnoses of malignant tumours with stabilised incidence rates C33, C34 (GI: +7.2%) C32 (GI: +1.7%) C91 C95 (GI: +0.7%) C23, C24 (GI: 1.0%) C56 (GI: 4.6%) Diagnoses of malignant tumours with decreasing incidence rates C53 (GI: 13.6%) C16 (GI: 24.5%) C81 (GI: 28.2%) Classification according to detection rates of less advanced stages of malignant tumours Low frequency of detection rates at stages (S) I and II C15 (S I+II: 19.1%) C33, C34 (S I+II: 15.0%) C23, C24 (S I+II: 15.0%) C25 (S I+II: 8.5%) C22 (S I+II: 4.7%) Middle frequency of detection rates at stages (S) I and II C61 (S I+II: 47.0%) C47, C49 (S I+II: 43.9%) C18 C21 (S I+II: 42.8%) C32 (S I+II: 39.4%) C56 (S I+II: 29.9%) C00 C14 (S I+II: 27.1%) C16 (S I+II: 25.0%) High frequency of detection rates at stages (S) I and II C44 (S I+II: 98.0%) C43 (S I+II: 76.8%) C62 (S I+II: 76.0%) C50 (S I+II: 70.6%) C67 (S I+II: 66.3%) C54, C55 (S I+II: 66.0%) C73 (S I+II: 65.6%) C53 (S I+II: 59.6%) C64 (S I+II: 50.1%) C00 C14: oral cavity and pharynx; C15: oesophagus; C16: stomach; C18 C21: colon and rectum; C22: liver and intrahepatic bile ducts; C23,C24: gallbladder and biliary tract; C25: pancreas; C32: pharynx; C33,C34: trachea, bronchus and lung; C43: skin melanoma; C44: other malignant neoplasm of the skin; C47,C49: connective and soft tissues; C50: breast (women); C53: cervix uteri; C54,C55: uterus; C56: ovary; C61: prostate; C62: testis; C64: kidney; C67: bladder; C70 C72: brain and spinal cord; C73: thyroid gland; C81: Hodgkin s lymphoma; C82 C85,C96: non-hodgkin s lymphoma; C90: multiple myeloma; C91 C95: leukaemia GI: growth index related to the period ; S I+II: proportion of clinical stages I+II in the overall incidence of the disease, data from period

66 Incidence rates men Other malignant neoplasm of the skin (C44) 1 Colon and rectum (C18 C21) 2 Trachea, bronchus and lung (C33,C34) 3 Prostate (C61) 4 Bladder (C67) 5 Kidney (C64) 6 Stomach (C16) 7 Oral cavity and pharynx (C00 C14) 8 Pancreas (C25) 9 Melanoma of skin (C43) 10 Leukaemia (C91 C95) 11 Non-Hodgkin's lymphoma (C82 C85, C96) 12 Liver and intrahepatic bile ducts (C22) 13 Larynx (C32) 14 Brain and spinal cord (C70 C72) 15 Testis (C62) 16 Oesophagus (C15) 17 Gallbladder and biliary tract (C23,C24) 18 Multiple myeloma (C90) 19 Connective and soft tissues (C47,C49) 20 Hodgkin's lymphoma (C81) 21 Thyroid gland (C73) 22 Other malignant neoplasms In situ neoplasms (D00 D09) Neoplasms of uncertain or unknown behaviour (D37 D48) Number of newly diagnosed malignant tumours per 100,000 men Incidence rates women Other malignant neoplasm of the skin (C44) 1 Breast (C50) 2 Colon and rectum (C18 C21) 3 Uterus (C54,C55) 4 Trachea, bronchus and lung (C33,C34) 5 Ovary (C56) 6 Cervix uteri (C53) 7 Kidney (C64) 8 Pancreas (C25) 9 Melanoma of skin (C43) 10 Stomach (C16) 11 Gallbladder and biliary tract (C23,C24) 12 Bladder (C67) 13 Non-Hodgkin's lymphoma (C82 C85, C96) 14 Thyroid gland (C73) 15 Leukaemia (C91 C95) 16 Brain and spinal cord (C70 C72) 17 Liver and intrahepatic bile ducts (C22) 18 Oral cavity and pharynx (C00 C14) 19 Multiple myeloma (C90) 20 Hodgkin's lymphoma (C81) 21 Connective and soft tissues (C47,C49) 22 Oesophagus (C15) 23 Larynx (C32) 24 Other malignant neoplasms Neoplasms in situ (D00 D09) Neoplasms of uncertain or unknown behaviour (D37 D48) Number of newly diagnosed malignant tumours per 100,000 women Figure 7 Incidence rates of malignant tumours in the Czech Republic in the period comparison among different diagnostic groups. Source: Czech National Cancer Registry.

67 Mortality rates men Trachea, bronchus and lung (C33,C34) 1 Colon and rectum (C18 C21) 2 Prostate (C61) 3 Pancreas (C25) 4 Stomach (C16) 5 Kidney (C64) 6 Liver and intrahepatic bile ducts (C22) 7 Bladder (C67) 8 Oral cavity and pharynx (C00 C14) 9 Leukaemia (C91 C95) 10 Brain and spinal cord (C70 C72) 11 Oesophagus (C15) 12 Non-Hodgkin's lymphoma (C82 C85, C96) 13 Gallbladder and biliary tract (C23,C24) 14 Larynx (C32) 15 Melanoma of the skin (C43) 16 Multiple myeloma (C90) 17 Other malignant neoplasm of skin (C44) 18 Connective and soft tissues (C47,C49) 19 Hodgkin's lymphoma (C81) 20 Testis (C62) 21 Thyroid gland (C73) 22 Other malignant neoplasms In situ neoplasms (D00 D09) Neoplasms of uncertain or unknown behaviour (D37 D48) Number of cancer related deaths per 100,000 men Mortality rates women Breast (C50) 1 Colon and rectum (C18 C21) 2 Trachea, bronchus and lung (C33,C34) 3 Pancreas (C25) 4 Ovary (C56) 5 Stomach (C16) 6 Gallbladder and biliary tract (C23,C24) 7 Uterus (C54,C55) 8 Kidney (C64) 9 Leukaemia (C91 C95) 10 Cervix uteri (C53) 11 Liver and intrahepatic bile ducts (C22) 12 Brain and spinal cord (C70 C72) 13 Non-Hodgkin's lymphoma (C82 C85, C96) 14 Bladder (C67) 15 Multiple myeloma (C90) 16 Melanoma of skin (C43) 17 Oral cavity and pharynx (C00 C14) 18 Oesophagus (C15) 19 Thyroid gland (C73) 20 Other malignant neoplasm of skin (C44) 21 Connective and soft tissues (C47,C49) 22 Hodgkin's lymphoma (C81) 23 Larynx (C32) 24 Other malignant neoplasms In situ neoplasms (D00 D09) Neoplasms of uncertain or unknown behaviour (D37 D48) Number of cancer related deaths per 100,000 women Figure 8 Mortality rates of malignant tumours in the Czech Republic in the period comparison among different diagnostic groups. Source: Czech National Cancer Registry

68 Prevalence rates men Other malignant neoplasm of the skin (C44) 1 Colon and rectum (C18 C21) 2 Prostate (C61) 3 Bladder (C67) 4 Kidney (C64) 5 Melanoma of the skin (C43) 6 Trachea, bronchus and lung (C33,C34) 7 Testis (C62) 8 Oral cavity and pharynx (C00 C14) 9 Non-Hodgkin's lymphome (C82 C85, C96) 10 Larynx (C32) 11 Leukaemia (C91 C95) 12 Stomach (C16) 13 Hodgkin's lymphoma (C81) 14 Brain and spinal cord (C70 C72) 15 Thyroid gland (C73) 16 Connective and soft tissues (C47,C49) 17 Multiple myeloma (C90) 18 Pancreas (C25) 19 Oesophagus (C15) 20 Gallbladder and biliary tract (C23,C24) 21 Liver and intrahepatic bile ducts (C22) 22 Other malignant neoplasms In situ neoplasms (D00 D09) Neoplasms of uncertain or unknown behaviour (D37 D48) Number of cancer survivors per 100,000 men Prevalence rates women Other malignant melanoma of the skin (C44) 1 Breast (C50) 2 Uterus (C54,C55) 3 Colon and rectum (C18 C21) 4 Cervix uteri (C53) 5 Melanoma of the skin (C43) 6 Ovary (C56) 7 Kidney (C64) 8 Thyroid gland (C73) 9 Bladder (C67) 10 Non-Hodgkin's lymphoma (C82 C85, C96) 11 Leukaemia (C91 C95) 12 Trachea, bronchus and lungs (C33,C34) 13 Hodgkin's lymphoma (C81) 14 Stomach (C16) 15 Oral cavity and pharynx (C00 C14) 16 Brain and spinal cord (C70 C72) 17 Connective and soft tissues (C47,C49) 18 Gallbladder and biliary tract (C23,C24) 19 Multiple myeloma (C90) 20 Pancreas (C25) 21 Larynx (C32) 22 Liver and intrahepatic bile ducts (C22) 23 Oesophagus (C15) 24 Other malignant neoplasms In situ neoplasms (D00 D09) Neoplasms of uncertain and unknown behaviour (D37 D48) Number of cancer survivors per 100,000 women Figure 9 Prevalence rates of malignant tumours in the Czech Republic in the period comparison among different diagnostic groups. Source: Czech National Cancer Registry

69 2.2 Sources of data on cancer mortality The data on mortality in the Czech Republic are stored within the Death Records Database, which is administered by the Czech Statistical Office (CSO). This data is processed in compliance with the international methodology, based on data from the Death Certificate. A single diagnosis is established as the primary cause of death of an individual; this information is later processed in official statistical outputs. These statistics are available in CSO reports and in international databases of Eurostat and WHO. The Czech National Cancer Registry (CNCR) is another source of information on mortality in the Czech Republic: individual records contain data on the cause of death according to the Death Certificate, which can be subsequently used to establish cause-specific mortality with respect to different cancer diagnoses. Records on causes of death are thus kept independently in two information systems, making it possible to check the correctness of CNCR data retrospectively, and to verify the validity of mortality data and survival data on cancer patients. Table 6.3 gives an overview of the Czech sources of cancer mortality data. 3 Total cancer load of the Czech population The cancer load of the Czech population ranks among the highest worldwide and has been growing continuously, as shown in Fi gure 6.1 and Table 6.4. Growth in cancer incidence rates can be also expected in near future, due to the demographic structure and overall ageing of the Czech population. In 1995, the average age was 35.6 years for men and 38.9 years for women. Within twelve years, these values shifted to 38.8 years for men and 41.8 years for women (data from 2007). During the period , the proportion of inhabitants aged over 50 years increased by 6.6%. The demographic trends in the Czech population are described in more detail in Chapter 1 (Figure 1.2, Table 1.1). Figures 6.2 and 6.3 show age-specific incidence rates of all malignancies, proving that ageing of the Czech population actually leads to an increase in cancer incidence rates, particularly in the age categories 70 years and above. The cancer load in those aged 50 to 59 has been also increasing in the long term. Overall data on the incidence trends have not shown any significant changes in the cancer load in people under 30, and no shift of malignant diseases to younger age categories has been observed either. Age-specific incidence and mortality rates sorted by cancer type are shown in Figures 6.4 and 6.5. Each year, more than 28,000 persons die of cancer in the Czech Republic. Malignant tumours in the age categories 50 years and above take the highest toll with respect to the overall cancer mortality rates (Table 6.5). The total cancer load of the Czech population can be characterised by the following summary statistics involving CNCR data (data from 2005, all malignancies): Overall incidence rate: 661 patients per 100,000 population. Absolute number of newly diagnosed patients: 67,782. Overall mortality rate: 273 patients per 100,000 population. Absolute number of deaths related to cancer: 28,033. Prevalence rate: 256,543 persons. Time trends in incidence rates in the period : growing (2.5% annual increase on average). Time trends in mortality rates in the period : stable (0.1% annual decrease on average). Figure 6.6 shows the long-term trends in incidence rates, mortality rates and prevalence rates of all malignant tumours excluding nonmelanoma skin cancer (C44). Since 2000, the mortality rates have visibly stabilised, and the diverging trends in the incidence and mortality rates thus contribute to a significant increase in

70 Table 8 Regional differences in the main epidemiological trends of malignant tumours excluding nonmelanoma skin cancer (C00 C97 excluding C44) in the Czech Republic. Indicator Data from period Average value over the whole population Range of values in regions (Min Max; n = 14 regions) Incidence rate per 100,000 population ( ) Incidence growth index related to % (10.4% 37.5%) Mortality rate per 100,000 population ( ) Mortality growth index related to % ( 9.2% 10.1%) Prevalence rate per 100,000 population ,678 (2,473 3,006) Prevalence growth index related to % (60.9% 88.3%) Mortality / incidence index ( ) Main diagnostic groups proportion of detection at clinical stages I+II Stomach (C16) 25.0% (16.8% 36.8%) Colon and rectum (C18 C21) 42.8% (35.3% 51.7%) Pancreas (C25) 8.5% (5.0% 14.6%) Trachea, bronchus and lung (C33,C34) 15.0% (10.7% 19.2%) Skin melanoma (C43) 76.8% (72.9% 86.5%) Breast women (C50) 70.6% (66.4% 75.8%) Uterus (C54,C55) 66.0% (57.4% 77.3%) Prostate (C61) 47.0% (35.3% 67.9%) Kidney (C64) 50.1% (42.5% 55.9%) Bladder (C67) 66.3% (55.6% 85.1%) Proportion of detection during autopsy and from DCO % (3.9% 16.0%) the overall prevalence rates. Figure 6.7 gives an overview of the cancer incidence rates in main diagnostic groups, showing that Czech men are most frequently affected by colorectal cancer (C18 C21), closely followed by trachea, bronchus and lung cancer (C33 C34) and prostate cancer (C61). Figure 6.7 also shows that Czech women are most frequently affected by breast cancer (C50), followed by colorectal cancer (C18 C21), cancer of the uterus (C54 C55), and trachea, bronchus and lung cancer (C33, C34). All these diagnoses also rank among the most frequent ones with respect to mortality rates in both men and women, as shown in Figure 6.8. High mortality rates have been also observed in pancreatic cancer (C25), stomach cancer (C16) and ovarian cancer (C56). Mortality rates have recently stabilised also in most frequent diagnoses, which is the reason why corresponding prevalence rates have significantly increased (Figure 6.9). The tracheal, bronchial and lung cancers (C33, C34) are an exception, as their very high mortality rate does not contribute to the growing prevalence despite the increasing incidence rate. The main epidemiological characteristics of malignant diseases in the Czech Republic, sorted by individual diagnostic groups, are summarised in Table 6.6. Furthermore, the Appendix to this chapter provides detailed incidence rates for individual cancer diagnoses in the period , for both overall values and age-specific rates. 4 Data on clinical stages of newly diagnosed malignant tumours The Czech National Cancer Registry (CNCR) contains complete and comprehensive records

71 on the clinical stage at the time of diagnosis, including detailed records on individual components of TNM classification. A complete CNCR record, therefore, must contain data on the stage of the disease. Unfortunately, some diagnoses (such as haematological malignancies, or malignant tumours of the brain and nervous system) require other indicators than clinical stage (such as risk type of the disease or grade), and this data is not adequately repres ented within CNCR. For this reason, population-based data in CNCR is inadequate for some diagnoses (haematological malignancies in particular) and is presented only in a limited extent in this publication. The Czech expert medical societies have been trying to solve this problem by keeping records on these malignancies in diagnosis-specific clinical registries. Continuous methodical developments in the classification of malignant tumours present another problem, leading to complications in the analyses of long-term trends. Close attention is paid to the validation and controlled update of records on the clinical stage within CNCR. The reason is evident: without knowledge of the clinical stage, oncologists are actually not able to interpret any changes in epidemiological parameters. The abovementioned increase in crude incidence rate of malignant tumours (Figure 6.1) can serve as textbook example to demonstrate that detailed analysis of individual diagnoses does not necessarily have to reflect high-risk time trends: when taking into account the clinical stage of breast cancer, close analysis of its growing incidence rate reveals that this trend has resulted from an increasing number of cases diagnosed in early clinical stages, which in turn is the result of continuously improving diagnostics and effective screening. Chapter 5 evidenced detailed audit of the quality of CNCR data with respect to correct identification of malignant tumours and the statement on clinical stage. The classification scheme in Chapter 5 is also employed in this chapter to describe the detected clinical stages of individual diagnoses. Special attention is paid to records where information on clinical stage is missing: records missing data on clinical stage due to non-existent TNM classification, records missing data on clinical stage for objective reasons (DCO records, diagnosis based on autopsy, unfinished diagnosis due to early death, patient s refusal to treatment, loss of contact with patient or very poor health condition of the patient), records missing data on clinical stage without objective reason (incorrect and incomplete records, mostly missing the morphological verification of malignant tumour). The overall CNCR assessment has revealed only 5.8% of records which unfoundedly miss information on both TNM classification and clinical stage (see Chapter 5). A very positive message is that the completeness of data increases with time and that the most recent period (which is the most relevant one for clinical analyses) provides high-quality data. This fact is documented in Figure 6.11 and Table 6.7, which can be concluded as follows: It is evident that the problem of incomplete diagnostic identification in the period after 2000 applies less to the most frequent diagnoses, such as the breast cancer, colorectal cancer and lung cancer. In some diagnoses, there is very satisfactory proportion of primarily detected, less advanced stages of the disease. Breast cancer can serve as an example, with more than 70% of cases detected in clinical stages I and II. In general, however, the early detection of malignant tumours in the Czech Republic is not very frequent, and the situation is very unsatisfactory in numerous epidemiologically important diagnoses: more than 60% of colorectal cancer cases are detected at stage III and higher, and similar rates have been reported for lung cancer, ovarian cancer, head and neck tumours in general, and several other diagnoses. Figure 6.12 documents that cancer incidence rates have been growing in a number of these diagnostic groups throughout the Czech population. The growing num-

72 bers of patients and late diagnosis logically result in worse treatment results. This conclusion also applies to preventable diagnoses, such as colorectal cancer, cervical cancer, and partly also prostate cancer. Our discussion about the early detection of malignant tumours is concluded by an overview of in situ tumours, as shown in Figure A significant increase in early detection can be seen particularly in women since 1992, which can definitely be explained by the growing intensity and availability of secondary cancer prevention programmes focused on breast cancer and cervical cancer. The number of early detected in situ tumours is significantly lower in men, although a slightly growing trend has been reported since The incidence rates of individual clinical stages are described in more comprehensive manner in detailed analyses of individual diagnostic groups of malignant tumours (see the Appendix to this chapter). The preceding paragraphs were aimed at proving that the Czech population-based data is available in a quality that is high enough to sufficiently monitor the cancer load of the Czech population, to assess the preventive programmes and to perform survival analyses, conducted separately for individual clinical stages. 5 Monitoring the epidemiological situation in individual regions of the Czech Republic The Czech Republic is administratively divided into 14 regions of various sizes and partly also of various demographic structure of population (see Chapter 1, Table 1.1). The Czech National Cancer Registry fully respects the administrative division of the country, making the collected data on malignant tumours regionally available. Detailed regional overviews are provided in the Appendix to this chapter, which focuses on cancer epidemiology sorted by individual diagnoses; Table 6.8, in contrast, provides an overview of the regional variability of the main epidemiological indicators. In addition, Table 6.8 clearly shows that there are significant differences among the regions with respect to early diagnosis of malignant tumours, confirming the conclusion from the preceding subsection (i.e., that early detection of in situ tumours needs to be improved significantly). Furthermore, complex regional overviews which provide the possibility of comparison among regions are available on-line at 6 Cancer patients with multiple malignancies The Czech National Cancer Registry collects data which make it possible to identify a specific patient; therefore, recurring malignancies in the same person can be found in the registry, whether it be the same location or another and the chronological order of recurring malignancies can also be established. Figure 6.13 and Table 6.9 sum up the overall data, showing that multiple malignancies are relatively common, although they differ markedly among individual diagnoses. If non-melanoma skin cancers (C44) and malignant neoplasms of uncertain behaviour (D37 D48) are not taken into account, the relative frequency of recurring malignancies ranges from 12 to 14 per cent, the overwhelming majority (96%) of recurring malignancies belonging to other diagnostic group than the primary tumour. Recurring malignancies of the same diagnostic group is more common in breast cancer (C50), bladder cancer (C67) and partly in testicular cancer (C62).

73 60 Number of newly diagnosed cases per 100,000 persons men; 18.2 cases per 100,000 men in 2005 women; 54.0 cases per 100,000 women in Proportion of carcinomas in situ diagnosed in Men N = 3, Year Women N = 12,868 D00 C. i. s. of oral cavity, oesophagus and stomach 2.4 % 0.4 % D01 C. i. s. of other and unspecified digestive organs 24.0 % 4.1 % D02 C. i. s. of middle ear and respiratory system 2.4 % 0.1 % D03 Melanoma in situ 13.4 % 5.1 % D04 C. i. s. of skin 50.7 % 13.5 % D05 C. i. s. of breast 0.1 % 9.3 % D06 C. i. s. of cervix uteri % D07 C. i. s. of other and unspecified genital organs 2.0 % 2.9 % D09 C. i. s. of other and unspecified sites 5.0 % 0.6 % C. i. s. - Carcinoma in situ Figure 10 Trends in incidence rates of carcinomas in situ (D00 D09) in the Czech Republic. Source: Czech National Cancer Registry. Figure 11 Malignant neoplasms reported in the period , sorted by occurrence of neoplasms in the same patient. Source: Czech National Cancer Registry.

74 140 Incidence growth index in the period [%] C43 40 C67 C62 C00-C14 C15 C25 C64 C50 20 C18-C21 C47,C49 C54,C55 C22 C33,C34 C32 0 C23,C24 C C53-20 C16 C61 C73 C44-40 Proportion of stages I+II among newly diagnosed malignant tumours in the period [%] Figure 12 Diagnostic groups of malignant tumours: comparison with respect to trends in the incidence rates and to the proportion of detection at early stages. C00 C14: oral cavity and pharynx; C15: oesophagus; C16: stomach; C18 C21: colon and rectum; C22: liver and intrahepatic bile ducts; C23,C24: gallbladder and biliary tract; C25: pancreas; C32: larynx; C33,C34: trachea, bronchus and lung; C43: melanoma of the skin; C44: other malignant neoplasm of the skin; C47,C49: connective and soft tissues; C50: breast women; C53: cervix uteri; C54,C55: uterus; C56: ovary; C61: prostate; C62: testis; C64: kidney; C67: bladder; C73: thyroid gland Figure 13 Proportion of stages in malignant tumours diagnosed in the period ; diagnoses sorted by proportions of stages I and II. Source: Czech National Cancer Registry.

75 Table 9 Malignant diseases diagnosed in period with respect to other neoplasms in the same patient Newly diagnosed neoplasms in total Primary neoplasms Other neoplasms: same diagnosis Other neoplasms: other diagnosis Patients with multiple neoplasms the most frequent other diagnosis in patients with respective diagnosis Oral cavity and pharynx (C00 C14) 5,880 5,137 (87.4%) 15 (0.3%) 728 (12.4%) C44; C33,C34; C18 C21; C00 C14; C61 Oesophagus (C15) 2,267 2,013 (88.8%) 1 (0.0%) 253 (11.2%) C44; C00 C14; C18 C21; C33,C34; C16 Stomach (C16) 8,613 7,500 (87.1%) 10 (0.1%) 1,103 (12.8%) C44; C18 C21; C61; C50; C33,C34 Colon and rectum (C18 C21) 40,040 34,609 (86.4%) 441 (1.1%) 4,990 (12.5%) C44; C18 C21; C61; C50; C33,C34 Liver and intrahepatic bile ducts (C22) 4,038 3,601 (89.2%) 1 (0.0%) 436 (10.8%) C44; C18 C21; C33,C34; C64; C61 Gallbladder and biliary tract (C23,C24) 5,038 4,454 (88.4%) (11.6%) C44; C18 C21; C50; C54,C55; C25 Pancreas (C25) 8,512 7,408 (87.0%) 10 (0.1%) 1,094 (12.9%) C44; C18 C21; C50; C64; C33,C34 Larynx (C32) 2,670 2,388 (89.4%) 2 (0.1%) 280 (10.5%) C33,C34; C44; C18 C21; C00 C14; C61 Trachea, bronchus and lung (C33,C34) 30,701 26,913 (87.7%) 104 (0.3%) 3,684 (12.0%) C44; C18 C21; C67; C61; C64 Melanoma of skin (C43) 8,248 7,013 (85.0%) 173 (2.1%) 1,062 (12.9%) C44; D00 D09; C18 C21; C50; C61 Other malignant neoplasm of skin (C44) 73,270 45,269 (61.8%) 22,672 (30.9%) 5,329 (7.3%) C44; C18 C21; D00 D09; C61; C50 Connective and soft tissues (C47,C49) 1,297 1,163 (89.7%) 1 (0.1%) 133 (10.3%) C44; C18 C21; C50; C33,C34; C64 Breast (in women) (C50) 27,921 24,664 (88.3%) 1,021 (37%) 2,236 (8.0%) C44; C50; C18 C21; C54,C55; D00 D09 Cervix uteri (C53) 5,192 4,877 (93.9%) 4 (0.1%) 311 (6.0%) C44; C50; C18 C21; C33,C34; D00 D09 Uterus (C54,C55) 8,727 7,804 (89.4%) 11 (0.1%) 912 (10.5%) C50; C44; C18 C21; C56; D00 D09 Ovary (C56) 6,161 5,401 (87.7%) 13 (0.2%) 747 (12.1%) C50; C44; C18 C21; C54,C55; D00 D09 Prostate (C61) 19,710 16,872 (85.6%) 4 (0.0%) 2,834 (14.4%) C44; C18 C21; C67; C64; C33,C34 Testis (C62) 2,069 1,998 (96.6%) 34 (1.6%) 37 (1.8%) C44; C62; C18 C21; C33,C34; C61 Kidney (C64) 13,035 11,131 (85.4%) 153 (1.2%) 1,751 (13.4%) C44; C18 C21; C61; C33,C34; C50 Bladder (C67) 11,301 9,207 (81.5%) 515 (4.6%) 1,579 (14.0%) C44; C61; C18 C21; C33,C34; C67 Brain and spinal cord (C70 C72) 3,888 3,623 (93.2%) 8 (0.2%) 257 (6.6%) C44; C18 C21; C61; C50; C64 Thyroid gland (C73) 3,233 2,879 (89.1%) 7 (0.2%) 347 (10.7%) C44; C50; C18 C21; C54,C55; C64

76 Newly diagnosed neoplasms in total Primary neoplasms Other neoplasms: same diagnosis Other neoplasms: other diagnosis Patients with multiple neoplasms the most frequent other diagnosis in patients with respective diagnosis Hodgkin s lymphoma (C81) 1,298 1,234 (95.1%) 0 64 (4.9%) C44; C33,C34; C18 C21; D00 D09; C50 Non-Hodgkin s lymphoma (C82 C85,C96) 5,754 4,982 (86.6%) 9 (0.2%) 763 (13.3%) C44; C18 C21; C50; D00 D09; C33,C34 Multiple myeloma (C90) 2,198 1,970 (89.6%) 1 (0.0%) 227 (10.3%) C44; C18 C21; C61; C33,C34; C50 Leukaemia (C91 C95) 5,779 5,017 (86.8%) 3 (0.1%) 759 (13.1%) C44; C18 C21; C33,C34; C61; C64 Other malignant neoplasms 13,615 11,632 (85.4%) 34 (0.2%) 1,949 (14.3%) C44; C18 C21; C67; C33,C34; C50 Neoplasms in situ (D00 D09) 16,222 13,662 (84.2%) 420 (2.6%) 2,140 (13.2%) C44; C18 C21; D00 D09; C50; C43 Neoplasms of uncertain or unknown behaviour (D37 D48) 4,997 3,991 (79.9%) 192 (3.8%) 814 (16.3%) C18 C21; C44; D37 D48; C50; D00 D09 Total 341, ,412 (81.5%) 25,859 (7.6%) 37,403 (10.9%) - Acknowledgements The validation of the Czech National Cancer Registry and the development of reporting system over population data have been supported by the Internal Grant Agency of the Czech Ministry of Health (the IGA project No ). We would also like to acknowledge the work of administrators of the Czech National Cancer Registry, namely the Institute of Health Information and Statistics of the Czech Republic (IHIS) and the Coordination Centre for Departmental Medical Information Systems (CCDMIS). References Binding instructions of the National Health Information System (NHIS): Czech National Cancer Registry instruction for the contents of data structure, version /2, Institute of Health Information and Statistics of the Czech Republic (IHIS), Prague On-line version: section IHIS, part Binding instructions Coordination Centre for Departmental Medical Information Systems (CCDMIS), Czech National Cancer Registry (CNCR), [ ], available on narodni-zdravotni-registry/nor/nor.html Curado. M. P., Edwards, B., Shin. H.R., Storm. H., Ferlay. J., Heanue. M. and Boyle. P., eds (2007) Cancer Incidence in Five Continents, Vol. IX. IARC Scientific Publications No. 160, Lyon, IARC. Demographic data of the Czech Republic and Death Records Database of the Czech Republic, Czech Statistical Office

77 Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Vyzula R. Epidemiology of malignant tumours in the Czech Republic [online]. Masaryk University, [2005], [cit ]. On-line available: Version 7.0 [2007], ISSN Dušek L., Mužík J., Kubásek M., Koptíková J., Šnajdrová L., Ondrušová M. (2007) National portal of epidemiology of malignant tumours in the Slovak Republic [online]. Masaryk University, Brno, Czech Republic [2007]. ISBN European health for all database (HFA-DB). World Health Organisation [cit ] Ferlay J., Bray F., Pisani P. and Parkin D.M. GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004, www-dep.iarc.fr. Institute of Health Information and Statistics of the Czech Republic (IHIS), National Health Information System (NHIS) legislation, [ ], available on info.php?article=32&mnu_id=7110 Institute of Health Information and Statistics of the Czech Republic (IHIS), National Health Information System (NHIS), Czech National Cancer Registry, [ ], available on uzis.cz/info.php?article=368&mnu_id=7300 Ondrušová M., Pleško I., Safaei-Diba Ch., Obšitníková A., Štefaňáková D., Ondruš D. Comprehensive analysis of incidence and mortality of malignant tumours in the Slovak Republic, [online]. Bratislava, National Cancer Registry of the Slovak Republic, NHIC, Parkin, D.M., Whelan, S.L., Ferlay, J., and Storm, H. Cancer Incidence in Five Continents, Vol. I to VIII. IARC CancerBase No. 7, Lyon,

78 Paper II. Mužík J., Koptíková J., Dušek L., Žaloudík J., Vyzula R., Abrahámová J. Historical data of the Czech National Cancer Registry: information value and risk of bias; Klinická onkologie 20, Supplement 1/2007, 63-76, ISSN X (original article in Czech)

79 Role of population-based registries in oncology and current situation in the Czech Republic. KLINICKÁ ONKOLOGIE 20, SUPPLEMENT 1/2007, 63 76, ISSN X (ORIGINAL ARTICLE IN CZECH) HISTORICAL DATA OF THE CZECH NATIONAL CANCER REGISTRY: INFORMATION VALUE AND RISK OF BIAS ANALÝZA HISTORICKÝCH DAT NÁRODNÍHO ONKOLOGICKÉHO REGISTRU ČR: INFORMAČNÍ POTENCIÁL A RIZIKA ZKRESLENÍ MUŽÍK J. 1, KOPTÍKOVÁ J. 1, DUŠEK L. 1, ŽALOUDÍK J. 2, VYZULA R. 2, ABRAHÁMOVÁ J. 3 1 FACULTY OF MEDICINE, MASARYK UNIVERSITY, BRNO 2 MASARYK MEMORIAL CANCER INSTITUTE, BRNO 3 THOMAYER UNIVERSITY HOSPITAL, PRAGUE Summary This paper is focused on the quality assessment of population-based cancer registries. The analyses are methodically based on the database of the Czech National Cancer Registry (CNCR) that offers a valuable model with epidemiological data of all diagnostic groups of cancer, all collected continuously since Quality control is necessary for applications of CNCR in the health care evaluation and for the international presentation of Czech cancer epidemiology. The audit resulted in a set of diagnostically completed and verified records and in a set of records with incomplete diagnostics (namely TNM) due to well distinguished objective reasons (DCO cases, cases diagnosed at autopsy, early death after diagnosis, etc.). All the objective reasons were recognized and the database was enriched by codes that identify the status of the record and its reason. In addition, we found 5.1 % of cases with wrong and/or incomplete diagnostics and without any acceptable explanation. These cases typologically correspond to the whole CNCR database and therefore they can be excluded from serious analyses with a minimum risk of bias. Increased probability of bias due to erroneously incomplete records exists only in regionally specific comparisons. Key words: cancer epidemiology, population registries, quality of data Souhrn Článek se zabývá otázkou kvality dlouhodobě sbíraných epidemiologických dat v populačních registrech. Jako model je využita databáze Národního onkologického registru ČR (NOR), která nabízí kompletní epidemiologická data všech diagnostických skupin zhoubných nádorů od roku Analýza kvality dat NOR je nutným předpokladem jejich využití v hodnocení výsledků léčebné péče a je také základem odpovědné prezentace českých dat v zahraničí. Výsledkem auditu dat NOR je soubor validních záznamů s řádně zaznamenanou diagnostikou a soubor neúplných záznamů s identifikovanou objektivní příčinou problémů (DCO, diagnóza při pitvě, velmi časná úmrtí, apod.). Nad rámec těchto dvou kategorií bylo zachyceno 5,1 % záznamů s chybně neuvedeným klinickým stadiem i TNM klasifikací. Chybné záznamy typologicky odpovídají celkové databázi NOR a lze je tedy v závažných analýzách vypustit, aniž bychom se dopustili systémového zkreslení výstupů. Možnost zkreslení existuje pouze při některých regionálních srovnáních. Klíčová slova: epidemiologie zhoubných nádorů, populační registry, kvalita dat Problems with retrospective analysis of data from population-based cancer registries Areal registries are of concern only if they are actually arearepresentative, long-term operated according to standards and regularly updated. All of this may be assured for cancer diseases as well, if a reliable organization background, adequate budget and sufficient range of collected data is available, although a universal cancer database cannot contain specific records of individual diagnostic groups. A much bigger problem is the development of oncology itself, changes in classification systems, development of diagnostic and therapeutic methods etc. These changes cannot naturally be avoided, but it is also impossible to constantly adjust the registry and change its structure. Area-collected data therefore accumulate a number of factors that may cause serious bias in results, especially in long-term trends. Due to heterogeneity of such problems there is no simple and direct solution available, as illustrated in the following survey: Problems with completeness of the registry and entered records Completeness of records is a fundamental requirement and in areal epidemiological registries, a basic condition of their meaningful existence. It should be stated that this is difficult to reach in cancer registries due to a diagnostic heterogeneity and a high number of reporting health care facilities. The term completeness in this case includes both a representative number of entered records and internal completeness of each individual record, in which fundamental data components cannot KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

80 Role of population-based registries in oncology and current situation in the Czech Republic. be missing, such as tumour diagnostic identification, clinical stage, and updated information about patient status. Even the most prestigious registries face delays in incidence reporting and searching tumours diagnosed at autopsy. Thus, an independent death records database is very valuable for comparison and retrograde verification of epidemiological data Problems with reliability of registered data Reliability is closely related to completeness; incomplete records are of course unreliable, too, as are the analytical outputs based on them. The solution lies in a registry design with an essential minimum of parameters that may be collected reliably at the population level. Any detailed parameters should be a subject of specialized clinical projects, in which data management is much easier to perform. Changes in regional structure in an area covered by the registry Very unpleasant problem complicating retrograde data transformation. All records should be explicitly assigned to region and place and these codes should allow for a simple transformation when the changes occur. Changes in number and type of reporting health care facilities Explicit assignment of diagnosed and treated patient to a given facility is a very important condition for future use of records. Each record may be assigned to a certain type of health care facility. The assignment, besides others, expresses responsibility of the facility for the patient and his records. In case that the patient changes to other health care facilities, it should be clearly stated what facility is responsible for diagnostics and primary treatment and what for follow-up care or treatment of relapses and progressions. Changes in therapeutic methods The objective factor, which is unavoidable, but significantly complicates retrospective assessment of records in cancer registries. Improvement of therapeutic methods surely affects trends in mortality, but this fact cannot be a reason for implementation of detailed clinical records into the epidemiological registries. Applied therapeutic modalities may be generally acceptable for areal registries, but any further observation belongs to specialized clinical databases. Changes in diagnostic techniques and approach Development in diagnostics changes, particularly the ratio of clinical stages of detected tumours. In a reliable areal registry the diagnostic development should not mean long-term an increase in incidence reporting, but a change internal structure of records. That is why diagnostic identification of records is so important, because knowledge of clinical stages is a condition for the correct interpretation of trends. Changes in tumour classification system Improvement the in tumour classification system represents a problem in categories, in which the way of clinical stage determination from TNM components changes. If both TNM parameters and clinical stages are recorded in the registry and the records are complete, the change in classification does not represent a problem in retrospective data assessment, since that data may be transformed. It is obvious that most of the problems may already precede the preparation of the registry (particularly determination of a reasonable amount of collected data), some problems must be reflected and registry must be adapted during its operation, and some problems cannot be objectively solved and analytical outputs must take them into account. The following strategic measures may be generally recommended to minimize future problems: A. The areal registry should be oriented mainly around epidemiological data without collection of specific diagnostic and therapeutic records, which often reflect progress within the frame of individual diagnoses. B. Each new record should be reliably and robustly assigned to the locality and health care facility, which is responsible for treatment, and later to the facility, which is responsible for the patient s follow-up. These measures make the registry ready for retrograde transformations after changes of regional organization and also simplify flexible sorting according to the category of the health care facility. C. Correctness and completeness of diagnostic records must be emphasized. For solid tumours it is namely diagnosis identification, primary tumour location, and determination of TNM and clinical stage. Particularly the last of the measures stated above should be considered as a fundamental rule determining the reliability of the whole registry. Complete diagnostic data may naturally be unavailable for all records (e.g. when the diagnosis is determined at autopsy), but without respect to this fact the diagnostic identification is a presumption for use of data from areal registries. Without knowledge of the diagnosis we cannot work even with such basic information as incidence and mortality, and the whole registry loses its practicality. For solid tumour, moreover, TNM classification and clinical stage are on the same level of importance as the diagnosis itself, because different stages of the disease have different therapeutic options and procedures, and finally patient survival. This paper further deals with the quality assessment of longterm collected epidemiological data. For methodological analyses and considerations, an ideal model is available in the form of the Czech National Cancer Registry (CNCR), which offers extensive epidemiological data of all cancer diagnostic groups collected since The following analyses will also be focused on the quality of the CNCR epidemiological data, its consistency and possible risks of bias. This analysis is necessary not only from the viewpoint of CNCR use for the assessment of health care results, but is also a basis for the responsible presentation of Czech epidemiological data abroad. 64 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

81 Role of population-based registries in oncology and current situation in the Czech Republic. Diagnostic identification of malignant neoplasms as a basis for quality of areal registries A missing or unreliable record on diagnosis disqualifies all further collected data and inhibits an exact inclusion of the record into epidemiological analyses. One may wonder why we emphasize such an apparent thing as stating diagnosis in the epidemiological registry. In expert terms let us say that identification of the diagnosis according to the MKN-O system by Cxx code is an essential condition, not sufficient in cancer registries. A reliable and usable record must also contain information about the spread of the disease, which is in case of solid tumours the TNM system and resulting clinical stage, for some other diagnoses (e.g. hematooncological) also other indicators (risk type of the disease, grade). These records confirm the diagnosis itself and determine the stage of the disease at the time of diagnosis. Table 1: Overview of malignant neoplasms diagnoses C00 C01 C02 C03 C04 C05 C06 C07 C08 C09 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 C21 C22 C23 C24 C25 C26 C30 C31 C32 C33 C34 C37 C38 C39 C40 C41 C43 C44 C45 Malignant neoplasm of lip Malignant neoplasm of base of tongue Malignant neoplasm of other and unspecified parts of tongue Malignant neoplasm of gum Malignant neoplasm of floor of mouth Malignant neoplasm of palate Malignant neoplasm of other and unspecified parts of mouth Malignant neoplasm of parotid gland Malignant neoplasm of other and unspecified major salivary glands Malignant neoplasm of tonsil Malignant neoplasm of oropharynx Malignant neoplasm of nasopharynx Malignant neoplasm of pyriform sinus Malignant neoplasm of hypopharynx Malignant neoplasm of other and ill-defined sites in the lip, oral cavity, and pharynx Malignant neoplasm of esophagus Malignant neoplasm of stomach Malignant neoplasm of small intestine Malignant neoplasm of colon Malignant neoplasm of rectosigmoid junction Malignant neoplasm of rectum Malignant neoplasm of anus and anal canal Malignant neoplasm of liver and intrahepatic bile ducts Malignant neoplasm of gallbladder Malignant neoplasm of other and unspecified parts of biliary tract Malignant neoplasm of pancreas Malignant neoplasm of other and ill-defined digestive organs Malignant neoplasm of nasal cavity and middle ear Malignant neoplasm of accessory sinuses Malignant neoplasm of larynx Malignant neoplasm of trachea Malignant neoplasm of bronchus and lung Malignant neoplasm of thymus Malignant neoplasm of heart, mediastinum, and pleura Malignant neoplasm of other and ill-defined sites in the respiratory system and intrathoracic organs Malignant neoplasm of bone and articular cartilage of limbs Malignant neoplasm of bone and articular cartilage of other and unspecified sites Malignant melanoma of skin Other malignant neoplasms of skin Mesothelioma C46 C47 C48 C49 C50 C51 C52 C53 C54 C55 C56 C57 C58 C60 C61 C62 C63 C64 C65 C66 C67 C68 C69 C70 C71 C72 C73 C74 C75 C76 C77 C78 C79 C80 C81 C82 C83 C84 C85 C88 C90 C91 C92 C93 C94 C95 C96 C97 Kaposi's sarcoma Malignant neoplasm of peripheral nerves and autonomic nervous system Malignant neoplasm of retroperitoneum and peritoneum Malignant neoplasm of other connective and soft tissue Malignant neoplasm of breast Malignant neoplasm of vulva Malignant neoplasm of vagina Malignant neoplasm of cervix uteri Malignant neoplasm of corpus uteri Malignant neoplasm of uterus, part unspecified Malignant neoplasm of ovary Malignant neoplasm of other and unspecified female genital organs Malignant neoplasm of placenta Malignant neoplasm of penis Malignant neoplasm of prostate Malignant neoplasm of testis Malignant neoplasm of other and unspecified male genital organs Malignant neoplasm of kidney, except renal pelvis Malignant neoplasm of renal pelvis Malignant neoplasm of ureter Malignant neoplasm of bladder Malignant neoplasm of other and unspecified urinary organs Malignant neoplasm of eye and adnexa Malignant neoplasm of meninges Malignant neoplasm of brain Malignant neoplasm of spinal cord, cranial nerves, and other parts of central nervous system Malignant neoplasm of thyroid gland Malignant neoplasm of adrenal gland Malignant neoplasm of other endocrine glands and related structures Malignant neoplasm of other and ill-defined sites Secondary and unspecified malignant neoplasm of lymph nodes Secondary malignant neoplasm of respiratory and digestive organs Secondary malignant neoplasm of other sites Malignant neoplasm without specification of site Hodgkin's disease Follicular [nodular] non-hodgkin's lymphoma Diffuse non-hodgkin's lymphoma Peripheral and cutaneous T-cell lymphomas Other and unspecified types of non-hodgkin's lymphoma Malignant immunoproliferative diseases Multiple myeloma and malignant plasma cell neoplasms Lymphoid leukemia Myeloid leukemia Monocytic leukemia Other leukemias of specified cell type Leukemia of unspecified cell type Other and unspecified malignant neoplasms of lymphoid, hematopoietic, and related tissue Malignant neoplasms of independent (primary) multiple sites Clinical stage and TNM values definitely affect the variability of overall survival and explain a number of interregional differences and time trends (Sant et al., 2003). Even the most simplified registries must contain records with all three TNM components; the system cannot be replaced by a mere statement of clinical stage, because alone it may be of a higher informational value than integrated data about the clinical stage (Sant et al., 2001). It was, for example, demonstrated that variability KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

