A COMBINATORIAL APPROACH TO ELECTRONIC HEALTHCARE RECORDS IN PHARMACOEPIDEMIOLOGY

Transcription

1 A COMBINATORIAL APPROACH TO ELECTRONIC HEALTHCARE RECORDS IN PHARMACOEPIDEMIOLOGY MARLOES BAZELIER

2 A COMBINATORIAL APPROACH TO ELECTRONIC HEALTHCARE RECORDS IN PHARMACOEPIDEMIOLOGY MARLOES BAZELIER 1

3 Cover design Cover illustrations Layout Photography Printed by Marloes Bazelier shutterstock.com Matthijs Ariëns, Ipskamp drukkers Heleen Klop fotografie, Utrecht Ipskamp drukkers, Enschede A COMBINATORIAL APPROACH TO ELECTRONIC HEALTHCARE RECORDS IN PHARMACOEPIDEMIOLOGY Het combineren van elektronische gezondheidszorggegevens in de farmaco-epidemiologie (met een samenvatting in het Nederlands) Proefschrift The work presented in this thesis was performed at: - Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Utrecht, the Netherlands - Aarhus University Hospital, Aarhus, Denmark - National Institute of Public Health, Copenhagen, Denmark CIP-gegevens Koninklijke Bibliotheek, Den Haag Bazelier, M.T. A combinatorial approach to electronic healthcare records in pharmacoepidemiology Thesis Utrecht University ISBN: ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. G.J. van der Zwaan, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op woensdag 19 december 2012 des middags te 2.30 uur door 2012 Marloes Bazelier For articles published or accepted for publication, the copyright has been transferred to the respective publisher. No part of this thesis may be reproduced, stored in a retrieval system, or transmitted in any form or by any means without the permission of the author, or when appropriate, the publisher of the manuscript. Marloes Tanneke Bazelier geboren op 11 augustus 1981 te Apeldoorn 2 3

4 Promotoren: Prof.dr. H.G.M. Leufkens Prof.dr. T.P. van Staa CONTENTS Chapter 1. Introduction 7 Co-promotor: Dr. F. de Vries Chapter 2. Risk of fracture with multiple sclerosis 2.1 The risk of fracture in patients with multiple sclerosis: the UK General Practice Research Database 2.2 A simple score for estimating the long-term risk of fracture in patients with multiple sclerosis 2.3 Risk of fractures in patients with multiple sclerosis: a populationbased cohort study 2.4 Incidence of fractures in patients with multiple sclerosis: the Danish National Health Registers 2.5 The risk of fracture in incident multiple sclerosis patients: the Danish National Health Registers Chapter 3. Risk of fracture with thiazolidinediones 3.1 Use of thiazolidinediones and risk of osteoporotic fracture: disease or drugs? 3.2 Risk of fracture with thiazolidinediones: disease or drugs? Chapter 4. Multi-country studies 4.1 Multiple sclerosis and fracture risk: meta-analysis versus megaanalysis of individual patient data 4.2 Risk of fracture with thiazolidinediones: an individual patient data meta-analysis Chapter 5. Discussion 187 Appendices. Summary Samenvatting Acknowledgements List of co-authors List of publications About the author

5 INTRODUCTION De rijkdom van de natuurwetenschappen bestaat niet uit de hoeveelheid, maar uit de combinatie van feiten. Alexander von Humboldt ( ) CHAPTER 1