82 Role of population-based registries in oncology and current situation in the Czech Republic. Table 2: Short overview of development in TNM classification of malignant neoplasms TNM CLASSIFICATION VERSION VALID TNM 2 nd edition, 1974 (1) TNM 3 rd edition, 1982 (2) TNM 4 th edition, 2 nd revision, 1992 (3) TNM 5 th edition, 1997 (4) TNM 6 th edition, 2002 (5) since 2005 MAIN CHANGES * only TNM classification implemented without discrimination into clinical stages Changes and supplements of classification of 23 diagnoses from the 1st edition (1968), other diagnoses added. The 2nd edition contains classification of the diagnoses C00*, C01 C06, C09*, C10*, C11*, C13*, C32, C15, C16*, C34, C50, C51*, C52*, C53, C54, C55*, C56, C57*, C64*, C67*, C60*, C61*, C62*, C73*, C44* Changes and supplements of classification of diagnoses from the 2nd edition, implementation of clinical stages for diagnoses C00, C09-C13, C16, C51, C52. New classifications for further diagnoses: C18-C20, C21*, C43, C47, C49. Changes and supplements of classification of diagnoses from the 3rd edition (particularly C18-C20), implementation of clinical stages for diagnoses C21, C60-C63, C64, C67, C44. New classifications for further diagnoses: C07, C08, C17, C31, C22-C25, C40, C41, C45, C65, C66, C69, C71. Most of diagnoses without changes or with small changes reflecting new information on prognosis and new methods for the prognosis determination (changes in C11, C58, C51, C53, C57, C61, C62, C64, C67, abandoned classification in C71). New classifications for further diagnoses: C58, C57.0 Most of diagnoses without changes or with small changes reflecting new information on prognosis (changes in C22-C25, C38.4, C40, C41, C50, C69.3, C69.4) 1. International Union Against Cancer (UICC): TNM Classification of Malignant Tumours. 2nd ed. Geneva, International Union Against Cancer (UICC): TNM Classification of Malignant Tumours.. 3rd ed. M.H. Harmer, ed. Geneva, Enlarged and revised International Union Against Cancer (UICC): TNM Classification of Malignant Tumours.. 4th ed. P. Hermanek, L.H. Sobin, eds. Berlin, Heidelberg, New York: Springer Verlag, Revised International Union Against Cancer (UICC): TNM Classification of Malignant Tumours.. 5th ed. Sobin LH, Wittekind Ch (ed.), New York, Wiley-Liss International Union Against Cancer (UICC): TNM Classification of Malignant Tumours.. 6th ed. Sobin LH, Wittekind Ch (ed.), New York, Wiley-Liss Figure 1. Progress of incidence and mortality of all malignant neoplasms (C00 C96, D03, D05, D06) in tumour size (T) within the frame of the N0/M0category may explain significant interregional differences in survival of women with breast carcinoma (Shetty and Reinman, 1997). Components of the TNM system are often mentioned in population-based and clinical prognostic models as independent predictors (Sant et al., 2003). Furthermore, partial TNM components have an important retrograde correction role, because on the basis of the TNM classification we can also retrospectively correct interpretation of results for various clinical stages. Table 1 displays a survey of cancer diagnoses, most of which are classified by the TNM system. The survey illustrates an apparent variety of the system, which moreover passes through methodological progress in time and covers an increasing range of diagnoses and stages (table 2). When performing a retrograde analysis of data from a long-term registry, such as the CNCR, we have to take these changes into account, since they affect the interpretation of trends even in the fundamental epidemiological data as incidence and mortality. It may be declared that without knowledge of the clinical stage no such change in epidemiological parameters can be reliably interpreted in oncology. We can often use a media-promoted increase or the raw incidence of malignant tumours in the Czech Republic (figure 1) as a textbook illustrative example, which however, after a detailed analysis of individual diagnoses, does not have to mean a mere increase of cancer frequency in the population. For example, in case of the dramatic increase in breast cancer incidence we can conclude from knowledge of the clinical stage and TNM that the trend is caused by a higher number of newly diagnosed less advanced stages, which is mainly the result of improved diagnostics. Implementation of mammography and 66 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

83 Role of population-based registries in oncology and current situation in the Czech Republic. reliably work only with data entered according to recent and more comprehensive rules of the TNM classification (table 2). Moreover, the background of older registries has been improving over time and they may produce data of higher quality. Significant improvement of data quality is also certainly the case with the CNCR, as documented by analysis identifying the random component of incidence data variability (figure 3). Interregional and temporal (annual) fluctuations are significantly lower for data entered after 1990 than for historical data and have recently reached an accepted level up to 5 10% of total variability. Improvement is apparent for almost all diagnostic groups with the exception of some less frequent diagnoses, where higher variability can be explained by too small sample size and sometimes difficult diagnostics in a given localization (e.g. oesophageal or rhinopharynx cancer, figure 3). Figure 2. Changes in the incidence and mortality progress on the example of breast cancer (C50) a recent breast carcinoma screening programme also play an important role (see data in figure 2). Methodological progress of the malignant tumours TNM classification is naturally an unpleasant complication for long-term trend analyses. Nevertheless, this is not fundamental from the viewpoint of clinically useful information, because due to the progress in diagnostic techniques and treatment options, we must focus on analyses of records performed during last 5-15 years. This does not reduce complications e.g. for the assessment of long-term survival, but the positive thing is that we can Methodological analysis of epidemiological data quality No one can naturally expect that analysis of historical data in a large population-based registry will be performed without problems. The development of the classification system itself represents a source of variability that cannot often be retrospectively corrected. Reliable publications of data from population-based registries also contain information about the ratio of various problematic records % reliability or on the contrary 5-10% error in diagnostic records of population-based cancer data is often cited (e.g. Eaker et al., 2006). Thus, even prestigious population-based databases face a certain amount of problematic records. For example the recent analysis of aggregated European registries in the EUROCARE-3 project displayed in 6.5 million records 1.1% loss during follow-up period, and further 4.2% DCO records and 1.2% tumours diagnosed at autopsy (Capocaccia et al., 2003). As minimal attributes of data, consistency may be considered succession of patient s age and age at diagnosis, and further a correct filing of diagnostic records. In any case the incomplete or incorrectly completed records must be excluded during the input audit or marked as irrelevant for further processing and their effect on the representativeness of the rest of the records must be considered. There may be more reasons that decrease information value of individual records and each of them may be of a different importance in various cases. Therefore, the data quality audit should lead to labelling primary records with codes that reflect origin, completeness, and reliability of the record: - Summary code defining overall reliability and usability of the record, e.g. with following code list items: o Complete record, full use o Record incomplete for objective reasons, otherwise consistent o Record with serious problems and low quality o Incorrect and unusable record - Code defining missing items in the primary record - Code defining record with excluded items due to inconsistence - Code defining main causes of record incompleteness (objective reasons / error) KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

84 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 3. Assessment of interregional and temporal fluctuation of cancer incidence in the Czech National Cancer Registry Such an audit must be comprehensive, because its results eliminate accessibility of some records for important analyses. The audit must therefore monitor all situations and causes that lead to the incomplete diagnostic identification of records. It is obvious from the following survey that lots of reasons are objective and lots of them have information value that was recognized (see e.g. Jemal et al., 2004): 1) Segregation of diagnostic groups for which the TNM classification does not make sense and their incompleteness is basically correct (in the Czech Republic e.g. case of the CNCR universal reporting form, which segregated diagnostic identification of e.g. leukaemia). 2) Segregation of those diagnostic groups for which the TNM system was not implemented at the time of diagnosis. 3) Incomplete diagnostics due to diagnosis based on DCO record or autopsy. 4) Incomplete diagnosis due to early patient s death after diagnosis before the anticancer therapy was initiated. 5) Incomplete diagnostics due to unexpected progress, patient s rehousing, therapy refusal etc. 6) Incomplete or otherwise problematic records caused by actual error, which do not belong to any of the categories mentioned above. It is evident that incomplete records do exist and their rate is relatively high particularly in older data. Nevertheless, a total erasure of the record from the database is an extreme intervention, which should be performed only in cases where we are not even sure that it was a malignant tumour. The code system suggested above gives administrators the option of sorting records according to various level of incompleteness without having to erase them. A space opens based on this system for the following corrective solutions: - Records including at least tumour diagnosis may be involved in basic epidemiological summarizations - In records in which only the clinical stage or TNM is stated, retrograde completion of the other is possible according to rules valid in time of diagnosis and based on expert opinion. - Records without clinical stage and TNM stated, whether by error or for objective reasons, may be classified based on probability methods. One of the standard solutions of this problem is the use of so called imputation statistical methods that replenish missing data with the most probable value by means of regression models or model distributions. If the final record is consistent and passes logical tests, it may be treated as complete (Schenker and Taylor, 1996; Horton and Lipsitz, 2001; Bernaards et al., 2003). However, only those techniques must be used for this purpose that may be retrospectively checked, and replenished records must be clearly labelled. 68 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

85 Role of population-based registries in oncology and current situation in the Czech Republic. All analyses stated above may be performed on records that were entered into the registry ; the audit of existing records does not solve representativeness of the registry itself. In this case, comparison with external data sources, demographic registries, and international databases is necessary. In the Czech Republic the Death Records Database is used to fulfil this task; its records may serve as a base for retrospective corrections of mortality estimates. Systematic bias of population-based data may be documented by following indirect methods (ordered for non-specific analyses to analyses requiring data from specific diagnoses): 1) Monitoring of interregional and temporal fluctuations. Increased ratio of this variability component indicates decreased quality of incidence and mortality data and probably registry organizational problems, as well. 2) Patients demographic typology, ration of gender or age categories according to published data or in comparison with other reference databases. 3) Monitoring of bias of incidence data caused by delayed registration. Very important fact that may lead to incidence underestimation in a number of diagnoses. In the study performed in the SEER database (NCI, 2001) Clegg et al. found that up to 12% of underestimation of some diagnoses caused by the delayed registration. It should be noted that this finding concerns a functional registry, in which the standard delay from diagnosis to registration is 1-2 years (Clegg et al., 2002). 4) Detection of the diagnosis at autopsy or DCO record. Figure 5. Assessment of interregional and temporal fluctuation of cancer incidence in the Czech National Cancer Registry Figure 4. Assessment of interregional and temporal fluctuation of cancer incidence in the Czech National Cancer Registry A higher proportion of such records may indicate nonrepresentativeness of the population-based registry. Nevertheless, we also have to take into account specifics of individual diagnoses, states, and historical context. For example, the number of autopsy diagnoses in the Czech Republic was significantly increased before 1989, because the health care system management led to a higher number of autopsies. 5) Number of repeated malignancies. This factor significantly fluctuates among registries; however, there is a rule that number of repeated malignancies should rise with the registry duration and size. Low occurrence of repeated malignancies in long-term registries (< 3%) may be considered as indicator of a worsened representativeness of the registry (Capocaccia et al., 2003). 6) Assessment of differences among diagnoses in records completeness. For example Capocaccia et al. (2003) demonstrated in the EUROCARE-3 database analysis that cancer diseases diagnosed and accessible with difficulty (lung, digestive tract) had the lowest ratio of verified diagnostic results, while records of more easily detectable diagnoses (breast, melanoma) were in a better status. 7) Survival analysis, which should in a representative registry elicit results comparable with international data, if standard therapeutic procedures comply. Rather diagnoses with bad prognosis are sufficient for this purpose, which offer more complete survival data within a shorter time period, e.g. acute leukaemia (Capocaccia et al., 2003) KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

86 Role of population-based registries in oncology and current situation in the Czech Republic. Table 3: CNCR diagnoses groups in light of determination of clinical stage and TNM. Period (TNM classification 4th and 5th edition). N Stage and TNM stated Stage or TNM stated Stage stated TNM stated Missing TNM classification Stage and TNM not stated DCO/ autopsy diagnosis Early death Patient not treated Stage and TNM not stated C00 C08 Oral cavity % 2.6% 3.3% 4.1% 1.1% 1.7% 7.8% C09 C10, C12 C14 Pharynx % 3.2% 2.7% 0.8% 4.3% 2.2% 1.5% 7.0% C11 Nasopharynx % 3.9% 1.9% 5.1% 0.6% 2.6% 6.9% C15 Esophagus % 3.0% 5.1% 10.8% 5.8% 11.8% 8.8% C16 Stomach % 2.3% 5.4% 10.1% 5.8% 9.6% 4.0% C18 C21 Colon, rectum, and anus % 2.0% 4.4% 6.4% 2.3% 3.7% 5.3% C22 Liver and intrahepatic bile ducts % 3.2% 8.8% 31.0% 11.3% 13.4% 3.6% C23, C24 Gallbladder and biliary tract % 3.1% 10.1% 17.6% 7.1% 9.6% 5.1% C25 Pancreas % 2.3% 12.4% 18.5% 7.7% 13.2% 3.4% C32 Larynx % 2.3% 2.2% 3.8% 1.1% 1.7% 4.5% C33, C34 Trachea, bronchus, and lung % 2.2% 6.9% 0.1% 12.0% 4.0% 5.1% 3.5% C43 Melanoma of skin % 1.6% 3.2% 1.6% 0.3% 0.5% 5.6% C44 Other malignant neoplasms of skin % 0.6% 3.7% 0.1% 0.0% 0.1% 1.8% C50 Breast % 2.2% 3.7% 3.5% 1.0% 1.0% 6.7% C51, C52 Vulva and vagina % 5.4% 3.6% 2.2% 1.8% 3.8% 13.7% C53 Cervix uteri % 8.0% 3.4% 2.0% 1.3% 2.6% 12.6% C54, C55 Uterus % 5.1% 3.5% 0.5% 2.8% 1.1% 2.7% 12.1% C56 Ovary % 4.9% 4.9% 5.6% 2.4% 2.3% 10.2% C61 Prostate % 5.2% 4.7% 6.4% 1.5% 7.1% 19.9% C62 Testis % 2.4% 3.4% 0.9% 0.2% 0.2% 9.9% C64 C66, C68 Kidney and other urinary organs % 2.8% 7.6% 9.1% 2.1% 4.5% 9.0% C67 Bladder % 5.2% 3.2% 3.6% 1.8% 3.2% 23.7% C70 C72 Brain and spinal cord % 23.9% 4.6% 31.6% 15.9% 7.9% 7.9% C73 Thyroid gland % 3.1% 3.2% 5.8% 1.1% 1.0% 7.9% C81 Hodgkin disease % 70.5% 0.1% 27.6% C82 C85 Non-Hodgkin lymfoma % 50.2% 0.9% 45.3% C90 Multiple myeloma % C91 C95 Leukaemia % Carcinoma in situ % 0.6% 0.9% Other malignant neoplasms % 17.2% 7.2% 9.0% 15.7% 9.2% 9.0% 3.7% All malignant neoplasms % 4.7% 4.7% 2.3% 6.8% 2.8% 4.3% 5.9% Figure 6. CNCR diagnoses groups in light of determination of clinical stage and TNM 70 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

87 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 7. CNCR diagnoses in light of determination of clinical stage and TNM KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

88 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 8. CNCR diagnoses in light stated clinical stage and TNM 72 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

89 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 9. Structure of CNCR records with unreasonably missing clinical stage and TNM proportion of diagnoses Figure 10. Typology of the CNCR records with unreasonably missing clinical stage and TNM age and sex KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

90 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 11. Typology of the CNCR records with unreasonably missing clinical stage and TNM proportion of regions Figure 12. Typology of the CNCR records with unreasonably missing clinical stage and TNM health care facilities 74 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

91 Role of population-based registries in oncology and current situation in the Czech Republic. The Czech National Cancer Registry: quality analysis of data for tumour identification All rules stated may be naturally applied to the CNCR data and acquire a clear idea about the quality of key data. The preceding discussion concerning figure 3 showed that the CNCR data quality improves in time and variability of the incidence and mortality estimates fluctuates in an acceptable range. The analysis stated below aims to document a diagnostic reliability of the CNCR database and thus its usability for important analyses, e.g. of overall survival. It is an analysis of extraordinary importance, since there is no similar data set available in the Czech Republic and if the CNCR data were shown not to be sufficiently reliable, Czech oncology would lose a unique chance to set up reference standards for a number of epidemiological and clinical indicators. Another fact should not be omitted that the CNCR is a result of the work of hundreds of physicians and experts,, which should be appropriately appreciated and used. Table 2: Short overview of development in TNM classification of malignant neoplasms SEX All records (n=469,235) Males 50.1% 54.3% Females 49.9% 45.7% AGE Median 68.0 let 68.0 let Mean 65.4 let 66.2 let REGION Prague 12.5% 15.7% Středočeský 10.3% 9.1% Jihočeský 6.5% 3.9% Plzeňský 6.1% 3.6% Karlovarský 2.9% 3.1% Ústecký 7.6% 6.6% Liberecký 4.0% 2.4% Královéhradecký 5.1% 10.1% Pardubický 4.8% 8.2% Vysočina 4.6% 4.7% Jihomoravský 11.4% 6.2% Olomoucký 6.8% 6.0% Moravskoslezský 12.1% 16.3% Zlínský 5.5% 4.2% FOLOW-UP FACILITY Large health care facilities 16.7% 21.4% Medium health care facilities 27.8% 37.1% Small health care facilities 35.9% 34.8% Facility not stated 19.6% 6.6% Stage and TNM not stated (n=27,833) Figure 4 shows a basic scheme of the CNCR record stratification according to their diagnostic part s completeness. The scheme has been proposed according to described and literature-reasoned categories stated above with the aim of discriminating partly incomplete records from records incomplete for objective reasons and records that must be marked as incorrect. Performed evaluations show evident improvement of the CNCR data quality in time (figure 5 and table 3), which cannot be substantiated by a mere development of the TNM classification, but also by development of the registry itself. The TNM classification maturation is also noticeable in figures 6 to 8, where the recent CNCR data include only 2.3% incomplete records because of missing TNM classification. It should be emphasized that there are specific groups of tumours even nowadays, for which the TNM system has not been defined. These diagnoses are objectively incompatible with the major standard of solid tumour registration and should be therefore assessed separately. This will mean no bias from a populationbased point of view, since they are diagnostic groups with a relatively low to very low incidence (hematooncological malignancies, tumours of brain and central nervous system, in situ tumours, other skin tumours...) (Davis et al., 1997; Fritz et al., 2000; Castillo et al., 2004). Assuming that we can supplement a clinical stage into records with known TNM, the CNCR data, particularly from the period , reach completeness comparable with international databases (Clegg et al., 2002; Capocaccia et al., 2003; Eaker et al., 2006). Incidence of DCO and autopsy records is also acceptable in recent data in comparison with international databases (Capocaccia et al., 2003; Jemal et al., 2004). The data presented in figure 5 further documents a decreasing number over time of untreated patients and patients with early death after diagnosis. We used a threshold value of 1 month from diagnosis for incomplete records due to early death, which is a criterion generally used in international projects analyses, e.g. EUROCARE-3 (Gatta et al., 2000; Capocaccia et al., 2003; Micheli et al., 2003). During the overall assessment only 5.1% of records were found in the CNCR, in which both missing TNM classification and clinical stage are not reasonable. Very positive news is that the data completeness has risen over time and the recent period, which is most needed for clinical analyses, provides data in a sufficient quality (figure 5 and 8). It is further obvious that the problem of incomplete identification less concerns the most frequent diagnoses, such as breast, colorectal, and lung carcinoma, after 1995 (figure 6 and 7). Although the incomplete records of these diagnoses represent a high absolute number due to their high incidence, in relative numbers they do not affect overall assessment (figure 7). According to the rules mentioned above, the records with unreasonably incomplete diagnostics cannot be used e.g. for the overall survival analysis and other clinically oriented outputs. Typology of these records shown in figures and in table 4 demonstrates that the records approximately correspond to the total CNCR database in KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

92 Role of population-based registries in oncology and current situation in the Czech Republic. all key parameters (patients demographic characteristics, health care facility etc.). Therefore, a relatively low ratio of incorrect records is not a reason to worry about a systematic bias of results from the CNR database as a whole. Possible bias exists only for some interregional comparisons. A relatively high ratio (> 15%) of unreasonably incomplete records was detected in Prague and the Moravian-Silesian region (figure 11), which could interfere with interregional comparison in some diagnostic groups, e.g. breast carcinoma. The result of the CNCR data quality audit (figures 4-11, tables 3 and 4) is a set of reliable records with correct diagnostics and a set of in incomplete records with clearly identified objective causes of problems. The filters stated above explain most of the records with missing TNM and clinical stage. A similar filtration procedure is used in most population-based studies that observe cancer patients survival in relation to diagnostics or treatment. Overall 5.1% of records with unreasonably missing both clinical stage and TNM classification were detected. However, even these records may be used in a routine incidence and mortality assessment, because the diagnosis is known. The detailed analysis documented that they typologically correspond to the whole CNCR database. Therefore, these records may be omitted in the health care assessment without risk of a systematic bias or decreasing information value of the outputs. These records may also be subject to so called imputation of the clinical stage by model estimate. This step is not, however, necessary in many populationbased analyses; on the contrary, a standard procedure mentioned in the literature is to assess the unstaged records separately and to calculate overall survival directly for them, which usually provides results with comparable stages 2 and 3 (e.g. Sant et al., 2003; Brenner a Arndt, 2005). Literature: Bernaards C.A., Farmer M.M., Qi K., Dulai G.S., Ganz P.A., Kahn K.L.: Comparison of two multiple imputation procedures in a cancer screening survey. J. Data Sci,1, 1-20, Brenner H., Arndt V.: Long-term survival rates of patients with prostate cancer in the prostate-specific antigen screening era: population-based estimates for the year 2000 by period analysis. J. Clin. Oncol., 23(3), , Capocaccia D., Gatta G., Roazzi P., Carrari E., Santaquilani M., de Angelis R., Tavilla A., Eurocare Working Group: The EUROCARE-3 database: methodology of data collection, standardization, quality control and statistical analysis. Annals of Oncology, Supplement 5: v14-v27, Castillo M.S., Davis F.G., Surawicz T., Bruner J.M., Bigner S., Coons S., Bigner D.D.: Consistency of primary brain tumor diagnoses and codes in cancer surveillance systems. Neuroepidemiology, 23, 85 93, Clegg L.X., Feuer E.J., Midthune D.N., Fay M.P.: Impact of reporting delay and reporting error on cancer incidence rates and trends. J. Nat. Cancer Inst., 94 (20), , Davis F.G., Bruner J.M., Surawicz T.S.: The rationale for standardized registration and reporting of brain and central nervous system tumors in population-based registries. Neuroepidemiology, 16(6), , Eaker S., Dickman P.W., Berquist L., Holmberg L.: Differences in management of older women influence breast cancer survival: results from a population based database in Sweden. PloS Med 3(3): e25, Fritz A., Percy C., Jack A. et al.: International classification of diseases of oncology. Geneva: World Health Organization, Gatta G., Capocaccia R., Ries L. et al.: Towards a comparison of survival in America and European cancer patients. Cancer 2000, 89: , Horton N.J., Lipsotz S.R.: Multiple imputation for the fatal accident reporting system. Appl Stat J. Royal Stat Soc., Series C, 40, , Jemal A., Clegg L.X., Ward E. et al.: Annual report to the nation on the status of cancer, , with a special feature regarding survival. Cancer, 101(1), 4 27, Micheli A., Baili P., Quinn M.: EUROCARE Working Group. Life expectancy and cancer survival in the EUROCARE-3 cancer registry areas. Ann Oncol, 14 (Suppl 5): v28-v40, NCI (National Cancer Institute): Surveillance, Epidemiology and End Results Sant M., Allemani C., Capocaccia R., Hakulinen T., Aareleid T. et al.: Stage at diagnosis is a key explanation of differences in breast cancer survival across Europe. Int. J. Cancer, 106, , Sant M. and EUROCARE Working Group: Differences in stage and therapy for breast cancer across Europe. Int. J. Cancer, 93, , Shetty M.R., Reinam H.M.: Tumour size and axillary metastasis, a correlative occurrence in 1244 cases of breast cancer between 1980 and Eur. J. Surg. Oncol., 23, , Schenker N., Taylor J.M.G.: Partially parametric techniques for multiple imputation. Comput. Stat. Data Anal., 22, , KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

93 Paper III. Mužík J., Faber E. Epidemiology of chronic myeloid leukemia. In Chronická myeloidní leukémie. 1st edition. Praha : Galén, ISBN , s (original article in Czech)

94 Published in Chronická myeloidní leukémie. 1st edition. Praha : Galén, ISBN , s (original article in Czech) 9 2. EPIDEMIOLOGY OF CHRONIC MYELOID LEUKEMIA JAN MUŽÍK, EDGAR FABER Not many epidemiological studies have been published on CML. In statistical studies and reviews dedicated to the occurrence of malignant diseases in general, patients with CML are often included in the category of myeloproliferative disorders, myeloid or chronic leukaemia. Even more often, it is not specified whether the diagnosis has been confirmed by cytogenetic or molecular tests of Ph chromosome or BCR-ABL gene presence or merely by blood and bone marrow tests, and Ph-positive CML is reported together with Ph- and BCR-ABL- negative cases. CML is usually reported to constitute 15-20% of all leukaemia cases. Total incidence figures differ widely and range from 0.6 to 2.0 new cases per persons per year (Ries LAG, 2008, Swedish Cancer Registry, Rohrbacher M, 2009). Unlike lymphoproliferative conditions, whose numbers have increased greatly in the last few years, CML incidence has not according to some sources significantly changed in the last few decades, thus showing rather loose etiological connection with civilisation factors such as environment or nutrition (Figure 1). Other authors, on the contrary, report a slight yet significant (max. 2.2%) decrease in CML incidence (between 1973 and 1998 in USA, Xie Y, 2003). The condition becomes more common with increasing age; it s very rare in childhood where it accounts for 3% of childhood leukaemias and represents 10% of leukaemias in older children and teenagers (see Chapter 8). The vast majority of patients are seniors (Figure 2). According to epidemiological registry data, the median age of diagnosis is in the late 50 s (Hehlmann R, 2007). In patients participating in Rate per 100, Year of diagnosis Figure 2-1. CML incidence in whites in the USA in period according to SEER. Males Females Both Sexes Source: SEER, 9 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, and Atlanta). Rates are per 100,000 and are age-adjusted to the 2000 US Std Population (19 age groups - Census P ). Available from:

95 10 Chronická myeloidní leukémie Rate per 100,000 in age category < Age at diagnosis Males Females Both Sexes Source: SEER, 17 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, Atlanta, San Jose- Monterey, Los Angeles, Alaska Native Registry, Rural Georgia, California excluding SF/SJM/LA, Kentucky, Louisiana and New Jersey). Available from: Figure 2-2. Age-specific incidence of CML in whites in the USA in period according to SEER. clinical studies, this figure may be lower by 10 to 20 years. The probability of participation in a clinical study is 3.8 times higher for younger patients compared to seniors. According to Scottish registry data, patients with CML have, regardless of their age, a 50% chance of being included in a clinical trial. Similarly it s reported that older patients are not sufficiently indicated for target treatment by tyrosine kinase inhibitors. There are no rational reasons for this approach (shorter life expectancy or comorbidity are not valid reasons, as explained in Chapter 7.3). The most probable reason is the institutional effort to minimise financial costs and a lack of medical knowledge. Study of mortality shows that before Imatinib was introduced, CML annually accounted for 1.5 deaths per persons. Mortality increases with age from less than 0.1 for children under 14, 1.0 for patients below 45, and up to 8 deaths per persons a year for patients above 80. American registries from the late 90 s report a 75 percent decrease in CML mortality since the introduction of targeted tyrosine kinase inhibitor therapy (Figure 3) and this favourable trend of new diagnoses and related deaths is observed for all age categories (Figure 4). A change of first line treatment from oral chemotherapy and Interferon to target therapy from the 70ies brought significantly higher probability of 5 year survival: from 23.9% between 1975 and 1977 to 50.2% between 1996 and 2004 (Ries LAG, 2008) (Figure 5). The introduction of tyrosine kinase inhibitors also brought significantly higher prevalence of CML. Usually, a slightly higher occurrence of CML in males compared to females is reported (from 1.4:1 to 2.2:1). Females, on the other hand, have (according to a German study) a better chance of survival than males (Berger U, 2005). Familial occurrence of CML is unusual; siblings get sick very rarely. Racial or geographical predispositions to CML have not been described, either. No precise figure on CML occurrence in the Czech Republic has been determined by an epidemiological study. A study published in 2000 (Faber E, 2000) aimed to compare the occurrence of duly reported CML cases in selected regions of middle and northern Moravia, based on records of the National Oncology Register (NOR), with an overview of patients treated in respective catchments haematological clinics and with a database of patients, whose diagnosis was confirmed in the only cytogenetic laboratory in the

96 Epidemiologie mortality / incidence ratio Rate per 100, Year of diagnosis Figure 2-3. Trend of incidence and mortality of CML in whites in the USA Incidence Mortality Sources: incidence SEER, 9 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, and Atlanta); mortality - US Mortality Files, National Center for Health Statistics, CDC. Rates are per 100,000 and are age-adjusted to the 2000 US Std Population (19 age groups - Census P ). Available from: region at the time between 1990 and Between these years, cytogenetically Ph-positive CML was confirmed in 52 patients between the ages of 19 and 83 with a median age of 47.5 yrs. The male to female ratio was 1:1.08. This number of patients means an incidence of newly cytogenetically diagnosed Rate per 100,000 in age category < Age Figure 2-4. Age-specific incidence and mortality of CML in whites in the USA in period Incidence Mortality Source: incidence SEER, 17 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, Atlanta, San Jose-Monterey, Los Angeles, Alaska Native Registry, Rural Georgia, California excluding SF/SJM/LA, Kentucky, Louisiana and New Jersey); mortality: US Mortality Files, National Center for Health Statistics, CDC. Available from:

97 12 Chronická myeloidní leukémie 60 5-year relative survival [%] Year of diagnosis Figure 2-5. Trend of 5-year relative survival of patients with CML in whites in the USA according to SEER. Males Females Both Sexes Source: SEER, 9 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, and Atlanta). Available from: CML patients per persons per year. However, there were 190 newly diagnosed CML patients entered into the National Oncology Registry in the same period. They were 111 males and 79 females (a male to female ratio of 1.41:1) between the age of 11 and 91, with median of 63 yrs. We confirmed that 12 cytogenetically verified CML patients were not reported to the National Oncology Registry. The final average incidence of CML was thus 1.46 new cases per persons per year and the male to female ratio was 1.4:1 (Faber E, 2000). Our analysis confirmed the incidence and age distribution of Ph-positive CML known from literary sources. We demonstrated that patients, especially in higher age categories, are not referred to cytogenetic tests for confirmation of CML diagnosis and sometimes not even to haematologists. Since the CML diagnosis was cytogenetically confirmed in only one-fourth of patients, it s likely that several patients registered in the National Registry could have Ph-negative myeloproliferative disorder or an atypical CML. Published information from the National Oncology Registry was restricted to mere categories of lymphatic and myeloid leukaemia for a long time. Anonymous NOR data was only recently made accessible to the Czech Oncology Society ČLS JEP to enable their own detailed analysis; the epidemiological portion is also published in the form of interactive analysis on the Epidemiology of Malign Tumours in the Czech Republic portal accessible at (Dušek L. 2007). CML data from NOR covering were analysed (Indrák K, 2007); results revealed differences from data known from literature and showed an insufficiency of these data, especially in regard to mortality and survival rate. According to NOR, CML incidence remained stabilised for a long time at newly diagnosed patients per persons. At the end of 90 s, however, the incidence was decreasing to in A similar trend was observed for specific CML mortality on NOR data (Figure 6), which is quite high in respect to incidence when compared to other population databases (see Figure 3). The ratio of CML deaths to newly diagnosed patients, the so called mortality incidence ratio, in was 0.81 in the Czech Republic (according to NOR), and 0.48 in the US (according to SEER); in it was 0.59 according to NOR and 0.24 according to SEER. The non-favourable values and generally lower incidence in NOR can be largely explained by incomplete and incorrect data in NOR, as supported by the high ratio of cases reported as diagnosed during autopsy and based on death certificates (DCO), which accounted

98 Epidemiologie 13 Rate per 100, mortality / incidence ratio Year of diagnosis Source: National Cancer Registry of the Czech Republic, IHIS CR Incidence Mortality Figure 2-6. Trends of incidence and mortality of diagnosis C chronic myeloid leukaemia according to the National Cancer Registry of the Czech Republic, crude rate per 100,000 persons. for 12.1% (!) CML in NOR between , while the average value of all malignant tumour records except skin tumours in this period was 6.8%, i.e. approximately half. Incidence and mortality by age from NOR data then show that high mortality in respect to incidence is related mostly to age categories above 60 (Figure 7). Other potential sources of data on CML epidemiology in the Czech Republic are clinical patient databases. One of the goals of CAMELLIA project of the Leukaemia Section of Czech Haematology Society and Slovak Haematology Society is to follow all newly diagnosed patients with CML in catchments regions of participating clinics regardless of the treatment used. These regions include the entire Slovak Republic, north-east Moravia, east, west and south Bohemia and partially north and central Bohemia and Prague (52% of Czech population). Patients from the remaining parts of the Czech Republic are followed in respective clinical projects of a work group called CELL (The CzEch Leukemia Study Group for Life). Because catchments areas in the Czech Republic are not equally covered from a haematological point of view and patients can be treated regardless of their residency, clinical databases in the Czech Republic can not be considered population databases; this role is played only by the CAMELLIA project in the Slovak Republic. Available data shows that if the CML incidence was calculated from the number of patients recorded in CAMELLIA database, it would be 0.63 new CML cases per persons per year in (a male to female ratio of 1.03:1), while in Czech it would be 0.79 per persons (a male to female ratio of 1.20:1). It is therefore likely that not all newly diagnosed patients are recorded in the database. This can also be demonstrated by the relatively low age median of patients a mere 51 years in 673 patients with Ph-positive CML entered into registry since Analysis of age divided by first line treatment type also shows that tyrosine kinase inhibitors are used significantly more for younger patients, while a major portion of older patients receive another treatment (Figure 8). This finding is in contrast with both current foreign expert recommendations (European LeukemiaNet or NCCN) and the recommendations of the Czech and Slovak Haematological Society for treatment of CML, who state Imatinib as the first line treatment regardless of age. What is the expected CML incidence in the Czech Republic? According to the reports of the National Oncology Registry, CML incidence remained stabilised for a long time at new cases per

99 14 Chronická myeloidní leukémie Rate per 100,000 in age category < Age at diagnosis Source: National Cancer Registry of the Czech Republic, IHIS CR Males Females Both Sexes Figure 2-7. Age-specific incidence and mortality of diagnosis C chronic myeloid leukaemia according to the National Cancer Registry of the Czech Republic in period Age at diagnosis Mean Median tyrosinkinase inhibitors (N = 545) 50 years 52 years transplantation (N = 80) 37 years 36 years other treatment (N = 48) 64 years 65 years Proportion according to age category [%] Age structure of patients according to treatment < Age at diagnosis Proportion of patients [%] Proportion of types of treatment according to age < Age at diagnosis Figure 2-8. Age and main treatment of patients with CML - clinical data from project CAMELIA; patients from the Czech Republic and the Slovak Republic diagnosed in period Main treatment is defined according to following key: transplantation > tyrosinkinase inhibitors > other treatment.

100 Epidemiologie 15 Proportion according to age category [%] Mean Median NCR 59 years 62 years CAMELIA 50 years 52 years Age at diagnosis Figure 2-2. Age structure of patients with CML in the Czech Republic according to National Cancer Registry of the Czech Republic (NCR) and according to the clinical data from project CAMELIA in period NCR data are only from regions of the Czech Republic that are in catchment area of the centres participating in project CAMELIA. persons per year (a male to female ratio of 1.06:1), but it decreased in the late 90 s to in (a male to female ratio of 1.09:1). Ironically, this decrease can indicate a better quality of NOR content, when after the introduction of cytogenetic and molecular-genetic diagnostics it is more usual for only BCR-ABL/Ph-positive patients to be registered as CML patients. This could only be verified by further study comparing NOR records with clinical patients records. A comparison of the age structure of patients duly reported to NOR and patients in the clinical database of CAMELLIA project in also shows that, in contrast with expected reality, there are less patients below 60 in NOR and, on the contrary, less patients above 60 diagnosed and treated outside of haematological centres in CAMELLIA clinical database (Figure 9). Incidence figures in both NOR and clinical databases are therefore underestimated and the real values can be expected to be 20-40% higher, i.e newly diagnosed CMLs per persons. Since 2009, there has been a prospective study carried out within a project of the international scientific society, European Leukemia Net, and the pharmaceutical company, Novartis, called EUTOS for CML (The EUropean Treatment Outcome Study for CML), aimed at creating population registries of newly diagnosed CML patients in geographically defined regions of Europe. Both clinical databases of former Czechoslovakia are a part of this project. It is reasonable to expect that the project will bring new interesting data on both CML epidemiology and current treatment practise. References Berger U, Maywald O, Pfirrman M, et al. Gender aspects in chronic myeloid leukemia: long term results from randomized studies. Leukemia 2005; 19: Dušek L, Mužík J, Kubásek M, Koptíková J, Žaloudík J, Vyzula R. Epidemiologie zhoubných nádorů v České republice [online]. Masarykova univerzita, [2005]. Avalilable from: Faber E, Beška F, Jarošová M, et al. K výskytu chronické myeloidní leukemie v severomoravském regionu. Prakt lékař 2000; 80:

101 16 Chronická myeloidní leukémie Helhmann R, Berger U, Pfirrmann M. Drug treatment is superior to allografting as first-line therapy in chronic myeloid leukemia. Blood 2007; 109: Indrák K, Papajík T, Faber E, et al. Kritická analýza dat o akutních a chronických leukemiích v národním onkologickém registru České republiky. Klin Onkol 2007; 20(Suppl.1): Ries LAG, Melbert D, Krapcho M, et al. SEER cancer statistics review, National Cancer Institute. Bethesda, MD Avalilable from: Rohrbacher M, Berger U, Hochhaus A, et al. Clinical trials underestimate the age of chronic myeloid leukemia (CML) patients. Incidence and median age of Ph/BCR-ABL-positive CML and other chronic myeloproliferative disorders in a representative area in Germany. Leukemia 2009; 23: Swedish Cancer Registry , Annual report publications of the Centre of Epidemiology at the National Board of Health and Welfare. Xie Y, Davies SM, Xiang Y, et al. Trends in leukemia incidence and survival in the United States ( ). Cancer 2003; 97:

102 Paper IV. Dušek L., Mužík J., Koptíková J., Žaloudík J., Klimeš D., Bourek A., Indrák K., Mihál V., Hajdúch M., Št rba J., Vyzula R., Abrahámová J. Data registries form indispensable information base of current oncology. Klinická onkologie 20, Supplement 1/2007, 53-62, ISSN X (original article in Czech)

103 Role of population-based registries in oncology and current situation in the Czech Republic. KLINICKÁ ONKOLOGIE 20, SUPPLEMENT 1/2007, 53 62, ISSN X (ORIGINAL ARTICLE IN CZECH) DATA REGISTRIES FORM INDISPENSABLE INFORMATION BASE OF CURRENT ONCOLOGY REGISTRACE DAT TVOŘÍ NEZBYTNOU INFORMAČNÍ ZÁKLADNU SOUČASNÉ ONKOLOGIE DUŠEK L. 1, MUŽÍK J. 1, KOPTÍKOVÁ J. 1, ŽALOUDÍK J. 2, KLIMEŠ D. 1, BOUREK A. 1, INDRÁK K. 3, MIHÁL V. 3, HAJDÚCH M. 3, ŠTĚRBA J. 1, VYZULA R. 2, ABRAHÁMOVÁ J. 4 1 FACULTY OF MEDICINE, MASARYK UNIVERSITY, BRNO 2 MASARYK MEMORIAL CANCER INSTITUTE, BRNO' 3 FACULTY OF MEDICINE, PALACKÝ UNIVERSITY IN OLOMOUC 4 THOMAYER UNIVERSITY HOSPITAL, PRAGUE Summary Registration of epidemiological and clinical data represents an essential source of information for the assessment of prognosis and health care results. However, determining factors must be respected to ensure functional registration on population-based data, i.e. data model of the registry, technological and organizational background of data collection, and quality management system. First of all, we have to consider objectives and range of registration, whether it is supposed to be generally epidemiological, or clinical. Previous experience shows that population-based registries covering a number of cancer diagnoses should be focused purely on epidemiology with a minimum of data about treatment. Responsible treatment assessment must include detailed information that is not available universally for all diagnoses: (1) prognostic risks of newly diagnosed patients, (2) record of treatment process and therapy response assessment, (3) schedule of follow-up care, (4) monitoring of short-term, overall, and disease-free survival, (5) time and cause of death. Each of these components is of different importance in various cancer diagnoses and cannot be omitted. Such an information standard may be assured only in a specialized clinical registry, preferably guaranteed directly by an appropriate expert medical society. On the contrary, a population-based epidemiological registration should be assured by automatic communication of hospital information systems and the registry database. This contribution comments on the main principles determining the functionality of cancer registries and methodologically summarizes recent findings from international literature. Key words: cancer epidemiology, data registration, clinical registries Souhrn Registrace epidemiologických a klinických dat představuje nepostradatelný zdroj informací pro hodnocení prognózy a výsledků léčebné péče. Pro zajištění funkční registrace populačních dat je však vždy nutné respektovat determinující faktory, tedy datový model registru, technologické a organizační zajištění sběru dat a systém kontroly kvality. V první řadě musíme zvážit cíl a rozsah registrace, zda má být obecně epidemiologická nebo i klinická. Z dosavadních zkušeností vyplývá, že plošné registry, jdoucí napříč onkologickými diagnózami, by měly být zaměřeny výhradně epidemiologicky s minimem záznamů o léčbě. Odpovědné hodnocení léčby musí totiž zahrnovat podrobné údaje, které není možné získávat univerzálně pro všechny diagnózy: (1) prognostická rizika nově diagnostikovaných pacientů, (2) záznam léčebného postupu a hodnocení léčebné odpovědi, (3) časový harmonogram dispenzární péče, (4) monitoring krátkodobého a celkového přežití a přežívání bez známek choroby (5) údaje o příčinách a datu úmrtí. Každá z těchto komponent má u různých onkologických diagnóz a klinických stadií jiný význam, žádnou nelze vypustit. Takový informační standard je možné zajistit pouze ve specializovaném klinickém registru, nejlépe vedeném pod přímou kontrolou k tomu příslušné odborné společnosti. Plošná epidemiologická registrace by naopak měla být zajišťována automatickou komunikací nemocničních informačních systémů a databáze registru. Příspěvek komentuje hlavní principy určující funkčnost onkologických registrů a metodicky shrnuje poznatky z mezinárodní literatury. Klíčová slova: epidemiologie nádorů, registrace dat, klinické registry Introduction The objective of this article is the methodological summarization of the main principles determining the quality of data registration in oncology and to provide an overview of international literature. A quick view into any international database reveals that phrases such as population-based study, survey, registry are very frequent. Registries include all diagnostic groups and deal with epidemiology and risk factors, as well as diagnostics and treatment at least with the same frequency as studies found under the key words clinical trial or evidence-based medicine. Although the retrospective data registration has frequently been employed, it is also often doubted as a source of reliable data. And there is also KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

104 Role of population-based registries in oncology and current situation in the Czech Republic. the fact that registries face a number of barriers, such as laborious data collection, the need for collaboration among various institutions in data centralization, processing of personal data etc. Therefore, it seems that there are enough reasons to look at the data registration suspiciously, which is, furthermore, supported by a frequently proclaimed opinion that registries cannot overcome information value of randomized clinical trials. bring a nearly unbiased picture of reality that may be used at all levels of the health care system, from a patient to health care facilities to regions and countries. If we based therapy assessment exclusively on published clinical trials, we would overlook a majority of patients and risk statuses that could not be included in clinical trials for various reasons. International sources demonstrate that published clinical trials contain only 5 10 % of cancer patients and the results Table 1: Selected international sources of oncological data and cancer registries DATA SOURCE GLOBOCAN Cancer Incidence in Five Continents (CI5 I-VIII) WHO mortality database Automated Childhood Cancer Information System (ACCIS) accis.htm The Surveillance, Epidemiology, and End Results (SEER) Program EBMT Registry org OPERATOR The International Agency for Research on Cancer (IARC), World Health Organization. The International Agency for Research on Cancer (IARC), World Health Organization World Health Organization, The International Agency for Research on Cancer (IARC) The International Agency for Research on Cancer (IARC), European Commission, European Network of Cancer Registries (ENCR) National Cancer Institute (NCI), U.S. National Institutes of Health The European Group for Blood and Marrow Transplantation (EBMT) In spite of this, we will attempt to convince you that registration projects are not less reliable in comparison with prospective clinical trials and that such assessment is tendentious and incorrect, respectively. Both of these information sources may not only exist in parallel, but they may also be synergistic and complementary to each other and form complex background information for clinical research and health care management. We will document that the data registration population on a population level has been one of the flagships of current medicine and a strategic goal of a number of important international institutions and activities (see selection of international projects based on registries in table 1). Say everything is a question of purpose, basic hypothesis and assurance of adequate data quality, as discussed below. Clinical and epidemiological registries represent a key information source of modern medicine. Reliable registries CONTENT, PARAMETERS, AND TARGET POPULATION Estimated national incidence, mortality, and prevalence of 27 main cancer diagnoses according to sex and age categories / Absolute numbers of cases, crude incidence on persons and age standardized data ASR(W) / All world countries as of Cancer incidence according to diagnose, sex and age categories and according to individual registries / Absolute numbers of cases, crude incidence on persons and age standardized data ASR(W) / Cancer registries from the whole world observed in ; data availability in time differs for individual registries. Cancer mortality in individual countries according to sex, age, and diagnosis / Absolute numbers of cases, crude incidence on persons and age standardized data ASR(W) / Further, populationbased data of individual countries, worldwide data of the WHO databank, WHO Statistical Information System (WHOSIS) ( Period is available according to individual national sources. Incidence and survival of children and adolescents with cancer according to individual registries and diagnoses. European cancer registries of children patients, available period is according to individual registries. Incidence and mortality on cancer according to age, sex, race, diagnosis / Absolute numbers of cases, crude incidence on persons and age standardized data / USA, period Database focused on patients with bone marrow transplantation (hematooncological diagnoses) / Numbers of patients, survival, basic clinical parameters / Hematooncological centres of the whole world. Period is available. Web pages containing very useful overview of cancer registries and presentations available online regarding cancer epidemiology. The registry section contains more than 60 functional links. are, therefore, unrepresentative from a population point of view (Goodwin et al., 1988; Cassileth, 2003; Satariano a Silliman, 2003). Moreover, it is known that trials with positive results are published more easily than statistically non-significant or problematic results, which have a lower chance of being published (Krzyzanowska et al., 2003). Thus, population-based data, although heterogenic and problematic, may also be considered as a valuable tool for the correction of information strategy dictated by clinical trials. However, these facts are not stated here as an overall critique of randomized clinical trials or evidencebased medicine. We could not accurately evaluate the effectiveness and safety of various therapeutic regimens, new drugs and technologies only on the basis of populationbased registries. Clinical trials have irreplaceable in this case, but they are neither the beginning nor end of clinical 54 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

105 Role of population-based registries in oncology and current situation in the Czech Republic. research. Data analysis of population-based registries provides a much needed extension of clinical projects, which are always limited to a certain degree by specific focus and patient inclusion criteria. We state the following examples selected from international literature as a document of importance of population-based registries, in which long-term collected population-based data represented the only information source that was irreplaceable by any other type of research. Information benefits of population-based registries documented on selected cases Epidemiological surveys. Population-based registries are naturally close to the assessment of epidemiological data, trends, and associated population risks. Epidemiological analyses are sometimes even considered as the only output of population-based registries, which is an unacceptable simplification. Even the simplest surveys on incidence and mortality reveal significant trends in results of diagnostics and treatment of malignant tumours and therefore contribute to the health care quality assessment on any level of health care management. Epidemiological data is also necessary for the optimization of health care for various age groups of patients and for the essential support for national programmes of anticancer prevention (e.g. Potosky et al., 2001; Scheiden et al., 2003). The role of Figure 1. Quality of worldwide cancer data according to the methodology of cancer incidence and mortality determination used in the international database of malignant tumours GLOBOCAN a. Methodology of incidence determination national cancer registries, however, has not been perceived the same way round the world. Even lots of developed countries build only regional registries as epidemiological models, subsequently extrapolated to the population level. Literature is very rich in this area; let us mention only the EUROCARE 3 study and the associated work of Capocaccio et al. (2003) on behalf of other studies, which summarizes the status and functionality of 67 populationbased and regional registries of 22 countries. Assessment of health care results and quality. This is definitely the most valuable contribution of populationbased registries requiring representative data collection in a target population and its comprehensive analysis without misrepresenting limitations (Bethell et al., 2004). Randomized trials contribute to health care assessment by prospective consideration of possible treatment strategies and methods. Unfortunately, however, this is usually without the feedback of data from real practice. Moreover, some serious oncological diseases are not even adequately monitored by comparative studies. This concerns particularly advanced stages of the disease, for which data comparing effectiveness of anticancer therapy and the best available palliative therapy are often completely missing. Population-based studies may bring fundamental findings in comparison of life prolongation by aggressive and expensive therapy and possible effects of palliative care (Shanafelt et al., 2004). The number of papers has been increasing, documenting that patients with advanced stages of malignant tumours are often treated excessively without visible effects on population level (Emanuel et al., 2003). Thus, populationbased data may contribute to the rationalization of treatment decision-making, reduction of costs, and improving quality of life, while keeping the achieved results related to survival (Sankila et al., 2003). Assessment of diagnostics and risk factors. Population-based registries provide valuable data on cancer diagnostics. Registration of clinical stages of specific disease may serve as an indicator of how the given unit (country, region, hospital, etc.) is able to detect early stages of the monitored disease. Early diagnostics are certainly the best way to improve health care results and simultaneously reduce treatment costs. Populationbased data may further KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

106 Role of population-based registries in oncology and current situation in the Czech Republic. 1b. Methodology of mortality determination Further literature sources: Bray et al. (2002), Ferlay et al. (2004), Parkin et al. (1999), Pisani et al. (1999), Pisani et al. (2002) contribute to understanding the effects of various risk factors, most of which affect disease occurrence and progress and subsequently treatment options. Populationbased registries are the perfect tool for monitoring the incidence of these and further intercurrent facts; they allow the study of their relationships, and the weighting of their prognostic effects (Yancik et al., 1998; Piccirillo et al., 2002; Hershman et al., 2004; Geraci et al., 2005). Research extending and correcting results of clinical trials. Due to representative coverage of a wide range of patients, the systematic analyse of population-based data may clarify problems that are interpreted vaguely or inconsistently in clinical trials. Vinh-Hung et al. published an article in 2002 that reacted to indefinite conclusions from studies focused on early stages of breast carcinoma and proving that radiotherapy after surgery has a low effect on patient survival. The summary analysis was based on survival modelling by Cox regression on aggregated records of 83,776 women from the SEER database (Surveillance, Epidemiology and End Results, NCI, 2001). Results unambiguously demonstrated that radiotherapy applied after surgery significantly improves survival regardless of type of surgery. The best results were achieved in patients without affected nodes, where the breast was maintained. We recommend this paper to read also because of a reliable discussion, in which the disadvantages of populationbased data are summarized, particularly updates of follow-up records, insufficient records on type of therapy and the manner of its application, etc. For this reason most of the population-based registries may be designated as monitoring projects, which reveal fundamental relationships in a real patient population, but cannot often be used as a reliable source for determination of treatment standards and guidelines. Health care optimization. Reliable population-based registries offer an authentic picture on the structure of the target patient cohort and are, therefore, an essential information source for the optimization of health care availability and effectiveness. Eaker et al. (2006) analyzed data from a regional registry of women with breast cancer in Uppsala (the registry had contained more than 12,000 records since 1992 with more than 10-year follow-up. The objective of the study was to find out whether there were differences in quality and extent of health care for older and younger women and whether these differences are reflected in resulting survival. The study, in which, among others, representativeness and quality of input data is demonstrated, proved that women above 70 years are provided impaired diagnostics (i.e. later detection of disease, less nodes examined...) and subsequently less intensive treatment. This fact resulted in decreased survival of this age category in comparison to others. The work further documents that observed differences cannot be explained by co-morbidity and recommends a solution in form of developing clinical standards for higher age categories of women. Population-based registries are background information for health care management The outputs listed above are certainly convincing proof of population-based data importance not only for oncology. The existence of information itself does not guarantee its actual employment for the benefit of patients and the health care system. Just like any other tool, the populationbased registries must document their contribution in real practice. Only the interest of management authorities (management of hospitals, private facilities, regional authorities, expert societies) in this data and its subsequent use can demonstrate the significance of population-based registries. Use of the information database in clinical practice may, however, be accelerated 56 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

107 Role of population-based registries in oncology and current situation in the Czech Republic. by reliable information service targeted on current hot topics of the health care system. The following points show an overview of possible information outputs that may only be representatively obtained from population-based databases and would be appreciated by all management levels. The list is adjusted to the needs of Czech oncology: Burden of health care facilities. Description of cancer epidemiology in selected attraction zone containing basic parameters, such as incidence, mortality, and prevalence of diagnostic groups. This information may be used for the optimization of a specialized facilities network or the systematic localization of sources within the regions. In the Czech Republic such information may be obtained from an analysis of Czech National Cancer Registry (CNCR) and demographic data. Assessment of differences among regions. Long-term regional differences in occurrence of some diagnoses may indicate influence of risk factors. Conversely, shortterm fluctuations indicate problems with quality of data collection. A representative epidemiological profile of the selected region is a valuable background for health care optimization. Assessment of interregional differences in time trends may also reveal organization problems, which can then be addressed and solved. Diagnostics accessibility and results. Very valuable information determining in what stage of the disease, in what types of health care facilities and by what methods malignant tumours have been diagnosed in defined target area. This data allows for reviewing work of diagnostic centres, optimizing their network or eventually targeted informing general practitioners. In the Czech Republic such assessment may be performed for most of the diagnoses based on data from the CNCR alone. Diagnostics accuracy. For some disease (e.g. malignant lymphomas) immunophenotypic and genetic examinations are the basic condition of accurate diagnostics, initiation of targeted therapy and monitoring of therapy response (including presence of minimal residual disease). So long as they exist, it is possible to use disease registries to assess the accessibility, use, and benefits, and thereby the legitimacy of a facility to diagnose and treat specific diseases (including rational use of expensive therapy). Feasibility studies of health care programmes and research projects. Characterization of the target population involved in the proposed project is often omitted in the Czech Republic. We obtain not only data on required cohort size e.g. for a clinical trial, but also a realistic estimation of duration, i.e. how long it will take to collect required amount of data at given incidence and health care organization. Analysis of possible demographic and regional factors biasing the results is also valuable. Such analysis should compulsorily precede all expensive health care projects and studies. Representative assessment of long-term health care results. Significant information output implementing reference standards based on achieved patients survival. Without representative population-based data this assessment is often limited to locally available estimates of survival, which are inaccurate when extrapolated to the whole population (e.g. Christakis and Lamont, 2000). Alternatively, overall survival may be replaced by short-term parameters, such as therapy response (e.g. ratio of patients with complete Figure 2. Possible connection of data sources in automatic collection of populationbased data. remission), which cannot, however, be considered as a full-value substitution (e.g. Shanafelt et al., 2004). The Czech Republic has the advantage from this point of view, because the CNCR data allows for analysis of overall survival for most diagnostic groups, moreover with retrospective verification against an independent population-based database (Death Records Database DRD). Therapeutic results in advanced and terminal stages. Very valuable information output, recently urged also in international literature (Emanuel et al., 2003; Shanafelt et al., 2004). Prevalence of patients in these serious stages itself is very valuable and should be monitored in time and according to regions or hospitals attraction zones. Monitoring of advanced malignant tumours in KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

108 Role of population-based registries in oncology and current situation in the Czech Republic. the Czech Republic is possible due to data from CNCR and DRD, including analysis of resulting survival. Overview of applied treatment strategy, drugs, and total costs of the treatment. Essential information output, which should be the basis for the optimization of anticancer treatment costs. In relation to diagnostic groups and various clinical stages, we can define reference standards for therapy procedures and monitor costs development and structure. Unfortunately, this type of output cannot be obtained by epidemiological data analysis alone; a representative database about therapeutic procedures and applied drugs is necessary. In the Czech Republic this means analytical data mining of health care payers databases in connection to CNCR (see also Figure 2) or simultaneous data mining of specific disease registries guaranteed by expert societies, providing they exist. Assessment of processes conditioning diagnostics and therapeutic care. Although it is not obvious from the title, this is the data to be obtained and analyzed most hardly. This information concerns the accessibility of appropriate health care to various groups of inhabitants and further migration of previously diagnosed patients among various types of health care facilities within or even between regions. Outputs may be defined as absolute numbers and description, or related to a target population of specific size. This type of output is based on population-based epidemiological data, but this data alone is not sufficient. The analysis would have to involve databases of health care payers and at least together with clinical registries of large facilities (e..g. Richards, 1996) or specific disease registries guaranteed by expert societies, assuming they exist. Information value of a registry is determined by structure and extent of collected data It is obvious that every registry may provide only information and outputs that are well-founded by registered parameters. The parametric extent of a registry is always a compromise between purposes of the registry initiator and acceptable work difficulty (price) of data collection for registry contributors. Low-cost registries are more likely to obtain valid data, but at the expense of a lower number of monitored parameters. There are generally three basic groups of parameters, of which only two are overall obligatory: 1) Minimal epidemiological data is an essential basis of any meaningful registration. Identification of patient and diagnose, record on date of diagnosis and death are entire minimum for population-based registration in oncology. 2) Continuous update of records in time ( follow-up ) is unfortunately often omitted, bud very important component of a functional registration. Updated data on patients survival is fundamental for relevant analysis of survival in a given time and further for estimates of mortality and prevalence. 3) Extended diagnostic and clinical data are often included in population-based registries due to research goals. Nevertheless, epidemiological registration does not require neither detailed analyses of risk and prognostic markers, or identification of therapeutic modalities and procedures. Therefore, there is always level to be considered, on which clinical and research goals may be fulfilled in a flat population-based registry, which has universal form for most of the cancer diagnoses. After consideration of all problematic aspects we can explicitly recommend maintenance of flat population-based registries above a minimal patient s record that includes basic identification data, diagnostics, and data on overall survival (the last with a regular update). Such a registry will be functional and data included can be sufficiently reliable. Detailed analysis of prognostic markers, patient risk typology and therapy assessment should be task for diagnose-specific clinical registries guaranteed by expert societies, which have more space to collect extended data in sufficient quality. These registries do not have to have epidemiological goals, may have narrow focus or may be operated only temporarily. Such strategy is very effective and does not reduce the amount of data available for health professionals. As it is obvious from the list of registration outputs, even the simplest epidemiological registration may be also used for the assessment of serious problems related to health care results and quality. Epidemiological registry with a minimal parametric record Records in the epidemiological registry should be reduced to the set of most essential parameters, always with respect to two basic principles: (1) principle of valid data and (2) principle of data availability. The following methodical notes may be stated on this topic (see also Dušek et al., 2004): 1) A flat registry should primarily monitor only data leading to basic epidemiological parameters and overviews. Such data is hoped to be collected accurately and reliably. In principle this includes identification of a patient and tumour and assurance of follow-up records. Even in such a reduced set, not all parameters are universal for all cancer diagnoses and it is necessary to create parametric space for all diagnostic groups, otherwise the areal registration loses meaning (some diagnoses do not have defined clinical stages, diagnostic records are very specific in all hematooncological diseases, etc). Information about the follow-up centre is important from practical point of view. It indicates that the patient is in charge of health care. The place should be a guarantee for continuous health care and source of further information about the patient, whose data may be assigned to the known type of a health care facility. 2) Epidemiological registries should provide demographic identification of the patient for eventual populationbased comparison (age, sex, region, occupation, education). On the other hand, it is pointless to observe subjectively affected data about risk factors and life 58 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

109 Role of population-based registries in oncology and current situation in the Czech Republic. style on this level, such as smoking. Such data will probably not be valid. 3) The highest attention must be paid to follow-up records, which are always filed in certain time periods in the follow-up facility. Their effective processing provides quick information about the number of patients with cancer in the population, how many are under followup, treated, relapsed (progressed), died of another disease and are lost from the statistics. Patients status in follow-up time must be recorded according to internationally valid code lists: complete or partial remission, stabilized disease, progression. Exact date of death must be recorded. Follow-up records must always include information, whether the patient was or has been treated in the previous period in relation to relapse/progression of the fundamental disease and dates of such events must be stated. A. Optimally adjusted registry provides particularly the following information: Epidemiological surveys: 1) Incidence of malignant tumours according to their location (diagnosis: topographic code) 2) Incidence of malignant tumours according to their histological type (diagnosis: morphological code) 3) Prevalence of malignant tumours 4) Mortality on a specific type of tumours 5) International epidemiological comparisons on population-standardized data 6) Epidemiological parameters in regional stratification Information related to individual records (with option to aggregate groups of records): 1) Diagnosis of the disease described by means of standard classification systems 2) Spread of the disease in time of diagnosis (TNM, ptnm, clinical stage, risk group...) 3) Method by which tumour was found 4) Basic time data: Date of the first visit Date of diagnosis (Date of the first treatment procedure) Date of the last follow-up or death 5) Basic demographic and social identification of the patient (education, occupation or prevailing occupation during his life), sex, and age at time of diagnosis 6) Information about patient s follow-up 7) Other diagnoses including other tumours 8) Patient s clinical status according to the WHO criteria B. Epidemiological registries further allow us to: 1) Compare changes in epidemiological parameters according to diagnostic codes (topographic and morphological), regions or other sorting criteria 2) Deduce phenomena in the population or medicine by means of changes in the disease spread. E.g. higher ratio of less advanced stages in individual diagnoses documents better prevention, diagnostics, cancer watchfulness and education of physicians, diagnostic methods etc. 3) Compare values of all parameters monitored according to health care facilities or their types and offer them the option to compare their own results with overall aggregated data. 4) Observe connections with the social sphere 5) Reveal indirectly possible delays or problems in diagnostics according to temporal data about the first visit to the physician and date of diagnosis determination 6) Compare changes in mortality on individual types of cancer and, therefore, assess effectiveness of the treatment programme and monitor overall survival. 7) Examine other connections explaining risk factors for occurrence and progress of cancer by comparison with the inhabitants registry and other registries (Death Records Database, natality, abortion rate etc.) 8) Perform region-specific analyses indirectly indicating broader connections, e.g. changes in environmental parameters etc. C. Epidemiological registry contributes to the assessment and planning on cancer care 1) Based on incidence curves and their extrapolations we can predict further development and estimate trends and thus plan long-term requirements of the regional health care for space, personnel, and equipment. 2) Based on temporal and space heterogeneity in epidemiological parameters, we can deduce changes in accessibility of diagnostic and therapeutic care. 3) Effectiveness of prevention programmes may be assessed according to trends in incidence and mortality. 4) Effectiveness of implemented screening pro-grammes may be assessed relatively accurately according to incidence, spread of the disease at time of diagnosis, and median mortality. Specialized diagnostic and clinical registries do not have to be neither minimized nor flat As mentioned above, we strictly recommend separation of epidemiological registration and specialized clinical registries. Specialized registries should be maintained only for diagnostic groups, where there is actual need, space, and will. Epidemiological registration may, therefore, fulfil its essential role more easily and with lower costs and output data will acquire appropriate weight. Aerial registration of detailed clinical data is definitely not necessary for all diagnoses and should be performed with respect to following requirements: 1) Relevant risk stratification of patients at diagnosis, including prognostic factors and treatment affecting examinations 2) Information on other diagnoses and patient s clinical status according to WHO 3) Decision and record of treatment plan (treatment strategy, definition of the procedure standard protocol and monitoring of its fulfilment 4) Records of therapeutic modalities with a clear KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

110 Role of population-based registries in oncology and current situation in the Czech Republic. statement of their schedule in treatment episodes (or eventually phases of the treatment protocol) 5) Assessment of patient s response after the primary therapy and designing the follow-up, clear identification of relapse (progression) time and, in such case, also repeated diagnostic records. Monitoring of treatment of these risk events. 6) Monitoring of applied procedures, drugs, eventually total costs of the treatment 7) Detailed description of causes of patient s death in relation to primary disease, complications, or treatment. Data quality as a factor determining usability of population-based registries If we want to use the registry data for the fundamental analyses mentioned above, their quality is a natural and essential condition. Nevertheless, this is unfortunately the weakest side of population-based registries, which usually collect data from multiple sources without the possibility of detailed inspection according to patient s documentation. Therefore, the population-based data must be processed with respect to this fact and each analysis should be supplemented with risk analysis of results bias. It should be noted that this is a worldwide problem; any aggregation of national or even international data logically brings a certain ratio of problematic records. Figure 1 documents sources of internationally available estimates of cancer incidence and mortality in various world countries. Data of the highest quality may be found in nationwide representative registries, which, however, are not available in all countries, including the developed states. Regarding incidence, estimates obtained from full regional registries may be considered as reliable if they are corrected by nationwide data on mortality. Data obtained from regional registries without possibility of populationbased correction, on the contrary, is very problematic (Figure 1). Population-based data may be generally assessed by three criteria of quality: Representativeness: assessed as completeness and representativeness of records Internal structure: temporal, diagnostic, and logical consistency of records Reliability: indicator assessed by inspection against independent data source These criteria define data quality more comprehensively than a simple number of newly diagnosed patients. Importance of individual criteria may change depending on the purpose of the specific analysis. For example, epidemiological surveys on a nationwide level will necessarily require documented representativeness of records, although all parameters do not have to be filled in (e.g. diagnosis found out during autopsy will be included in incidence, although the record will not be complete due to objective reasons). On the contrary, analyses focused on health care assessment may be performed with a separated part of all records (e.g. patients treated with a certain way) and logical consistence, completeness, and reliability of these records will be particularly emphasized. There are naturally minimal criteria of records consistency for oncological data, too, such as logical linking age age at diagnosis diagnostic record. Key properties of diagnostic records are the consistence of TNM classification, clinical stage and correct morphological code (WHO, 2000; Capocaccia et al., 2003). Representativeness is usually verified e.g. by comparison of ratio of sex or age categories with published data or with other reference databases. Quality and completeness of data is also demonstrated by ratio of tumours detected during autopsy. Occurrence of recurrent malignant tumours is also a significant factor, which fluctuates among registries, but their number should increase with a longer follow-up and registry size. Low number of recurrent malignancies in long-term registries (< 3 %) can be considered as an indicator of lower registry quality and representativeness (Capocaccia et al., 2003). Reliability of records is often verified against external data sources, some studies of population-based data are based on national standards of quality and document data inspection by an independent group of experts (e.g. Swedish Data Inspection Board cited in Eaker et al., 2006). When assessing population-based data quality, we cannot omit the fact that record quality is also given by the registration process itself, its management and error rate. Clegg et al. studied the effects of delay in reporting of incidence of main cancer types in the database of the SEER programme (NCI, 2001) and found out that commonly reported incidence may be significantly underestimated in comparison to the real situation. The continuously reported incidence for various diagnoses reached % of the final number of records obtained after completion of records in the following years. The delay significantly differed among diagnoses, but it would take an average of 4-17 years to achieve 99 % completeness. Authors therefore suggested correction of recent data on possible bias as a result of incomplete reporting (Clegg et al., 2002). Quality of population-based registries is certainly affected by the technological and software background of data collection, as well. Nowadays, we can explicitly refuse data collection for population-based registries by means of paper forms with delayed collection and subsequent, even more delayed, validation. Modern population-based registry must be implemented in electronic collection and management of clinical data and operated without the workload of health professionals and physicians. Status of population-based registries in the system of oncological data collection Everyone who has ever organized collection of clinical data without support of hospital information systems knows that long-term, successful data registration cannot run in a vacuum and without databases that have already been operating and routinely filled. Only those registries may be functional that minimize the workload of physicians in data collection. Although epidemiological registration is relatively simple, regular filling forms, searching and update of data quickly fatigue motivation of already overburdened health care personnel. And on the other 60 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

111 Role of population-based registries in oncology and current situation in the Czech Republic. hand, because it is relatively simple documentation, there is no reason not to export it automatically from hospital information systems. Such a system would get physicians to the appropriate position, i.e. expert data inspectors. Current information technologies enable it, even with a required level of data safety. Internet and online technologies have been progressing in medicine and increase users productivity and comfort. (Tarlov et al., 1989; Fischbacher et al., 2000; Ruland et al., 2003; Bethell et al., 2004). Figure 2 indicates the logical connection of more sources that synergically bring higher information value. Besides epidemiology, we can also imagine the centralization of the most important clinical data, which would in combination with health care payers databases acquire an essential economical dimension, as well. That is how all of the three dimensions obligatory for a responsible assessment of health care results and quality can be comfortably completed (Figure 3). Although the schemes could look as a far future for the Czech health care system, in Figure 3. Three basic dimensions essential for a comprehensive health care assessment confrontation with the international literature we will soon find out that this is highly up-to-date reality. Automatized data collection from hospital information systems must be perceived as an essential condition for further development. These problems have also been solved by a number of developed European countries (e.g. Chouillet et al., 1994; Sant et al., 2003). Conclusion Whether you like it or not, the current health care system has been separated into different parts that are limited by lack of mutual fact-based communication. Concerns of health care facilities and expert medical societies do not always have to be in agreement with the system management of health care payers politics. Collection and analysis of reliable population-based data is probably the only tool for credible reasoning and supporting one s purposes in such a heterogenic environment. It is, however, also a real tool of self-control and a visible mirror of achieved results. Literature: Bethell C., Fiorillo J., Lansky D., Hendryx M., Knickman J.: Online consumer surveys as a methodology for assessing the quality of the United States health care system. J. Med Internet Res., 6(1), e2, Bray, F., Sankila, R., Ferlay, J. and Parkin, D.M.: Estimates of Cancer Incidence and Mortality in Europe in Eur. J. Cancer 38, , Capocaccia D., Gatta G., Roazzi P., Carrari E., Santaquilani M., De angelis R., Tavilla A.: Eurocare Working Group: The EUROCARE-3 database: methodology of data collection, standardization, quality control and statistical analysis. Annals of Oncology, Supplement 5: v14-v27, Cassileth, B.R.: Clinical trials: time for action (editorial). J. Clin. Oncol. 21: , Eaker S., Dickman P.W., Berquist L., Holmberg L.: Differences in management of older women influence breast cancer survival: results from a population based database in Sweden. PloS Med 3(3): e25, Clegg L.X., Feuer E.J., Midthune D.N., Fay M.P.: Impact of reporting delay and reporting error on cancer incidence rates and trends. J. Nat. Cancer Inst., 94 (20), , Dušek L., Abrahámová J., Indrák K., Vyzula R., Žaloudík J., Vorlíček J.: Registration of epidemiological data in oncology and its importance for the assessment of health care quality [in Czech]. Klinická onkologie, Supplement 2004, 39 44, Emanuel E.J., Young-Xu Y., Levinsky N.G. et al.: Chemotherapy use among Medicare beneficiaries at the end of life. Ann Intern Med 138, , European Network of Cancer Registries, International Agency for Research on Cancer; Ferlay J., Bray F., Pisani P. and Parkin D.M.: GLOBOCAN KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

112 Role of population-based registries in oncology and current situation in the Czech Republic. 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004; Feuer E.J., Merrill R.M., Hankey B.F.: Cancer surveillance series: Interpreting trends in prostate cancer Part II: Cause of death misclassification and the recent rise and fall in prostate cancer mortality. J. Natl Cancer Inst., 91, , Fischbacher C., Chappel D., Edwards R., Summerton N.: Health surveys via the Internet: quick and dirty or rapid and robust? J R Soc Med., 93(7), , Geraci J.M., Escalante C.P., Freeman J.L., Goodwin J.S.: Comorbid disease and cancer: the need for more relevant conceptual models in health services research. J. Clin. Oncol., 23(30), , Goodwin J.S., Hunt W.C., Key C.R.: Cancer treatment protocols: Who gets chosen? Arch. Intern. Med., 148, , Hankey B.F., Feuer E.J., Clegg L.X. et al.: Cancer surveillance series: Interpreting trends in prostate cancer. Part I: Evidence of the effects of screening in recent prostate cancer incidence, mortality, and survival rates. J Natl Cancer Inst., 91, , Hershman D., Fleischauer A.T., Jacobson J.S. et al.: Patterns and outcomes of chemotherapy for elderly patients with stage II ovarian cancer: A population based study. Gynecol. Oncol. 92, , Chouillet A.M., Bell C.M.J., Hiscox J.C.: management of breast cancer in southeast England. BMJ, 308, , Christakis N.A., Lamont E.B.: Extent and determinant of error in doctor s prognoses in terminally ill patients: prospective cohort study. BMJ 320, , International Association of Cancer Registries, c/o International Agency for Research on Cancer; iacr.com.fr/ Krzyzanowska M., Pintilie M., Tannock I.: Factors associated with failure to publish large randomized trials presented at an oncology meeting. JAMA 290, , NCI (National Cancer Institute): Surveillance, Epidemiology and End Results Parkin D.M., Pisani P., Ferlay, J.: Estimates of the worldwide incidence of twenty-five major cancers in Int J. Cancer: 80, , Parkin D.M., Whelan S.L., Ferlay J., Storm, H.: Cancer Incidence in Five Continents, Vol. I to VIII. IARC CancerBase No. 7, Lyon, Piccirillo J.F., Tierhey R.M., Costas I. et al.: Prognostic importance of comorbidity in a hospital-based cancer registry. JAMA 291, , Pisani P., Parkin D.M., Bray F., Ferlay, J.: Estimates of the worldwide mortality from twenty-five cancers in Int. J. Cancer: 83, 18-29, Pisani, P., Bray, F., Parkin, D.M.: Estimates of the worldwide prevalence of cancer for twenty-five sites in the adult population. Int. J. Cancer: 97,72-81, Potosky A.L., Feuer E.J., Levin D.L.: Impact of screening on incidence and mortality of prostate cancer in the United States. Epidemiol Rev., 23, , Richards M.A.: Tertiary cancer services in Britain: benchmarking study of activity and facilities at 12 specialist centres. BMJ, 313, , Ruland C.M., White T., Stevens M., Fangiullo G., Khilani S.M.: Effects of a computerized system to support shared decision making in symptom management of cancer patients: preliminary results. J. Am. Med. Inform. Assoc., 10(6), , Sankila R., Black R., Coebergh J.W.C., Démaret E., Forman D., Gatta G., Parkin D.M.: Evaluation of Clinical Care by Cancer Registries. IARC Technical Publication No. 37, ISBN , Sant M., Allemani C., Capocaccia R., Hakulinen T., Aareleid T. et al.: Stage at diagnosis is a key explanation of differences in breast cancer survival across Europe. Int. J. Cancer, 106, , Satariano W.A., Silliman R.A.: Comorbidity: implications for research and practice in geriatric oncology. Crit. Rev. Oncol. Hematol., 48, , Scheiden R., Sand J., Weber J., Turk P., Wagener Y., Capesius C.: Rectal cancer in Luxembourg. a national populationbased data report, BMC Cancer, 3, 27 36, Shanafelt T.D., Loprinzi C., Marks R., Novotny P., Sloan J.: Are chemotherapy response rates related to treatmentinduced survival prolongations in patients with advanced cancer? J. Clin. Oncol. 22(10), , Tarlov A.R., Ware J.E., Greenfield S., Nelson E.C., Perrin E., Zubkoff M.: The medical outcomes study. An application of methods for monitoring the results of medical care. JAMA 1989, 262(7), , Vinh-Hung V., Burzykowski T., Van de Steene J., Storme G., Soete G.: post-surgery radiation in early breast cancer: survival analysis of registry data. Radiotherapy and Oncology, 64: , World Health Organization. International Classification of Diseases for Oncology (ICD-O), 3rd edition, Geneva, Switzerland, WHO, Yancik R., Wesley M.N., Ries L.A., Havlik R.J., Long S., Edwards B.K., Yates J.W.: Comorbidity and age as predictors of risk for early mortality of male and female colon carcinoma patients. A population-based study. Cancer, 82, , KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

113 Paper V. Dušek L., Pavlík T., Koptíková J., Mužík J., Gelnarová E., Žaloudík J., Vyzula R., Hajdúch M., Abrahámová J. Czech National Cancer Registry and reference standards for health care assessment. Klinická onkologie 20, Supplement 1/2007, 77-95, ISSN X (original article in Czech)

114 Role of population-based registries in oncology and current situation in the Czech Republic. KLINICKÁ ONKOLOGIE 20, SUPPLEMENT 1/2007, 77 95, ISSN X (ORIGINAL ARTICLE IN CZECH) CZECH NATIONAL CANCER REGISTRY AND REFERENCE STANDARDS FOR HEALTH CARE ASSESSMENT NÁRODNÍ ONKOLOGICKÝ REGISTR ČR JAKO ZDROJ REFERENČNÍCH STANDARDŮ PRO HODNOCENÍ VÝSLEDKŮ LÉČEBNÉ PÉČE DUŠEK L. 1, PAVLÍK T. 1, KOPTÍKOVÁ J. 1, MUŽÍK J. 1, GELNAROVÁ E. 1, ŽALOUDÍK J. 2, VYZULA R. 2, HAJDÚCH M. 3, ABRAHÁMOVÁ J. 4 1 FACULTY OF MEDICINE, MASARYK UNIVERSITY, BRNO 2 MASARYK MEMORIAL CANCER INSTITUTE, BRNO 3 FACULTY OF MEDICINE, PALACKÝ UNIVERSITY, OLOMOUC 4 THOMAYER UNIVERSITY HOSPITAL, PRAGUE Summary This paper examines the value of the Czech National Cancer Registry (CNCR) for health care assessment. More than 1.3 million records collected since 1977 were audited from the viewpoint of comprehensiveness and correctness. The audit proved steadily increasing quality of the CNCR over time, diagnostic error rate (including incorrectly unstaged cases) decreased below 6 % after The most recent CNCR records are therefore fully usable for the evaluation of health care results. For that purpose, reference data set with > valid records was defined as a source of population standards for overall survival modelling. The reference data set covers a recent period ( ) and is clinically relevant (it contains only fully diagnosed and then treated patients). So called complete analysis of 5-year survival was proposed as optimal for benchmarking of cancer centres against population-based reference. If the benchmarking should reflect survival experienced by recently diagnosed patients, the period analysis was proposed for such a more modernised assessment (Brenner and Gefeller, 1996). All calculated reference standards of overall survival are recommended for selfbenchmarking of cancer centres. The values should not be applied for mutual comparisons of hospitals. Key words: population-based cancer registry, overall survival, reference standards, benchmarking Souhrn Článek se zabývá Národním onkologickým registrem ČR (NOR) a jeho využitím pro hodnocení výsledků léčebné péče. U více než 1,3 milionu záznamů zhoubných nádorů hlášených za období byl proveden audit úplnosti a správnosti. Analýza prokázala postupně se zlepšující kvalitu dat NOR, po roce 1990 klesla neúplnost a chybovost diagnostických záznamů pod 6 %. Data NOR jsou tedy využitelná pro hodnocení výsledků léčebné péče. Pro tento účel byl vyčleněn soubor více než validních záznamů vhodných pro definici referenčních standardů celkového přežití pacientů. Referenční soubor zahrnuje klinicky aktuální období ( ) a týká se pouze pacientů, kteří byli kompletně diagnostikováni a následně léčeni. Pro srovnávání výsledků zdravotnických zařízení s referenčním standardem byla jako optimální navržena tzv. kompletní metoda odhadu 5ti-letého absolutního a relativního přežití. Pro situace vyžadující odhad x-letého přežití u pacientů diagnostikovaných především v posledních letech byla doporučena analýza časových period dle Brennera (Brenner a Gefeller, 1996). Vypočítané referenční hodnoty 5ti-letého přežití jsou doporučeny výhradně k sebehodnocení onkologických pracovišť, neměly by sloužit k vzájemnému porovnávání zdravotnických zařízení. Klíčová slova: populační onkologický registr, celkové přežití, referenční standardy Introduction Health care quality and results assessment must always be based on the analysis of real data, since it is the only way to fulfil basic requirements for: - objectivity given by reference standards - currentness reflecting recent trends - righteousness for all involved subjects and types of health care facilities - flexibility reflecting regional or other specifics These requirements naturally apply for all fields of medicine, not only for oncology. Current oncology is rather more aware of the necessity of the health care results assessment, which is given by properties of cancer diseases and also economical press leading to the system rationalization. At any rate, data for the fulfilment of the attributes stated above do not have to be very complicated. The fundamental condition is, however, that the parameters used are available in the same form in all assessed hospitals and also at both regional and national KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

115 Role of population-based registries in oncology and current situation in the Czech Republic. levels. The only way to ensure such permeability through all levels of organization is a standardized populationbased registry. By the way, the existence of a reliable population-based registry itself may be considered as one of the health care quality indicators. It is surprising how few developed European countries may elicit comprehensive populationbased cancer data collection. The prestigious study that assessed databases of the EUROCARE-3 project (Capocaccia et al., 2003) summarizes data from 67 registries of 22 countries and claims approximately 25% coverage of the whole population. Although the Czech data registration is not highly acknowledged in this publication, probably because of poor communication abroad, there is no doubt that the Czech National Cancer Registry (CNCR) belongs to a select few fully representative and longterm operated standard databases. This fact has obliged us to use this data as a reference standard, undoubtedly interesting even for countries that do not dispose of such data (Capocaccia et al., 2003; Ferlay et al., 2004). However, one should ask whether the CNCR areal data is of a sufficient quality for setting up reliable and meaningful reference standards of health care results. We attempt to answer this question in this paper, which aims to bring the CNCR data near use in individual health care facilities in the Czech Republic. The discussion on reference standards should by no means lead to setting up any centrally enforced limits or comparison of health care facilities. A retrospective population-based registry is not suitable for mutual comparison of individual facilities, even if it was reliable and full up to the last record. The term reference standards is related to an aggregated description of health care results typical from the whole population. Such output may be used by regions or health care facilities, particularly for analysis of their own results. Population-based data and assessment of cancer care results Implementation of population-based registries in the assessment of health care results is of course limited by their relatively simple parametric structure. The data can hardly be expected to provide information service complying with the operation of a specific clinic or hospital in details. Continuous assessment in health care facilities must be ensured by hospital information systems, which allow the recording of therapy course, eventual complications and their solution. Population-based data should be used only as background information for comparison of own results with those reached on regional or national levels (Sankila et al., 2003). This self-assessment may be focused on a wide range of parameters. Already basic epidemiological data allow the health care facilities to assess load of cancer, analyze detection of tumours in their attraction zone or assess diagnostics results. Nevertheless, a key analysis in oncology has been and will be survival analysis, which may be provided by a very reliable registry with ensured continuous update of records. The following text is focused on CNCR validation in light of information on survival. This does involves not only data reliability, but also the methodology used for the assessment and interpretation of obtained results. Information value of overall survival cannot be replaced by another parameter and assessment based e.g. on therapy response, used in some studies, may be misleading (e.g. Shanafelt et al., 2004). Population-based estimates of overall survival also cannot be replaced by subjective assessment of results of individual teams or assessment on a regional level; all such data is subject to a high probability of bias, which applies particularly to advanced or terminal stages of cancer (Christakis and Lamont, 2000; Glare et al., 2003). Although the term survival reference standard is often connected to scoring of therapy fruitfulness, its use is actually much wider and rather manager goals represent the smaller part of applications. Reference standards of overall survival have been used for the assessment of specific and non-specific risk factors and co-morbidity effects (Piccirillo et al., 2004; Geraci et al., 2005), comparative assessment of various treatment strategies (Vinh-Hung et al., 2002), or study of relationships among continuously monitored clinical parameters (Shanafelt et al., 2004). The overall survival reference assessment thus belongs to the basic information equipment of current oncology and clinical research. On the other hand it should be emphasized that the overall survival is a time-dependent parameter integrating a number of influences that do not have to be associated to the original anticancer treatment at the moment of death. Besides the information on health care results this parameter is, therefore, also an indicator of very complex population-based trends. Survival prolongation itself does not have to be an explicit proof of more successful therapy. It may be a short-term effect of diagnostics improvement and detection of less advanced stages with better therapy results (Welch et al., 2000). Interregional differences in survival may also be explained by various ratios of clinical stages as a result of different availability of health care or a diagnostic activity (Berrino, 2003). A complicated interpretation of overall survival lays big stress on standardization of calculation, but also on an adequate definition of the reference data set. Survival alone may even be an indirect indicator of data quality in the source registry. Based on international data it is known for many diagnoses, what survival may be expected and achieved in the Czech population. If the registry analysis shows significantly higher or lower survival in comparison to international references, the problem very probably lies in data representativeness. Diagnosed with rather worse survival are used for these analyses, such as acute leukaemia, liver and lung cancer etc (Capocaccia et al., 2003). Definition of the reference data set for population-based analysis of overall survival Although representativeness of a population-based registry is one of the infallible signs of its quality, in health care assessment it is rather a handicap. Population-based representativeness means almost unseizable heterogeneity 78 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

116 Role of population-based registries in oncology and current situation in the Czech Republic. of primary data, which group all patients age categories and treatment procedures. Therefore, if we aim to generate a clinically usable standard from the population-based database, we have to proceed as follows: a) It is necessary to define target population, which will be compared with the standard. Different target populations will have various standards, scilicet in both epidemiological and clinical attributes. b) It is necessary to define parameters that will be a subject of reference comparison c) It is necessary to document completeness and reliability of source data in key parameters. d) It is necessary to separate the basic reference set from the source population-based database, which will correspond to selected target population and key parameters. The definition of reference data set is not only a matter of record type, but also of their quality: a) Only those records may enter the analysis that are consistent and complete. b) The analysis must treat only records relevant for a given purpose to avoid a systematic bias of results. c) The reference data set must be sufficiently larger and its structure must correspond to the target population. In practice the analysis may easily shift to two extremes that in principle cannot provide relevant outputs: (1) all registry records are included in the reference analysis in a naïve belief that this is the right representativeness or (2) records are anxiously selected, which leads to the reliable data set that, nevertheless, does not correspond to the target population and cannot therefore represent its reference standard. Everything stated above applies two-fold for the survival analysis, which is analytically complicated and requires further inputs: - correctly entered date of diagnosis, date of treatment initiation, and correctly recorded follow-up with a relevant date of the last follow-up - adequate number of patients surviving in time of the last follow-up (so called censored points), which is associated with the record quality on follow-up and long enough follow-up duration. The definition of the reference set in principle means that all records are omitted from the database that do not correspond to the target population and their inclusion in the reference standard would unacceptably increase the risk of bias. The whole process must, however, be retrospectively verifiable. The method is often used in literature. For example, Eaker et al. (2006) analyzed data of the Uppsala population-based registry in order to assess differences in treatment care available for different age categories of women with breast carcinoma. The basic hypothesis concerned effects of different care on survival of two age groups between 50 and 84 years. Of the primary database containing 12,163 women, women under 50 years (17.8%) and above 84 years (7.1%) were excluded; the latter mainly due to more frequent co-morbidities. Further 0.8% of women were excluded for insufficiently long duration of follow-up or other problems in records. The final set of 9,059 women was used for population-based analyses. Figure 1a. Cancer incidence in the Czech Republic ( ) Figure 1b. Cancer incidence in the Czech Republic ( ) patients with anticancer therapy Reference set usable for the survival analysis in the CNCR database The CNCR database certainly represents a basis for finding a meaningful reference set with a sufficient number of records. The definition was performed according to prestigious population-based studies, i.e. we restricted the searched set to a clinically relevant target group. The first condition was an updated reference set, since recent clinically analysis cannot contain historic trends. For this reason we limited the range of analyzed data to the period from , in which we can operate with valid CNCR records according to updated TNM classification. The treatment results assessment in this period also reflects conditions of the Czech health care system set up after Data from this period certainly represent a sufficiently large sample for a population-based analysis, since more than 520,000 records are available about newly diagnosed malignant neoplasms. As shown in figure 1a, incidence recorded in all main diagnostic groups is KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

117 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 2. Cancer prevalence in the Czech Republic in 2003 sufficient. If the selected time period would not suit a specific diagnostic group, the input data may be extended without problems and all subsequent steps remain valid. Next step is restriction of the reference set to patients, who actually underwent anticancer treatment. Figure 1b documents that the treated patients form a significant majority in all diagnostic groups and therefore we still work with a sufficient number of records. Figure 2 closes the input definition with an estimate of prevalence at the end of the observed period, i.e. in Figure 3 displays a scheme of all steps that led to the selection of the final reference data set from the original 521,745 records assigned to 442,254 uniquely identified patients (figure 4). Stratification shown in figure 4 treats only unique patients and is necessary, because repeated reported malignancies of the same patients would cause bias in the reference set definition. The procedure displayed in figures 3 and 4 can be methodologically summarized and commented as follows: 1. Filtering of the diagnoses, for which the exact assessment of treatment results is not possible from the CNCR data. They are namely diagnoses with a specific treatment or diagnoses, in which no TNM classification exists or has a limited validity, such as hematooncological diagnoses, in situ tumours and CNS tumours (Davis et al., 1997; Fritz et al., 2000; Howe et al., 2003). o Assessment of health care results in these diagnoses requires collection of a specific data, which is not available in the CNCR. If these diagnostic groups remained in the reference data set, the resulting reference standard would not correspond to a major patient population. o Childhood tumours are for the most part classified and included in the analysis in a way that is compatible with records of adult patients, although the final diagnosis does not fully correspond to international classification of childhood malignant neoplasms (Kramarova a Stiller, 1996). 2. Filtering of the records with unclear date of diagnosis that cannot be included in the calculation of overall survival, i.e. records diagnosed at autopsy and DCO records. o Ratio of these records is an indirect indicator of data quality. Ratio of so-called death certificateinitiated cases should feature long-term stability; increased rate indicates incompleteness of other treated patients records. On the contrary, low rate indirectly indicates data reliability (Berrino et al., 1995). The performed CNCR analysis did not reveal any fundamental problems (figure 3 and 4). 3. Filtering of records with incomplete diagnostics due to early death or another event that prevented completion of diagnostics and treatment initiation. These records would disable stratified survival analysis for individual clinical stages. o In accordance with literature, a 1-month threshold from diagnosis was used for determination of early death (Micheli et al., 2003). Figure 5 documents a relatively low ratio of these problematic records in the CNCR database. 4. Filtering of records with problematic diagnostic identification, particularly not stated clinical stage and TNM classification. o These records cannot be considered as valid and correct. This problem is faced in a number of robust international registries as well, as demonstrated by the EUROCARE-3 database analysis (Capocaccia et al., 2003). Clinical stage is a basic factor explaining the differences in treatment results and its statement must be obligatory for reference data (Berrino, 2003). Figure 5 confirms that the period was selected properly, since these CNCR records are of a very sufficient quality in light of TNM and stage. 5. Patients included in the reference set must undergo the last follow-up in a relevant timeframe and sufficiently long follow-up duration. Therefore, the calculation is not biased by predominance of patients with recently initiated treatment, who could not yet have affected the survival calculation. o This requirement, nevertheless, restricts sample size again, because if we want to prevent bias as a result of incomplete follow-up reports, we cannot use the most recent data. In the case of the CNCR, the temporal limit up to 2003 appears to be sufficient and the reference data set is not biased due to delayed incidence reporting 6. Furthermore, the exclusion of patients who have more than one proven malignant neoplasm at time of their last follow-up should be considered as well. We left these patients in the data set used for our analysis, but for reference analyses only the first diagnosed tumour is considered. Figure 3 therefore defines the reference set for records of malignant neoplasms, but figure 4 is fundamental, because it operates with unique patients. In case two tumours were diagnosed at the same time, the one with worse clinical stage and prognosis is taken for the analysis. The summarized overview of repeated malignant neoplasms in one patient is shown in figure KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

118 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 3. Overview of the CNCR records in the proposed reference period * Figure 3. Proposed process to the reference set for population-based survival analyses ( )* The procedure described above may be naturally criticized, since exclusion of records with incomplete diagnostics due to early death overestimates the estimate of overall survival. The objection is relevant from epidemiological point of view, but aim of the analyses described here is setting up a reference standard for health care assessment. The chosen procedure may also be defended by the following arguments: a) A population-based reference set valid for health care facilities cannot operate with patients who were not involved in treatment at all. b) The ratio of DCO and other problematic records should be stable within the population; if incidence of such patients changes or fluctuates in time, it is likely a problem of quality and completeness of registry data, not a problem of reference analyses. c) Proposed filtration system is fully reversible and no record is erased forever. General population-based analyses may include all records on survival without regard to their quality and origin. Nevertheless, designed criteria of quality are essential for the health care assessment and ensure, among other things, the interpretational value of the result. d) Focus of the reference set on data of treated patients naturally leads to better values of survival than it would in case of the whole population analysis. Everything is a matter of purpose and meaning of the analysis. When we want to provide a representative picture about mortality and prevalence of the given disease in the whole population, it is certainly correct to include all records in the analysis, including untreated patients or patients with noninitiated treatment due to early death. In such case the analysis has rather an epidemiological character. If we want to provide a comparative standard for health care facilities or document change in survival caused by change in treatment, we have to focus on patients who had a chance to undergo it. A similar procedure of filtration is used in many confirmative populationbased studies (e.g. Brenner and Arndt, The result is a set of reliable and verified records that represent survival of treated patients with correctly completed diagnostics. As demonstrated in figures 4 and 5, subsequent separation of treated and untreated patients still provides a sufficiently large sample for population- KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

119 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 5. Usability of the CNCR records for population-based survival assessment Figure 6. Treatment of patients according to clinical stages ( ) on the example of three diagnostic groups based analyses. In treated patients, less advanced clinical stages logically prevail, as documented in figure 7. Figure 8 further confirms a very satisfying equal distribution of the reference set among Czech regions, based on data from The only real loss in case of the CNCR database and definition of the reference set is the group of patients in which the TNM classification and clinical stage were not stated without objective reasons. However, these patients reach rather medium values of survival for most diagnoses and their exclusion is not a source of systematic bias. The ratio of problematic records decreases with increasing registry quality over time; their rate is only 6% in the selected period (figure 3). These unstaged records are usually assessed separately in the literature and survival comparable with clinical stages 2 and 3 is often stated, which is in agreement with our conclusions (Gatta et al., 2000; Sant et al., 2003; Brenner and Arndt, 2005). In conclusion, we can state that the CNCR database currently contains records sufficiently valid for the reference analysis of overall survival. In agreement with international standards we successfully defined the reference set, which may serve as a basis for clinical comparison and provides a large enough number of records even for detailed stratifications. The selected period ( ) may be extended if needed, especially for less frequent diagnoses, while the quality criteria remain valid. Figures 8 and 9 document that the proposed reference set covers all size categories of health care facilities, at least at the level enabled by the CNCR database. It is therefore possible to define population-based standards for various facilities according to their size and eventually sort records according to place of treatment. Selection of reference records according to the place of treatment is not rare in the literature, since reference analyses may involve only data from reliable sources. That is optimally from larger hospitals, where the patient is treated comprehensively in one place without transfer to other health care facilities (e.g. Vinh-Hung et al., 2002). This dimension of the assessment is, nevertheless, not simple, because definition of so called high-volume hospitals cannot be performed generally for all types of tumours. For this reason sorting of health care facilities was proposed that reflects the number of treated patients within the frame of aggregated diagnostic groups of cancer (figure 8). 82 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

120 Role of population-based registries in oncology and current situation in the Czech Republic. Method for assessment of survival from the populationbased data and CNCR Overall survival is the most frequently assessed characteristic of oncological data. The best known output of the method is a classical curve of survival and estimate of median survival. Nevertheless, these outputs generate results of survival that have the following interpretation problems: 1. Problems associated with insufficient data. The analysis according to Kaplan and Meier provides description of survival achieved within the given patients cohort, whereas the cohort size and duration of follow-up are fundamental. This time frame determines degree of the overall survival estimate reliability. Less frequent diagnoses cannot often be assessed in some hospitals due to a low number of records. A similar problem is in newly established facilities, where the short time of their operation does not enable survival analysis. The problem lies mainly in survival comparison of two patient groups with different sizes and durations of follow-up. An alternative solution may be temporarily found in health care assessment from more points of view. Other, short-term available parameters are emphasized, such as therapy response, occurrence of adverse effects etc. 2. Survival measured on the given patients cohort is not associated only with treatment, but there is also a number of other factors and patient characteristics, at least e.g. age. Mutual comparability of smaller cohorts is then affected by patient characteristics that cannot be controlled. Direct comparison of survival curves in more health care facilities or regions without population-based standardization may therefore be very misleading. 3. Overall survival belongs to parameters which we refer to as integrating. That means they reflect various factors that are not associated only with health care quality. Particularly in early stages of a number of diagnoses the median survival gets very high and is related to patients who were diagnosed and treated according to standards different from those that have been valid at present. This fact may totally deteriorate analysis of long-term trends. Figure 5. Treated cancer patients in 2003 comparison of Czech regions Figure 6. Overview of Czech health care facilities according to number of cancer patients under follow-up in the period KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

121 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 9. Size of the proposed reference set for the population-based survival analysis ( ) Figure 10. Repeated malignant neoplasms of the same diagnosis in the CNCR records As arising from facts stated above, for the survival analysis of population-based data we need a robust method rather than direct analysis of survival curves and estimates of median survival. Such analyses may be performed in clinical studies or anywhere, where comparability is guaranteed; in population-based data, unfortunately, we have to face a significantly higher heterogeneity of compared cohorts. 1. Effects of random temporal fluctuations and sample heterogeneity are partly solved by estimates of x-year survival (typically 5-, 10-, 20-year survival). The larger time period covered, the more stable the results are, since the calculation aggregates more full-year records of follow-up. The output is stated in percent of patients that reached x-year survival. The following paragraphs define methods suitable for these population-based estimates (see also scheme in figure 11). 2. A needed population-based standardization of outputs is implemented by weighted estimates of x-year survival in the assessed cohort with regard to status of the whole population, i.e. calculation of so called relative survival. The calculation refers the x-year survival in a selected group of patients to survival achieved at the same period in a normal population that corresponds to the patient cohort in age and sex ratio (so called matched population ). The method proposed by Hakulinen (Hakulinen, 1982; Hakulinen and Abeywickrama, 1985) may be recommended for a specific calculation and the resulting relative survival estimate should be supplemented by a variability indicator, e.g. standard error or recalculated confidence interval (Greenwood, 1926; Altman, 1991). The relative survival is, therefore, a populationweighted estimate of the given patients cohort survival. This parameterization is widely used and there is no problem to find reference data (Henson and Ries, 1995; Brenner, 2002) or algorithms and software tools (Pohar and Stare, 2006). 3. A common method for calculation of the x-year observed or relative survival is so called cohort analysis (Cutler and Ederer, 1958; Kaplan and Meier, 1958; Ederer et al., 1961; Piccirillo, 2004). The term refers to the way in which the analysis defines patient cohorts for the x-year survival estimate. E.g. 5-year survival assessed for a period is estimated from follow-up data of patients diagnosed in years The analyzed patient cohort is defined retrospectively according to date of diagnosis and returns the analysis outputs back, which is particularly noticeable for 10-year and more survival (figure 11). The method output, therefore, does not fully record recent changes, e.g. as a result of diagnostics or therapeutic progress. 4. A solution for reflecting more recent data may be the period analysis, which was recently validated (Brenner 84 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

122 Role of population-based registries in oncology and current situation in the Czech Republic. Figure 11. Methodological scheme of the 5-year survival assessment on the example of period and Gefeller, 1996; Brenner et al., 2004). The cohort analysis calculates 5-year survival for the period from the cohort of patients diagnosed in , i.e. patients diagnosed x years retrospectively are needed. The period analysis, on the contrary, includes only the follow-up data from the recent period , i.e. including the last year, as of which the survival is calculated. 1-year survival after diagnosis may be estimated for patients diagnosed in , patients from period contribute to the 2-year survival etc. (figure 11). This method may be helpful especially for data in which the survival significantly changes due to progress in diagnostics and therapeutic procedures. After adequate validation this method may be recommended for the survival analysis in health care facilities, since it reflects recent health care results (Brenner a Hakulinen, 2002). A disadvantage exists in the optimistic overestimation of the 1-year survival, which was also documented in literature, particularly for diagnoses where the changes in diagnostics bring positive change in rate of clinical stages, e.g. in prostate carcinoma (Talbäck et al., 2004; Brenner et al., 2004). 5. A compromise between somewhat extreme approaches of the cohort and period analyses is so called complete analysis (Brenner and Gefeller, 1996). The time frame for calculation again involves all diagnosed years including the recent, the complete analysis therefore has similar characteristics as the period analysis from this point of view. In contrast to the period analysis, however, it does not limit the time frame from left Table 1. 5-year observed and relative survival calculated by COMPLETE analysis on the CNCR reference data set from the period A summary analysis of results in patients with anticancer therapy. 1a. Data analysis of patients with anticancer therapy for all types of health care facilities All stages Stage 1+2 Stage 3 Stage 4 5-year 5-year 5-year 5-year Diagnostic group 5-year 5-year 5-year 5-year N relative N relative N relative N relative survival survival survival survival survival survival survival survival C00-C08 Oral cavity ,0 50, ,8 78, ,9 36, ,8 21,5 C09-C14 Pharynx and nasopharynx ,3 35, ,4 59, ,2 43, ,8 27,3 C15 Esophagus ,2 10, ,6 19, ,0 7, ,4 2,8 C16 Stomach ,0 32, ,8 52, ,9 14, ,7 4,8 C18-C21 Colon and rectum ,0 57, ,0 75, ,3 46, ,0 12,8 C22 Liver and intrahepatic bile ducts ,6 14, ,1 32, ,1 26, ,1 6,9 C23-C24 Gallbladder and biliary tract ,1 22, ,2 41, ,2 15, ,1 6,3 C25 Pancreas ,8 7, ,6 21, ,5 9, ,6 1,9 C32 Larynx ,7 55, ,2 78, ,3 52, ,5 25,3 C34 Bronchus and lung ,2 13, ,5 33, ,1 9, ,8 3,3 C43 Melanoma of skin ,8 84, ,8 92, ,6 55, ,4 21,2 C50 Breast ,0 80, ,6 91, ,2 59, ,5 24,8 C51-C52 Vulva and vagina ,3 60, ,1 75, ,9 33, ,4 16,3 C53 Cervix uteri ,5 73, ,5 85, ,1 46, ,6 14,3 C54 Corpus uteri ,3 84, ,4 90, ,3 50, ,5 26,9 C56 Ovary ,8 51, ,3 84, ,7 35, ,7 19,0 C61 Prostate ,6 74, ,1 97, ,7 80, ,8 37,0 C62 Testis ,3 90, ,1 96, ,5 57,3 * * * C64 C66, C68 Kidney and other urinary organs ,4 69, ,7 89, ,2 64, ,7 19,1 C67 Bladder ,7 78, ,6 86, ,7 32, ,9 21,4 C73 Thyroid gland ,4 94, ,6 101, ,3 84, ,5 56,9 Other malignant neoplasms ,6 58, ,9 73, ,1 51, ,0 21,5 * Stage 4 is not defined for the C62 diagnosis in the TNM classification KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

123 Role of population-based registries in oncology and current situation in the Czech Republic. 1b. Data analysis of patients with anticancer therapy for health care facilities with a high number of patients of the given diagnostic group (see also fig. 8) Diagnostic group N Stage 1 Stage 2 Stage 3 5-year 5-year 5-year 5-year 5-year relative N relative N survival survival survival survival survival C00-C08 Oral cavity ,6 73, ,3 44, ,6 25,1 C09-C14 Pharynx and nasopharynx 73 65,1 74, ,1 55, ,0 34,1 C15 Esophagus ,3 26, ,6 9,6 77 6,9 8,0 C16 Stomach ,0 65, ,5 22, ,5 4,8 C18-C21 Colon and rectum ,6 82, ,9 50, ,7 16,4 C22 Liver and intrahepatic bile ducts 19 32,9 35, ,0 42,2 63 6,6 6,7 C23-C24 Gallbladder and biliary tract ,4 53, ,7 22, ,3 14,4 C25 Pancreas ,4 23, ,9 20, ,2 5,4 C32 Larynx ,7 77, ,6 56, ,4 34,1 C34 Bronchus and lung ,1 39, ,6 11, ,8 4,5 C43 Melanoma of skin ,2 93, ,0 55, ,1 28,4 C50 Breast ,9 93, ,9 63, ,8 29,4 C51-C52 Vulva and vagina ,2 81, ,5 49, ,6 33,2 C53 Cervix uteri ,3 84, ,2 57, ,9 19,0 C54 Corpus uteri ,6 93, ,8 57, ,6 38,9 C56 Ovary ,6 92, ,9 45, ,8 25,8 C61 Prostate ,4 108, ,4 90, ,2 50,1 C62 Testis ,9 97, ,3 58,9 * * * C64-C66,C68 Kidney and other urinary organs ,2 91, ,6 68, ,4 20,6 C67 Bladder ,5 91, ,7 44, ,8 22,1 C73 Thyroid gland ,5 95, ,6 90, ,7 61,2 Other malignant neoplasms ,3 67, ,4 64, ,3 27,5 * Stage 4 is not defined for the C62 diagnosis in the TNM classification 5-year relative survival Health care facilities were sorted according to number of patients on the basis of CNCR records on the follow-up facility. It is therefore only the first step of analyses that should be further performed in the Czech Republic. There is no other option given by population-based data, since records about reporting facility and primary treatment facility have been recorded since The solution lies in the inclusion of health care payers central data into the analyses, which would enable an exact determination of treatment procedure volume. and includes the follow-up data of patients diagnosed earlier as well (figure 11). Since this method does not return time frame strictly x years back and also includes results of recent follow-ups, it is often marked as more objective and accurate than other methods in the literature (Brenner et al., 2002; Brenner et al., 2004). Final proposals of standards for the survival assessment based on the CNCR data It may be concluded that there are no methodological doubts in question of survival assessment on the basis of population-based data. The standard method according to Kaplan and Meier can be applied only in the comparison of differently treated patient groups inside the health care facilities, always if the groups are of similar size and under follow-up for a similar time. Once these local results are to be generalized on a population level, it is necessary to vote for estimate of observed and relative x-year survival. The relative survival estimate alone is populationstandardized and shows whether the given patient group does not have increased mortality in a health care facility in comparison to the whole population. For the assessment of clinically relevant data, (treated patients only, updated data) in comparison to the reference standard we suggest the so called complete method of the x-year survival estimate (figure 11). If the x-year survival estimate may include results achieved in patients diagnosed particularly in recent years, the period analysis according to Brenner (Brenner and Gefeller, 1996; Brenner et al., 2004) may be recommended. We applied the survival analysis to the reference data set extracted from the CNCR database (figure 4 and 9) and calculated estimates of 5-year observed and relative survival. Results and final proposals of reference population-based survival values for main diagnostic groups are shown in table 1. The data in table 1 may be commented as follows: Table 1 summarizes estimates of 5-year observed and relative survival calculated according to clinical stages in patients with anticancer therapy. The calculations were performed by means of complete analysis of survival, which we consider as more objective than other methods for Czech data as well, as stated in the discussion above. Table 1 is based exclusively on validated CNCR data from the reference data set (see also figure 4 and 9). Because the standards are for health care facilities, the survival estimates are calculated for the sub-population of patients, who actually underwent anticancer therapy. Table 1A summarizes results of the calculation for individual malignant neoplasms clinical stages of patients from all types of health care facilities. Estimates 86 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

124 Role of population-based registries in oncology and current situation in the Czech Republic. in table 1B represent detailed survival results only for facilities with a high number of patients from the given diagnostic group (see also figure 8). Usability of records about health care facilities in the CNCR reference data The above mentioned table 1B gives 5-year survival reference standards of cancer patients, but only for health care facilities with a high number of patients from the given diagnostic group. Data for the analysis were obtained for individual diagnostic groups separately (figure 8). These analyses should be perceived as the first step toward improving estimates calculated with the whole sample, since the results achieved in high-volume facilities should theoretically be the best achievable treatment results. In our case it is really only the first step, because the CNCR reference data from the period do not offer unambiguous identification of the facility responsible for the patient s primary treatment. Records on the followup care facility had to be used for the analysis presented in table 1B, because it is the best recorded information in the whole CNCR history with increasing number of records (> 84% of records in the assessed period). Information on facilities providing treatment is usable since 1995; the facility is stated in 65 72% of surgery records, 15 17% of radiotherapy records, 16 18% of chemotherapy records, 7 8% of hormonal therapy, and 1 3% of other types of treatment. The most complete data is about facility of the reporting physician; however, this information has been available since 2000 (completeness 92 97%), which is too short of a period to be used in the survival analysis. Information about the GP s facility is also stated in the CNCR, its usability for the health care assessment is rather controversial (completeness since 1995 is 38 48%). Due to mentioned reasons, the use of records about followup care facility was the only option and our identification of facilities with a high number of patients is, therefore, only approximate (figure 8, table 1B). Nevertheless, we can hypothetically declare that the presented results are not significantly biased, because the large facilities much more likely perform follow-ups on patients from their attraction zones, who also underwent therapy in the same facility. These analyses should necessarily continue and there are two options offered: In time (2 3 years) the period in which the reporting physicians have been recorded in the CNCR will be long enough to perform the survival analysis. The CNCR data may be confronted with the health care payers data, from which both sides would benefit. The payers would obtain diagnostic data and data about the stage of the disease, while the CNCR would obtain very accurate identification of health care facilities involved in the treatment. Discussion on further methods of survival assessment in the CNCR data The relative survival assessment based on the CNCR data naturally offers much more possibilities than the mere calculation for patients with anticancer therapy, as shown in table 1. Besides calculations for other patient sub-groups, different methods of calculation may also be compared. For this reason we also added the following results to this article in the form of attachments: Attachment 1 summarizes results of survival for all patients included in the reference data set (valid patient data in the CNCR, ) with subsequent discrimination to patients with and without anticancer therapy. The attachment does not stratify resulting values according to clinical stages and thus serves as a global illustration of patient survival in the recent CNCR reference standard. Attachment 2 contains survival estimates of patients from the reference data set calculated according to clinical stage at diagnosis (stage I+II, III, IV). Values for both patients with and without anticancer therapy are aggregated. It is therefore a rather general indicator of all patient survival in reference data, in contrast to the attachment 1 according to clinical stages. Attachment 3 is similar to attachment 2 and contains estimates of 5-year survival according to clinical stages at diagnosis, but only patients with complete diagnostics and initiated anticancer therapy are included. These are reference estimates for patients who actually underwent anticancer therapy, calculated for the reference data set without discrimination of type and size of the health care facility. Attachment 4 concludes the set of analyses with estimates performed on the total population of all patients in the CNCR, including DCO records and patients with early death. The estimates stated in this attachment, therefore, cannot be used for the health care quality assessment in the health care facilities, since they include data on patients that could not even meet treatment effects. It is a survival estimate of the whole population of patients with malignant neoplasms and as such has a mainly epidemiological character. Analyses of observed and relative survival in attachments 1 3 are performed by all three methods available, i.e. the cohort, complete, and period analysis. It is obvious that the period analysis operates with a smaller patient cohort and its time frame gives higher weight to the more recent years of the observed period, in our case Final estimates of the relative survival performed by the period analysis are 15 25% higher than estimates calculated from the whole observed period by the complete analysis for most diagnoses and clinical stages (attachment 1 3). These results are in agreement with international literature (Brenner, 2002; Brenner et al., 2004). The tendency of the period analysis to overestimate is in our case probably even strengthened by preceding data set filtration and focusing mainly on treated patients. Discussion on the size of patients cohort for the reference survival assessment A high attention should be paid to cohort size in the assessment of patients survival in individual health care KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

125 Role of population-based registries in oncology and current situation in the Czech Republic. Attachment 1. 5-year observed and relative survival calculated by COHORT, COMPLETE, and PERIOD analysis on the CNCR reference data set from the period (see also figures 4 and 9). Attachment 1a. Analysis of patients with valid CNCR records: patients both with and without anticancer treatment Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,6 39,1 55,5 45,6 48,5 68,9 C09-C14 Pharynx and nasopharynx ,0 28,4 42,5 30,5 32,8 49,1 C15 Esophagus ,6 6,5 12,6 7,7 8,1 15,6 C16 Stomach ,8 16,1 30,3 19,2 20,9 39,4 C18-C21 Colon and rectum ,3 38,1 53,4 47,1 49,5 69,5 C22 Liver and intrahepatic bile ducts ,1 4,8 9,4 5,2 6,0 11,8 C23-C24 Gallbladder and biliary tract ,7 9,3 18,6 9,9 12,0 24,2 C25 Pancreas ,5 3,7 7,8 4,4 4,6 9,7 C32 Larynx ,5 45,1 61,0 51,6 53,6 72,5 C34 Bronchus and lung ,7 8,2 16,6 9,2 9,9 20,0 C43 Melanoma of skin ,8 71,0 80,1 82,7 83,3 94,1 C50 Breast ,6 68,6 78,5 76,8 78,8 90,3 C51-C52 Vulva and vagina ,9 45,1 60,0 51,9 55,7 74,2 C53 Cervix uteri ,0 66,8 78,5 70,8 70,5 83,0 C54 Corpus uteri ,9 72,3 82,7 80,6 82,5 94,4 C56 Ovary ,1 42,5 57,9 45,5 46,1 62,9 C61 Prostate ,5 48,5 61,2 69,9 72,4 91,6 C62 Testis ,8 87,8 93,2 88,9 89,8 95,4 C64-C66,C68 Kidney and other urinary organs ,5 51,6 66,1 62,0 62,1 79,6 C67 Bladder ,7 56,8 69,2 73,7 75,0 91,6 C73 Thyroid gland ,2 84,8 90,9 90,8 91,3 97,8 Other malignant neoplasms ,0 44,6 58,8 53,9 53,2 70,2 Period analysis Attachment 1b. Analysis of patients with valid CNCR records: patients with anticancer treatment Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,4 41,0 57,2 47,7 50,8 70,8 C09-C14 Pharynx and nasopharynx ,8 30,3 44,4 32,6 35,1 51,5 C15 Esophagus ,3 9,2 16,0 12,1 10,9 18,7 C16 Stomach ,1 25,0 39,8 31,0 32,1 51,1 C18-C21 Colon and rectum ,2 44,0 57,6 55,9 57,1 74,8 C22 Liver and intrahepatic bile ducts ,4 12,6 19,1 15,2 14,6 22,2 C23-C24 Gallbladder and biliary tract ,8 18,1 30,1 19,7 22,5 37,9 C25 Pancreas ,6 6,8 11,1 9,0 7,9 12,8 C32 Larynx ,9 46,7 62,3 53,1 55,3 73,8 C34 Bronchus and lung ,7 11,2 20,5 12,5 13,1 24,0 C43 Melanoma of skin ,7 71,8 80,5 83,7 84,1 94,6 C50 Breast ,0 70,0 79,4 78,2 80,3 91,2 C51-C52 Vulva and vagina ,2 49,3 63,1 57,3 60,9 78,0 C53 Cervix uteri ,9 69,5 80,0 73,7 73,2 84,3 C54 Corpus uteri ,8 74,3 83,8 82,6 84,6 95,6 C56 Ovary ,9 47,8 61,5 51,8 51,7 66,6 C61 Prostate ,7 50,6 62,8 72,6 74,9 93,2 C62 Testis ,7 88,3 93,3 89,8 90,3 95,5 C64-C66,C68 Kidney and other urinary organs ,2 58,4 71,0 69,7 69,9 85,0 C67 Bladder ,8 59,7 71,0 77,7 78,7 94,0 C73 Thyroid gland ,0 87,4 92,2 93,8 94,0 99,1 Period analysis Other malignant neoplasms ,4 49,6 62,6 59,9 58,7 74,3 88 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

126 Role of population-based registries in oncology and current situation in the Czech Republic. Attachment 1c. Analysis of patients with valid CNCR records: patients without anticancer treatment Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,3 10,2 19,7 11,1 14,0 27,1 C09-C14 Pharynx and nasopharynx ,5 11,5 22,5 9,6 12,6 24,8 C15 Esophagus ,4 1,9 4,9 1,5 2,9 7,7 C16 Stomach ,0 2,2 5,9 3,0 3,2 8,7 C18-C21 Colon and rectum ,9 4,0 10,0 5,6 5,9 14,8 C22 Liver and intrahepatic bile ducts ,2 1,5 3,6 2,1 2,3 5,6 C23-C24 Gallbladder and biliary tract ,5 1,8 4,6 2,5 2,9 7,4 C25 Pancreas ,2 2,2 5,5 2,9 3,0 7,4 C32 Larynx ,8 28,0 44,2 33,9 34,4 54,5 C34 Bronchus and lung ,4 2,4 6,3 3,5 3,5 9,2 C43 Melanoma of skin ,9 21,2 35,9 29,6 27,0 46,4 C50 Breast ,5 10,1 17,9 16,9 15,4 27,9 C51-C52 Vulva and vagina ,2 4,7 12,5 7,6 5,7 15,4 C53 Cervix uteri ,6 25,5 45,1 28,6 28,8 51,5 C54 Corpus uteri ,3 19,4 33,5 23,7 24,1 43,0 C56 Ovary ,3 3,7 9,2 2,6 4,8 12,1 C61 Prostate ,3 35,6 50,1 53,2 56,8 80,1 C62 Testis ,7 8,6 * 7,8 8,7 * C64-C66,C68 Kidney and other urinary organs ,0 6,3 13,1 9,5 9,4 19,9 C67 Bladder ,9 15,9 28,8 25,2 21,3 38,8 C73 Thyroid gland ,5 10,4 21,8 12,2 12,1 25,3 Other malignant neoplasms ,7 8,2 16,4 9,6 12,0 23,9 * Analysis not performed duet to a small cohort size Period analysis Attachment 2. 5-year observed and relative survival calculated by COHORT, COMPLETE, and PERIOD analysis on the CNCR reference data set from the period (see also figures 4 and 9). The analysis is stratified according to clinical stages. Attachment 2a. Analysis of patients with valid CNCR records: patients both with and without anticancer treatment in clinical stage 1+2 Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,4 61,2 72,9 74,0 78,0 92,9 C09-C14 Pharynx and nasopharynx ,3 49,8 62,3 55,4 57,7 72,6 C15 Esophagus ,2 13,8 23,3 17,0 16,7 27,9 C16 Stomach ,9 37,0 54,0 45,2 48,0 70,2 C18-C21 Colon and rectum ,6 56,5 69,0 71,0 73,6 90,2 C22 Liver and intrahepatic bile ducts ,7 15,7 24,8 14,8 19,0 30,0 C23-C24 Gallbladder and biliary tract ,9 30,3 46,2 33,5 38,4 59,4 C25 Pancreas ,1 12,8 23,4 14,2 15,2 28,7 C32 Larynx ,0 64,8 76,2 76,3 78,2 92,0 C34 Bronchus and lung ,1 23,9 41,5 25,1 28,4 49,5 C43 Melanoma of skin ,1 79,7 86,2 91,8 92,7 100,4 C50 Breast ,3 80,3 86,0 90,5 91,4 98,0 C51-C52 Vulva and vagina ,8 59,5 70,9 70,5 73,5 87,5 C53 Cervix uteri ,6 80,8 87,7 85,0 84,9 92,2 C54 Corpus uteri ,0 78,6 86,5 87,4 89,6 98,7 C56 Ovary ,7 77,9 85,7 83,8 84,2 92,8 C61 Prostate ,4 65,0 73,1 95,7 96,7 108,7 C62 Testis ,8 94,0 96,5 96,2 96,2 98,8 C64-C66,C68 Kidney and other urinary organs ,2 73,3 80,9 87,5 87,6 96,9 C67 Bladder ,5 64,7 75,1 84,3 85,4 99,5 C73 Thyroid gland ,6 94,4 96,3 100,4 101,0 103,3 Period analysis Other malignant neoplasms ,8 61,0 72,9 72,6 73,3 87,7 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

127 Role of population-based registries in oncology and current situation in the Czech Republic. Attachment 2b. Analysis of patients with valid CNCR records: patients both with and without anticancer treatment in clinical stage 3 Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,3 30,5 47,3 32,2 35,0 54,5 C09-C14 Pharynx and nasopharynx ,0 36,4 49,6 39,5 42,8 58,0 C15 Esophagus ,0 5,2 10,7 5,7 5,9 12,2 C16 Stomach ,6 10,2 19,9 11,0 12,8 25,1 C18-C21 Colon and rectum ,0 34,2 47,3 40,8 43,9 60,9 C22 Liver and intrahepatic bile ducts ,8 11,3 20,4 10,1 13,3 24,1 C23-C24 Gallbladder and biliary tract ,2 8,7 17,1 9,1 11,3 22,0 C25 Pancreas ,8 5,1 11,2 5,9 6,0 13,1 C32 Larynx ,2 42,8 59,6 46,4 50,2 69,9 C34 Bronchus and lung ,3 6,8 13,6 7,6 8,2 16,4 C43 Melanoma of skin ,6 43,5 55,7 54,6 55,1 70,5 C50 Breast ,5 49,4 63,2 56,4 58,8 75,6 C51-C52 Vulva and vagina ,3 22,9 36,8 25,7 30,4 49,0 C53 Cervix uteri ,1 41,5 56,3 43,9 44,9 60,9 C54 Corpus uteri ,4 41,9 57,0 45,0 48,0 65,3 C56 Ovary ,5 31,1 42,7 32,4 33,7 46,4 C61 Prostate ,5 53,0 63,6 78,4 79,8 96,3 C62 Testis ,7 55,4 70,8 53,3 56,3 71,9 C64-C66,C68 Kidney and other urinary organs ,0 51,5 66,0 64,0 62,4 80,0 C67 Bladder ,9 21,7 33,3 21,2 29,0 45,7 C73 Thyroid gland ,3 73,8 81,3 81,7 81,6 89,9 Other malignant neoplasms ,0 41,4 52,4 45,5 49,2 62,3 Period analysis Attachment 2c. Analysis of patients with valid CNCR records: patients clinical both with and without anticancer treatment in stage 4 Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,0 17,3 30,3 19,5 19,8 34,7 C09-C14 Pharynx and nasopharynx ,0 21,8 35,3 23,4 25,0 40,5 C15 Esophagus ,6 1,8 4,1 1,7 2,1 4,8 C16 Stomach ,4 2,1 5,1 2,0 2,9 7,1 C18-C21 Colon and rectum ,6 6,7 13,0 7,4 8,7 17,0 C22 Liver and intrahepatic bile ducts ,4 2,2 4,3 3,6 3,0 5,9 C23-C24 Gallbladder and biliary tract ,5 2,5 5,3 2,0 3,3 7,1 C25 Pancreas ,9 1,4 2,9 1,1 1,7 3,6 C32 Larynx ,8 20,8 34,7 22,1 23,4 39,1 C34 Bronchus and lung ,8 2,1 4,6 2,2 2,6 5,7 C43 Melanoma of skin ,2 14,6 25,2 14,5 17,8 30,6 C50 Breast ,9 18,1 30,6 20,7 22,0 37,0 C51-C52 Vulva and vagina ,0 10,2 17,9 10,4 11,5 20,2 C53 Cervix uteri ,8 9,8 18,3 11,4 10,2 19,2 C54 Corpus uteri ,3 17,7 30,8 21,6 20,6 36,5 C56 Ovary ,6 13,6 24,4 14,5 14,7 26,6 C61 Prostate ,5 22,3 35,1 34,3 33,4 52,6 C62 Testis * * * * * * * * * C64-C66,C68 Kidney and other urinary organs ,8 11,7 20,3 16,8 14,4 24,9 C67 Bladder ,2 12,6 21,5 17,6 16,1 27,5 C73 Thyroid gland ,2 43,5 60,4 56,3 48,2 66,8 Other malignant neoplasms ,2 14,3 22,7 18,4 16,6 26,4 * Stage 4 is not defined for the C62 diagnosis in the TNM classification Period analysis 90 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

128 Role of population-based registries in oncology and current situation in the Czech Republic. Attachment 3. 5-year observed and relative survival calculated by COHORT, COMPLETE, and PERIOD analysis on the CNCR reference data set from the period (see also figures 4 and 9). The analysis of patients with anticancer treatment is stratified according to clinical stages. Attachment 3a. Analysis of patients with valid CNCR records: patients with anticancer treatment in clinical stage 1+2 Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,5 61,8 73,2 75,4 78,8 93,3 C09-C14 Pharynx and nasopharynx ,9 51,4 63,3 57,4 59,8 74,0 C15 Esophagus ,6 16,6 26,2 20,0 19,9 31,3 C16 Stomach ,6 40,8 57,2 49,7 52,6 74,0 C18-C21 Colon and rectum ,4 58,0 69,9 73,2 75,5 91,1 C22 Liver and intrahepatic bile ducts ,2 28,1 35,5 27,0 32,6 41,1 C23-C24 Gallbladder and biliary tract ,0 33,2 48,4 36,3 41,4 61,0 C25 Pancreas ,8 17,6 26,7 21,3 21,3 32,1 C32 Larynx ,0 65,2 76,5 76,0 78,4 92,2 C34 Bronchus and lung ,9 28,5 45,5 30,1 33,2 53,2 C43 Melanoma of skin ,1 79,8 86,2 91,7 92,6 100,4 C50 Breast ,6 80,6 86,2 90,6 91,6 98,0 C51-C52 Vulva and vagina ,7 61,1 71,7 73,0 75,4 88,4 C53 Cervix uteri ,2 81,5 88,1 85,6 85,6 92,5 C54 Corpus uteri ,7 79,4 86,9 88,1 90,4 99,1 C56 Ovary ,0 78,3 86,0 84,2 84,7 93,1 C61 Prostate ,1 66,1 74,3 96,3 97,6 109,8 C62 Testis ,0 94,1 96,6 96,4 96,3 98,9 C64-C66,C68 Kidney and other urinary organs ,5 74,7 82,0 88,7 89,0 97,8 C67 Bladder ,5 65,6 75,6 85,5 86,6 100,1 C73 Thyroid gland ,0 94,6 96,4 100,8 101,2 103,4 Other malignant neoplasms ,8 61,9 73,4 73,1 73,9 87,8 Period analysis Attachment 3b. Analysis of patients with valid CNCR records: patients with anticancer treatment in clinical stage 3 Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,4 31,9 48,8 33,5 36,6 56,1 C09-C14 Pharynx and nasopharynx ,2 37,2 50,0 39,6 43,7 58,5 C15 Esophagus ,6 7,0 12,8 8,6 7,9 14,5 C16 Stomach ,5 11,9 22,3 12,0 14,8 27,7 C18-C21 Colon and rectum ,5 36,3 48,8 43,9 46,5 62,6 C22 Liver and intrahepatic bile ducts ,5 23,1 36,8 19,4 26,6 42,7 C23-C24 Gallbladder and biliary tract ,3 12,2 22,4 13,0 15,3 28,1 C25 Pancreas ,8 8,5 13,3 10,4 9,5 15,0 C32 Larynx ,7 44,3 60,9 48,1 52,0 71,5 C34 Bronchus and lung ,6 8,1 15,5 9,0 9,6 18,4 C43 Melanoma of skin ,7 43,6 55,8 54,7 55,2 70,7 C50 Breast ,0 50,2 64,1 56,9 59,8 76,7 C51-C52 Vulva and vagina ,7 25,9 40,3 28,9 33,9 53,0 C53 Cervix uteri ,5 43,1 57,1 46,5 46,4 61,5 C54 Corpus uteri ,4 44,3 58,6 48,4 50,6 67,0 C56 Ovary ,9 32,7 43,8 35,0 35,5 47,4 C61 Prostate ,2 54,7 65,3 79,8 80,8 96,8 C62 Testis ,3 56,5 71,3 54,9 57,3 72,4 C64-C66,C68 Kidney and other urinary organs ,9 53,2 67,1 66,1 64,2 81,0 C67 Bladder ,9 23,7 35,5 22,4 32,5 48,7 C73 Thyroid gland ,1 76,3 82,6 84,8 84,3 91,3 Period analysis Other malignant neoplasms ,2 43,1 54,0 46,7 51,0 64,1 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

129 Role of population-based registries in oncology and current situation in the Czech Republic. Attachment 3c. Analysis of patients with valid CNCR records: patients with anticancer treatment in clinical stage 4 Diagnostic group Cohort analysis N 5-year survival (%) 5-year relative survival (%) Complete analysis Period analysis Cohort analysis Complete analysis Period analysis Cohort analysis Complete analysis C00-C08 Oral cavity ,4 18,8 32,3 21,1 21,5 37,0 C09-C14 Pharynx and nasopharynx ,9 23,8 37,8 25,7 27,3 43,5 C15 Esophagus ,5 2,4 5,1 2,7 2,8 5,9 C16 Stomach ,4 3,7 7,2 4,3 4,8 9,4 C18-C21 Colon and rectum ,1 10,0 17,1 11,8 12,8 21,9 C22 Liver and intrahepatic bile ducts ,1 6,1 8,7 10,3 6,9 9,8 C23-C24 Gallbladder and biliary tract ,9 5,1 9,2 2,3 6,3 11,4 C25 Pancreas ,2 1,6 3,6 1,4 1,9 4,0 C32 Larynx ,7 22,5 36,7 24,3 25,3 41,3 C34 Bronchus and lung ,4 2,8 5,5 2,8 3,3 6,4 C43 Melanoma of skin ,4 17,4 28,3 18,3 21,2 34,3 C50 Breast ,8 20,5 33,7 23,1 24,8 40,8 C51-C52 Vulva and vagina ,0 14,4 23,5 15,0 16,3 26,5 C53 Cervix uteri ,4 13,6 23,8 16,4 14,3 24,9 C54 Corpus uteri ,1 23,5 38,9 28,4 26,9 44,9 C56 Ovary ,2 17,7 29,2 19,4 19,0 31,4 C61 Prostate ,3 24,8 37,8 38,0 37,0 56,4 C62 Testis * * * * * * * * * C64-C66,C68 Kidney and other urinary organs ,4 15,7 26,0 22,1 19,1 31,5 C67 Bladder ,8 16,9 26,3 24,6 21,4 33,6 C73 Thyroid gland ,6 51,5 66,5 65,5 56,9 73,4 Other malignant neoplasms ,9 19,0 28,2 24,7 21,5 32,1 * Stage 4 is not defined for the C62 diagnosis in the TNM classification Period analysis Attachment 4. 5-year observed and relative survival calculated by COMPLETE analysis of all patients recorded in the CNCR in the period , including DCO records and early deaths All stages Stage 1+2 Stage 3 Stage 4 5-year 5-year 5-year Diagnostic group 5-year 5-year 5-year N relative N N relative N N relative survival survival survival survival survival survival N C00-C08 Oral cavity ,0 47, ,2 78, ,5 35, ,3 19,8 C09-C14 Pharynx and nasopharynx ,1 32, ,8 57, ,4 42, ,8 25,0 C15 Esophagus ,9 7, ,8 16, ,2 5, ,8 2,1 C16 Stomach ,2 18, ,0 48, ,2 12, ,1 2,9 C18-C21 Colon and rectum ,1 47, ,5 73, ,2 43, ,7 8,7 C22 Liver and intrahepatic bile ducts ,2 5, ,7 19, ,3 13, ,2 3,0 C23-C24 Gallbladder and biliary tract ,3 10, ,3 38, ,7 11, ,5 3,3 C25 Pancreas ,8 4, ,8 15, ,1 6, ,4 1,7 C32 Larynx ,1 52, ,8 78, ,8 50, ,8 23,4 C34 Bronchus and lung ,9 9, ,9 28, ,8 8, ,1 2,6 C43 Melanoma of skin ,7 81, ,7 92, ,5 55, ,6 17,8 C50 Breast ,6 77, ,3 91, ,4 58, ,1 22,0 C51-C52 Vulva and vagina ,2 52, ,5 73, ,9 30, ,2 11,5 C53 Cervix uteri ,6 66, ,8 84, ,5 44, ,8 10,2 C54 Corpus uteri ,5 79, ,6 89, ,9 48, ,7 20,6 C56 Ovary ,4 43, ,9 84, ,1 33, ,6 14,7 C61 Prostate ,5 74, ,0 96, ,0 79, ,3 33,4 C62 Testis ,5 89, ,0 96, ,4 56,3 * * * C64 C66, C68 Kidney and other urinary organs ,2 59, ,3 87, ,5 62, ,7 14,4 C67 Bladder ,4 72, ,7 85, ,7 29, ,6 16,1 C73 Thyroid gland ,8 89, ,4 101, ,8 81, ,5 48,2 Other malignant neoplasms ,5 51, ,0 73, ,4 49, ,3 16,6 * Stage 4 is not defined for the C62 diagnosis in the TNM classification 92 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

130 Role of population-based registries in oncology and current situation in the Czech Republic. facilities, especially the number of patients with complete follow-up. Attention given to this topic in evaluation of clinical trials is extremely high (Schoenfeld and Richter, 1982; Ahnn and Anderson, 1998; Halabi and Singh, 2004). On the other hand, this topic somewhat recedes into the background in the assessment of epidemiological data, as most authors presume á priori sufficient cohort size, if not fairly an exhaustive representativeness. Nevertheless, in standardization of the observed and relative survival estimates according to age and sex, we can fall into the situation after discrimination of patients into subcategories, when some groups do not include enough patients. As mentioned above, estimate of the survival probability and its variability is associated particularly with patients with complete records. It may be stated that censoring decreases the number of patients that are available for the estimate of the survival probability. Dickman and Hakulinen (2003) therefore define so called effective sample size in a given time from the follow-up beginning, which is the estimate of the real patient number that really contribute to the obtained value of survival probability (.e.g. effective cohort size contributing to 5-year survival estimate). We assume that this characteristic, although it is only estimated, says more about the patient cohort than their absolute number. The effective patient cohort size should correspond to mathematic presumptions (especially asymptotic normality), which are a basis for the calculation of survival estimates and variability indicators, including individual sub-categories for the estimate standardization according to age and sex. Based on the simulation studies (Halabi and Singh, 2004) we suggest a minimal effective cohort size of 20 patients in each standardizing category. Discussion on interpretation of the reference survival assessment outputs A number of methodological recommendations should be followed during the application of population-based reference standards to data of individual health care facilities. In the analysis above we suggested processing the data of hospitals by the complete survival analysis and the period analysis, which allows for calculation including recent data from the given year. This procedure is not generally usable, however, and has several methodological presumptions and interpretation limitations. At the level of individual hospitals, we must always analyze survival within the frame of cancer diagnostic groups. Analysis of cancer patients as a whole is used in literature, though, but it serves only for description of populationbased burden of demographic groups (Jemal et al., 2004). Such an approach in hospitals would lead to uncontrolled bias as a result of various burdens by different diagnoses, which may moreover change in time. Further, record management according to cause of death should be considered. The proposed method of relative survival assessment does not require detailed coding of cause of death and may thus be used for all cancer patients. Assessment of the disease-specific survival strongly depends on the records reliability on cause of death, which may be discussible in a population-based registry. For this reason we suggest, in accordance with literature, the inclusion of all deaths for reference purposes without discrimination of causes (Percy et al., 1990; Brown et al., 1993). Characteristics of patients who enter the reference comparison in a selected health care facility create so called case-mix. The parameters must be examined, so that no bias occurs as a result of non-representative distribution of some factor (Elliot, 2001). First it is necessary to make sure that the patient cohort corresponds to the reference set in basic characteristic and effects of individual patients characteristics on survival may be further examined directly in local data. Another problem of analyses inside hospitals is the interpretation of changes in survival in recent years. These changes may represent fluctuations resulting from a changing spectrum of incoming patients, but may also reflect an important progress in quality. This particularly involves the improvement of diagnostics leading to increased detection of early stages of cancer. Although it is naturally a positive trend, it may bias information about the survival of patients diagnosed before. Improved diagnostics may contribute to the bias by earlier disease detection without real prolongation of patients survival and further by a selective detection of slowly growing or hidden and painless malignant tumours that would not be detected at all. In literature these events are described as led-time bias, length-time bias and overdiagnosis bias (Black and Welch, 1993; Patz et al., 2000; Ries et al., 2001). This bias may also concern the survival assessed for early stages of cancer (Feinstein et al., 1985). The relative survival assessment in specific health care facilities may even lead to values above 100% for some less dangerous diagnoses, e.g. as a result of increased health care. Further, cancer patients may live healthier lives and the effects of low numbers of patients cannot be of course excluded in cohorts from individual hospitals. This phenomenon is often affected by a diagnostic procedure, because not all diagnostic techniques are equally available within the whole population. Particularly cancer detection in screening programmes may be accumulated in a conscious, more socially-developed part of the population, which attends preventive programmes more frequently and has a generally lower mortality than the overall population (Swan et al., 2003). A big problem regarding the relative survival assessment in hospitals is that we are about to use a method verified on high-volume population-based data in significantly smaller units, moreover with a uncertain data structure. Aggregation of different cohorts from various health care facilities may be a substantial source of variability (Berrino et al., 1995; Coleman et al., 2003). The relative survival calculation alone does not guarantee comparability of different populations. For this reason an age-adjusted relative survival calculation has been used, which operates with age distribution in several standard age categories (e.g. five). A problem occurs if a high enough number KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

131 Role of population-based registries in oncology and current situation in the Czech Republic. of patients is not available for a specific age category in smaller hospitals. However, here we return to the previous chapter, in which we recommended using reference standards rather for larger facilities and separately aggregate and analyze data of smaller facilities. Population-based reference survival values may be further biased by factors associated with tumour biology, or rather with tumour biological identity. Proliferation ability, metastatic potential, chemoresistance, genetic basis all these factors play a more important role in the assessment of specific health care facilities than at the population level, and their effects grow with decreasing cohort size. Therefore, even when working with a correctly set up reference standard, the difference in survival of a certain patient cohort may be affected by the distribution of these important prognostic parameters (Shetty and Reiman, 1994; Louwman et al., 2003). It is evident that practical employment of the survival reference standards in individual facilities is definitely not simple. The process is accompanied by a number of methodical problems from the initial data targeting over calculation methodology to interpretation. Reference data should therefore be used very sensitively, always considering the specifics of the assessed unit. The reference standard should serve particularly for self-assessment in hospitals and cancer centres. By means of comparison of own results with adequate population standard, the responsible experts obtain valuable feedback, which can be used for the improvement of health care. Conclusion When looking for reference standards for health care quality we have to be aware of the fact that they are targeted to individual health care facilities. This certainly does not represent a problem in diagnostic groups, whose treatment is localized into a limited number of centres with valid standards. Nevertheless, the most frequent diagnoses are treated in many different health care facilities. This fact applies to Czech conditions as well and complicates the reference standard definition setting up the standard on a simple whole republic data averaging is not correct. The solution lies in the derivation of the reference standard from results of large and medium health care facilities. These hospitals elicit a higher reliability of reported population-based data and a high burden by cancer diagnoses has required the standardization of certain processes. Furthermore, the large facilities are also responsible for patient s treatment and follow-up. Moreover, it is known from the literature that a high number of patients (so called high hospital procedure volume ) is a presumption of better treatment results (Hannan et al., 1989; Yao and Lu-Yao, 1999; Schrag et al., 2000; Meyerhardt et al., 2003; Panageas et al., 2003). Volume of procedures, practice, and experience are known as significant predictors of the cancer care quality and achieved survival (Hannan et al., 1989; Rohan et al., 1998; Harmon et al., 1999; Kee et al., 1999; Hillner et al., 2000). For this reason we propose the Czech reference data set for survival assessment as an open system that allows the narrowing of patient selection to defined types of health care facilities if circumstances require (figure 8 and 9). Although the CNCR database does not allow to fully close the analysis according to type of the health care facility, it offers very valuable data, without which further continuation of these analyses is not possible. We consider the values shown in table 1A-B as final reference standards of overall survival. These standards have been derived from a reliable and updated reference set and have been calculated using internationally validated methodology. The same methodology will be suitable to obtain survival estimates from local data of all newly established comprehensive cancer centres in the Czech Republic. Literature: Altman D.G.: Practical Statistics for Medical Research. Chapman and Hall, London, 619p, Ahnn S., Anderson J.S.: Sample size determination in complex clinical trials comparing more than two groups for survival endpoints. Statist. Med. 17, , Berrino F., Esteve J., Coleman M.P.: Basic issues in estimating and comparing the survival of cancer patients. In Berrion F., Sant M., Verdecchia A. et al. (eds): Survival of Cancer Patients in Europe: the EUROCARE Study. IARC Scientific Publication No Lyon, France: International Agency for Research on Cancer, 1 14, Berrino F.: The EUROCARE Study: strengths, limitations and perspectives of population-based comparative survival studies. Ann. Oncol., 14, Suppl. 5., v9 v13, Black W.C., Welch H.G.: Advances in diagnostic imaging and overestimation of prevalence and the benefits of therapy. N Engl J Med, 328, , Brenner H., Gefeller O.: An alternative approach to monitoring cancer patient survival. Cancer, 78, , Brenner H.: Long-term survival rates of cancer patients achieved by the end of the 20th century: A period analysis. Lancet, 360, , Brenner H., Hakulinen T.: Up-to-date long-term survival estimates of patients with cancer by period analysis. J. Clin. Oncol., 20, , Brenner H., Soderman B., Hakulinen T.: Use of period analysis for providing more up-to-date estimates of long-term survival rates: empirical evaluation among cancer patients in Finland. Int. J. Epidem., 31, , Brenner H., Gefeller O., Hakulinen T.: Period analysis for up-to-date cancer survival data: Theory, empirical evaluation, computational realization and applications. Eur. J. Cancer, 40, , Brenner H., Arndt V.: Long-term survival rates of patients with prostate cancer in the prostate-specific antigen screening era: population-based estimates for the year 2000 by period analysis. J. Clin. Oncol., 23(3), , Brown B.W., Brauner C., Minnotte M.C.: Noncancer deaths in white adult cancer patients. J. Natl. Cancer Inst., 85, , Capocaccia D., Gatta G., Roazzi P., Carrari E., Santaquilani M., De angelis R., Tavilla A., Eurocare Working Group: The EUROCARE-3 database: methodology of data collection, standardization, quality control and statistical analysis. Ann. Oncol., 14, Suppl. 5., v14 v27, Coleman M.P., Gatta G., Verdecchia A.: EUROCARE Working Group. EUROCARE-3 Summary: Cancer survival in Europe at the end of the 20th century. Ann. Oncol, 14, Suppl. 5., v128 v149, Cutler S.J., Ederer F.: Maximum utilization of the life table method in analyzing survival. J. Chron. Dis., 8, , Davis F.G., Bruner J.M., Surawicz T.S.: The rationale for standardized 94 KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/2007

132 Role of population-based registries in oncology and current situation in the Czech Republic. registration and reporting of brain and central nervous system tumors in population-based registries. Neuroepidemiology, 16(6), , Dickman P.W., Hakulinen T. (2003). Population-based cancer survival analysis. Available from URL com/teaching/tampere2004/book_draft.pdf Eaker S., Dickman P.W., Berquist L., Holmberg L.: Differences in management of older women influence breast cancer survival: results from a population based database in Sweden. PloS Med., 3(3), e25, Ederer F., Axtell L.M., Cutler S.J.: The relative survival rate: a statistical methodology. Natl. Cancer Inst. Monogr., 6, , Elliot M.N., Swartz R., Adams J. et al.: Case-mix adjustment of the national CAHPS benchmarking data 1.0: a violation of model assumptions? Health Serv. Res., 36(3), , Feinstein A.R., Sosin D.M., Wells C.K.: The Will Rogers phenomenon. Stage migration and new diagnostic techniques as a source of misleading statistics for survival in cancer. N. Engl. J. Med., 312, , Ferlay J., Bray F., Pisani P. and Parkin D.M.: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC Cancer Base No. 5. version 2.0, IARCPress, Lyon, 2004 Fritz A., Percy C., Jack A. et al.: International classification of diseases of oncology. Geneva: World Health Organization, Gatta G., Capocaccia R., Sant M., Bell C.M., Coebergh J.W.W. et al.: Understanding variations in survival for colorectal cancer in Europe: a EUROCARE high resolution study. Gut, 47, , Geraci J.M., Escalante C.P., Freeman J.L., Goodwin J.S.: Comorbid disease and cancer: the need for more relevant conceptual models in health services research. J. Clin.Oncol., 23(30), , Glare P., Virik K., Jones M. et al.: A systematic review of physicians survival predictions in terminally ill cancer patients. BMJ, 327, , Greenwood M.: The Natural Duration of Cancer (Report on Public Health and Medical Subjects, No. 33). London, UK: His Majesty s Stationery Office, Hakulinen T.: Cancer survival corrected for heterogeneity in patient withdrawal, Biometrics, 38, , Hakulinen T., Abeywickrama K.H.: A computer program package for relative survival analysis. Comput. Programs Biomed.,19, , Halabi S., Singh B. (2004). Sample size determination for comparing several survival curves with unequal allocations. Statistics in Medicine, 23: Hannan E.L. O Donell J.F:, Kilburn H.Jr, Bernard H.R., Yazici A.: Investigation of the relationship between volume and mortality for surgical procedures performed in New York State hospitals. JAMA, 262, , Harmon J.W., Tang D.G., Gordon T.A., Bowman H.M., Choti M.A., Kaufman H.S. et al.: Hospital volume can serve as a surrogate for surgeon volume for achieving excellent outcomes in colorectal resection. Ann. Surg., 230, , Hillner B.E., Smith TJ., Desch C.E.: Hospital and physician volume or specialization and outcomes in cancer treatment: importance in quality of cancer care. J. Clin.Oncol., 18, , Henson D.E. and Ries L.A.: The relative survival rate. Cancer, 76, , Howe H.L., Edwards B.K., Young J.L. et al.: A vision for cancer incidence surveillance in the United States. Cancer Causes Control, 14, , Christakis N.A., Lamont E.B.: Extent and determinants of error in doctor s prognoses in terminally ill patients: prospective cohort study. BMJ, 320, , Kaplan E.L., Meier P.: Nonparametric estimation from incomplete observations. J. Am. Stat. Assoc., 58, , Kee F., Wilson R.H., Harper C., Patterson C.C., McCallison K., Houštin R.F. et al.: Influence of hospital and clinician workload on survival from colorectal cancer: cohort study. BMJ, 318, , Kramarova E., Stiller C.A.: The international classification of childhood cancer. Int. J. Cancer., 68, , Louwman W.J., van Diest P.J., van Beek M.W., Schapers R.F.M. et al.: Trends in breast cancer aggressiveness before the introduction of mass screening in southeastern Netherlands Breast Cancer Res. Treat., 73, , Meyerhardt J.A., Catalano P.J., Schrag D., Ayanian J.Z., Haller D.G., Mayer R.J., Macdonald J.S., Benson A.B., Fuchs C.S.: Association of hospital procedure volume and outcomes in patients with colon cancer at high risk for recurrence. Ann. Intern. Med., 139, , Micheli A., Baili P., Quinn M, EUROCARE Working Group. Life expectancy and cancer survival in the EUROCARE-3 cancer registry areas. Ann. Oncol., 14, Suppl. 5., v28-v40, Panageas K.S., Schrag D., Riedel E., Bach P.B., Begg C.B.: The effect of clustering of outcomes on the association of procedure volume and surgical outcomes. Ann. Itern. Med., 139, , Patz E.F. Jr, Goodman P.C., Bepler G.: Screening for lung cancer. N. Engl. J. Med., 343, , Percy C.L., Miller B.A., Ries L.A.G.: Effects of changes in cancer classification and the accuracy of cancer death certificates on trends in cancer mortality. Ann NY Acad Sci, 609, 87 97, Piccirillo J.F., Tierhey R.M., Costas I. et al.: Prognostic importance of comorbidity in a hospital-based cancer registry. JAMA 291, , Pohar M., Stare J.: Relative survival analysis in R. Computer Methods and Programs in Biomedicine, 8, , Ries L.A.G., Eisner M.P., Kosary C.L. et al.: SEER cancer statistics review, Available from URL gov/csr/ Roohan P.J., Bickell N.A., Baptiste M.S., Therriault G.D., Ferrara E.P., Siu A.L.: Hospital volume differences and five-year survival from breast cancer. Am. J. Public Health, 88, , Sankila R., Black R., Coebergh J.W.C., Démaret E., Forman D., Gatta G., Parkin D.M.: Evaluation of Clinical Care by Cancer Registries. IARC Technical Publication No. 37, ISBN , Sant M., Allemani C., Capocaccia R., Hakulinen T., Aareleid T. et al.: Stage at diagnosis is a key explanation of differences in breast cancer survival across Europe. Int. J. Cancer, 106, , Shanafelt, T.D., Loprinzi, C., Marks R., Novotny P., Sloan J.: Are chemotherapy response rates related to treatment-induced survival prolongations in patients with advanced cancer? J. Clin. Oncol. 22(10), , Shetty M.R., Reiman H.M.: Tumour size and axillary metastasis, a correlative occurrence in 1244 cases of breast cancer between 1980 and Eur. J. Surg Oncol, 23, , Schoenfeld D.A., Richter J.R. (1982). Nomograms for calculating the number of patients needed for a clinical trial with survival as an endpoint. Biometrics, 38: Schrag D., Frajer L.D., Bach P.B., Cohen A.M., Warren J.L., Begg C.B.: Influence of hospital procedure volume on outcomes following surgery for colon cancer. JAMA, 284, , Swan J., Breen N., Coates R.J. et al.: Progress i cancer screening practices in the United States. Results from the 2000 National Health Interview Survey. Cancer, 97, , Talbäck M., Rosén M., Stenbeck M., Diskman P.W.: Cancer patient survival in Sweeden at the beginning of the third millenium predictions using period analysis. Cancer Causes and Control, 15, , Vinh-Hung V., Burzykowski T., Van de Steene J., Storme G., Soete G.: Post-surgery radiation in early breast cancer: survival analysis of registry data. Radiotherapy and Oncology, 64, , Welch H.G., Schwartz L.M., Woloshin S.: Are incresing 5-year survival rates evidence of success against cancer? JAMA, 283, , Yao S. L., Lu-Yao G.: Population-based study of relationship between hospital volume of prostatectomies, patient outcomes, and length of hospital stay. J. Natl. Cancer Inst., 91, , KLINICKÁ ONKOLOGIE 20 SUPPLEMENT 1/

133

134 Paper VI. Zavoral M., Suchánek Š., Závada F., Dušek L., Mužík J., Seifert B., Fri P. Colorectal cancer screening in Europe. World J Gastroenterol, Beijing, China, the WJG Press and Baishideng. vol. 15, no. 47, s

135 Online Submissions: wjg.wjgnet.com World J Gastroenterol 2009 December 21; 15(47): [email protected] World Journal of Gastroenterology ISSN doi: /wjg The WJG Press and Baishideng. All rights reserved. Colorectal cancer screening in Europe EDITORIAL Miroslav Zavoral, Stepan Suchanek, Filip Zavada, Ladislav Dusek, Jan Muzik, Bohumil Seifert, Premysl Fric Miroslav Zavoral, Stepan Suchanek, Filip Zavada, Premysl Fric, Clinic of Medicine, Central Military Hospital, 1st Faculty of Medicine, Charles University, Prague, CZ16902, Czech Republic Ladislav Dusek, Jan Muzik, Institute of Biostatistics and Analyses, Masaryk University, Brno, CZ602 00, Czech Republic Bohumil Seifert, Institute of General Medicine, 1st Faculty of Medicine, Charles University, Prague, CZ128 00, Czech Republic Author contributions: Zavoral M, Suchanek S and Zavada F carried out the data collection and wrote the initial draft of the manuscript; Dusek L and Muzik J wrote the epidemiology section of the manuscript; Fric P and Seifert B performed the overall scientific direction and revision. Supported by International Agency for Research on Cancer (Lawrence von Karsa, MD); International Digestive Cancer Alliance (Professor Meinhard Classen, MD, Professor Sidney J Winawer, MD) Correspondence to: Miroslav Zavoral, MD, PhD, Professor of Medicine, Chief, Clinic of Medicine, Central Military Hospital, 1st Faculty of Medicine, Charles University, U vojenske nemocnice 1200, Prague 6, CZ16902, Czech Republic. [email protected] Telephone: Fax: Received: October 15, 2009 Revised: November 12, 2009 Accepted: November 21, 2009 Published online: December 21, 2009 Abstract Colorectal cancer (CRC) is the second most frequent malignant disease in Europe. Every year, people are diagnosed with this condition, and patients die of it. In 2003, recommendations for screening programs were issued by the Council of the European Union (EU), and these currently serve as the basis for the preparation of European guidelines for CRC screening. The manner in which CRC screening is carried out varies significantly from country to country within the EU, both in terms of organization and the screening test chosen. A screening program of one sort or another has been implemented in 19 of 27 EU countries. The most frequently applied method is testing stool for occult bleeding (fecal occult blood test, FOBT). In recent years, a screening colonoscopy has been introduced, either as the only method (Poland) or the method of choice (Germany, Czech Republic) The WJG Press and Baishideng. All rights reserved. Key words: Colorectal cancer; Europe; Fecal occult blood test; Screening colonoscopy; Screening programs Peer reviewer: Dr. James Hardwick, Department of Gastroenterology, Leiden University Medical Center, Albinusdreef 2, 2300RC, Leiden, The Netherlands Zavoral M, Suchanek S, Zavada F, Dusek L, Muzik J, Seifert B, Fric P. Colorectal cancer screening in Europe. World J Gastroenterol 2009; 15(47): Available from: URL: DOI: dx.doi.org/ /wjg INTRODUCTION Colorectal cancer (CRC) poses a serious health problem in countries with a Westernized lifestyle. Over the last decade, a whole range of new technologies have been introduced in clinical practice to diagnose and treat the disease, with therapeutic modalities extending to advanced stages of the disease. Nevertheless, prevention undoubtedly remains the key to reducing morbidity and mortality. The introduction of national or transnational population-wide screening programs is a priority for the healthcare policy of individual states, and this is also being addressed at the highest level by European Union (EU) administrators. The approach of individual countries to screening programs varies significantly because of differences in health insurance systems and budgets. This summary article focuses on a brief description and comparison of these programs. EPIDEMIOLOGY CRC is the second most frequent malignant disease in developed countries. The incidence of CRC is generally higher for men, and the risk of the disease increases with age, as the majority of cases are diagnosed in patients more than 50 years of age [1]. European countries rank highest in the global statistics, both in terms of incidence and mortality. In 1998 to 2002, the incidence of CRC in the USA for men and women was 38.6 and 28.3, respectively; in Europe, it was 38.5 and 24.6 [world age standardization (ASR-W)], as calculated per inhabitants [2]. However, mortality over the same period of time was much higher in Europe than in the US, both for men and women: in the USA, the figures were 13.5 and 9.2, respectively, while in Europe, they were 18.5 and 10.7 (ASR-W), as calculated per inhabitants [3]. A detailed comparison of data for European countries is made difficult because of the absence of a unified data

136 5908 ISSN CN /R World J Gastroenterol December 21, 2009 Volume 15 Number 47 A Sorted by mean value ASR (World) of males and females: B Sorted by mean value ASR (World) of males and females: Czech Republic Hungary Slovakia Norway Germany Luxembourg Denmark The Netherlands Croatia Ireland Austria Slovenia Switzerland France Italy United Kingdom Belgium Iceland Sweden Spain Portugal Bosnia Herzegovina Poland Estonia Malta Serbia and Montenegro Albania Finland Macedonia Russian Federation Moldova Ukraine Belarus Lithuania Bulgaria Latvia Romania Greece Males Females Hungary Czech Republic Slovakia Denmark Slovenia Ireland Norway Croatia Germany Austria The Netherlands Belgium Russian Federation Luxembourg Portugal Bosnia Herzegovina Estonia Latvia Ukraine France United Kingdom Spain Poland Belarus Lithuania Malta Bulgaria Serbia and Montenegro Albania Italy Moldova Iceland Sweden Switzerland Romania Finland Macedonia Greece Males Females Figure 1 Epidemiology of colorectal cancer in European countries. A: Incidence in international comparison-european countries; B: Mortality in international comparison-european countries. Adapted from: Ferlay J, Bray F, Pisani P, Parkin DM. GLOBOCAN 2002: Cancer incidence, mortality and prevalence worldwide. IARC Cancer Base No. 5 version 2.0. Lyon: IARC press, Available from: URL: section CI5 I-VIII (Detailed). Last accessed on August 8, Table 1 Colorectal cancer incidence in European countries in 2006 Parameter Countries with the highest incidence Countries with the lowest incidence Incidence > 70/ men (ASR-E): Hungary (106), Czech Republic (94.4), Slovakia (87.1), Switzerland (79.1), Germany (70.2) > 45/ women (ASR-E): Switzerland (55.6), Norway (51.2), Hungary (50.6), Denmark (48), Czech Republic (46), Germany (45.1) < 40/ men (ASR-E): Albania (13.6), Greece (31), Bosnia Herzegovina (34.6), Republic of Moldova (38.7), Finland (39.2) < 26/ women (ASR-E): Greece (21.3), Albania (21.4), Romania (25.1), Spain (25.4) Adapted from: Ferlay J, Autier P, Boniol M, Heanue M, Colombet M, Boyle P. Estimates of the cancer incidence and mortality in Europe in Ann Oncol 2007; 18: ASR-E: European age standard. source. Not all countries maintain sophisticated population and cancer registers, and it is sometimes necessary to obtain input data by projecting aggregated data. In this outline, figures available from international studies summarizing global and European epidemiologic data have been used [4,5]. A detailed comparison of countries within Europe using the ASR-W of incidence and mortality is presented in Figure 1. Most recent epidemiologic data on CRC for 2006 recalculated to the European age standard are given in Tables 1 and 2. CRC comprises 12.9% of all newly-diagnosed carcinomas in the European population (men 12.8%, women 13.1%) and account for 12.2% of deaths caused by malignancy. CRC is the second most frequent malignancy, after breast carcinoma (13.5% of all malignancies) and bronchogenic carcinoma (12.1% of all malignancies). It has been estimated that in 2006, people were diagnosed with CRC in Europe, and of them die of the disease [6]. The average incidence has shown a tendency to increase in recent years ( ), with a year-on-year growth of 0.5%. Available data on time trends of CRC incidence and mortality are shown in Figures 2 and 3. A detailed analysis of individual diagnoses confirms that malignant disease of the colon is the most frequent, accounting for 57% of all cases (> 35 cases/10 5 inhabitants), followed by malignant diseases of the rectum

137 Zavoral M et al. Colorectal cancer screening in Europe 5909 Table 2 Colorectal cancer mortality in European countries in 2006 Parameter Countries with the highest mortality Countries with the lowest mortality Mortality > 40/ men (ASR-E): Hungary (54.4), Czech Republic (51), Slovakia (43.3), Croatia (40.7) > 20/ women (ASR-E): Hungary (26.7), Slovakia (24.4), Czech Republic (24.1), Denmark (24.1), Norway (21.4) < 20/ men (ASR-E): Albania (7.3), Greece (15.5), Finland (17.9), Switzerland (19.1), Cyprus (19.3), Bosnia Herzegovina (19.5) < 12/ women (ASR-E): Albania (9.9), Greece (10.8), Finland (11.3), Switzerland (11.6) Adapted from: Ferlay J, Autier P, Boniol M, Heanue M, Colombet M, Boyle P. Estimates of the cancer incidence and mortality in Europe in Ann Oncol 2007; 18: Maximum 35 Maximum ASR (W)-men Median Minimum Median Minimum t /yr Maximum ASR (W)-women Median Minimum ASR (W)-men t /yr Maximum ASR (W)-women Median Minimum t /yr t /yr Figure 2 Incidence trends of colorectal cancer in Europe. Thirty nine cancer registries in , 37 cancer registries in 1997, 96 cancer registries in Adapted from: Parkin DM, Whelan SL, Ferlay J, Storm H. Cancer Incidence in Five Continents, Vol. I to VIII. IARC CancerBase No. 7, Lyon, Available from: URL: section CI5 I-VIII (Detailed). Last accessed on August 8, 2009; Curado MP, Edwards B, Shin HR, Storm H, Ferlay J, Heanue M, Boyle P, editors. Cancer Incidence in Five Continents, Vol. IX. IARC Scientific Publications No. 160, Lyon: IARC, Available from: URL: www-dep.iarc.fr/ section CI5 IX. Last accessed on August 8, and rectosigmoid (> 22 cases/10 5 inhabitants) and tumors of the anus and anal channel (> 1.0 cases/10 5 inhabitants) (Table 3). According to recently published data, CRCrelated mortality has stabilized or shown a slight decrease over recent years. The most extensive population study monitoring the relative survival rate (RSR) is the EUROCARE program [7], which takes registers of patients suffering from malignant diseases as a basis. Data have been gathered and evaluated since The most recent version, EUROCARE-4, Figure 3 Mortality trends of colorectal cancer in Europe. As available in WHO database, countries with cancer registry (Cancer Incidence in Five Continents, Vol. IX). Adapted from: CancerMondial - WHO, International Agency for Research on Cancer, Available from: URL: fr/; World Health Organization (2006), mortality database whosis/whosis/, United Nations, World Population Prospects, the 2006 revision. Available from: URL: Last accessed on August 8, uses comparative analyses of data from the year 1995 to 1999, while data are also available for the years 2000 to 2002 [8]. Data from the European population carcinoma register are also used in the CONCORD study [9], which focuses on a systematic comparison of statistical data between Europe and Northern America. Apart from these two studies, data from population registers of carcinoma have been published for some European countries. Data available regarding the 5-year RSR show high variability across European countries, with borderline values in the Czech Republic (50%) on the one hand and Germany

138 5910 ISSN CN /R World J Gastroenterol December 21, 2009 Volume 15 Number 47 Table 3 Epidemiology of colorectal cancer in the Europe (96 individual cancer registries, ) Parameter incidence Sex C18-C21 Individual diagnoses C18 C19-C20 C21 Crude incidence (cases/ inhabitants) Men Women All ASR-E Men Women All ASR-W Men Women All Mean age (yr) Men Women All Ratio (males: females) (based on No. of cases) 1.1:1 1.0:1 1.4:1 0.6:1 Adapted from: Curado MP, Edwards B, Shin HR, Storm H, Ferlay J, Heanue M, Boyle P, editors. Cancer Incidence in Five Continents, Vol. IX. IARC Scientific Publications No. 160, Lyon: IARC, Available from: URL: section CI5 IX. Last accessed on August 8, ASR-W: World age standardization. (60%) on the other hand [7-16] (Table 4). Several studies have confirmed a favorable time trend in the 5-year RSR; however, these results have to be interpreted carefully with respect to the hidden reasons leading to such positive conclusions. Evaluation of survival rate based on clinical studies of CRC is, unfortunately, rather rare, and therefore, it is impossible to make a representative evaluation of this indicator. This fact should be seen as a challenge when improving population registers of malignant diseases. SCREENING METHODS CRC screening focuses on asymptomatic individuals more than 50 years of age. Age is a low (average) risk factor for sporadic CRC, that is, carcinoma in patients with negative family or case history of CRC or chronic inflammatory bowel disease; this type of carcinoma accounts for 70 to 95% of all CRC cases. Three groups of screening methods are currently used as indicated in Table 5. Guaiac-based fecal occult bleeding test (gfobt) is at present the most frequently used method in screening programs. It detects the peroxidase reaction of hemoglobin, which causes the detection paper impregnated with guaiac resin to turn blue. Dietetic provisions are necessary to exclude false-positive results. A recent study showed limited sensitivity of this test for both, advanced adenomas (11%) and carcinomas (13%) [17]. With the use of gfobt, a decrease in mortality for CRC by 15 to 33% has been proved [18]. Immunochemical fecal occult bleeding test (ifobt) reacts exclusively to human hemoglobin, so no dietetic restrictions are necessary. Taking and assessing the stool samples are easier than is the case with gfobts, which may explain a higher participation rate in the target group. A wide range of qualitative and quantitative tests is presently available, with varying levels of sensitivity and specificity. The advantage of quantitative tests is the possibility to set cut-off limits; the most frequently used values are 75 or 100 ng/ml. The disadvantage of ifobt is its cost; however, the price is now approaching that of gfobt, particularly for qualitative tests [19]. New screening methods include tests which examine the stool for the presence of abnormal DNA. Studies published to date focused on the characteristics of the test rather than the reduction in CRC incidence or mortality. Generally, these tests have higher sensitivity but lower specificity than gfobt. The major obstacle to their implementation in screening programs is price [20]. Flexible sigmoidoscopy (FS) is an endoscopic examination with maximum reach to the splenic flexure. On the basis of the information available, this is a promising screening test, although the studies published to date do not show sufficient statistical significance to determine reduction in CRC mortality. The recommended interval varies from 3 to 5 years. The risk of serious complications is 0% to 0.03% [21]. Unlike FS, colonoscopy also detects lesions in the proximal colon. Its biggest advantage is the possibility of removing pathological lesions within a single examination. It is more sensitive in detecting both adenomas and carcinomas, although limited information is available on reducing CRC incidence and mortality, and on the recommended interval between examinations. The risk of serious adverse events is higher than for FS, at 3 to 5 events per 1000 colonoscopies [22]. To date, no prospective, randomized, multicenter study has been published supporting a reduction in CRC incidence and mortality with the use of screening colonoscopy. Nevertheless, its implementation in screening programs is one of the most widely discussed topics and the American College of Gastroenterology recommends screening colonoscopy as a preferred screening test [23]. On the other hand, no study addressing reductions in the incidence and mortality rates through stool analyses would have been completed without the gold standard of colonoscopy. Computed tomographic colonography (CTC) shows lesions in the colorectum by reconstructing two- and three-dimensional images. To date, no studies have been published assessing reduction in CRC incidence or mortality. The majority of studies have focused on comparing the characteristics of this method with colonoscopy. For larger polyps (over 10 mm), the sensitivity of the methods is comparable; for smaller polyps (less than 5 mm), flat and depressed adenomas, the sensitivity is much higher for optical colonoscopy. Results of studies assessing the effect in terms of reduction in incidence and mortality, cost-effectiveness, and the potential risk of radiation are awaited [24]. Double contrast barium enema shows the entire colorectum, although with significantly lower sensitivity and specificity than colonoscopy or CTC. The percentage of undetected carcinomas is up to 22%. The test is no longer widespread and available, but still has a purpose in countries lacking sufficient resources for other examinations [25]. CRC screening is a complex process which, to function properly, requires the coexistence of a number of factors, such as a functioning invitation-reminder system,

139 Zavoral M et al. Colorectal cancer screening in Europe 5911 Table 4 Five-year relative survival rate (RSR) for colorectal cancer for selected European countries Country Diagnoses Assessment period Five-year RSR (%) Change in time (%) Stage-specific estimates EUROCARE pool [7] C18-C NA EUROCARE pool [8] C18-C NA NA England & Wales [10,11] C M; 47.4 F 5.6 M; 5.6 F NA C19-C M; 51.3 F 7.4 M; 8.1 F NA Germany [12] C18-C NA 85.4 L; 58.1 R; 10.7 M Finland [13] C18-C NA Norway [14] C18-C NA Slovenia [15] C18-C NA Sweden [16] C M; 59.7 F 1.8 M; 2.6 F NA C19-C M; 59.1 F 2.5 M; -1.7 F NA M: Estimate for males; F: Estimate for females; NA: Not available; L: Localized; R: Regional; M: Metastatic; NA: Not available. Numbers in brackets represents source of data available at references section. Table 5 Screening methods Type of method Method Stool tests For presence of occult blood (FOBT) Guaiac-based (gfobt) Immunochemical (ifobt) For presence of abnormal DNA Endoscopic examinations Flexible sigmoidoscopy (FS) colonoscopy Radiologic examinations Computed tomographic colonography (CTC) Double contrast barium enema (DCBE) media campaigns targeted at the general public, the development of recommendations for general practitioners, patient compliance, sufficient funding, stratification of risks, and last but not least the selection of the most suitable screening test. Of the above described tests, only the fecal occult blood tests meet the WHO criteria for screening. As published recently, most CRC screening strategies lead not only to a reduction in CRC incidence and mortality, but also to better control of the costs of CRC treatment, especially with increased chemotherapy costs for advanced CRC [26]. GENERAL ONCOLOGY PREVENTATIVE PROGRAMS IN EUROPE In 1985, the Europe Against Cancer program was initiated, which aimed at a reduction of 15% in the number of deaths caused by tumors (from to ) by The program was implemented, thanks to the cooperation of professional and lay public, charities and anti-smoking groups, healthcare media, and healthcare workers. The project focused on three major areas: prevention, screening, and education. Results published show that although the planned goal was not achieved, the mortality due to tumors was reduced by 10% in the EU. In some countries (Austria and Finland), the desired reduction of 15% was achieved, while in others (Portugal and Greece), the mortality reduction was much lower [27]. The experience gained in this program served as a basis for the Recommendations of the Council of the EU for screening programs following comprehensive European quality assurance guidelines. In December 2003, these recommendations were unanimously approved by the health ministers of the individual EU states. European guidelines for quality assurance of breast and cervical cancer screening have been developed by experts and published by the European Commission; quality assurance guidelines for CRC screening are currently under preparation [28]. CRC SCREENING IN EUROPE In 2008, the Report on the Implementation of the Council Recommendation on Cancer Screening [29], which provides the most comprehensive available data, was published; giving the definitions of program screening as requiring public responsibility by law or official regulation and supervision in contrast to wild screening outside of any program. In program screening, the screening test, the examination interval and the eligible group of persons should be specified. Organized screening should generally include a regional or national team responsible for the implementation, quality assurance and reporting of results. Comprehensive guidelines, rules and a quality assurance structure should be available. Populationbased screening requires the identification and personal invitation of each person in the eligible target population (by an office or special agency). According to this report CRC screening is running or being established in 19 of 27 EU countries. The target group contains approximately 136 million individuals suitable for CRC screening (aged 50 to 74 years). Of this number, 43% individuals come from 12 countries where CRC population screening is performed or being prepared on either national or regional levels; 34% come from 5 countries where national population screening has been implemented (Finland, France, Italy, Poland, and United Kingdom). In 7 EU countries, national non-population based screening is carried out, which covers 27% of the target population. In 2007, gfobt (which in 2003 was the only test recommended by the Council of the European Union) was used as the only screening method in twelve countries (Bulgaria, Czech Republic, Finland, France, Hungary, Latvia, Portugal, Romania, Slovenia, Spain, Sweden, and United Kingdom). Colonoscopy was the only screening method used in Poland. In six countries, two types of tests were used: ifobt and FS in Italy, and gfobt and

140 5912 ISSN CN /R World J Gastroenterol December 21, 2009 Volume 15 Number 47 Table 6 Colorectal cancer screening programs in 2007 Program Test type Screening interval Age eligible national population Type Status years or times in LT Age (yr) Persons ( 1000) Austria NonPB Natw FOBT 1 or 2 > NonPB Natw CS 10 > Belgium No Prog 2880 Bulgaria NonPB Natw FOBT 1 > Cyprus PB Natw-plan FOBT 1 in LT PB Natw-plan CS 1 in LT Czech Republic NonPB Natw FOBT 2 > Denmark No Prog 1540 Estonia No Prog 370 Finland PB Natw-roll ong FOBT France PB Natw-roll ong FOBT Germany NonPB Natw FOBT 1 and 2 > NonPB Natw CS 10 (2 in LT) Greece NonPB Natw FOBT 5 > NonPB Natw CS 5 > Hungary PB Natw-pilot FOBT Ireland No Prog 940 Italy PB Natw-roll ong FOBT (70-75) PB Reg-roll ong FS 1 in LT 58 or Latvia NonPB Natw FOBT 1 > Lithuania No Prog 870 Luxembourg No Prog 120 Malta No Prog 120 Netherlands No Prog 4460 Poland PB Natw-roll ong CS Portugal PB Natw-plan FOBT Romania PB Natw-plan FOBT Slovak Republic NonPB Natw FOBT > NonPB Natw-plan CS 10 > Slovenia PB Natw-plan FOBT Spain PB Reg-pilot FOBT Sweden PB Reg-plan FOBT UK PB Natw-roll ong FOBT 2 (50) (74) 7600 Dual prog/test Subtotal Excluded pop Total PB: Population based; Prog: Program; Natw: Nationwide; Reg: Regional; Plan: Planning; Roll ong: Rollout ongoing; Pilot: Piloting; CS: Colonoscopy; LT: Lifetime. dual prog/test: Individuals entered twice due to screening programs of different implementation or using different screening tests. excluded pop.: Individuals excluded from national target populations due to regional or national variations in the age group targeted for screening, or due to lack of screening programs in some regions of countries with regional implementation status. Adapted from: von Karsa L, Anttila A, Ronco G, Ponti A, Malila N, Arbyn M, Segnan N, Castillo-Beltran M, Boniol M, Ferlay J, Hery C, Sauvaget C, Voti L, Autier P. Cancer screening in the European Union. Report on the implementation of the Council Recommendation on cancer screening - First Report. ISBN European Communities (publ.) Printed in Luxembourg by the services of the European Commission, Available from: URL: pdf. Last accessed on August 4, colonoscopy in Austria, Cyprus, Germany, Greece, and Slovak Republic. In the remaining eight states (Belgium, Denmark, Estonia, Ireland, Lithuania, Luxembourg, Malta, and the Netherlands), CRC screening has not been implemented yet. The age limit for the target population varies across EU countries (Table 6). In 2007, it was estimated that a total of 12 million individuals participated in CRC screening. In the United Kingdom, a screening program was announced in 2004 and initiated in 2006, with the prospect of national coverage in It has been designed in two stages, with gfobt examinations at 2-year intervals and colonoscopy for positive tests. In 2007, compliance was 52%. The program is carried out through regional centers falling under one of five national hubs. The role of general practitioners is less significant here [30]. In France, a screening program was initiated in 2003, based on gfobt tests at 2-year intervals with colonoscopy for positive results. The role of general practitioners as coordinators is of crucial importance. The major advantage of the French program is its good organization, with a call-recall system comprising central management at national level and individual steps taken by centers in individual departments. Asymptomatic individuals aged from 50 to 74 are mailed gfobt tests, with a reminder at three-monthly intervals for nonparticipants. Compliance in referred districts achieved 42%, and the overall positive test rate was 2.7% [31].

141 Zavoral M et al. Colorectal cancer screening in Europe 5913 In Italy, a nation-wide campaign was initiated in 2005; the implementation was entrusted entirely to 21 regional centers, including choice of the testing method. With state financial support, screening has been initiated in 11 regions to date, mostly in the industrial areas of northern Italy. In the Piedmont region, FS is the method of choice, in other regions immunochemical FOBT, with colonoscopy for positive tests. Compliance in ifobt and FS programs was 44.6% and 51.4%, respectively. Positivity rate of ifobt was 5.3% at first and 3.9% at repeat screening [32]. In Spain, no screening program has taken place as yet. The main obstacle to its implementation is the highly heterogeneous healthcare system, in terms of organization and insurance coverage in individual selfgoverning units. Catalonia, for instance, considers implementation of country-wide screening in 2010, while in other regions only limited pilot studies have been held so far. In Finland, a structured screening program was initiated in The target population, aged from 60 to 69 years ( individuals), was randomized into two groups. Individuals in the screening group were mailed a gfobt test at intervals of 2 years. The Finnish program shows a high level of compliance of the target population (70.8%), particularly for females [33]. In the Netherlands, the optimum screening strategy is still being developed. It will be based on the results of studies currently taking place at major academic workplaces, comparing the effect of endoscopic procedures, various types of FOBTs, and fecal DNA analysis. Poland is the only state at the moment using colonoscopy as the only screening method, without the alternative of FOBT. An opportunistic screening program was initiated in 2000, and by 2005, this had grown to 57 centers across Poland. The program is financed by the Ministry of Health, independent of the overall healthcare system. The target population (asymptomatic individuals aged years) is recruited through general practitioners. High emphasis is placed on the quality control of colonoscopies, with complications reported for 0.1% of procedures, and no patient mortality. The advantage of the program is thorough monitoring and evaluation, including monitoring of interval cancers [34]. Germany was the first country to introduce a population screening program (in 1976) based on annual gfobt for individuals more than 44 years of age. Starting from 2002, it has been offering participants a choice between colonoscopy at 55 years of age and FOBT at annual intervals between 50 and 55 years of age. After 55 years of age, examinations are carried out at 2-year intervals. If the test results are positive, colonoscopy is indicated. Those who undergo a screening colonoscopy with no neoplasia detected at the initial examination are recommended reexamination in 10 years time if the first colonoscopy was carried out before they were 65 years. The positive feature of the screening and data gathering in Germany is the emphasis on staging the disease at the time of its diagnosis. Recent cost analyses have proven that this type of screening is cost-efficient [35]. In the Czech Republic, CRC screening has many years of tradition [36,37]. The country was the second in the world to start screening nation-wide, in In the initial years, gfobt was the first method offered to asymptomatic individuals more than 50 years of age by their general practitioners at preventative medical checks, followed by colonoscopy if tests were positive. From 2000 to 2008, gfobts were carried out, of which were positive (3.76%). In 2006, a central database for online data input was established. Between 2006 and 2008, colonoscopies were carried out, indicated as a result of a positive FOBT; carcinoma was diagnosed in 1047 (5.9%) individuals, and 5362 (30.1%) adenomas were removed by endoscopic polypectomy. The participation of the target group, however, was only 20% [38]. In order to achieve a higher compliance rate, screening colonoscopy was added to current FOBT screening as an alternative method, in the same intervals as in the German program. Both, gfobt and ifobt are offered as well. The implementation of the newly designed program is supported by an intensive media campaign ( The first study which focused on monitoring the effect of colonoscopy screening on reducing CRC incidence and mortality is NordICC (The Nordic-European Initiative on Colorectal Cancer), which is currently underway in northern states of Europe (Norway, Sweden, and Iceland), Poland, and the Netherlands. It will involve a minimum of individuals aged 55 to 64 years. Individuals in the screening group will undergo a screening colonoscopy once in a lifetime. The primary objective is to compare incidence and mortality against the control group (with no screening) after 10 years [39]. CONCLUSION CRC presents a serious public healthcare issue for the population of Europe. Understandably, the number of countries introducing population screening has been growing constantly. Although epidemiologic data differ in various European countries, implementation of screening programs in accordance with the principles spelled out in the Council Recommendation on Cancer Screening of 2 December 2003 may be expected to have a favorable effect on the burden of this disease in the population. Countries in the EU may benefit from unified policy, knowhow and central oncology registers, while economically less developed countries may draw on special funding for the development of preventative programs. At the same time, varying epidemiologic situations, economic conditions, and different systems of health insurance and organization of healthcare are factors that may limit the implementation of a unified screening program. Therefore, to respond to the needs of the member countries, the EU should consider adopting the recommendation of the World Gastroenterology Organization for CRC screening, possibly even in a modified form [40]. This is a cascade concept in which recommendations for individual countries are graded into six levels, depending on the resources available (financial

142 5914 ISSN CN /R World J Gastroenterol December 21, 2009 Volume 15 Number 47 Table 7 Cascade concept Level Average risk High risk 1 Colonoscopy in 10 years interval, from 50 years of age Special procedure, for individual groups 2 Colonoscopy once in a lifetime, at 50 years of age Special procedure, for individual groups 3 Flexible sigmoidoscopy in 5 years interval, from 50 years of age; Special procedure, for individual groups colonoscopy to follow if positive 4 Flexible sigmoidoscopy once in a lifetime, at 50 years of age; colonoscopy to follow if positive Special procedure, for individual groups 5 Flexible sigmoidoscopy once in a lifetime, at 50 years of age; colonoscopy to follow only if advanced adenoma is detected 6 FOBT in annual interval after 50 years of age; if positively tested, colonoscopy or double contrast barium enema Same as individuals with average risk, if resources are not available for colonoscopy Same as individuals with average risk, if resources are not available for colonoscopy Adapted from: World Gastroenterology Organization/International Digestive Cancer Alliance. Practice Guidelines: Colorectal cancer screening. Available from: URL: Last accessed on August 4, and professional) (Table 7). In the case of lack of funds, FOBT at intervals of 1 or 2 years for individuals with average risk is a realistic possibility. This open concept best fulfils the simple recommendation by Sydney Winawer, Co-Chair of IDCA (International Digestive Cancer Alliance): The best screening test is the one that gets done...and gets done well. Do what you can with what you have. In most European countries, fortunately, the majority of the population is covered by some form of health insurance, meaning that economic aspects need not critically affect the availability of screening programs. Although at the end of 2007, CRC screening was still not running or being established in 8 of 27 EU member states, some of which rank among the most developed economies of the world, additional programs are currently under development. Given the substantial burden of the disease, implementation and continuous improvement in CRC screening programs should remain high on the healthcare agenda in Europe. REFERENCES 1 Spann SJ, Rozen P, Young GP, Levin B. Colorectal cancer: how big is the problem, why prevent it, and how might it present? In: Rozen P, Young GP, Levin B, Spann SJ, editors. Colorectal Cancer in Clinical Practice. London: Martin Dunitz Ltd, 2002: Curado MP, Edwards B, Shin HR, Storm H, Ferlay J, Heanue M, Boyle P, editors. Cancer Incidence in Five Continents, Vol. IX. IARC Scientific Publications No. 160, Lyon: IARC, Available from: URL: section CI5 IX. Last accessed on August 8, World Health Organization (2006), mortality database United Nations, World Population Prospects, the 2006 revision. Available from: URL: Last accessed on August 8, Ferlay J, Bray F, Pisani P, Parkin DM. GLOBOCAN 2002: Cancer incidence, mortality and prevalence worldwide. IARC Cancer Base No. 5 version 2.0. Lyon: IARC press, Available from: URL: section CI5 I-VIII (Detailed). Last accessed on August 8, Parkin DM, Whelan SL, Ferlay J, Storm H. Cancer Incidence in Five Continents, Vol. I to VIII. IARC CancerBase No. 7, Lyon, Available from: URL: fr/, section CI5 I-VIII (Detailed). Last accessed on August 8, Ferlay J, Autier P, Boniol M, Heanue M, Colombet M, Boyle P. Estimates of the cancer incidence and mortality in Europe in Ann Oncol 2007; 18: Berrino F, De Angelis R, Sant M, Rosso S, Bielska-Lasota M, Coebergh JW, Santaquilani M. Survival for eight major cancers and all cancers combined for European adults diagnosed in : results of the EUROCARE-4 study. Lancet Oncol 2007; 8: Verdecchia A, Francisci S, Brenner H, Gatta G, Micheli A, Mangone L, Kunkler I. Recent cancer survival in Europe: a period analysis of EUROCARE-4 data. Lancet Oncol 2007; 8: Coleman MP, Quaresma M, Berrino F, Lutz JM, De Angelis R, Capocaccia R, Baili P, Rachet B, Gatta G, Hakulinen T, Micheli A, Sant M, Weir HK, Elwood JM, Tsukuma H, Koifman S, E Silva GA, Francisci S, Santaquilani M, Verdecchia A, Storm HH, Young JL. Cancer survival in five continents: a worldwide population-based study (CONCORD). Lancet Oncol 2008; 9: Mitry E, Rachet B, Quinn MJ, Cooper N, Coleman MP. Survival from cancer of the colon in England and Wales up to Br J Cancer 2008; 99 Suppl 1: S26-S29 11 Mitry E, Rachet B, Quinn MJ, Cooper N, Coleman MP. Survival from cancer of the rectum in England and Wales up to Br J Cancer 2008; 99 Suppl 1: S30-S32 12 Gondos A, Arndt V, Holleczek B, Stegmaier C, Ziegler H, Brenner H. Cancer survival in Germany and the United States at the beginning of the 21st century: an up-to-date comparison by period analysis. Int J Cancer 2007; 121: Gondos A, Bray F, Brewster DH, Coebergh JW, Hakulinen T, Janssen-Heijnen ML, Kurtinaitis J, Brenner H. Recent trends in cancer survival across Europe between 2000 and 2004: a model-based period analysis from 12 cancer registries. Eur J Cancer 2008; 44: Gondos A, Bray F, Hakulinen T, Brenner H. Trends in cancer survival in 11 European populations from 1990 to 2009: a model-based analysis. Ann Oncol 2009; 20: Janssen-Heijnen ML, Houterman S, Lemmens VE, Brenner H, Steyerberg EW, Coebergh JW. Prognosis for long-term survivors of cancer. Ann Oncol 2007; 18: Talback M, Rosen M, Stenbeck M, Dickman PW. Cancer patient survival in Sweden at the beginning of the third millennium--predictions using period analysis. Cancer Causes Control 2004; 15: Imperiale TF, Ransohoff DF, Itzkowitz SH, Turnbull BA, Ross ME. Fecal DNA versus fecal occult blood for colorectalcancer screening in an average-risk population. N Engl J Med 2004; 351: Mandel JS, Bond JH, Church TR, Snover DC, Bradley GM,

143 Zavoral M et al. Colorectal cancer screening in Europe 5915 Schuman LM, Ederer F. Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. N Engl J Med 1993; 328: van Rossum LG, van Rijn AF, Laheij RJ, van Oijen MG, Fockens P, van Krieken HH, Verbeek AL, Jansen JB, Dekker E. Random comparison of guaiac and immunochemical fecal occult blood tests for colorectal cancer in a screening population. Gastroenterology 2008; 135: Ahlquist DA, Sargent DJ, Loprinzi CL, Levin TR, Rex DK, Ahnen DJ, Knigge K, Lance MP, Burgart LJ, Hamilton SR, Allison JE, Lawson MJ, Devens ME, Harrington JJ, Hillman SL. Stool DNA and occult blood testing for screen detection of colorectal neoplasia. Ann Intern Med 2008; 149: , W81 21 Single flexible sigmoidoscopy screening to prevent colorectal cancer: baseline findings of a UK multicentre randomised trial. Lancet 2002; 359: Lieberman DA. Clinical practice. Screening for colorectal cancer. N Engl J Med 2009; 361: Rex DK, Johnson DA, Lieberman DA, Burt RW, Sonnenberg A. Colorectal cancer prevention 2000: screening recommendations of the American College of Gastroenterology. American College of Gastroenterology. Am J Gastroenterol 2000; 95: Johnson CD, Chen MH, Toledano AY, Heiken JP, Dachman A, Kuo MD, Menias CO, Siewert B, Cheema JI, Obregon RG, Fidler JL, Zimmerman P, Horton KM, Coakley K, Iyer RB, Hara AK, Halvorsen RA Jr, Casola G, Yee J, Herman BA, Burgart LJ, Limburg PJ. Accuracy of CT colonography for detection of large adenomas and cancers. N Engl J Med 2008; 359: Toma J, Paszat LF, Gunraj N, Rabeneck L. Rates of new or missed colorectal cancer after barium enema and their risk factors: a population-based study. Am J Gastroenterol 2008; 103: Lansdorp-Vogelaar I, van Ballegooijen M, Zauber AG, Habbema JD, Kuipers EJ. Effect of rising chemotherapy costs on the cost savings of colorectal cancer screening. J Natl Cancer Inst 2009; 101: Boyle P, d Onofrio A, Maisonneuve P, Severi G, Robertson C, Tubiana M, Veronesi U. Measuring progress against cancer in Europe: has the 15% decline targeted for 2000 come about? Ann Oncol 2003; 14: European Council. Council Recommendation of 2 December 2003 on cancer Screening (2003/878/EC). OJ L 327 (16 December 2003), 3, Available from: URL: eurlex.europa.eu/lexuriserv/lexuriserv.do?uri=oj: L:2003:327:0034:0038:EN:PDF. Last accessed on August 4, von Karsa L, Anttila A, Ronco G, Ponti A, Malila N, Arbyn M, Segnan N, Castillo-Beltran M, Boniol M, Ferlay J, Hery C, Sauvaget C, Voti L, Autier P. Cancer screening in the European Union. Report on the implementation of the Council Recommendation on cancer screening - First Report. ISBN European Communities (publ.) Printed in Luxembourg by the services of the European Commission, Available from: URL: eu/health/ph_determinants/genetics/documents/ cancer_screening.pdf. Last accessed on August 4, West NJ, Boustiere C, Fischbach W, Parente F, Leicester RJ. Colorectal cancer screening in Europe: differences in approach; similar barriers to overcome. Int J Colorectal Dis 2009; 24: Goulard H, Boussac-Zarebska M, Ancelle-Park R, Bloch J. French colorectal cancer screening pilot programme: results of the first round. J Med Screen 2008; 15: Zorzi M, Falcini F, Fedato C, Grazzini G, de Bianchi PS, Senore C, Vettorazzi M, Visioli C, Zappa M. Screening for colorectal cancer in Italy: 2006 survey. Epidemiol Prev 2008; 32: Malila N, Oivanen T, Malminiemi O, Hakama M. Test, episode, and programme sensitivities of screening for colorectal cancer as a public health policy in Finland: experimental design. BMJ 2008; 337: a Regula J, Rupinski M, Kraszewska E, Polkowski M, Pachlewski J, Orlowska J, Nowacki MP, Butruk E. Colonoscopy in colorectal-cancer screening for detection of advanced neoplasia. N Engl J Med 2006; 355: Pox C, Schmiegel W, Classen M. Current status of screening colonoscopy in Europe and in the United States. Endoscopy 2007; 39: Fric P, Zavoral M, Dvorakova H, Zoubek V, Roth Z. An adapted program of colorectal cancer screening--7 years experience and cost-benefit analysis. Hepatogastroenterology 1994; 41: Zavoral M, Zavada Filip, Salek C. Czech Society of Gastroenterology: Colorectal Cancer Screening in the Czech Republic. Endoscopy 2006; 38: Seifert B, Zavoral M, Fric P, Bencko V. The role of primary care in colorectal cancer screening: experience from Czech Republic. Neoplasma 2008; 55: NordiCC Study Protocol. Version MB Available from: URL: Projects/NordICC. Last accessed on August 4, World Gastroenterology Organization/International Digestive Cancer Alliance. Practice Guidelines: Colorectal cancer screening. Available from: URL: worldgastroenterology.org/colorectal-cancer-screening.html. Last accessed on August 4, 2009 S- Editor Tian L L- Editor Webster JR E- Editor Ma WH

144

145 Paper VII. Dusek L., Abrahamova J., Lakomy R., Vyzula R., Koptikova J., Pavlik T., Muzik J., Klimes D. Multivariate analysis of risk factors for testicular cancer: a hospital-based case-control study in the Czech Republic. Neoplasma Vol.55, No.4, p , 2008

146 356 NEOPLASMA 55, 4, 2008 Multivariate analysis of risk factors for testicular cancer: a hospital-based case-control study in the Czech Republic L. DUSEK 1*, J. ABRAHAMOVA 2, R. LAKOMY 3, R. VYZULA 3, J. KOPTIKOVA 1, T. PAVLIK 1, J. MUZIK 1, D. KLIMES 1 1 Institute of Biostatistics and Analyses, Masaryk University, Kamenice 126/3, Brno, Czech Republic, [email protected]; 2 Thomayer University Hospital, Prague, Czech Republic; 3 Masaryk Memorial Cancer Institute, Brno, Czech Republic Received December 14, 2007 Growing incidence of testicular cancer around the world stimulates research attempting to explain the trends. This study quantified the contribution of different types of potential risk factors for testicular germ-cell cancer (TGCC) with differentiation between seminoma and non-seminoma. A standardized questionnaire containing demographic data, pre- and perinatal factors, social, lifestyle and occupational parameters was prepared. The data file consists of n = 356 TGCCs (seminoma: n = 195; non-seminoma: n = 161) and n = 317 controls, frequency matched on age to cases. The following factors were significantly associated with the risk of TGCCs in univariate analyses (ORs): atrophic testis (5.3), smoking over 12 pack-yr (4.9), cryptorchidism (2.9), testicular trauma (2.0), birth weight under 3,000 g (1.6), low degree of education (3.0) in correlation with manual occupation (2.3) and finally, overall familial cancer history (1.5) and familial history of breast (1.8) and prostate cancer (3.9). On the other hand, maternal age over 20 yr (OR < 0.4) and moderate recreational sport activity (OR = 0.5) significantly reduced the risk of TGCCs. A significant risk was associated with cryptorchidism (OR = 2.9; 95% CI = ) where orchidopexy was delayed after 5 yr of age (OR = 5.2; 95% CI = ). Delayed orchidopexy was associated namely with the risk of seminomas (OR = 7.5; 95% CI = ). Only some of the variables were retained in multivariate model for TGCCs as well as for histological subtypes (multivariate adjusted OR for all TGCCs): atrophic testis (5.9), family history of prostate cancer (4.8), cryptorchidism (3.8) and interaction term low degree of education & manual occupation (3.0). Familial history of breast cancer elevated risk of TGCCs and of seminomas (OR: ). Birth weight under 3,000 g was retained in a multivariate model for TGCCs with a borderline significance (OR = 1.67). We could not rule out any type of risk factors, as each one was significantly represented in the final multivariate models. Familial cancer history remained to be an influential risk factor, altogether with some lifestyle and occupational parameters. This suggests that both environmental exposures and genetic inheritance can play role in the moderation of the risk of TGCC. Keywords: testicular cancer, risk factors, case-control study Testicular cancer (TC) is the most frequent malignancy in young men, mostly diagnosed from 20 to 45 years of age. These neoplasms represent a diverse group with dominant occurrence of testicular germ cell cancers (TGCC; 95% of all TC). TC incidence has increased significantly around the world in the past few decades [1 3]. Recent trends in the Czech Republic also suggest an increase in TC age-standardized incidence rate (by 22.4% in the period ; [4]). The growing incidence stimulates epidemiological research that attempts to explain the trends. Although many studies focused on risk factors for TC, the aetiology of this cancer * Corresponding author remains largely unknown and future research is suggested [5, 6]. Only a few risk factors for TC are consistently established, including cryptorchidism, carcinoma in situ and exposure to estrogen in utero [7]. Although the exact mechanisms are still not known, many studies proved critical importance of pregnancy and prenatal characteristics for the development of TC [8 11]. Of course, many other factors affecting the individual s later life may also play a role in the aetiology of TC; however, the evidence is still unclear or even conflicting in different studies. Our study attempts to contribute to the current discussion on aetiology of TC [5, 6, 12]. We seek to quantify the contribution and mutual interactions of very different types of

147 RISK FACTORS FOR TESTICULAR CANCER, MULTIVARIATE ANALYSIS 357 potential risk factors for TGCCs with further differentiation between seminoma and non-seminoma. The risk power of hereditary, pre- and perinatal factors in combination with a wide range of social and lifestyle factors is assessed, using data from epidemiologically well-defined Czech population. Patients and methods Parameters, subjects and data. For the purpose of this study, we prepared a standardized questionnaire containing demographic data and a complex set of risk factors including maternal health, pre- and perinatal factors, social parameters, personal lifestyle and occupational history. The key parameters identifying the TGCC patient are as follows: diagnosis (coded according to the International Classification of Diseases, ICD-10), date of diagnosis, age at diagnosis, cancer location and histology, TNM classification and staging [13]. The type of TGCCs was binary classified as pure seminoma and as other histological type, coded as non-seminoma. The latter group thus includes also tumors with mixed histology. The questionnaires were always filled in with the assistance of an experienced physician. The histological types were additionally extracted from pathology reports and hospital information systems. If needed, the characteristics of the patient s health history were verified using his health care documentation. All participants were asked to give permission for the anonymous processing of collected data. The questionnaires were digitized using an on-line database where data can be used for further clinical and epidemiological research. The validity of the registry can be regarded as high due to double digitization and subsequent control of key variables. Criteria for the exclusion of a record from the analysis were as follows: incorrect or insufficient identification of cancer (12 cases), missing information on important covariates (18 cases, 6 controls) or another inconsistency in parameters (24 cases, 13 controls). Additionally, 2 cases with TC of non-germ cell type were excluded. The study was also strictly limited to men whose parents were known and described in terms of age, health status and cancer burden in order to eliminate bias in the familial risk analysis. After applying all inclusion/exclusion criteria, the resulting file consists of n = 356 TGCCs and n = 317 matched controls. Participants were recruited in two important Czech cancer centers Thomayer University Hospital in Prague (181 cases, 191 controls) and Masaryk Memorial Cancer Institute in Brno (175 cases, 126 controls). Testing for effects of participating centers revealed no significant difference in recruitment frequency, histological types of TC and in age distribution of cases and controls (data not shown). All participants were Caucasians. Cases were diagnosed with primary TGCC in the period (aged yr at the time of diagnosis). The age of parents was not limited. Controls were men with verified health status, frequency matched on age to cases. Controls were recruited among blood donors, among healthy persons accompanying the patients (not relatives) and partially among hospital personnel. The recruitment of controls was facilitated by quoting the age groups. The quoting was partially skewed to include adequate number of men younger than 24 years in order to reach comparable age distribution with cancer cases. The control recruitment was fully random within each age stratum. The recruitment took place in the period of 01/ /2007. Statistical data analysis. All analyses were carried out separately for TGCCs and for seminomas and non-seminomas. For each histological group, we carried out exactly the same set of analyses as for all cancer cases. The risk factors were coded as binary variables and if necessary, the quantitative factors were categorized according to quartiles in their sample distribution. Frequency analysis or robust summary statistics (median, percentiles) were used to summarize values of examined variables. Standard univariate statistical techniques were used to test the differences between groups of patients and controls, ML chisquare test for ordinal variables and Mann-Whitney U test for continuous variables. One-way ANOVA model was applied to compare means of quantitative variables among histological subtypes of TGCCs and controls. Unconditional logistic regression was used for all analyses. All univariate models were adjusted only for age of men as the matching variable (age taken as continuous variable). Only model for smoking intensity quantified as pack-years was further adjusted for years since quitting. Estimated odds ratio (OR) was supported with 95% confidence intervals (CI). The significance tests were always two-sided, based on the likelihood ratio test [14]. In the case of categorized quantitative variables, the test for linear trends was performed taking median of each category and performing logistic regression evaluated by Wald s 2 statistics. The same statistics was applied in the heterogeneity test based on multinominal logistic regression comparing histological subtypes with controls for groups defined by risk factors [15]. In addition to age as adjusting variable, multivariate ORs were adjusted for all the variables entering the model. It ensured that the outcome from multivariate regression measured the effect of a particular parameter, after adjusting for all important attributes. The significance of interaction terms in the multivariate models was successively tested comparing loglikelihood difference between models with and without the terms. Significant and contributing interactions were incorporated as independent variables. All p values were derived from two-tailed testing with universal limit 0.05 for the statistical significance. In agreement with many previous studies, we decided to present the original outcomes without additional corrections for multiple comparisons. Instead of it, achieved ORs were correctly sorted according to level of significance and borderline significance was carefully interpreted. The study was also supported by prospectively optimized power analysis. It ensures 85% power to detect and odds ratio of 2.0 or more for factors with a prevalance of 5% and 90% power to detect OR of 2.0 or more for factors with a prevalence of 10%.

148 358 L. DUSEK, J. ABRAHAMOVA, R. LAKOMY, R. VYZULA et al. Table 1. Distribution of the matching variables and social covariates among TGCCs and controls All TGCC cases (N = 356) Controls (N = 317) Seminoma cases (N = 195) Non-seminoma cases (N = 161) Age group (years) (11.0 %) 36 (11.4 %) 11 (5.6 %) 28 (17.4 %) (21.6 %) 88 (27.8 %) 33 (16.9 %) 44 (27.3 %) (20.8 %) 81 (25.6 %) 39 (20.1 %) 35 (21.8 %) (16.6 %) 40 (12.5 %) 37 (19.0 %) 22 (13.7 %) (13.5 %) 31 (9.8 %) 33 (17.0 %) 15 (9.3 %) (6.7 %) 18 (5.7 %) 19 (9.7 %) 5 (3.1 %) (7.0 %) 14 (4.3 %) 19 (9.7 %) 6 (3.7 %) (1.7 %) 6 (1.9 %) 2 (1.0 %) 4 (2.5 %) (1.1 %) 3 (1.0 %) 2 (1.0 %) 2 (1.2 %) p level 1 p = p = Living place Village 131 (36.8 %) 61 (19.2 %) 77 (39.5 %) 54 (33.5 %) Small town 126 (35.4 %) 104 (32.8 %) 69 (35.4 %) 57 (35.4 %) Big town (city) 99 (27.8 %) 152 (48.0 %) 49 (25.1 %) 50 (31.1 %) p level 1 p < p = Education Primary/skilled 130 (36.5 %) 50 (15.8 %) 70 (35.9 %) 60 (37.3 %) Secondary 152 (42.7 %) 190 (59.9 %) 77 (39.5 %) 75 (46.5 %) University 74 (20.8 %) 77 (24.3 %) 48 (24.6 %) 26 (16.2 %) p level 1 p < p = Legend to Table 1. 1 Main lifetime living place 2 Significance of ML- 2 test for differences between TGCCs and controls or between seminoma and non-seminoma cases The age statistics of the histological subtypes is calculated in Table 2. On average, non-seminoma cases were younger than seminoma cases (mean age 32 yr and 37 yr). Among TGCC patients, there appeared to be a significantly increased prevalence of men with low degree of education (pri- Table 2. Distribution of age, histological types and clinical stages in the Czech National Cancer Registry (NCR) and in the examined sample of TGCCs NCR Examined TGCCs (Diagnostic period ) (Diagnostic period ) Variables and categories Nonseminoma All TGCCs Seminoma All TGCCs Seminoma Non-seminoma No. of cases Prevalence of clinical stages I 60.3 % 65.6 % 53.7 % 60.9 % 67.2 % 53.4 % II 18.0 % 17.5 % 18.8 % 24.7 % 22.0 % 27.9 % III 15.1 % 9.8 % 21.7 % 14.3 % 10.8 % 18.6 % Unknown 6.5 % 7.1 % 5.8 % Prevalence of T categories T0 0.3 % 0.2 % 0.4 % 0.3 % 0.5 % 0 % T % 65.5 % 60.2 % 67.7 % 69.7 % 65.2 % T % 20.0 % 25.3 % 23.8 % 20.5 % 27.9 % T % 11.0 % 11.5 % 8.1 % 9.2 % 6.8 % Tx 3.1 % 3.3 % 2.7 % Age at diagnosis (years) Mean Median /75 percentile th 27/41 31/45 24/35 27/40 29/42 25/36 10/90 th percentile 21/55 25/57 19/48 23/48 25/49 22/43 Results The Table 1 documents the age distribution as matching variable between cases and controls. The case (control) series had a mean age of 35 yr (34 yr) and a median of 33 yr (32 yr).

149 RISK FACTORS FOR TESTICULAR CANCER, MULTIVARIATE ANALYSIS 359 mary or skilled) than controls. There was also a significant asymmetry in the type of living place between TGCCs and controls, the latter living more frequently in big towns. Neither education nor type of living place appeared to affect the distribution between seminomas and non-seminomas (Table 1). All relevant risk factors from the questionnaire were listed in Table 3. All factors were retained in the analyses provided that they occurred in at least one TGCC case. Nevertheless, some variables were strongly limited in their discrimination potential due to very low prevalence (diseases in personal health history, hereditary defects, etc.) or, on the other hand, due to ubiquitous occurrence (declared vegetable and fruit consumption, dietary fat preferences). The Tables 4 6 document a complex set of univariate logistic regression-derived odds ratios (ORs) with 95% confidence intervals. Only associations that reached minimum level of significance (p < 0.05) were listed according to the type of variable (maternal and reproductive health history in Table 4, social and lifestyle factors in Table 5 and family cancer history in Table 6). As for the reproductive history and health-related factors (Table 4), we found elevated ORs among men with positive history of cryptorchidism, atrophic testis and declared testicular trauma. Newborns with low birth weight (< 3,000 g) revealed only borderline significance of risk for TGCCs and for seminomas. Maternal age in all categories above 20 yr of age significantly reduced risk of the development of TGCCs. The protective effect of maternal age > 20 yr was slightly more pronounced in non-seminomas than in seminomas but without significant difference (p of trend test > 0.173, p of heterogeneity test = 0.202). The significant risk association of cryptorchidism (OR = 2.94; 95% CI = ) further increased when the time of orchidopexy was taken into account. The cut-off time was identified at 5 years, orchidopexy after 5 yr of age achieved highly significant OR = 5.24 (95% CI = ). Odds ratio for seminomas was significantly elevated among the cryptorchidic men with orchidopexy after 5 yr of age (OR = 7.47, 95% CI = ) while this risk factor was not significant for non-seminomas (OR = 2.66, 95% CI = ) (Table 4). No significant risk was detected in hypospadia, phimosis, epidydimitis, orchitis and other diseases in the men s health history. Among examined dietary factors, only smoking reached significantly elevated ORs for the whole group of TGCCs, as well as for seminomas and non-seminomas (Table 5). Neither categorized smoking history nor adjusted pack-years contributed to the discrimination between histological subtypes (p of the heterogeneity test ranged from to 0.594). Smoking history reached weak significance, separating only pure non-smokers from both former and active smokers. Quantified pack-years appeared to be a more sensitive risk indicator, namely in the category of heavy smokers with more than 12 pack-yr (OR = 4.93; p < 0.001). The increasing ORs with increasing pack-yr revealed consistently significant trend both for seminomas and non-seminomas. There were no significant differences between seminomas and non-seminomas in the risk profiles of occupational, social and lifestyle factors (Table 5). In addition to already described significantly increased risk association of low degree of education, we identified elevated ORs among men working manually (OR = 2.87, 95% CI = ), with occupational physical activity (OR = 2.26, 95% CI = ) and among those who declared regular night work for at least 3 years prior to the diagnosis (OR = 1.48, 95% CI = ). The analysis of contingency tables (data not shown) proved a significant inter-correlation among all these social and occupational attributes. These characteristics at least partially reflect the same situation, i.e. occupation associated with manual work and relatively low degree of education (manual workers, drivers). The type of main living place stands in a rather different position: a significant reduction of ORs among inhabitants of big towns was confounded with an increased prevalence of men with office-like work and students. The elevated OR among men living in rural places (OR = 1.63, 95% CI = ) achieved very borderline significance only for TGCCs (Table 5). As for factors related to passive or active life style, the only significant association was proved in the case of the protective effect of a moderate recreational sport activity (defined as non-professional activity carried out max. 1-2 times per week). Here, age adjusted OR for all TGCCs as well as for histological subtypes was significantly reduced in comparison with people without any sport activity as referent category (ORs < 0.4; Table 5). Neither the type of professional activity nor any specific sport activity (biking, motor biking, horse riding) changed significantly OR values. Similarly, no significant effect was observed adjusting any type of sport activity for life periods, including professional sport at puberty. Summed counts of neoplasms in family history appeared to be associated with a significantly increased risk of TGCC (Table 6). It refers to the occurrence of any neoplasms among relatives, positive history in close relatives being slightly more significant (OR = 1.72; 95% CI = ) than in distant relatives (OR = 1.42; 95% CI = ). Familial cancer history did not contribute significantly to the differentiation of histological types, although there were numerically increased ORs in seminomas as compared with non-seminomas (Table 6). Among all specifically questioned cancer diagnoses (breast, prostate, GIT, ovary, testes), only positive familial history of breast and prostate cancer elevated significantly risk of TGCC. History of prostate cancer significantly increased ORs in all examined groups and although OR was higher for non-seminomas (OR = 4.68; 95% CI = ) than for seminomas (OR =3.25; 95% CI = ), we could not distinguish the groups (heterogeneity test: p = 0.526). Familial history of breast cancer achieved only borderline significance of a global OR (OR = 1.83; 96% CI = ) and of OR for seminomas (OR = 1.93; 96% CI = ).

150 360 L. DUSEK, J. ABRAHAMOVA, R. LAKOMY, R. VYZULA et al. Table 3. Potential risk factors in the study that occurred in at least one TGCC case MATERNAL HEALTH, REPRODUCTIVE HISTORY, DIETARY FACTORS CHILDBIRTH AND NEWBORN CHILD Control TGCC Control TGCC Prevailing type of fat in diet Any abnormality during the pregnancy 10.7 % 21.1 % Vegetable 73.8 % 74.5 % (in binary code) 1 Animal 26.2 % 25.5 % Nausea 0.6 % 1.7 % Vegetable and fruit consumption 98.4 % 93.5 % Hypertension 0.3 % 0.8 % Intensive alcohol consumption 22.7 % 19.1 % Bleeding 0.0 % 0.6 % Smoking history (any) 41.6 % 53.5 % Proteinuria 0.4 % 0.6 % Diabetes 0.3 % 0.0 % Pack-years of smoking ( ) 7.6 ( ) Maternal age (yrs) 2 26 (23-32) Birth weight (kg) ( ) Birth length (cm) 2 51 (42-55) Birth order 24 LIFESTYLE FACTORS (20-29) Occupational physical activity 31.9 % 51.4 % 3.2 Occupational night work 29.3 % 39.0 % ( ) Recreational physical activity 84.6 % 71.9 % 50 Professional sport (38-52) Overall 24.2 % 24.3 % At puberty 18.5 % 19.6 % Special types of sport 1 st 55.5 % 57.3 % Biking 78.2 % 71.9 % 2 nd 32.5 % 30.6 % Motor biking 21.1 % 24.2 % 3 rd 22.0 % 12.1 % Horse riding 6,3 % 6.4 % Cryptorchidism 3.4 % 9.6 % Orchidopexy after 5 yr of age 0.9 % 4.8 % SOCIO-ECONOMIC STATUS AND RESIDENT HISTORY 4 Main lifetime occupation Office-like work 49.1 % 38.8 % DISEASES AND COMPLICATIONS Student 24.8 % 4.4 % Infectious diseases Manual worker 22.1 % 48.3 % Rubella 29.9 % 26.9 % Driver 3.9 % 8.5 % Morbilli 36.9 % 37.6 % Mumps 42.9 % 46.6 % FAMILY CANCER HISTORY History of any neoplasms Any defect of urogenital Close relatives (parents, siblings) 14.8 % 22.2 % 4.7 % 2.0 % system Other relatives % 38.2 % Varicocele 3.8 % 1.4 % History of specified cancer Phimosis 4.1 % 1.2 % diagnoses (in all relatives) Orchitis 1.6 % 0.6 % Testicular cancer 1.3 % 2.8 % Epididymitis 1.6 % 2.5 % Breast cancer 5.0 % 9.3 % Atrophic testis 1.3 % 6.8 % Prostate cancer 0.8 % 3.7 % Inguinal hernia 13.3 % 10.9 % Ovarial cancer 5.4 % 3.9 % Testicular trauma 15.8 % 26.4 % GIT cancer 13.6 % 17.1 % Legend to Table 3. 1 Binary code that aggregates both specified problems during the pregnancy (bleeding, nausea, proteinuria, diabetes, hypertension) and not-specified reporting on risk events 2 Quantitative variables expressed as median and 10%-90% percentiles 3 Median and 10%-90% percentiles of pack-yrs adjusted for time since quitting 4 Categories of education and living place are described in table 1 5 Includes grandparents, uncles and aunts, nieces, nephews

151 RISK FACTORS FOR TESTICULAR CANCER, MULTIVARIATE ANALYSIS 361 Table 4. Prenatal, perinatal and reproductive health history of men with TGCC and of controls in univariate logistic regression analysis 1 Variables and categories All TGCCs (n = 356) Seminoma (n = 195) - Adjusted OR (95% confidence interval) 2 Non-seminoma (n = 161) Heterogeneity test ( 2, p level) Any abnormity during the 2.22 (1.43; 3.44) * 2.49 (1.53; 4.07) * 1.91 (1.12; 3.26) (p = 0.305) pregnancy 3 Maternal age (yrs) (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) (0.13; 0.76)* 0.40 (0.14; 1.06) 0.26 (0.09; 0.69)* (0.22; 0.89) (0.15; 1.17) 0.23 (0.09; 0.63)* 1.62 (p = 0.202) (0.11; 0.74) (0.39; 0.91) (0.07; 0.66)* (0.07; 0.81) (0.29; 0.75) (0.08; 1.31) Test for trend p = p = p = Birth weight (g) < 3, (1.03; 2.68) (1.04; 3.16) (0.88; 2.84) 3,000 3,999 (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 0.03 (p = 0.867) 4, (0.49; 1.62) 0.95 (0.46; 1.99) 0.81 (0.38; 1.73) Cryptorchidism 2.94 (1.46; 5.91)* 3.78 (1.98; 6.80)* 2.65 (1.17; 5.99)* 0.25 (p = 0.617) Orchidopexy after 5yr of age 5.24 (1.52; 18.12)* 7.47 (2.09; 26.66)* 2.66 (0.58; 12.10) 3.90 (p = 0.048) Atrophic testis 5.27 (1.80; 15.47)* 4.69 (1.47; 14.85)* 6.23 (1.93; 20.20)* 0.03 (p = 0.951) Testicular trauma 2.02 (1.37; 2.98)* 2.07 (1.31; 3.26)* 1.96 (1.23; 3.12)* 0.13 (p = 0.719) Legend to Table 4. 1 Only associations with p < 0.05 for at least one subgroup are included. (+) mark for OR significant at the level p< 0.05; (*) mark for OR significant at the level p< OR: odds ratio (adjusted for men s age) with 95% confidence interval 3 Binary code that aggregates both specified problems during the pregnancy (bleeding, nausea, proteinuria, diabetes, hypertension) and not-specified reporting on risk events Following are the variables that were retained in the multivariate model (Table 7) as significant for TGCCs as well as for both histological subtypes (listed with OR for all TGCCs): atrophic testis (OR = 5.88), family history of prostate cancer (OR = 4.81), cryptorchidism (OR = 3.83) and interaction variable low degree of education & manual occupation (OR = 3.01). Family history of breast cancer remained in multivariate model for TGCCs and seminomas with OR values Finally, low birth weight category (< 3,000 g) was retained only in a model for all TGCCs with a borderline significance of OR = TheTable 8 summarizes the survey of testicular self-examination (TSE) that was administered simultaneously with the main study questionnaire. Majority of control men (65%) were uninformed about the TSE and only minor part of them practiced TSE at least irregularly (22.1%). In comparison with controls, the TGCC patients declared significantly higher awareness of TSE (69.7%) as well as more regular practicing (irregularly 17.9%; regularly as recommended 46.9%). Awareness and practicing of TSE were not associated with age, education or cryptorchidism. Significantly increased awareness and practicing of TSE were found among men with history of testicular trauma or atrophic testis. Discussion The epidemiology of testicular cancer in the Czech Republic is defined in the Czech National Cancer Registry (NCR) that ensures an obligatory notification of all newly diagnosed cases of cancer since 1976 including regular follow-up [4, 16, 17]. Age standardized incidence rate (ASR) in the Czech population increased from 3.9/100,000 in 1980 to 8.2/100,000 in Mean annual increase of ASR reached 0.15/100,000 in the period Very similar growing trend has been reported from many European countries [1, 3, 18 21]. Similarly increasing ASR of testicular cancer for the Czech and neighboring Slovak population is also documented in national epidemiological portals of both countries [4, 22]. The prevalence of main histological types of TC in the Czech NCR is similar to commonly published profiles [13, 23]. As for the period , the Czech NCR contains 3,800 records on primary testicular cancer. TGCCs form 96.2 % (n = 3,655) of all the histologically verified testicular tumors with the following structure: seminomas (53.8 %) and other TGCCs (42 4 %). Non-seminomas of one histological type (embryonal carcinoma, choriocarcinoma, teratoma, yolk sac tumor, polyembryoma) form 24 % of all TGCCs. Primary non-germ cell testicular tumors form 2.2 % of all TC records (n = 84). Of course, our hospital-based data cannot be fully representative to all patients with testicular cancer in the Czech population. Therefore, all detected risk associations are discussed here with respect to potential sources of selection bias. We however documented satisfying comparability with the Czech National Cancer Registry (NCR) both in the age distribution and in the prevalence of histological subtypes of TGCCs. As documented in Table 2 and in Figure 1, there was

152 362 L. DUSEK, J. ABRAHAMOVA, R. LAKOMY, R. VYZULA et al. Table 5. Lifestyle and social factors of men with TGCC and of controls in univariate logistic regression analysis. 1 Variables and categories All TGCCs (n = 356) Seminoma (n = 195) - Adjusted OR (95% confidence interval) 2 Non-seminoma (n = 161) Heterogeneity test ( 2, p level) Smoking history Non smoker (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) Former smoker 1.50 (1.02; 2.03) (0.65; 1.77) 1.42 (1.01; 2.62) (p = 0.594) Active smoker 1.60 (1.10; 2.33) (1.02; 2.41) (1.04; 3.23) + Smoking (pack years) 3 Non smoker (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) (0.65; 2.82) 1.21 (0.48; 2.99) 1.51 (0.63; 3.65) (1.69; 6.42)* 2.59 (1.21; 5.56) (1.96; 9.02)* 1.68 (p = 0.199) (1.63; 12.58)* 3.68 (1.75; 8.78) (2.56; 18.21)* Test for trend p = p = p = Occupational physical activity 2.26 (1.65; 3.10)* 2.25 (1.55; 3.25)* 2.27 (1.54; 3.36)* 0.02 (p = 0.976) Occupational night work 1.48 (1.07; 2.06) (1.03; 2.23) (1.09; 2.47) (p = 0.951) Occupation Office-like work (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) Students 0.15 (0.07; 0.32)* 0.08 (0.03; 0.29)* 0.18 (0.07; 0.45)* Manual workers 2.87 (1.88; 4.41)* 2.36 (1.46; 3.84)* 3.62 (2.13; 6.17)* 1.94 (p = 0.164) Drivers, servicemen 2.70 (1.19; 6.14) (0.91; 5.82) 3.31 (1.17; 8.63) + Education Primary and skilled 3.04 (2.05; 4.52)* 2.89 (1.81; 4.60)* 3.07 (1.92; 4.88)* Secondary (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 0.28 (p = 0.593) University 1.11 (0.75; 1.67) 1.29 (0.81; 2.07) 0.89 (0.51; 1.53) Living place Village 1.63 (1.04; 2.45) (0.96; 2.49) 1.55 (0.93; 2.71) Small town (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) 0.42 (p = 0.515) Big town, city 0.54 (0.37; 0.77)* 0.49 (0.31; 0.76)* 0.60 (0.38; 0.94) + Sport activities No sport activity (referent) 1.00 (referent) 1.00 (referent) 1.00 (referent) Recreational sport activity 0.49 (0.32; 0.72)* 0.57 (0.36; 0.89)* 0.43 (0.27; 0.69)* 0.01 (p = 0.932) Legend to Table 5. 1 Only associations with p < 0.05 for at least one subgroup are included. (+) mark for OR significant at the level p< 0.05; (*) mark for OR significant at the level p< OR: odds ratio (adjusted for men s age) with 95% confidence interval 3 Adjusted for years since quitting; categorized according to quartiles of the control sample distribution. no significant discrepancy or tendency of bias. The case series included 54.8% of seminomas (n = 195) and 45.2% of non-seminomas (n = 161). This corresponded to the NCR that comprise 56.0% seminomas and 44.0% non-seminomas (Table 2). Age and clinical stage distribution of seminomas and non-seminomas in the study corresponded to the NCR data (Figure 1, Table 2). Our histology-specific analyses of risk factors should therefore give reliable outcomes. Comparing controls and cases in demographic and social factors, we recognized decreased proportion of TGCCs with high education degree and decreased proportion of TGCCs living in big towns (Table 1). Because education and living place were not matching variables, we checked their sample distribution against official Czech population statistics [24, 25]. Aggregated data of controls and TGCCs corresponded to the official statistics (residents living in big towns: study sample = 36.8 %, Czech population data = 32.4 %; people with primary education or skilled: study sample = 26.4 %, Czech population data = 29.1 %). Therefore, there should be no systematic bias and both education and living place were included in the risk analyses as potentially influencing factors. There is a remarkable inconsistency in the literature comparing histological types of TGCCs from the viewpoint of risk factors. For example, increased risk of seminomas was found in newborns with birth weight < 2.5 kg [11, 26], while in other studies, it was rather attributed to non-seminomas [9, 27]. The general conclusion is that there is only a little variation in risk factors between seminomas and non-seminomas [28]. Our study suggests also only a few variables that significantly differentiated OR values between histological subtypes. It was the case of orchidopexy after 5 yr of age that was significantly associated only with seminomas (OR = 7.47; 95% CI = ). Low birth weight (< 3,000 g) achieved a slightly higher OR values for seminomas (OR = 1.74) than for non-seminomas (OR = 1.58),

153 RISK FACTORS FOR TESTICULAR CANCER, MULTIVARIATE ANALYSIS 363 however without any significant differentiation of histological subtypes. Similar profile was observed for positive familial history of breast cancer that was associated with seminomas (OR = 1.93; p < 0.05) and not associated with non-seminomas (OR = 1.55, p = 0.296). In our multivariate outcomes, seminomas differed from non-seminomas in significant risk association with breast cancer and increased OR values of cryptorchidism (Table 7). All documented regression analyses indicated relatively increased variability of risk associations within non-seminomas than within seminomas. Although it is rather inconsistently reported in literature, seminomas appear to be increasingly more associated with risk factors than nonseminomas [12, 27]. Possible explanation could be found in the origin and structure of histological types. Seminomas developed from germ cells and are relatively more homogeneous than non-seminomas with higher compositional diversity [29]. Furthermore, non-seminomas can be developed from carcinoma in situ independently or sequentially via seminomas. This can be hypothetically regarded as another source of heterogeneity in the non-seminoma group and it might also indicate the processes requiring some additional risk factors [12, 30]. Our study suggests that different histological groups have different degree of association with some risk factors, whose differentiation may be obscured by combining the groups in one analysis. Numerous studies were focused on pre- and perinatal factors. As it is excellently summarized in reviews given by Garner et al. [5] and Richiardi et al. [6], several already published results generate rather inconsistent knowledge in this field. Namely low birth weight, birth order, nausea and bleeding during the pregnancy have been associated with increased risk [31-33]. We found significant risk association only for the nonspecific binary score aggregating any problem during the pregnancy (Table 4). However, none of the complications achieved significantly elevated OR if analyzed separately. At this point, we should mention the potential influence of recall bias in such retrospective searching for % of patients % of patients Seminoma cases < Non-seminoma cases < Age at diagnosis Age at diagnosis >= >=58 Patients in the National Cancer Registry (years ; n=1077) Patients in the study (n=195) Patients in the National Cancer Registry (years ; n=8 7) Patients in the study (n=161) Figure 1. Distribution of age at diagnosis in the study sample of TGCCs in comparison with the Czech National Cancer Registry as population background events that happened 20-30yrs ago. Finally, our score aggregating complications during the pregnancy was not retained in multivariate adjusted analysis and was excluded as confounding variable of low birth weight (< 3,000 g). Table 6. Family cancer history of men with TGCC and controls in univariate logistic regression analysis. 1 Variables and categories All TGCCs (n = 356) Seminoma (n = 195) - Adjusted OR (95% confidence interval) 2 Non-seminoma (n = 161) Heterogeneity test ( 2, 1d.f.; p level) History of any neoplasm All relatives 1.48 (1.09; 2.02) (1.13; 2.45) (1.01; 1.99) (p = 0.526) Close relatives 1.72 (1.02; 2.49) (1.19; 3.19) (1.02; 1.91) (p = 0.561) Distant relatives 1.42 (1.04; 1.95) (0.97; 2.15) 1.34 (0.83; 2.15) 0.01 (p = 0.914) History of specified diagnoses: in all relatives Breast cancer 1.83 (1.02; 3.42) (1.02; 3.85) (0.72; 3.39) 1.17 (p = 0.279) Prostate cancer 3.96 (1.11; 14.09) (1.09; 11.21) (1.19; 18.44) (p = 0.526) Legend to Table 6. 1 Only associations with p < 0.05 for at least one subgroup are included. (+) mark for OR significant at the level p< 0.05; (*) mark for OR significant at the level p< 0.01; 2 OR: odds ratio (adjusted for men s age) with 95% confidence interval

154 364 L. DUSEK, J. ABRAHAMOVA, R. LAKOMY, R. VYZULA et al. Table 7. Results of multivariate logistic regression models for all TGCCs and the subtypes of seminoma and non-seminoma with all risk factors 1 All TGCCs (n = 356) Seminoma (n = 195) Non-seminoma (n = 161) Risk factors OR (95% CI) 2 Risk factors OR (95% CI) 2 Risk factors OR (95% CI) 2 Atrophic testis Family history of prostate cancer History of cryptorchidism Low education degree & manual occupation Family history in breast cancer Birth weight < 3,000 g Legend to Table (2.04; 16.78) 4.81 (2.33; 16.83) 3.83 (2.46; 5.70) 3.01 (2.15; 5.41) 2.01 (1.05; 3.66) 1.67 (1.03; 2.60) Atrophic testis History of cryptorchidism 6.25 (2.41; 17.52) 4.11 (2.42; 7.10) Family history of prostate cancer 4.01 (1.61; 17.11) Low education degree & manual occupation Family history of breast cancer 1 Only associations with p < 0.05 for at least one subgroup are displayed (1.92; 4.40) 2.18 (1.06; 4.48) Atrophic testis Family history of prostate cancer Low education degree & manual occupation History of cryptorchidism 6.31 (1.96; 19.81) 4.91 (2.18; 18.18) 3.44 (2.17; 5.47) 2.29 (1.56; 5.59) 2 OR: odds ratio with 95% confidence interval (adjusted for men s age and all the other characteristics in the table in a multivariate analysis) Low birth weight was kept in multivariate model only for all TGCCs. It is in agreement with prevailing findings of the other authors [5, 11, 34]. Cryptorchidism is widely accepted as risk factor associated with a two- to eight-fold elevated risk of TC [35]. Our study also found a significant risk associated with cryptorchidism for all TGCCs (OR = 2.94) as well as for seminomas (OR = 3.78) and non-seminomas (OR = 2.65). We further found even more increased ORs if orchidopexy is performed at the age of 5 years or later (all TGCCs: OR = 5.24; seminoma: OR = 7.47; non-seminoma: OR = 2.66). These results correspond to several studies that indicated increasing risk of TC and more likely of seminoma in undescended testes with orchidopey after yr of age or not performed (OR ranging typically from 3 to 10; [36-38]). There is however sparse data on the effect of orchidopexy before age of 10 years. A few studies that analyzed orchidopexy in such young boys did not report significant effect [37, 39, 40]. Our data suggests that there can be a significant risk effect of orchidopexy before the age of 10 years with a significant cut-off point for arising risk in the age of 5 years. Late orchidopexy appears to modulate the risk, although the concrete value of OR is probably partially influenced by the clinical selection bias (disputable comparability of patients treated in very young age with the others). Our study presents an increased prevalence of low degree of education among TGCC patients, regardless of the histological subtype (OR = ). Some previous studies showed risk association of high level of education for TC, others found no effect [5, 38]. Expectedly, we found low degree of education to be associated with manual occupation in plants and in rural areas. Therefore, the factor of education cannot be discussed separately. There is no relevant explanation why either education or type of occupation should be an independent risk factor for TGCC. Instead of it, we argue for the combined effect of these factors as it is reflected by highly significant interaction term of them (p < 0.001). The combined effect of low degree of education and of manual work was retained in all multivariate risk models, even if adjusted for the other variables including inherited defects and familial cancer history (Table 7). Correlation of low education and manual occupation indicates even more complex underlying lifestyle association because the same group of men claimed no sport activity, more sedentary lifestyle and increased intensity of long-term smoking. All these factors tend to separate individuals with specific lifestyle that corresponds to relatively hard manual occupation. Such exposures can start relatively early in life, even immediately after the puberty in age of years. Typical schedule of education of skilled men in the Czech Republic covers 8-9 years in any type of primary school (starting in the age of 6 yr) plus 3-4 years in some training college where they are already exposed to the occupational conditions. The age distribution of testicular cancer allows for effective exposures in early adulthood and the age around puberty has been suggested as a period with probably increased risk of TC promotion [6,41]. Lifestyle factors are closely related to the risk effect of smoking [38, 42]. Although we recognized smoking as a risk factor in univariate analyses, there is lack of supportive multivariate result and we can speculate that smoking can operate as confounder of some type of lifestyle or occupation. At this point, we employed data of the Czech National Cancer Registry, where the smoking history is coded in all records except

155 RISK FACTORS FOR TESTICULAR CANCER, MULTIVARIATE ANALYSIS 365 Number of % of smokers patients Larynx (C32) Pharynx and other parts (C09-C10,C12-C14) Lung, bronchus and trachea (C33,C34) Oral cavity (C00-C08) Esophagus (C15) Nasopharynx (C11) Bladder (C67) Testis (C62) Cervix uteri (C53) Hodgkin's disease (C81) Non-Hodgkin's lymphoma (C82-C85) Stomach (C16) Kidney and organs of urinary tract (C64-C66,C68) Liver (C22) Prostate (C61) Colon and rectum (C18-C21) Pancreas (C25) Melanoma (C43) Multiple myeloma (C90) Leukemia (C88,C91-C96) Thyroid gland (C73) Other skin tumour (C44) Central nervous system (C70-C72) Ovary (C56) Breast (C50) Gallbladder and biliary tract (C23-C24) Vulva and vagina (C51-C52) Uterus (C54,C55) Other malignant tumours n=2644 n=2141 n=30405 n=3252 n=2162 n=273 n=10732 n=2001 n=5236 n=1332 n=5446 n=8784 n=13563 n=3938 n=17536 n=39334 n=8270 n=7823 n=2168 n=5920 n=2977 n=68604 n=3900 n=6272 n=27346 n=5048 n=1192 n=8607 n=12185 Figure 2. Prevalence of smokers in records of the Czech National Cancer Registry (period ). for DCO cases (Figure 2). NCR data documents the position of testicular cancer in the middle of all cancer diagnoses, with only 27.3% prevalence of smokers. This supports the conclusion that smoking can only weakly modulate the risk level for TC, highly probably in association with the other risk factors. Independent risk effect of smoking is very disputable. We have no exact explanation for the borderline significance of risk of TGCCs among men living in rural areas. The effect was observed only for the whole TGCC group and lost its significance both for seminomas and non-seminomas. Furthermore, no such effect was retained in the multivariate analysis. Therefore, it might be related to the composition of the study sample. A clear evidence for independent risk influence of rural/urban areas and living places does not exist [5, 43]. Although we detected remarkably elevated ORs among men that claimed testicular trauma in past, we finally dropped out this parameter from the multivariate models due to the following reasons, also discussed in literature [5, 44]. Firstly, the history of trauma can be influenced by recall bias and indeed we found increasing frequency of traumatic events reported for the periods very close to the time of cancer diagnosis (data not shown). This suggests that the trauma could make the injured patients more careful and facilitate the diagnosis of already growing tumor. Secondly, patients with reported testicular trauma declared significantly increased awareness, as well as practicing intensity of testicular selfexamination (TSE, Table 8). This again indirectly indicates their higher interest in health surveillance. Our survey however showed that proportion of Czech men practicing TSE is very small. Approximately 50 % of TGCCs do not perform TSE monthly and more than 35 % do not practice TSE at all. Similar results were reported by other authors [45, 46]. Based on a retrospective study of 1,832 patients with TC diagnosed in the period , Ondrusova and Ondrus [21] also concluded that Slovak men are poorly informed about the possibility of TC occurrence. Many familial studies proved evidence of the risk role of inherited factors. It holds namely for testicular cancer, brothers of TC patients have 8-fold increased risk of TC and sons of fathers with TC have 4-fold elevated risk [47]. Familial TC history can be also associated with increased occurrence of cryptorchidism suggesting that there is a relationship between urogenital maldevelopment and predisposition to testicular neoplasia [48]. However, the prevalence of testicular cancer among relatives of our sample was too low to allow statistically significant discrimination of cases and controls. Nevertheless, we found a significant risk association with overall familial cancer history and with prostate and breast carcinoma (Table 6). Several studies already described the

156 366 L. DUSEK, J. ABRAHAMOVA, R. LAKOMY, R. VYZULA et al. Table 8. Awareness and practice of testicular self-examination (TSE) declared by TGCCs and controls in relation to selected other factors Groups Awareness Practicing of TSE of TSE Irregularly Regularly TGCCs vs. controls Controls (n = 317) 35.0 % * 13.3 % 8.8 % * TGCCs (n = 356) 69.7 % * 17.9 % 46.9 % * Atrophic testis No (n = 645) 52.7 % * 15.7 % 28.1 % * Yes (n = 28) 67.9 % * 14.3 % 50.0 % * Testicular trauma No (n = 529) 49.7 % * 15.5 % 25.1 % * Yes (n = 144) 66.7 % * 16.7 % 43.1 % * Cryptorchidism No (n = 628) 53.2 % 15.8 % 28.5 % Yes (n = 45) 55.6 % 15.6 % 35.6 % Education Primary or skilled (n = 180) 49.4 % 16.1 % 31.7 % Secondary (n = 342) 56.1 % 16.4 % 28.7 % University (n = 151) 51.7 % 13.9 % 26.5 % * Mark of statistically significant difference between controls and TGCCs or between yes/no categories of the other categorizing factors (M-L 2 test; p< 0.05) risk of familial cancer history for testicular cancer, including specific position of breast cancer and prostate cancer [49, 50]. Similarly to our study, Walschaerts et al. [50] have actually published increased ORs for breast cancer in family (OR = 1.77, 95% CI = ). The authors also confirmed this effect in a multivariate model constructed in the same way as we present here. Corresponding findings were presented also by other authors [51, 52]. Walschaerts et al [50] also found increased OR for familial history of prostate cancer (OR = 1.80; 95% CI = ), but without subsequent confirmation in multivariate system. In our study, we met significant risk elevation due to familial history of prostate cancer without any confounding influence of other variables. Such result seems to be in contrast with negative findings published by Hemminki and Chen [49] or by Westergaard et al. [53]. Conclusion Based on the employed multivariate strategy, we could not rule out effects of any category of risk factors for TGCC. Each type of factors had its statistically significant representative in the multivariate risk model. Our study documented that a wide set of risk factors generates numerous confounding effects that cannot be detected in univariate systems. Multivariate modeling with a special focus on interactions among variables is therefore highly recommended. Familial cancer history remains an influential risk factor in multivariate models altogether with lifestyle and occupational parameters. It documents that both environmental exposures and genetic inheritance can play some role in moderation of the risk of TGCC. Our study suggests that seminomas and non-seminomas may have different degree of association with some risk factors, whose differentiation may be obscured by combining the groups in one analysis. This work was supported by a grant project of the Grant Agency of the Czech Ministry of Health Care, project no. NR8442-3/2005. References [1] MOLLER H. Trends in incidence of testicular cancer and prostate cancer in Denmark. Hum Reprod. 2001; 16: [2] McGLYNN KA, GRAUBARD BI, DEVESA SS et al. Increasing incidence of testicular germ cell tumors among black men in the United States. J. Clin. Oncol. 2005; 23: [3] BRAY F, RICHIARDI L, EKBOM A et al. Trends in testicular cancer incidence and mortality in 22 European countries: continuing increases in incidence and declines in mortality. Int J. Cancer 2006; 118: [4] DUSEK L, MUZIK J, KUBASEK M et al. Czech National Portal for Cancer Epidemiology. Masaryk University [cit ]. Available on-line: ISSN [5] GARNER MJ, TURNER MC, GHADIRIAN P et al. Epidemiology of testicular cancer: an overview. Int J. Cancer 2005; 116: [6] RICHIARDI L, PETTERSSON A, AKRE O. Genetic and environmental risk factors for testicular cancer. Int. J.Androl. 2007; 30: [7] MOLLER H. Clues to the aetiology of testicular germ cell tumors from descriptive epidemiology. Eur. Urol. 1993; 23: 8 15.

157 RISK FACTORS FOR TESTICULAR CANCER, MULTIVARIATE ANALYSIS 367 [8] WEIR HK, MARRETT LD, KREIGER N et al. Pre-natal and peri-natal exposures and risk of testicular germ-cell cancer. Int. J. Cancer 2000; 87: [9] RICHIARDI L, AKRE O, BELLOCO R et al. Perinatal determinants of germ-cell testicular cancer in relation to histological subtypes. Brit. J Canc. 2002; 87: [10] RICHIARDI L, AKRE O, LAMBE M et al. Birth order, sibship size and risk of germ-cell testicular cancer. Epidemiology 2004; 15: [11] ENGLISH PB, GOLDBERG DE, WOLFF C et al. Parental and birth characteristics in relation to testicular cancer risk among males born between 1960s and 1995 in California (US). Cancer Causes Cont. 2003; 14: [12] ASCHIM EL, HAUGEN TB, TRETLI S et al. Risk factors for testicular cancer differences between pure non-seminoma and mixed seminoma/non-seminoma? Int J. Androl. 2006; 29: [13] LOOIJENGA LH, OOSTERHUIS JW. Pathogenesis of testicular germ cell tumours. Reviews of Reproduction 1999; 4: [14] BLAND JM, ALTMAN DG. The odds ratio. British Medical Journal 2000; 320: [15] HOSMER DW, LEMESHOW S, editors. Applied logistic Regression. New York: Wiley, [16] Institute of Health Information and Statistics of the Czech Republic, Council of the National Cancer Registry of the Czech Republic. Cancer Incidence in the Czech Republic. Prague, 2004; 244 s. ISSN X. ISBN Available on-line: [17] ABRAHÁMOVÁ J. Testicular tumours. Management, treatment and results. Acta Univ. Carolinae Medica, monogr., 138, Praha [18] LEVI F, LUCCHINI F, BOYLE P et al. Cancer incidence and mortality in Europe, J. Epidem. Biostat. 1998; 3: [19] PLESKO I, OBSITNIKOVA A, CUNINKOVA M et al. Increasing occurrence of urological cancers in Slovakia. Neoplasma, 2004; 51: [20] ONDRUSOVA M, PLESKO I, SAFAEI-DIBA CH et al. Cancer incidence in the Slovak Republic National Cancer Registry, NHIC, Bratislava 2006; 208 s. ISBN [21] ONDRUSOVA M, ONDRUS D. Epidemiology and treatment delay in testicular cancer patients: a retrospective study. Int. Urol. Nephrol. 2008; 40: [22] ONDRUSOVA M, PLESKO I, SAFAEI-DIBA CH et al. Comprehensive analysis of incidence and mortality of malignant tumours in the Slovak Republic, Bratislava, National Cancer Registry of the Slovak Republic, NHIC, 2007 [2cit ]. Available on-line: [23] CUNINKOVA M, ONDRUS D, PLESKO I et al. Epidemiology of testicular tumors in Slovakia ( ): Preliminary report. Int. Urol. Nephrol. 2001; 32: [24] Czech Statistical Office. Focus on Women and Men 1-2. Population structure by sex, age and education attainment (LFSS). Czech Statistical Office Available on-line: [25] Czech Statistical Office. Basic final results of the Population and Housing Census 2001 (from Statistical Yearbook of the CR 2002): Population: by age, sex and municipality size group (1 March 2001); Czech Statistical Office, Available on-line: [26] WANDERAS EH, GROTMOL T, FOSSA SD et al. Maternal health and pre- and perinatal characteristics in the etiology of testicular cancer: a prospective population- and register-based study on Norwegian males born between 1967 and Cancer Causes and Control 1998; 9: [27] AKRE O, EKBOM A, HSIEH CC et al. Testicular nonseminoma and seminoma in relation to perinatal characteristics. J. Natl. Cancer Inst. 1996; 88: [28] BRAY F, RICHIARDI L, EKBOM A et al. Do testicular seminoma and nonseminoma share the same etiology? Evidence from an age-period-cohort analysis of incidence trends in eight European countries. Cancer Epidem Biom Prev. 2006; 15: [29] MOSHARAFA AA, FOSTER RS, LEIBOVICH BC et al. Histology in mixed germ cell tumors. Is there a favorite pairing? J. Urol. 2004; 171: [30] SMIRAGLIA DJ, SZYMANSKA J, KRAGGERUD SM et al. Distinct epigenetic phenotypes in seminomatous and nonseminomatous testicular germ cell tumors. Oncogene 2002; 20: [31] DEPUE RH, PIKE MC, HENDERSON BE. Estrogen exposure during gestation and risk of testicular cancer. J. Natl. Cancer Inst. 1983; 71: [32] EKBOM A, TRICHOPOULOS D, ADAM HO et al. Evidence of prenatal influences on breast cancer risk. Lancet 1992; 340: [33] PRENER A, HSIEH CC, ENGHOLM G et al. Birth order and risk of testicular cancer. Cancer Causes Control 1992; 3: [34] RASMUSSEN F, GUNNELL D, EKBOM A et al. Birth weight, adult height and testicular cancer: cohort study of 337,249 Swedish young men. Canc. Causes cont. 2003; 14: [35] DIECKMANN KP, PICHLMEIER U. Clinical epidemiology of testicular germ cell tumors. World J. Urol. 2004; 22: [36] STRADER CH, WEISS NS, DALING JR et al. Cryptorchism, orchidopexy, and the risk of testicular cancer. Am J. Epidem. 1988; 127: [37] PRENER A, ENGHOLM G, JENSEN OM. Genital anomalies and risk for testicular cancer in Danish men. Epidemiology 1996; 7: [38] MOLLER H, SKAKKEBAEK NE. Risks of testicular cancer and cryptorchidism in relation to socio-economic status and related factors: case-control studies in Denmark. Int. J. Cancer 1996; 66: [39] POTTERN LM, BROWN LM, HOOVER RN et al. Testicular cancer risk among young men: role of cryptorchidism and inguinal hernia. J Nat Cancer Inst. 1985; 74:

158 368 L. DUSEK, J. ABRAHAMOVA, R. LAKOMY, R. VYZULA et al. [40] SWERDLOW AJ, HIGGINS CD, PIKE MC. Risk of testicular cancer in cohort of boys with cryptorchidism. Brit Med. J. 1997; 314: [41] McGLYNN KA, SAKODA LC, RUBERTONE MV et al. Body size, dairy consumption, puberty, and risk of testicular germ cell tumors. Am J Epidemiol. 2007; 165: [42] SRIVASTAVA A, KREIGER N. Relation of physical activity to risk of testicular cancer. Am J Epidemiol 2000; 151: [43] SCHOUTEN LJ, MEIJER H, HUVENEERS JA et al. Urban-rural differences in cancer incidence in the Netherlands Int. J. Epidemiol. 1996; 25: [44] STONE JM, CRUICKSHANK DG, SANDEMAN TF et al. Laterality, maldescent, trauma and other clinical factors in the epidemiology of testis cancer in Victoria, Australia. Br. J. Cancer 1991; 64: [45] MOORE RA, TOPPING A. Young men s knowledge of testicular cancer and testicular self-examination: a lost opportunity? Eur. J. Cancer Care 1999; 8: [46] GASCOIGNE P, WHITEAR B. Making sense of testicular cancer symptoms: a qualitative study of the way in which men sought help from the health-care services. Eur. J. Oncol. Nurs. 1999; 3: [47] DONG C, HEMMINKI K. Modification of cancer risks in offspring by sibling and parental cancers from 2,112,616 nuclear families. Int. J. Cancer 2001; 92: [48] ONDRUS D, KUBA D, CHRENOVA S et al. Familial testicular cancer and developmental abnormalities. Neoplasma 1997; 44: [49] HEMMINKI K, CHEN B. Familial risks in testicular cancer as aetiological clues. Int. J. Androl. 2006; 29: [50] WALSCHAERTS M, MULLER A, AUGER J et al. Environmental, occupational and familial risks for testicular cancer: a hospital-based case-control study. Int J Androl. 2007; 30: [51] BAJDIK CD, PHILIPS N, HUCHCROFT S et al. Cancer in mothers and siblings of testicular cancer patients. Can. J. Urol. 2001; 8: [52] DONG C, LONNSTEDT I, HEMMINKI K. Familial testicular cancer and second primary cancers in testicular cancer patients by histological type. Eur. J. Cancer 2001; 37: [53] WESTERGAARD T, OLSEN JH, FRISCH M et al. Cancer risk in fathers and brothers of testicular cancer patients in Denmark. A population-based study. Int. J. Cancer 1996; 66:

159

160 Paper VIII. Mužík J., Dušek L., Pavliš P., Koptíková J., Žaloudík J., Vyzula R. Analysis of population cancer risk factors in national information system SVOD. Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J.

161 Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J. Analysis of Population Cancer Risk Factors in National Information System SVOD Mužík J. 1, Dušek L. 1,2, Pavliš P. 1, Koptíková J. 1, Žaloudík J. 3, Vyzula R. 3 Abstract Human risk assessment requires analysis of multiple data sources to make correct interpretations of the results. One of these sources is population epidemiologic databases. In this presentation we introduce System for Visualization of Oncological Data (SVOD, version 6.0) built up over the database of National Cancer Registry of Czech Republic. Today in this system are accessible data of all malignant diagnoses (C00 - C97) from time period (over 1.2 million cases). Such a database represents a great pool of information that can be used in retrospective evaluation of population risk. We will present tools and functions of the system, which make this information easily accessible for wide spectrum of users. 1. Introduction Human risk assessment in standard terminology simply means to estimate the incidence of death and disease resulting from exposure to hazardous agents. However this definition does not fully cover all the aspects associated with risk assessment studies. In practice, many times we have to analyze trends and epidemiological data without evidence that changes are directly related to some type of environmental exposure. Instead of causal analysis of population data (that are frequently limited in extent, quality and accessibility) we often enter environmental databases that are focused on hazardous compounds and exposure and analyze the problem from this point of view. Thus, ecological risk assessment serves as scientifically credible evaluation of environmental exposure that obviously can affect also human population. Many toxic and hazardous events in environment can be used as early warning entry to human risk assessment studies. It would be however counterproductive to argue against standard epidemiological analyses and to propose some single methodical approaches based on ecological risk assessment procedures. Our presentation is aimed to demonstrate how to solve problems with accessibility of epidemiology data and how to solve often very complicated analysis and interpretation of these data. We would like to present here population epidemiology data as quantitative base for comparing and prioritizing risks, both for retrospective and prospective risk assessment studies. 1 Centre of Biostatistics and Analyses, Faculty of Medicine and Faculty of Science Masaryk University in Brno, Czech Republic 2 Research Centre for Environmental Chemistry and Ecotoxicology, Faculty of Science, Masaryk University in Brno, Czech Republic 3 Masaryk Memorial Cancer Institute, Brno, Czech Republic 1

162 The study will introduce widely accessible expert system (SVOD, version 6.0) built up over the database of National Cancer Registry of Czech Republic, that consists of more than cases and the data are guaranteed by standardized methodology of collection since Such a database represents a great pool of information that can be used in retrospective evaluation of population risk, both in absolute terms (incidence, mortality) and in relative, population-adjusted indices. The system is developed as analytical tool that provides 4 types of automated expert analyses - standard epidemiology survey and trend analyses (incidence, prevalence, mortality, trend indices) - risk factors recognition and quantification (analyses standardized with respect to structure of target population, evaluation of demographic and social risk factors) - regionally specific analyses that transform epidemiology data into regionally or locally specified cohorts (these analyses are opened for aggregation with environmental data) - analyses for management of health care facilities (analysis of survival, diagnostic and therapeutic algorithms, etc.) Architecture and functions of the software is presented in chapter 4 of this article. Presentation of the expert system is available at 2. Cancer epidemiology as a model for risk assessment studies Human risk assessment studies often consider carcinogenicity because cancer is more dreaded by the public. Apart from its societal value, cancer epidemiology provides ideal environment for development of population model and risk assessment algorithms. Cancer incidence and mortality can be studied from the viewpoint of some inherent risk factors associated with genetic constitution of each individual, but in addition to that we must take into consideration many other influential factors like life style factors, demographic and social structure of population and of course environmental impacts. That is why epidemiological studies sometimes very resemble retrospective ecological risk assessment as it frequently starts by observing of some degradation, i.e. decline in population size of some species. Cancer epidemiology has become one of the most important topics in internationally guaranteed information systems focused on human risk data. Recent international data about cancer epidemiology are available in these databases: CI5 I-VIII (Cancer Incidence in Five Continents, Vol. I - VIII), GLOBOCAN 2002 (Cancer Incidence, Mortality and Prevalence Worldwide) and ACCIS (Automated Childhood Cancer Information System). These databases are accessible on pages of International Agency for Research on Cancer - IARC ( Czech national information and expert system presented here was developed to be compatible with these international systems. 3. Data model and structure of database Data model of expert system SVOD consists of two main parts: a) population demographic data - absolute numbers of men and women in five-year age groups in regions and districts in years These data are used in all epidemiologic calculations. b) data of National Cancer Registry of Czech Republic These data represents collection of all cases of malignant neoplasms detected in years (completely 91 diagnoses C00-C97, D03, D05 and D06 over 1.2 million cases). Data are separated into indi- 2

163 vidual databases according to diagnose of malignant neoplasm, each record represents newly detected malignant neoplasm in population. Individual parameters are related to these main areas: - basic parameters about patient sex, year of birth, age at diagnosis, region and district, social status, main area of employment, smoking and occurrence of malignant neoplasm in relatives - diagnostic parameters diagnosis, year of diagnosis, laterality of neoplasm, detailed TNM classification, clinical stage, topography of neoplasm, histology of tumour and grading, - clinical parameters used diagnostic methods, used therapeutic methods (surgery, radiotherapy, chemotherapy, hormonal therapy, other types of therapies) - parameters about patient status current status (alive/death), cause of death, date of death Updates of all databases are made yearly. 4. Architecture and functions of the software 4.1 Used technology Software SVOD is working with Windows operation systems (98/ME and above). It is written in C++ programming language, communication between software and database is done via SQL language and OLEDB interface. Data of individual diagnoses are stored in the Microsoft Access 97 database format. Software is also prepared for communication with Microsoft SQL Server database. 4.2 Function of the software The main function of the SVOD software is focused on the analysis and the interpretation of the population cancer data. Diagnosis selection is the first and basic step of each analysis. User can select one of 91 available diagnoses sorted into these groups: I. Tumours of head and neck (diagnoses C00, C01, C02, C03, C04, C05, C06, C07, C08, C09) II. Tumours of digestive organs (diagnoses C10, C11, C12, C13, C14, C15, C16, C17, C18, C19, C20, C21, C22, C23, C24, C25, C26) III. Tumours of respiratory and intrathoracic organs (diagnoses C30, C31, C32, C33, C34, C37, C38, C39) IV. Tumours of bone and soft tissues (diagnoses C40, C41, C45, C46, C47, C48, C49) V. Tumours of the skin (diagnoses C43, D03, C44) VI. Breast tumours (diagnoses C50, D05) VII. Gynecologic tumours (diagnoses C51, C52, C53, D06, C53, D06, C54, C55, C54, C55, C56, C57, C58) VIII. Urogenital tumours (diagnoses C60, C61, C62, C63, C64, C65, C66, C67, C68) IX. Tumours of central nervous system and eye (diagnoses C69, C70, C71, C72) X. Tumours of lymphoid, haematopoietic and related tissues (diagnoses C81, C82, C83, C84, C85, C88, C90, C91, C92, C93, C94, C95, C96) XI. Tumours of endocrine glands (diagnoses C73, C74, C75) XII. Other tumours (diagnoses C76, C77, C78, C79, C80, C97) 3

164 After diagnosis selection user can analyze data by three types of analytical tools, which are accessible by the standardized module COBRA (Comprehensive Data Browser): a) Presentation is predefined comprehensive set of analytical outputs, which describes selected diagnosis in various points of view: time trends of incidence and mortality, gender comparison, age structure of patients and deaths caused by diagnosis, regional comparisons and clinical stages. Presentation includes comments, additional information and access to specialized analytical tools of the system (fig. 1). Figure 1. Example of comprehensive presentation windows in software SVOD. b) Data browser this tool enables analysis of individual parameters of the database. User can perform his own analysis of selected group of patients (filtering parameters are region, period, age, sex, detailed diagnosis ICD-10, clinical stage, and other parameters related to diagnostics and treatment) and make stratified outputs. All outputs are available in tabular and graphical form (fig. 2). 4

165 Figure 2. Data browser and outputs of analyses. c) Expert services analytical tools fully controlled by user and focused on specific area: EPIDEMIOLOGY: - Epidemiology: incidence and mortality time trends of incidence, mortality and mortality/incidence ratio (absolute numbers, crude rate - number of cases per people in population, age standardized ratio (ASR) - European or World standard) - Epidemiology: changes in time - changes of incidence and mortality in time (growth index related to selected year and between-years changes - absolute numbers or percents) - Age structure of patient population - age structure of patients or deaths caused of diagnose (absolute numbers of cases in age categories, % of cases in age categories, age specific rate - number of cases in age category per people in population cohort of the same age) - Age specific trends in time time trends of age-specific incidence or mortality - Clinical stages - time trends of proportion of patients in specific clinical stages (absolute numbers, percentage or crude rate) - Time trends of ASR time trends of ASR with confidence intervals for regional comparisons - Regional data: districts tabular outputs for selected district (absolute numbers, crude rate and ASR with confidence intervals) COMPARATIVE ANALYSIS - Comparative analysis: epidemiology - time trends of incidence or mortality in selected region in comparison with situation in whole Czech Republic - Comparative analysis: age structure age structure of patients or deaths caused by diagnosis in selected region in comparison with situation in whole Czech Republic - Comparative analysis: survival analysis comparison of survival analysis outputs for selected region and Czech Republic - Comparison of ASR in regions comparison of ASR for all regions in selected time period HEALTH MANAGEMENT 5

166 - Diagnostics and treatment: time trends time trends of application of diagnostic and therapeutic methods (absolute number and percentage) - Diagnostics and treatment: age of patients - application of diagnostic and therapeutic methods according to age of patients (age structure of patients with applied method(s) absolute numbers and percentage, percentage of application of method according to age) - Diagnostics and treatment: combinations of methods tabular outputs of used combinations of diagnostic and treatment methods, enables comparison with user defined health care standards - Survival analysis tool for performing Kaplan-Meier survival analysis on selected groups of patients All analyses can be performed on selected groups of patients (filtering parameters are region, time period, age, sex, detailed diagnosis ICD-10, TNM, clinical stage, and other parameters related to diagnostics and treatment). Outputs are available in graphical and tabular form, graphs are fully editable (fig. 3). Figure 3. Example of expert services analytic window. Left window: basic setting of the analysis and selection of target group of patients; right: analytic window with graphic and tabular outputs All graphical and tabular outputs of Data browser and Expert services can be exported into other formats for use in other Windows applications (TXT and XLS for tables, BMP, JPG and EMF for graphs). Moreover, software SVOD serves not only as an analytic tool for the assessment of population cancer data, but offers additional communication and information services: - access to the National web portal for cancer epidemiology in the Czech Republic and pages of the software SVOD (see at - on-line internet discussion club this club is assigned for SVOD users, serves as place for discussion about actual problems, enables communication with software developers and offers on-line help. - tutorials for analyses and user manual. Development of system SVOD is supported by Ministry of Health Care Czech Republic (and research and developmental teams are granted by research project MZO solved in Masaryk Memorial Cancer Institute, Brno). Risk assessment analyses are supported by research project INCHEMBIOL, Ministry of Education Czech Republic project no

167 Paper IX. Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Vyzula R. Epidemiology of Malignant Tumours in the Czech Republic [online]. Masaryk University, Czech Republic, [2005], Version 7.0 [2007], Approved by National Technical Library, ISSN Published as Dušek L., Mužík J., Koptíková J., Brabec P., Žaloudík J., Vyzula R., Kubásek M. The National Web Portal for Cancer Epidemiology in the Czech Republic. Proceedings of the 19 th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J.

168 Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Edited and reviewed by H ebí ek J., Rá ek J. The national web portal for cancer epidemiology in the Czech Republic Dusek L. 1,2, Muzik J. 1, Kubasek M. 3, Koptikova J. 1, Brabec P. 1, Zaloudik J. 4, Vyzula R. 4 Abstract The aim of this article is to inform about newly issued automated web portal focused on population risk analyses related to cancer epidemiology. Portal was built up on very representative database of National Cancer Register (Czech Republic, Ministry of Health Care, standardized collection of cancer data from ) that provides fully representative long-term trends. Nowadays the database consists of more than cases stratified according to main risk factors and diagnostic descriptors including TNM classification of tumours. The automated system of on-line analyses offers unique access to cancer registries, demographic and environmental databases. Portal is principally developed as tool increasing the information potential of risk assessment studies. Portal is available at 1. Introduction Cancer epidemiology can be regarded as one of the most important and most frequently analysed topic in the field of human risk assessment. It is not only due to remarkable public concern about growing population risk, cancer incidence and mortality is evident and clearly reachable endpoint for risk assessment studies. We can enter this problem from the viewpoint of risk factors as agents initiating carcinogenesis, but epidemiological parameters can retrospectively indicate hazardous impact on population of a large scale. The bioindication from epidemiological data of course requires sufficient data sources. It means representative long-term profiles of incidence and mortality and a very good awareness of most important risk factors. Analysing epidemiological trends we must be able to distinguish between statistically significant trends and random fluctuations. For risk assessors it would be most important to recognize environmentally related cases that can be attributed to external factors like pollution of air, drinking water and/or food. And again, the influence of these factors must be filtered on the substantial background of the other natural risk factors like age structure of the population, genetic factors or frequently omitted life style. Apart from these complicated circumstances we can concentrate our attention to some cohorts of oncological patients that might probably indicate external impact if they increase in incidence and non randomly in some regional or temporal scales. It means namely occurrence of less advanced disease stages in age groups that 1 Centre of Biostatistics and Analyses, Faculty of Medicine and Faculty of Science Masaryk University in Brno, Czech Republic 2 Research Centre for Environmental Chemistry and Ecotoxicology, Faculty of Science, Masaryk University in Brno, Czech Republic 3 Department of Information Technologies, Faculty of Informatics, Masaryk University in Brno, Czech Republic 4 Masaryk Memorial Cancer Institute, Brno, Czech Republic 1

169 are commonly out of main risk (for example breast carcinoma cases in pre-menopausal women younger than years). In that context, we can take information benefit mainly from diagnostic groups that might be related to some external influence (colorectal carcinoma, kidney and lung carcinoma, breast cancer, malignant melanoma, ). All these analyses require easily available large data sets that are itself very expensive and typically not directly available epidemiological cancer registries. In addition to it, we must aggregate cancer data with demographic data in order to attain for example age-specific profiles of incidence. That is why there is growing interest of many professional groups (health care managers, environmental experts, risk assessors) in accessibility of these data. In our experience however, the demand for data cannot be easily fulfilled by blind databases of primary population data. The analyses are very time-consuming and finally, the outputs might be ambiguous and not safe to be communicated with public. Therefore, we developed professional web portal that offers automatically generated and verified epidemiological analyses on cancer incidence and mortality in Czech Republic. This presentation is aimed to introduce potential users to the technology, architecture and functioning of the portal. Web portal is accessible through the following address: It is developed in Czech and English version. 2. Aims of the portal and accessible data sources The portal is principally aimed to provide several automated analyses that are in detail specified in chapter 4. The analyses were developed over several data sources that are automatically aggregated according to user s choice (cancer epidemiology data, demographic data, data on risk status of Czech population ). All analytical functions result in fully colored graphical and table protocols that are opened for further editing and modification. The portal functions are targeted primarily for health care managers and risk assessors working in the field of human and ecological risk assessment. Some types of outputs (namely incidence profiles) are prepared in a very safe way and are widely accessible to general public. Portal was built up on very representative database of National Cancer Register (Czech Republic, Ministry of Health Care, standardized collection of cancer data from 1997) that provides fully representative longterm trends. Nowadays the database consists of more than cases stratified according to main risk factors and diagnostic descriptors including TNM classification of tumours. Database provides longterm trends in epidemiology of the following diagnostic groups: I. Tumours of head and neck (diagnoses C00, C01, C02, C03, C04, C05, C06, C07, C08, C09) II. Tumours of digestive organs (diagnoses C10, C11, C12, C13, C14, C15, C16, C17, C18, C19, C20, C21, C22, C23, C24, C25, C26) III. Tumours of respiratory and intrathoracic organs (diagnoses C30, C31, C32, C33, C34, C37, C38, C39) IV. Tumours of bone and soft tissues (diagnoses C40, C41, C45, C46, C47, C48, C49) V. Tumours of the skin (diagnoses C43, D03, C44) VI. Breast tumours (diagnoses C50, D05) VII. Gynecologic tumours (diagnoses C51, C52, C53, D06, C53, D06, C54, C55, C54, C55, C56, C57, C58) VIII. Urogenital tumours (diagnoses C60, C61, C62, C63, C64, C65, C66, C67, C68) IX. Tumours of central nervous system and eye (diagnoses C69, C70, C71, C72) 2

170 X. Tumours of lymphoid, haematopoietic and related tissues (diagnoses C81, C82, C83, C84, C85, C88, C90, C91, C92, C93, C94, C95, C96) XI. Tumours of endocrine glands (diagnoses C73, C74, C75) XII. Other tumours (diagnoses C76, C77, C78, C79, C80, C97) 3. Architecture and description of the portal 3.1 Menu and related information services The portal has following structure: About project: contains basic information about the project and its aims. News: news about the project and related activities in cancer epidemiology regularly updated information service Epidemiologic analyses: main part of web portal which offers set of analytical tools with graphical and tabular outputs. Publications, reports: electronic publications and short reports about specific fields of cancer epidemiology (detailed analyses of specific diagnoses etc.) Software tools: information related to software SVOD (System for Visualisation of Oncological Data) updates, tutorials, news, user s discussion etc. Analysis tutorial: comprehensive information and tutorial for epidemiologic analytic tools 3.2 Technological background, security and access rights The whole web portal is placed on two servers. The first is database server with operation system Linux and data management system MySQL and the second is Linux based web server with PHP scripting module. Both servers are connected via reserved network interface with private IP addresses, so database server is separated and secured from whole Internet. Special functions of web portal are accessible in secure area (access requires appropriate login and password) and all communication is based on Web Services standards (XML, SOAP, WSDL) and realized via SSL secure protocol. User interface was designed and realized by a combination of interrelated techniques like HTML templates, JavaScript and CSS. The main scripting language used for building dynamic web pages is PHP. Management of web portal has two levels: Web-Master with full access to all functions of the portal (main administrator) and administrators with access to specific parts of the portal (management of news, software SVOD section etc.) 4. Automated analyses Automated analyses form the core functions of the system. The analyses are accessible both directly and in mutual combination according to strategy of the user. Through these functions, you can easily analyze epidemiological trends examined over three decades, stratify and filter cohorts of patients and extract population risk in absolute or age-specific values. Some of analyses offer benchmarking with respect to clinical status of the disease. Major epidemiological trends are available in comparison with international data (GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide). 3

171 4.1 Automated analytical tool These tools enable to analyze data about specified diagnostic group: Incidence and mortality: time trends of incidence, mortality and mortality/incidence ratio. Available parameters are absolute numbers, crude rate (number of cases per people in population) and age standardized ratio (ASR - European or World standard) Time trends: changes of incidence and mortality in time. Available parameters are growth index related to selected year and between-years changes. Both parameters could be viewed as absolute numbers or as relative proportions. Regional overview: comparison of incidence and mortality in regions of the Czech Republic. Available parameters are crude rate and age standardized ratio (ASR - European or World standard), output could be bar-plot or map. Age analyses: age structure of population of patients with selected diagnose Available parameters are absolute numbers of cases in age categories, % of cases in age categories and age specific rate (number of cases in age category per people in population cohort of the same age). Clinical stages: time trends of proportion patients in specific clinical stages. Available parameters are absolute numbers, % and crude rate of patients in specific clinical stages; available outputs are time trend bar-plot, time trend line-plot or pie chart of selected time period. International data: comparison of incidence and mortality in the Czech Republic (CZ) with other countries. All these analyses are based on data obtained from IARC database GLOBOCAN Comparative standards: time trend of incidence or mortality in selected region in comparison with situation in whole Czech Republic. Available parameters are crude rate and age standardized ratio (ASR - European or World standard). Typology of patients: comprehensive overview of group of patient with specific diagnose Examples of automated analytical tools settings and outputs are on figures

172 Figure 1 Diagnose selection and consequent output window with tools for setting of the analysis, selection of specific group of patients and viewing tabular outputs and reports. Figure 2 Left window: example of analysis setting selection of requested epidemiologic parameter (incidence and/or mortality) and setting of appropriate units (absolute numbers, crude rate and age standardized ratio 5

173 - European or World standard); Right window: example of tools for selection of target group of patients available parameters are sex, age group, region, time period, clinical stage, TNM classification and other parameters related to the status of the patient. Figure 3 Example of comprehensive report of the analysis 4.2 Reporting system Results of all automated analytical tools are available in graphical and tabular form, but these outputs are viewed in separate windows. Thus, there was developed reporting system, which enables to make comprehensive outputs of every analysis (fig. 3). These reports contain graphical and tabular outputs with detailed description of setting of the analysis and are downloadable in pdf format. Please feel free to contact us at [email protected] if you would like to get more detailed information or if there is a potential to collaborate in similar type of project. Development of Web portal SVOD is supported by Ministry of Health Care Czech Republic (and research and developmental teams are granted by research project MZO solved in Masaryk Memorial Cancer Institute, Brno. Risk assessment analyses are supported by research project INCHEMBIOL, Ministry of Education Czech Republic project no

174 Bibliography J. Ferlay, F. Bray, P. Pisani and D.M. Parkin.: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004 Czech National Cancer Registry, 7

175

176 Paper X. Ondrušová M., Ondruš D., Dušek L., Mužík J.: What is the information availability to the cancer epidemiology data? Iranian Red Crescent Medical Journal 2008, 10(3):

177 Archive of SID Iranian Red Crescent Medical Journal What is the information availability to the cancer epidemiology data? REVIEW ARTICLE M Ondrusova 1,2,3*, D Ondrus 1,4, L Dusek 3, J Muzik 3 1 National Cancer Registry of the Slovak Republic, National Health Information Centre, 2 Department of Cancer Epidemiology, Cancer Research Institute, Bratislava, Slovak Republic, 3 Institute of biostatistics and analyses, Masaryk University, Brno, Czech Republic, 4 1 st Department of Oncology, Comenius University Medical School and St. Elisabeth Cancer Institute, Bratislava, Slovak Republic Abstract National Cancer Registry of the Slovak Republic, National Health Information Centre, would like to respond to many requests for easy and comprehensible access to the national and international data on cancer epidemiology. The working group created a new analytic web-page called "National portal on cancer epidemiology". All the data are valid, adapted for publications and quotation and the access to the web-page is free for the wide professional public. Keywords: Cancer, Epidemiology; Information system Introduction There has recently been a huge upsurge in demand for valid data on cancer epidemiology, particularly for the purposes of research effort and/or publications of experts in the fields of oncology, epidemiology, pathology, etc. The outputs released so far in regular yearly analyses of cancer incidence in the Slovak Republic, as well as in domestic and foreign publications, have often not been sufficient for the needs of the wide public. Therefore, it was decided to initiate the creation of a freely accessible information system, which would make the standard outputs of the National Cancer Registry of the Slovak Republic available to the public in an appropriate publishable form comparable to the international cancer data. The National Cancer Registry (NCR) at the National Health Information Centre (NHIC) of the Slovak Republic (hereinafter referred to as "the NCR SR/NHIC") is a specialized and methodical centre for the registration and processing of reports on cancer *Correspondence: RNDr. Martina Ondrusova, PhD, National Cancer Registry of the Slovak Republic, National Health Information Centre, Lazaretska 26, Bratislava, Slovak Republic. [email protected], Received: May 7, 2008 Accepted: June 8, 2008 incidence. In a close cooperation with the main experts in oncology from the Slovak Ministry of Health, as well as regional experts in medical oncology, the NCR SR ensures and guarantees the completeness andhighqualityofdataoncancerepidemiologyin the Slovak Republic. All the records are maintained in an internationally unified form, and stored in the NCR SR database. The data form the background for statistical processing and analyses of oncoepidemiological situation in the Slovak Republic, as well as on the European and global level - in cooperation with international organizations (e.g. IARC WHO, ENCR). Thanks to the high reliability and quality, the data processed and published by NCR SR have been included in all international periodical publications, overviews and databases of the World Health Organization (WHO), forming a constituent part of international projects focusing on cancer epidemiology. The main objective of the project "National Portal of Cancer Epidemiology in Slovakia" (hereinafter referred to as "the NCR SR web portal", Figure 1) has been to develop an information system to support scientific analyses of cancer data registered in the NCR SR/NHIC. The system has been prepared to aggregate, analyse and present epidemiological, clinical and demographic data from the NCR SR, IRCMJ 2008; 10(3): Iranian Red Crescent Society

178 Archive of SID Ondrusova et al. which have been so far regularly published in special monographs or in scientific reports (Figure 2). Fig 1: National Portal of Cancer Epidemiology in Slovakia. Fig 2: Present epidemiological, clinical and demographic data from the NCR SR. The NCR SR web portal processes mainly epidemiological data on patients registered in the NCR SR. Anonymized and validated data from 1978 to the present day are available in the registry, which is a uniquely representative dataset, at the European level at least (the database currently contains nearly records from the period ). The epidemiological trends were processed with relevant demographic data about the population of interest. The project initiation has been driven by the effort to make these representative and valuable data available to a broad spectrum of interested individuals. The project is based on the presumption that general information about cancer epidemiology and related population risks should be freely available. In this way, the user gets a direct and unique access to complex information from available resources, which can be further used in subsequent analyses and interpretations. The launch of the NCR SR web portal has provided a free access to all standard data on cancer epidemiology registered in the NCR SR/NHIC, thus ensuring that both professional and lay public in the Slovak Republic and abroad can get relevant information about cancer incidence and diagnosis. Here is an overview of information service at the NCR SR web portal: 1. News: regularly updated information on recent developments in population risk assessment and cancer epidemiology. 2. Interactive analyses: freely available software tools, allowing a direct examination of trends in cancer epidemiology. 3. Repors: commented outputs and presentations, prepared by leading experts in the field. The information service of the NCR SR web portal will be further improved, taking into consideration the users' remarks as well. At present, an opposition procedure is under way, in cooperation with experts in oncology and pathology. The data can be analysed according to basic demographic characteristics and diagnostic information, and also sorted by region or time. The system is ready for automatic updates of epidemiological data once a year at minimum as soon as the data are validated within the NCR SR/NHIC. Future objectives involve the extension of the information service over the population risk assessment, in relation to available data on environmental conditions in the Slovak Republic. The presented version of the NCR SR web portal makes available standard epidemiological data from the NCR SR, which can only be Vol 10 July

179 Archive of SID Information availability to cancer epidemiology used in conformity with the Copyright Act. Any reproduction, copying, editing, publication or further distribution is strictly forbidden without the appropriate citation of the source. 1, 2 All services of the portal are freely available. Time trends in cancer incidence and mortality in the Slovak Republic (Figure 3). Technical solution to the web portal The web portal is run on two computers: a database server with aggregated data from the NCR SR and other data sources (OS Linux with database control system MySQL v.4.0), and a web server with user interface (OS Linux supporting the PHP scripting modules), thereby guaranteeing the physical separation of the database server from the internet. The technical solution involves a secured backup of the processed data and scripts, ensuring a quick recovery of necessary data after a possible system breakdown. Technical requirements for the use of the web portal include a computer connected to internet and a web browser supporting HTML 4.01 and JavaScript. 2 Main benefits of the NCR SR web portal 1. The NCR SR web portal spares your time The web portal NCR SR makes available standard data from the NCR SR database in an aggregated form, providing end results of analyses and graphical outputs. 2. The NCR SR web portal is fully automated The software enables users to perform automatic analyses even to those without mathematical background, as the communication tools offer a simple selection of analyzed items. 3. The NCR SR web portal is a comprehensible and graphically-oriented tool The web portal contains an independent graphic module with more than 30 types of graphic outputs. All the results of analyses are directly visualized in the form of a graph. 4. The NCR SR web portal is a full-fledged information source The system provides its users with all data inputs needed for a comprehensive analysis of population risks related to cancer formation, i.e. aggregated epidemiological data as well as demographic population data. The following analytical tools are available: INCIDENCE AND MORTALITY Fig 3: Time trends in cancer incidence and mortality in the Slovak Republic. Fig 4: Age structure of living and deceased cancer patients. Vol 10 July

180 Archive of SID Ondrusova et al. TIME TRENDS Changes in trends in cancer incidence and mortality over time (growth index and year-on-year changes). AGE OF PATIENTS Age structure of living and deceased cancer patients (Figure 4). REGIONAL OVERVIEWS Comparison of cancer incidence and mortality in individual regions of the Slovak Republic (Figure 5) COMPARATIVE ANALYSES Time trends in epidemiological parameters in individual regions of the Slovak Republic in comparison to reference standards. CLINICAL STAGES Time trends in distribution of clinical stages (Figure 6). COMPARISON WITH FOREIGN COUNTRIES Comparison of cancer epidemiology in the Slovak Republic and worldwide (source: IARC - GLOBO- CAN 2002). SUMMARY PRESENTATION Comprehensive presentations of basic analyses for individual diagnoses. The authors of the project believe in a widespread use of these freely available data as provided by the NCR SR web portal. Fig 5: Comparison of cancer incidence and mortality in individual regions of the Slovak Republic. Fig 6: Time trends in distribution of clinical stages. References 1 Ondrusova M, Plesko I, Safaei-Diba Ch, Obsitnikova A, Stefanakova D, Ondrus D. Complex analysis of cancer incidence and mortality in the Slovak Republic, Bratislava, NCR SR, NHIC, 2007; (online), 2 Dusek L, Muzik J, Kubasek M, Koptikova J, Snajdrova L, Ondrusova M. National portal of Cancer Epidemiology in Slovakia. Masaryk University, Brno, Vol 10 July

181 Paper XI. Mužík J., Dušek L., Babjuk M., Kubásek M., Fínek J., Petruželka L. UROWEB - web portal for analysis and visualization of epidemiology, diagnosis and treatment of urologic malignancies [online]. Masarykova univerzita, Brno, Version 1.6d. Reviewed and guaranteed by the Czech Urological Society and the Czech Society for Oncology. Approving by National Technical Library in process. (original article in Czech)

182 UROWEB - web portal for analysis and visualization of epidemiology, diagnosis and treatment of urologic malignancies. Mužík J., Dušek L., Babjuk M., Kubásek M., Fínek J., Petruželka L. [online]. Masarykova univerzita, Brno, Version 1.6d. Reviewed and guaranteed by the Czech Urological Society and the Czech Society for Oncology. Approving by National Technical Library in process. (original article in Czech) Project Uroweb.cz ( is an information web portal dedicated to all who are interested in malignancies of urogenital system. The primary objective of the project is to build an information portal guaranteed by the Czech Urological Society (CUS) and the Czech Society for Oncology (CSO), which will provide important information about importance of urology in the treatment of urological malignancies in the Czech Republic (CR). Another goal is presentation of existing diagnostic and therapeutic procedures and regional presentations of urological departments. Dominant part of the portal is interactive analytical software that provides detailed epidemiological information for each diagnosis. For kidney, bladder, prostate and testicles following thematic interactive epidemiological analyses were prepared: - Epidemiology and population data of CR - trends of incidence, mortality and prevalence, age structure of patients, incidence in regions of CR, proportion of stages and incidence according to stage over time and by age, tumour morphology (Figure 1) - International epidemiological data - comparison of global data about incidence and mortality - data from publication GLOBOCAN 2002; comparison of incidence, patients' age and the cumulative risks in Europe - data from publication Cancer Incidence in Five Continents, Vol IX - Regional reports - comparison of trends in the incidence and mortality in regions with the situation in CR, comparison of age of patients in regions with the situation in the CR, proportion of disease stages and trends in the region compared with the situation in CR In addition to the interactive analytical tools focused on the epidemiology there are also available information on prediction, diagnosis, treatment and evaluation of the treatment: - 1 -

183 - Prediction of the number of patients treated in CR - Diagnosis and early detection - information on best practices of the European Association of Urology; overview of TNM classification (TNM classification III.-VI. Edition) - Therapeutic procedures and standards of treatment Czech recommendations and international standards, the outputs from the Library of chemotherapy regimens - Monitoring of results and the quality of care - reference standard of survival of patients with tumors of the urogenital system - Overview of current projects guaranteed by CUS and CSO The informative part of the portal further includes details about the project, available data sources, information about current research projects and presentations of urological departments connected to the Comprehensive Cancer Care Centres. The portal Uroweb.cz was launched in April References Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Vyzula R. Epidemiology of Malignant Tumours in the Czech Republic [online]. Masaryk University, Czech Republic, [2005], Version 7.0 [2007], ISSN Ferlay J., Bray F., Pisani P. and Parkin D.M.: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, Curado. M. P., Edwards, B., Shin. H.R., Storm. H., Ferlay. J., Heanue. M. and Boyle. P., eds (2007) Cancer Incidence in Five Continents, Vol. IX, IARC Scientific Publications No. 160, Lyon, IARC

184 Figure 1. Uroweb - example of outputs in the section of Epidemiology and population data of Czech Republic. Trends of incidence and mortality Trends of prevalence Age of patients Regional distribution Proportion of stages in time Trends of incidence according to stages Stages according to age Overview of morphology - 3 -

185 Figure 2. Uroweb - example of outputs in the section of International epidemiological data (source data: GLOBOCAN 2002, Cancer Incidence in Five Continents, Vol IX - European registries). Incidence according to European registries Age of patients (European registries) - 4 -

186 Figure 3. Uroweb - example of outputs in the section of Regional reports. Overview of incidence / mortality in regions Trends of incidence / mortality in regions Age of patients in regions Proportion of stages in regions - 5 -

187

188 XII. Appendix to paper I. Mužík J. Summary of cancer epidemiology in the Czech Republic. Published In Czech Cancer Care in Numbers Praha: Grada Publishing, a.s., ISBN (only selected examples of the most prevalent diagnoses)

189 Epidemiology of all malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) Summary All malignant tumours excluding nonmelanoma skin cancer (C00 C97 excluding C44) are responsible for 24.9% of the total number of deaths in the entire Czech population. The annual absolute number of newly diagnosed cases is 51,107, while the annual absolute mortality rate stands at 26,855 deaths (most recent data from 2005). The recent trend in incidence rate has been upward ( : annually +2.3%). An overall summary of all diagnoses revealed that the incidence rates are approximately equal in the male and female population. The analyses are based on the Czech National Cancer Registry ( verified information available for period ). Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) Proportion of all malignant tumours (y. 2005) Trend Age of patients Median (25 75% quartile) Males: Females: Males: 27,261 Females: 23,846 Males: 75.9% Females: 74.9% Males: annually +2.3% Females: annually +1.4% Males: Females: Ratio males:females 1.1:1 Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend Males: Females: Males: 14,975 Females: 11,880 Males: 27.7% Females: 22.1% Males: annually 0.1% Females: annually 0.3% Prevalence rate: 2005: 274,129 persons Mortality / incidence index: ( ): 0.56 Cumulative incidence risk: Males (0 74 years): 33.50% Females (0 74 years): 24.53% Recent time trends ( ): The incidence rate has continuously increased (annually +2.3%, males: +3.2%, females: 1.3%), while the mortality rate has slightly decreased. As a consequence, the prevalence rate has seen considerable growth (annually: +5.4%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: The overall summary of data from the Czech National Cancer Registry has confirmed that the Czech Republic ranks among countries with very high population burden caused by malignant tumours. With regard to both incidence and mortality rates, the Czech population occupies top positions in comparison to other European countries. Nevertheless, although the overall cancer incidence rate has seen steady growth, the trend for mortality rate appears to be stable or even slightly downward in recent times.

190 Epidemiology of all malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

191 Epidemiology of all malignant tumours excluding non-melanoma skin cancer (C00 C97 excluding C44) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

192 Epidemiology of malignant tumours of the colon, rectum and anus (C18 C21) Summary Colorectal cancer (C18 C21) is the second most common cancer in the Czech population, representing 11.8% of all newly diagnosed neoplasms. Colorectal cancer is responsible for 15.8% of all cancer-related deaths and for 4.0% of all mortalities. The annual absolute number of newly diagnosed cases is 7,982, while the annual absolute mortality rate stands at 4,312 deaths. The recent trend in incidence rate has been upward ( : annually +0.6%). The analyses are based on the Czech National Cancer Registry ( verified information available for period ). The main epidemiological parameters have been predicted for Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) Proportion of all malignant tumours (y. 2005) Trend Age of patients Median (25 75% quartile) Ratio males:females 1.4:1 Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend Males: 94.9 Females: 61.7 Males: 4,746 Females: 3,236 Males: 13.2% Females: 10.2% Males: annually +2.6% Females: annually +0.8% Males: 68 (60 75) years Females: 72 (62 79) years Males: 51.3 Females: 33.2 Males: 2,567 Females: 1,745 Males: 4.7% Females: 3.2% Males: annually +0.9% Females: annually 0.9% Prevalence rate: 2005: 40,580 persons 2009: 43,068 persons Mortality / incidence index: ( ): 0.56 Cumulative incidence risk: Males (0 74 years): 7.10% Females (0 74 years): 3.52% Recent time trends ( ): The incidence rate has shown a continuous slight increase as a predominant trend in the male population (annually +0.6%, males: +1.6%, females: 0.6%); the annual increments, however, appear to be lower than in the 1990s. The trend in mortality rate has been stable. As a consequence, the trend in prevalence rate has seen a significant increase (annually: +5.0%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: Colorectal cancer is the most common and the most dangerous cancer in the Czech Republic. The Czech population actually occupies an undesirable 1 st or 2 nd position in international comparisons of colorectal cancer incidence rate, and the number of newly diagnosed cases has seen a continuous increase. The prevalence rate is growing significantly, representing substantial and economically demanding burden for the Czech health care system; most of the patients are diagnosed and treated in advanced clinical stages.

193 Epidemiology of malignant tumours of the colon, rectum and anus (C18 C21) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

194 Epidemiology of malignant tumours of the colon, rectum and anus (C18 C21) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

195 Epidemiology of malignant tumours of the colon, rectum and anus (C18 C21) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I 1, % Stage II 2, % Stage III 1, % Stage IV 1, % Unstaged: objective reasons * % incomplete records % Total 8, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The epidemiological situation presents a challenge for a more effective colorectal cancer screening programme in the Czech Republic. Recently, 18.9% of patients have been diagnosed in stage III and a further 22.3% in primarily metastatic stage. An additional 10.7% of patients remain unstaged due to reasons excluding any therapy or effective therapy (diagnosed as DCO or at autopsy: 6.4%; patients with very advanced disease incompatible with therapy: nearly 3%). Nevertheless, the quality of data in the Czech National Cancer Registry is acceptable, only 5.3% records being incomplete (unstaged without reason) and this proportion has steadily decreased in time.

196 Epidemiology of malignant tumours of the trachea, bronchus and lung (C33, C34) Summary Cancer of the trachea, bronchus and lung (C33, C34) is the third most common cancer in the Czech population, representing 9.2% of all newly diagnosed neoplasms. This diagnostic group is responsible for 18.9% of all cancer-related deaths and for 4.8% of all mortalities. The annual absolute number of newly diagnosed cases is 6,249 in total, while the annual absolute mortality rate stands at 5,147 deaths. The disease is more common in males that in females (ratio 3.1:1). The analyses are based on the Czech National Cancer Registry ( verified information available for period ). The main epidemiological parameters have been predicted for Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) Proportion of all malignant tumours (y. 2005) Trend Age of patients Median (25 75% quartile) Ratio males:females 3.1:1 Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend Males: 92.6 Females: 30.8 Males: 4,632 Females: 1,617 Males: 12.9% Females: 5.1% Males: annually 0.2% Females: annually +3.8% Males: 66 (59 73) years Females: 68 (58 76) years Males: 77.7 Females: 24.0 Males: 3,886 Females: 1,261 Males: 7.2% Females: 2.3% Males: annually 1.4% Females: annually +2.7% Prevalence rate: 2005: 9,045 persons Mortality / incidence index: ( ): 0.86 Cumulative incidence risk: Males (0 74 years): 7.56% Females (0 74 years): 1.80% Recent time trends ( ): The overall incidence rate in the Czech population has slightly increased (annually +1.0%). However, this rise in incidence can be particularly attributed to the female population (+2.9%) rather than to men (+0.3%, females: +2.9%). In contrast to the incidence rate, the trend in mortality rate has been stable. As a consequence, the trend in prevalence rate has increased (annually: +4.3%). Specific comments: Lung cancer ranks among the most common cancer types in the Czech population, with further growing incidence rate as noted recently. Although the mortality rate appears to be stable, it is still remarkably high (mortality/incidence index: 0.86) and thus results in a relatively low prevalence rate in comparison to other common neoplasms (> 9,000 patients). Both the incidence and prevalence have shown significant regional differences, with the highest rates having been recorded in the western and northern parts of the Czech Republic.

197 Epidemiology of malignant tumours of the trachea, bronchus and lung (C33, C34) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

198 Epidemiology of malignant tumours of the trachea, bronchus and lung (C33, C34) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

199 Epidemiology of malignant tumours of the trachea, bronchus and lung (C33, C34) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I % Stage II % Stage III 1, % Stage IV 2, % Unstaged: objective reasons * 1, % incomplete records % Total 6, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. Recently, 22.9% of patients have been diagnosed in stage III and another 36.7% in the primary metastatic stage. However, 20.8% of patients remain unstaged due to reasons excluding any therapy or effective therapy (diagnosed as DCO or at autopsy: 12.0%; patients with very advanced disease incompatible with therapy: 6.3%). Nevertheless, the quality of data in the Czech National Cancer Registry is acceptable, only 4.6% records being incomplete (unstaged without any given reason), and this proportion has steadily decreased in time. With regard to the high proportion of cases diagnosed in advanced clinical stages, we cannot expect positive trends in patients survival.

200 Epidemiology of melanoma of the skin (C43) Summary Melanoma of the skin (C43) is the eight most common cancer in the Czech population, representing 2.7% of all newly diagnosed neoplasms. Melanoma is responsible for 1.4% of all cancer-related deaths and for 0.4% of all mortalities. The disease almost equally affects men and women. The annual absolute number of newly diagnosed cases is 1,824, while the annual absolute mortality rate stands at 391 deaths. Although the median age of newly diagnosed patients is 61 years, the disease also significantly affects the population under 50. The analyses are based on the Czech National Cancer Registry ( verified information available for period ). Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) Proportion of all malignant tumours (y. 2005) Trend Age of patients Median (25 75% quartile) Ratio males:females 1.0:1 Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend Males: 17.9 Females: 17.7 Males: 894 Females: 930 Males: 2.5% Females: 2.9% Males: annually +4.7% Females: annually +3.4% Males: 63 (52 73) years Females: 59 (48 72) years Males: 4.2 Females: 3.4 Males: 211 Females: 180 Males: 0.4% Females: 0.3% Males: annually 0.2% Females: annually +1.7% Prevalence rate: 2005: 15,617 persons Mortality / incidence index: ( ): 0.25 Cumulative incidence risk: Males (0 74 years): 1.27% Females (0 74 years): 1.05% Recent time trends ( ): The incidence has continuously and significantly increased (annually +5.4%, males: +4.3%, females: +6.6%), while the mortality has remained stable. As a consequence, the trend in the prevalence rate has seen a sharp rise (annually: +6.9%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: Melanoma of the skin ranks among cancer diagnoses with alarmingly growing incidence and prevalence rates in recent times, both in men and women. The mortality rate has been stable and a very positive profile of newly diagnosed clinical stages (> 57% of stage I) suggest a favourable prognosis with respect to survival of melanoma patients.

201 Epidemiology of melanoma of the skin (C43) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

202 Epidemiology of melanoma of the skin (C43) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

203 Epidemiology of melanoma of the skin (C43) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I % Stage II % Stage III % Stage IV % Unstaged: objective reasons * % incomplete records % Total 1, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The profile of newly diagnosed clinical stages indicates very effective diagnostics which results in nearly 60% of cases being detected at early stages. Stage IV only forms 5.1% of all newly diagnosed cases and its proportion has even slightly decreased in time. A low proportion of unstaged cases due to objective reasons (DCO and diagnosis at autopsy account for only 1.6% and early deaths only 0.2%) indirectly indicate that the diagnostics and treatment results are satisfactory.

204 Epidemiology of malignant tumours of the breast in women (C50) Summary Breast cancer (C50) is the fourth most common cancer in the Czech population (second in females), representing 8.2% of all newly diagnosed malignant tumours (17.4% in females). Breast cancer is responsible for 7.4% of all cancer-related deaths and for 1.9% of all mortalities (16.8% and 3.8% in females). The annual absolute number of newly diagnosed cases is 5,533, while the annual absolute mortality rate is 2,028 deaths. The recent trend in incidence rate has been stable ( : annually +1.5%, fluctuating). The analyses are based on the Czech National Cancer Registry ( verified information available for period ). The main epidemiological parameters have been predicted for Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) 5,533 Proportion of all malignant tumours (y. 2005) 17.4% Trend annually +2.1% Age of patients Median (25 75% quartile) 62 (53 73) years Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend , % annually 0.6% Prevalence rate: 2005: 49,539 females 2009: 55,204 females Mortality / incidence index: ( ): 0.37 Cumulative incidence risk: Females (0 74 years): 6.90% Recent time trends ( ): The incidence has remained stable, although there have been fluctuations (annually +1.5%), while the mortality rate has slightly decreased. As a consequence, the trend in prevalence rate has seen a sharp rise (annually: +5.6%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: The epidemiology of breast cancer in the Czech population has been recently influenced by the growing impact of mammography screening. The screening network was established in 2002 and every woman aged between 45 and 69 is eligible to attend free examination every 2 years. The screening has revealed a number of breast cancer cases at clinical stage I, which have significantly contributed to the overall incidence.

205 Epidemiology of malignant tumours of the breast (C50) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

206 Epidemiology of malignant tumours of the breast (C50) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

207 Epidemiology of malignant tumours of the breast (C50) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I 1, % Stage II 2, % Stage III % Stage IV % Unstaged: objective reasons * % incomplete records % Total 5, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The profile of newly diagnosed clinical stages is encouragingly asymmetric towards less advanced stages I and II. In addition, the proportion of newly diagnosed less advanced stages has steadily increased, while the relative contribution of stage II and III to the overall incidence has decreased in time. The proportion of primary diagnosed metastatic disease has remained stable in time, accounting for 8 9%. The proportion of unstaged records in the Czech National Cancer Registry has been stable in time; incomplete records accounting for 5.3% of all newly recorded cases.

208 Epidemiology of malignant tumours of the uterus (C54, C55) Summary Cancer of the uterus (C54, C55) is the ninth most common cancer in the Czech population (fourth in females), representing 2.6% of all newly diagnosed neoplasms (5.6% in females). This cancer is responsible for 1.9% of all cancer-related deaths and for 0.5% of all mortalities (4.4% and 1.0% in females). The annual absolute number of newly diagnosed women is 1,782 and the annual absolute mortality rate is 526 deaths. The recent trend in the incidence rate has been rather upward ( : annually +1.2%). Most tumours occur after the menopause, the median age at diagnosis being 65 years. The analyses are based on the Czech National Cancer Registry ( verified information available for period ). Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants 34.0 Absolute number of new cases (y. 2005) 1,782 Proportion of all malignant tumours (y. 2005) 5.6% Trend annually +1.3% Age of patients Median (25 75% quartile) 65 (57 73) years Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend % annually 0.4% Prevalence rate: 2005: 20,736 females Mortality / incidence index: ( ): 0.30 Cumulative incidence risk: Females (0 74 years): 2.30% Recent time trends ( ): The incidence rate has continuously increased (annually +1.2%), while the mortality rate has remained stable. As a consequence, the trend in prevalence rate has seen considerable growth (annually: +3.1%). The mortality/ incidence index is expected to decrease further in the years to come. Specific comments: Cancer of the uterus is a relatively common cancer, contributing significantly to the overall mortality of the Czech female population. The incidence rate has slowly increased with long-term annual increment of %; the Czech Republic currently occupies 2 nd position among the European countries with respect to incidence and mortality rates. Although no specific screening test has been designed to detect uterine cancer, > 58% of cases are diagnosed in clinical stage I, while the metastatic stage only accounts for 4.6%.

209 Epidemiology of malignant tumours of the uterus (C54, C55) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

210 Epidemiology of malignant tumours of the uterus (C54, C55) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

211 Epidemiology of malignant tumours of uterus (C54, C55) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I 1, % Stage II % Stage III % Stage IV % Unstaged: objective reasons * % incomplete records % Total 1, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The proportion of disease stages has been stable in time apart from sudden fluctuations in the quality of records which is being currently verified (significant increase of incomplete records in 2005). More than 58% of newly diagnosed women belong to stage I; primary diagnosed metastatic disease accounting for 4.6% only. Various objective reasons are responsible for 6.5% of cases which remain unstaged (2.0% as DCO or diagnosed at autopsy, 1.3% due to very early death and further > 3% due to objective clinical reasons).

212 Epidemiology of malignant tumours of the prostate (C61) Summary Prostate cancer (C61) is the fifth most common type of cancer in the Czech population (second in males), representing 7.1% of all newly diagnosed neoplasms (13.5% in males). Prostate cancer is responsible for 5.2% of all cancer-related deaths and for 1.3% of all mortalities (9.4% and 2.6% in males). The annual absolute number of newly diagnosed cases is 4,846, while the annual absolute mortality rate stands at 1,426 deaths. The prevalence rate reported in 2005 exceeded 20,000 males. The analyses are based on the Czech National Cancer Registry ( verified information available for period ). Selected epidemiological parameters have been predicted for Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants 96.9 Absolute number of new cases (y. 2005) 4,846 Proportion of all malignant tumours (y. 2005) 13.5% Trend annually +8,0% Age of patients Median (25 75% quartile) 72 (65 77) years Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants 28.5 Absolute number of deaths (y. 2005) 1,426 Proportion of total mortality (y. 2005) 2.6% Trend annually +1.3% Prevalence rate: 2005: 20,429 males 2009: 26,933 males Mortality / incidence index: ( ): 0.36 Cumulative incidence risk: Males (0 74 years): 5.95% Recent time trends ( ): The incidence rate has increased very significantly (annually +11.0), while the mortality rate has remained stable. As a consequence, the prevalence rate has seen a sharp rise (annually: +11.6%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: The incidence rates of prostate cancer vary widely across the world; the highest reported occurrence possibly resulting from the ever increasing number of cases detected at the early stages. This evidently applies to the Czech Republic, where a steep rise in the incidence rate (annually since 2000) has mostly resulted from the detection of less advanced stages I or II. Nevertheless, prostate cancer ranks among the leading causes of cancer deaths among men, similar to colorectal and lung cancers. The increasing rate of early detection of prostate cancer has resulted from the widely adopted PSA examination and from the generally improving cancer prevention programmes in the Czech Republic.

213 Epidemiology of malignant tumours of the prostate (C61) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

214 Epidemiology of malignant tumours of the prostate (C61) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

215 Epidemiology of malignant tumours of prostate (C61) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I % Stage II 1, % Stage III % Stage IV % Unstaged: objective reasons * % incomplete records % Total 3, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The proportion of stages has significantly changed over time, the proportion of early detected stages I and II having risen sharply (> 47% of all newly diagnosed cases). This trend is most probably associated with increasing accessibility to regular PSA examination in the Czech Republic. The sudden exchange of stage I and II in was due to the implementation of a new version of TNM classification in the Czech National Cancer Registry. Apart from the positive trends, nearly 26% of patients are still diagnosed at advanced stages III and IV. Another 13.5% of cases remain unstaged due to objective reasons (DCO or diagnosis at autopsy: 6.5%; refused care, very advanced age: 7.1%, early deaths: 1.5%).

216 Epidemiology of malignant tumours of the kidney (C64) Summary Kidney cancer (C64) is the sixth most common cancer in the Czech population, representing 4.1% of all newly diagnosed neoplasms. Kidney cancer is responsible for 4.2% of all cancer-related deaths and for 1.1% of all mortalities. The annual absolute number of newly diagnosed cases is 2,756, while the annual absolute mortality rate is 1,147 deaths. The majority of reported kidney cancers are diagnosed between 55 and 70 years of age. The disease is more prevalent in males than in females (ratio 1.7:1). The analyses are based on the Czech National Cancer Registry ( verified information available for period ). Selected epidemiological parameters have been predicted for Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) Proportion of all malignant tumours (y. 2005) Trend Age of patients Median (25 75% quartile) Ratio males:females 1.7:1 Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend Males: 34.8 Females: 19.3 Males: 1,742 Females: 1,014 Males: 4.8% Females: 3.2% Males: annually +2.8% Females: annually +1.0% Males: 64 (56 72) years Females: 69 (59 76) years Males: 14.5 Females: 8.1 Males: 723 Females: 424 Males: 1.3% Females: 0.8% Males: annually +0.7% Females: annually 1.1% Prevalence rate: 2005: 15,368 persons 2009 (prediction): 21,839 persons Mortality / incidence index: ( ): 0.46 Cumulative incidence risk: Males (0 74 years): 2.64% Females (0 74 years): 1.20% Recent time trends ( ): The incidence rate has continuously increased (annually +4.2%, males: +5.5%, females: +2.3%), while the mortality rate has stabilised and recently has even begun to decrease slightly. As a consequence, the prevalence rate has seen a sharp rise (annually: +6.5%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: Renal cell carcinoma is the most common type of kidney cancer, representing 72 74% of all kidney cancer cases newly diagnosed in the Czech population. The ever growing incidence rate of kidney cancer has been so high that the Czech Republic now occupies the unmatched 1 st position in international comparison. Unfortunately, the same applies for the mortality rate. This extraordinary position of the Czech Republic with regard to kidney cancer rates has recently initiated the effort to validate the correctness of reported Czech data on this cancer type.

217 Epidemiology of malignant tumours of the kidney (C64) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

218 Epidemiology of malignant tumours of the kidney (C64) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

219 Epidemiology of malignant tumours of the kidney (C64) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I % Stage II % Stage III % Stage IV % Unstaged: objective reasons * % incomplete records % Total 2, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The proportion of disease stages has significantly changed over time, with the increasing proportion of new cases diagnosed at stage I or II (recently exceeding 50%) being the most dominant trend. This trend might suggest improvements in early diagnostics of kidney cancer and in accessibility to examination. Nearly 15% of newly reported cases remain unstaged without objective reasons (DCO or diagnosis at autopsy: > 9%; early deaths: > 2%, very advanced disease or patient s refusal of treatment: > 3%).

220 Epidemiology of malignant tumours of the bladder (C67) Summary Bladder cancer (C67) is the seventh most common cancer in the Czech population, representing 3.7% of all newly diagnosed neoplasms. Bladder cancer is responsible for 2.8% of all cancer-related deaths and for 0.7% of all mortalities. The annual absolute number of newly diagnosed cases is 2,477, while the annual absolute mortality rate is 752 deaths. The incidence is significantly higher in men than in women (ratio 2.8:1). The risk increases with age, peak incidence being between 60 and 70 years of age. The analyses are based on the Czech National Cancer Registry ( verified information available for period ). Incidence rate Crude incidence (y. 2005) Cases/100,000 inhabitants Absolute number of new cases (y. 2005) Proportion of all malignant tumours (y. 2005) Trend Age of patients Median (25 75% quartile) Ratio males:females 2.8:1 Mortality rate Crude mortality (y. 2005) Deaths/100,000 inhabitants Absolute number of deaths (y. 2005) Proportion of total mortality (y. 2005) Trend Males: 36.5 Females: 12.4 Males: 1,827 Females: 650 Males: 5.1% Females: 2.0% Males: annually +3.5% Females: annually +3.8% Males: 69 (61 76) years Females: 71 (62 77) years Males: 11.1 Females: 3.7 Males: 556 Females: 196 Males: 1.0% Females: 0.4% Males: annually 0.2% Females: annually +1.2% Prevalence rate: 2005: 14,532 persons Mortality / incidence index: ( ): 0.35 Cumulative incidence risk: Males (0 74 years): 2.56% Females (0 74 years): 0.70% Recent time trends ( ): The incidence rate has continuously and significantly increased (annually +5.0%, males: +5.1%, females: +4.8%), while the mortality rate has remained stable. As a consequence, the prevalence rate has seen a sharp rise (annually: +6.0%). The mortality/incidence index is expected to decrease further in the years to come. Specific comments: With regard to the steadily and significantly growing incidence rate, bladder cancer ranks among the most important malignancies affecting the Czech population. The prevalence rate is also remarkably high (> 14,500 persons). However, the growing population burden mostly results from an increasing number of cases diagnosed at stage I, which suggests a rather favourable prognosis for the Czech population.

221 Epidemiology of malignant tumours of the bladder (C67) Incidence and mortality trend Regional incidence in the Czech Republic (years ) Age of the patients (years )

222 Epidemiology of malignant tumours of the bladder (C67) Incidence in international comparison European countries Mortality in international comparison European countries Source: J. Ferlay, F. Bray, P. Pisani and D.M. Parkin: GLOBOCAN 2002: Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5. version 2.0, IARCPress, Lyon, 2004,

223 Epidemiology of malignant tumours of the bladder (C67) Proportion of clinical stages in newly diagnosed patients Incidence according to disease stage Clinical stages in summary ( : annual average) N % Stage I 1, % Stage II % Stage III % Stage IV % Unstaged: objective reasons * % incomplete records % Total 2, % * DCO, cases diagnosed by autopsy, early deaths, therapy had not been started due to objective reasons, etc. The recent epidemiological profile of bladder cancer in the Czech Republic is dominated by the significantly growing trend in the incidence of stage I. This trend is accompanied by decreasing incidence of unstaged cases due to objective reasons, which again is a trend contributing to favourable prognosis in terms of survival. Currently, DCO reporting and diagnosis at autopsy accounts for < 4% of newly reported cases. The proportion of stages II IV has remained unchanged since The population database contains too many records with incomplete staging, which is a challenge to improve data management in this field.

224

225 XIII. Curriculum Vitae

226 Jan Mužík Curriculum vitae JAN MUŽÍK Curriculum vitae PERSONAL DETAILS Nationality: Czech Date of birth: 28 th July 1973 Place of birth: Brno, Czech Republic Marital status: Unmarried Phone: Cell: Address: Uzbecká 28, , Brno, Czech Republic Workplace: Institute of Biostatistics and Analyses, Masaryk University Kamenice 126/3, Brno, Czech Republic EDUCATION Since 2004: Ph.D. candidate, Oncology Masaryk University, Faculty of Medicine, Brno, Czech Republic 2008: RNDr. (advanced state examination), Mathematical Biology Masaryk University, Faculty of Science, Brno, Czech Republic : M.Sc., Molecular genetics Masaryk University, Faculty of Science, Brno, Czech Republic EMPLOYMENT Since 2002: Institute of Biostatistics and Analyses, Masaryk University, Brno, Czech Republic - research assistant, analyst of clinical and population-based medical data, focus on cancer epidemiology and haematooncology : Dept. of Pathophysiology, Faculty of Medicine, Masaryk University, Brno, Czech Republic - research worker, focus on genetics of multifactorial diseases

227 Jan Mužík Curriculum vitae SKILLS Biostatistics and data analysis: descriptive univariate and multivariate statistics, hypotheses testing, general and generalized linear models, survival analysis, cancer registration and epidemiology, management of large data sets Computer skills: Statistica, SPSS, Relational database systems (MS Access, Oracle), office software (MS Word, MS Excel, MS Power Point) Language skills: Czech native, English advanced, German basic, Russian - basic SELECTED PUBLICATIONS AND PRESENTATIONS Journal articles Mužík J., Koptíková J., Dušek L., Žaloudík J., Vyzula R., Abrahámová J. Historical data of the Czech National Cancer Registry: information value and risk of bias; Klinická onkologie 20, Supplement 1/2007, 63-76, ISSN X (original article in Czech) Mužík J., Dušek L., Kubásek M., Koptíková J., Schwarz D., Zemánek P., Žaloudík J., Vyzula R. Cancer epidemiology in the Czech Republic on-line; Klinická onkologie 20, Supplement 1/2007, , ISSN X. (original article in Czech) Dušek L., Mužík J., Koptíková J., Žaloudík J., Klimeš D., Bourek A., Indrák K., Mihál V., Hajdúch M., Št rba J., Vyzula R., Abrahámová J. Data registries form indispensable information base of current oncology. Klinická onkologie 20, Supplement 1/2007, 53-62, ISSN X. (original article in Czech) Klimeš D., Mužík J., Kubásek M., Koptíková J., Brabec P., Abrahámová J., Dušek L. Educational on-line version of data records for Czech National Cancer Registry; Klinická onkologie 20, Supplement 1/2007, , ISSN X. (original article in Czech) Dušek L., Pavlík T., Koptíková J., Mužík J., Gelnarová E., Žaloudík J., Vyzula R., Hajdúch M., Abrahámová J. Czech National Cancer Registry and reference standards for health care assessment. Klinická onkologie 20, Supplement 1/2007, 77-95, ISSN X (original article in Czech) Zavoral M., Suchánek Š., Závada F., Dušek L., Mužík J., Seifert B., Fri P. Colorectal cancer screening in Europe. World J Gastroenterol, Beijing, China, the WJG Press and Baishideng. vol. 15, no. 47, s Dusek L., Abrahamova J., Lakomy R., Vyzula R., Koptikova J., Pavlik T., Muzik J., Klimes D. Multivariate analysis of risk factors for testicular cancer: a hospital-based case-control study in the Czech Republic. Neoplasma Vol.55, No.4, p , 2008 Ondrušová, M., Ondruš, D., Dušek, L., Mužík, J.: National portal of cancer epidemiology in the Slovak Republic. 2008, Bratislavské lekárske listy 109 (7): s ISSN Ondrušová, M., Ondruš, D., Dušek, L., Mužík, J.: What is the information availability to the cancer epidemiology data?, Iranian Red Crescent Medical Journal 2008, 10(3): Stefancikova L., Moulis M., Fabian P., Ravcukova B., Vasova I., Muzik J., Malcikova J., Falkova I., Slovackova J., Smardova J. Loss of the p53 tumor suppressor activity is

228 Jan Mužík Curriculum vitae associated with negative prognosis of mantle cell lymphoma. International Journal Of Oncology Vol. 36 (3), p , Mar 2010 Holzerova M., Faber E., Veselovska J., Urbankova H., Balcarkova J., Rozmanova S., Voglova J., Muzik J., Chroust K., Indrak K., Jarosova M. Imatinib mesylate efficacy in 72 previously treated Philadelphia-positive chronic myeloid leukemia patients with and without additional chromosomal changes: single-center results. Cancer Genetics and Cytogenetics Vol. 191 (1), p. 1-9, May 2009 Doubek M., Muzik J., Szotkowski T., Koza V., Cetkovsky P., Kozak T., Zak P., Voglova J., Struncova S., Dusek L., Indrak K. Is FLT3 internal tandem duplication significant indicator for allogeneic transplantation in acute myeloid leukemia? An analysis of patients from the Czech Acute Leukemia Clinical Register (ALERT). Neoplasma Vol. 54 (1), p , 2007 Smejkal P., Brabec P., Matyskova M., Bulikova A., Slechtova M., Kissova J., Chlupova G., Muzik J., Penka M. FEIBA in treatment of acute bleeding episodes in patients with haemophilia A and factor VIII inhibitors: a retrospective survey in regional haemophilia centre. Haemophilia Vol. 15 (3), p , May 2009 Holla LI., Fassmann A., Muzik J., Vanek J., Vasku A. Functional polymorphisms in the matrix metalloproteinase-9 gene in relation to severity of chronic periodontitis. Journal Of Periodontology, 77 (11), , Nov 2006 Smrcka M., Horky M., Otevrel F., Kuchtickova S., Kotala V., Muzik J. The onset of apoptosis of neurons induced by ischemia-reperfusion injury is delayed by transient period of hypertension in rats. Physiological Research, 52 (1): Vasku A., Spinarova L., Goldbergova M., Muzik J., Spinar J., Vitovec J., Toman J., Vacha J. The double heterozygote of two endothelin-1 gene polymorphisms (G8002A and -3A/- 4A) is related to big endothelin levels in chronic heart failure. Experimental And Molecular Pathology, 73 (3): Dec 2002 Kankova K., Jansen EHJM., Marova I., Stejskalova A., Pacal L., Muzik J., Vacha J. Relations among serum ferritin, C282Y and H63D mutations in the HFE gene and type 2 diabetes mellitus in the Czech population. Experimental And Clinical Endocrinology & Diabetes, 110 (5): Aug 2002 Vasku V., Kankova K., Vasku A., Muzik J., Holla LI., Semradova V., Vacha J. Gene polymorphisms (G82S, 1704G/T, 2184A/G and 2245G/A) of the receptor of advanced glycation end products (RAGE) in plaque psoriasis. Archives Of Dermatological Research, 294 (3): May 2002 Benes P., Muzik J., Benedik J., Znojil V., Vacha J. The relationship among apolipoprotein(a) polymorphisms, the low-density lipoprotein receptor-related protein, and the very low density lipoprotein receptor genes, and plasma lipoprotein(a) concentration in the Czech population. Human Biology, 74 (1): Feb 2002 Jurajda M., Muzik J., Holla LI., Vacha J. A newly identified single nucleotide polymorphism in the promoter of the matrix metalloproteinase-1 gene. Molecular And Cellular Probes, 16 (1): Feb 2002 Jurajda M., Kankova K., Muzik J., Unzeitig V., Drabkova M., Izakovicova-Holla L., Vacha J. Lack of an association of a single nucleotide polymorphism in the promoter of the matrix metalloproteinase-1 gene in Czech women with pregnancy-induced hypertension. Gynecologic And Obstetric Investigation, 52 (2):

229 Jan Mužík Curriculum vitae Kankova K., Marova I., Zahejsky J., Muzik J., Stejskalova A., Znojil V., Vacha J. Polymorphisms 1704G/T and 2184A/G in the RAGE gene are associated with antioxidant status. Metabolism-Clinical And Experimental, 50 (10): Oct 2001 Kankova K., Zahejsky J., Marova I., Muzik J., Kuhrova V., Blazkova M., Znojil V., Beranek M., Vacha J. Polymorphisms in the RAGE gene influence susceptibility to diabetesassociated microvascular dermatoses in NIDDM. Journal Of Diabetes And Its Complications, 15 (4): Jul-Aug 2001 Kankova K., Muzik J., Karaskova J., Beranek M., Hajek D., Znojil V., Vlkova E., Vacha J. Duration of non-insulin-dependent diabetes mellitus and the TNF-beta Ncol genotype as predictive factors in proliferative diabetic retinopathy. Ophthalmologica, 215 (4): Jul-Aug 2001 Benes P., Kankova K., Muzik J., Groch L., Benedik J., Elbl L., Izakovicova-Holla L., Vasku A., Znojil V., Vacha J. Methylenetetrahydrofolate reductase polymorphism, type II diabetes mellitus, coronary artery disease, and essential hypertension in the Czech population. Molecular Genetics And Metabolism, 73 (2): Jun 2001 Benes P., Muzik J., Benedik J., Elbl L., Vasku A., Siskova L., Znojil V., Vacha J. The C766T low-density lipoprotein receptor related protein polymorphism and coronary artery disease, plasma lipoproteins, and longevity in the Czech population. Journal Of Molecular Medicine-Jmm, 79 (2-3): Apr 2001 Benes P., Muzik J., Benedik J., Elbl L., Znojil V., Vacha J. Apolipoprotein B signal peptide polymorphism in relation to lipids and diabetes in male CAD patients. Atherosclerosis, 152 (1): Sep 2000 Beranek M., Kankova K., Muzik J. Identification of novel common polymorphisms in the promoter region of the TIMP-3 gene in Czech population. MOLECULAR AND CELLULAR PROBES, 14 (4): AUG 2000 Benes P., Muzik J., Benedik J., Elbl L., Znojil V., Vacha J. Relation between the insertion/deletion polymorphism in the gene coding for receptor associated protein (RAP) and plasma apolipoprotein AI (apoai) and high-density lipoprotein cholesterol (HDL) levels. Clinical Genetics, 57 (4): Apr 2000 Benes P., Muzik J., Benedik J., Frelich M., Elbl L., Vasku A., Znojil V., Vacha J. Single effects of apolipoprotein B, (a), and E polymorphisms and interaction between plasminogen activator inhibitor-1 and apolipoprotein(a) genotypes and the risk of coronary artery disease in Czech male Caucasians. Molecular Genetics And Metabolism, 69 (2): Feb 2000 Book chapters Mužík J., Faber E. Epidemiology. In Chronická myeloidní leukémie. 1 st edition. Praha : Galén, ISBN , s (original article in Czech) Mužík J., Dušek L., Koptíková J., Fínek J. Cancer Epidemiology in the Czech Republic: Current Load and the Time Trends. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN Dušek L., Mužík J., Pavlík T., Koptíková J., Šnajdrová L., Fínek J., Vyzula R. Data and Background Information on Czech Cancer Care. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN

230 Jan Mužík Curriculum vitae Sláma O., Mužík J., Sk i ková J., Koptíková J., Dušek L. Management and Monitoring of Palliative Care in the Czech Republic. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN Dušek L., Pavlík T., Májek O., Koptíková J., Gelnarová E., Mužík J., Vyzula R., Fínek J. Information System for Predictive Evaluation of Cancer Epidemiology and the Number of Cancer Patients in the Czech Republic. In Czech Cancer Care in Numbers Praha : Grada Publishing, a.s., ISBN Electronic publications, software Dušek L., Mužík J., Kubásek M., Koptíková J., Žaloudík J., Vyzula R. Epidemiology of Malignant Tumours in the Czech Republic [online]. Masaryk University, Czech Republic, [2005], Version 7.0 [2007], Approved by National Technical Library, ISSN Mužík J., Dušek L., Babjuk M., Kubásek M., Fínek J., Petruželka L. UROWEB - web portal for analysis and visualization of epidemiology, diagnosis and treatment of urologic malignancies [online]. Masarykova univerzita, Brno, [cit ]. Version 1.6d. Approving by National Technical Library (ISSN) in process. (original article in Czech) Conference presentations with abstracts in journals Zackova D., Klamova H., Dusek L., Muzik J., Dobesova B., Rysava J., Racil Z., Doubek M., Dvorakova D., Jurcek T., Razga F., Moravcova J., Polakova K. Machova, Rulcova J., Cetkovsky P., Mayer J. Imatinib in the first line chronic myeloid leukemia (CML) treatment. Can we compare the real-life data to clinical trial results? Haematologica-The Hematology Journal vol. 95 Suppl. 2, p , JUN th Annual Meeting of the European-Hematology-Association, Jun 10-13, 2010 Barcelona, Spain Brychtova Y., Krejci M., Muzik J., Doubek M., Mayer J. Long-term results of allogeneic stem cell transplantation after reduced-intensity conditioning regimen fludarabin, busulfan and ATG. Bone Marrow Transplantation Vol. 45, Suppl. 2, s294, Mar th Annual Meeting of the EBMT. Mar , Vienna, Austria Faber E., Voglov J., Koza V., Kotoucek P., Demetrvicova L., Tothova E., Chudej L., Zak P., Jarosova M., Demeckova E., Muzik J., Dusek L., Indrak K. CAMELIA: a new international registry for patients with chronic myeloid leukemia. Blood Reviews Vol. 21 Suppl. 1, s126, Aug 2007 Dusek L., Brabec P., Klimes D., Koptikova J., Chroust K., Muzik J. Multi-diagnostic on-line information system for clinical and epidemiological registries in haematology and haemato-oncology in the Czech Republic. Blood Reviews Vol. 21 Suppl. 1, S143-S144, Aug 2007 Brychtova Y., Doubek M., Krejci M., Muzik J., Navratil M., Mayer J., Vorlicek J. Morbidity and mortality with non-myeloablative compared to myeloablative conditioning before stem cell transplantation. Bone Marrow Transplantation Vol. 39 Suppl. 1, S93-S94, Apr rd Annual Meeting of the EBMT, Mar 25-28, 2007 Lyon, France

231 Jan Mužík Curriculum vitae Conference presentations with articles in conference proceedings Mužík J., Dušek L., Pavliš P., Koptíková J., Žaloudík J., Vyzula R. Analysis of population cancer risk factors in national information system SVOD. Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Dušek L., Mužík J., Koptíková J., Brabec P., Žaloudík J., Vyzula R., Kubásek M. The National Web Portal for Cancer Epidemiology in the Czech Republic. Proceedings of the 19th International Conference of Informatics for Environmental Protection (Enviroinfo 2005). Brno: Masaryk University, Brno, s ISBN Mužík J., Májek O., Dušek L. Early stage detection of cancer in the Czech Republic according to NCR data. In EDUKA NÍ SBORNÍK - XXXIII. Brn nské onkologické dny a XXIII. Konference pro sestry a laboranty. Brno : Masaryk v onkologický ústav, ISBN (original article in Czech) Mužík J., Abrahámová J., Pavlík T., Koptíková J., Gelnarová E., Dušek L. Epidemiology of malignant testicular tumours in the Czech Republic - detailed analysis of NCR data in In Vybrané otázky onkologie XI. 15. onkologicko-urologické sympozium a 11. mammologické sympozium. Praha : Galén, ISBN , s , Praha. (original article in Czech) Mužík J., Dušek L., Babjuk M. UROWEB a new information system providing free access to the epidemiology of urological malignancies in regions of CR. In Vybrané otázky onkologie XIII. 17. onkologicko-urologické sympozium a 13. mammologické sympozium. Praha : Galén, ISBN , Praha. (original article in Czech) Mužík J., Dušek L., Abrahámová J. Quality of data of the National Cancer Registry of the Czech Republic in the period In EDUKA NÍ SBORNÍK - XXXIV. Brn nské onkologické dny a XXIV. Konference pro sestry a laboranty. Brno : Masaryk v onkologický ústav, ISBN (original article in Czech)