6 CHAPTER 1 INTRODUCTION Pharmacoepidemiology is the study of the effects of drugs in human populations that uses epidemiological methods [1]. The winter of 1961 may be regarded as the start of the field of pharmacoepidemiology, when the world experienced the infamous thalidomide disaster [2]. Shortly after thalidomide -a mild hypnotic drug- had entered the market, a dramatic increase was seen in the number of newborn babies without limbs or parts of their limbs; a birth defect called phocomelia. This tragedy triggered large regulatory changes in the approval process of drugs, with the ultimate goal for safer and more effective drugs to enter the market. The current drug approval process in most developed countries includes preclinical animal testing followed by three phases of clinical testing. In the third phase randomised clinical trials are conducted, which usually include between 500 and 3000 patients who are exposed to a drug [2]. Although premarketing studies form an important phase in the drug safety cycle nowadays, they are thus limited in size. They are also limited in follow-up time and clinical trial populations are usually different from patients in daily practice. As a consequence, adverse effects that are uncommon or long-term or that only occur in selective populations are often not detected before marketing. Postmarketing observational studies are therefore a valuable complement to clinical studies in the assessment of beneficial and harmful effects of drugs. Large electronic healthcare registries that contain electronic healthcare records from large populations are often used for these postmarketing studies. Regulators often rely on the combination of results from multiple independent studies, also known as meta-analyses, to make decisions. Over the last 30 years, the number of metaanalyses as published in literature has increased from approximately 250 in 1990 to 2300 in 2006 (based on counts in Pubmed) [3]. Meta-analyses are cited more often than all other types of research papers [4] and they are increasingly accepted as the best way to summarise the results from separate randomised controlled trials (RCTs). Moreover, regulatory decisions also depend on the combination of results from observational studies. An example of a regulatory action that was based on the pooled analysis of several observational studies is a recent decision from the Food and Drug Administration (FDA). On 25 May 2010, the FDA added a warning to the labelling of proton pump inhibitors (PPIs), mentioning a possible increased risk of fractures [5]. This decision was based on the FDA s internal review of seven epidemiological studies, including two studies that used data from the GPRD. Interestingly, the two GPRD studies reported conflicting results. Researchers of studies that use electronic healthcare registries have to make a large number of decisions when performing the analyses, such as the time-window of exposure, code lists for the outcomes and risk factors and selection of the comparison groups. Over the past decade, there have been several examples where different study designs have led to different results, even for studies conducted in the same database. These include the risk of venous thromboembolism with third generation pills [6], the association between PPIs and fracture [7], the use of oral bisphosphonates and risk of oesophageal cancer [8,9] and the risk of fracture for patients using statins [10,11]. For the last example, a third study examined the reasons for the discrepant results [12]. It was found that the age-band for matching cases and controls, the selection of potential confounders, the exclusion of high-risk patients and different definitions for exposure time-windows explained different results in the two studies [12]. This illustrates that the choice of the study design can substantially influence the results of an observational study. There are various ways to combine data from several independent studies, which all have advantages and disadvantages. The first strategy is to combine aggregate data (summary results), which can be obtained from the published reports. The major advantages of this approach are the ease in obtaining the requisite data and the simplicity of the analyses [13]. However, the results of aggregate meta-analyses can be misleading [14]. Drawbacks include publication bias, poor quality of the primary studies, the use of different end-points and the limited ability to explore heterogeneity between the studies [15]. An alternative strategy is to combine individual patient data from separate studies. From a scientific point of view, this approach has many advantages, such as the ability to use similar definitions, to adjust for the same variables across studies, to explore heterogeneity at a patient level and to perform subgroup analyses of patient level data [15,16]. Furthermore, a collaboration with the original authors may lead to a better understanding of the data. But there are also disadvantages of this approach, such as the complexity of the statistical techniques for combining individual patient data. In addition, these types of meta-analyses are usually time consuming and more expensive than an aggregate data meta-analysis. Meta-analyses can be based on data from clinical trials or observational studies. The key feature of clinical trials is randomisation, which minimises bias in the comparison between the treatment groups. Nevertheless, trial participants are often not representative of the patients in actual clinical practice and the duration of follow-up time is limited. Observational studies follow real-life patients, often for a longer duration, but the lack of random allocation may lead to biases in the comparisons between the treatment groups. Confounding by indication is one of those potential biases [17]. In actual clinical practice, there will always be a reason why some patients receive a particular treatment and others do not. Factors that may influence the decision to treat (or not treat) a patient include the stage or severity of the disease, symptoms, concomitant medical conditions, concurrent therapies, past response to other treatments and the age or sex of the patient. The underlying risk profile 1 8 9

7 CHAPTER 1 INTRODUCTION of treated individuals is thus likely to differ from non treated individuals, rendering the two groups incomparable. Possible bias in observational studies is a major concern when pooling together results from separate studies. For example, Egger and co-workers are very skeptical of meta-analyses of observational studies and wrote that the analyst may well be simply producing tight confidence intervals around spurious results [18]. Over the last five years, several large initiatives have been launched to develop methods for combining data from different electronic healthcare registries. In the U.S., the OMOP and the Mini-Sentinel program have been initiated. The OMOP project is a public private partnership among the FDA, academia, data owners, and the pharmaceutical industry [19,20]. The goal is to define and test analytic methods that can be used to explore the relationships between drugs and health-related conditions across multiple types of observational data. The Mini-Sentinel program is a collaboration between the FDA and 31 academic and private organisations [21,22]. It aims to develop methods, tools, resources, policies and procedures to facilitate the use of routinely collected electronic healthcare data to perform active surveillance of the safety of marketed medical products. In Canada, the DSEN and CNODES projects also involve multi-site studies [23]. In Europe, several projects have been set up as well. The EU-ADR Project was launched in 2008 [24,25]. It aims to exploit information from various electronic healthcare record databases in Europe to produce a computerised integrated system for the early detection of drug safety signals. As of 2010, the EU-ADR database platform comprised healthcare data from eight European databases located in four countries. Another European initiative is the PROTECT project, which is coordinated by the European Medicines Agency (EMA) and GlaxoSmithKline [26]. The project consists of a multi-national consortium of 32 partners including academics, regulators and pharmaceutical companies. The goal is to develop a set of innovative tools and methods to enhance the early detection and assessment of adverse drug reactions from different data sources, and to enable the integration of data on benefits and risks. The VAESCO project focuses on vaccine safety information in Europe by standardizing methodologies, facilitating data comparability and building collaborative networks [27]. In addition, one of the goals of the SOS-NSAIDS project is to design and conduct commonprotocol multi-country database studies [28]. Beside these large initiatives, there have been various studies that combined data from several healthcare databases [29-31]. There is a need to know when multi-country observational studies are useful and how they should be performed and interpreted. There are no well-established methods to deal with the combination of data from several independent electronic healthcare registries. As previously discussed, even studies from the same database can provide different results due to different study designs. Hence, sound methodological knowledge is crucial when combining data from different electronic healthcare registries. Specifically, there is a need for methods to deal with differences in study design, inclusion criteria, definitions of exposure and outcome, calendar time and adjustment for confounders. This thesis explores various ways to combine data from three different electronic healthcare registries: - The British General Practice Research Database (GPRD) (which has recently been renamed the Clinical Practice Research Datalink), which comprises patient registration information and all care events that general practitioners (GPs) record as part of their usual medical practice [32]. Currently, information is collected for over 10 million patients under the care of GPs in the United Kingdom (UK). For approximately 45% of all practices, GPRD data have been linked to the national Hospital Episode Statistics (HES) in England, which capture information on in-patient hospital diagnoses [33]. - The Dutch PHARMO Record Linkage System (RLS), which links individual pharmacy dispensing data and hospital diagnoses [34]. Drug dispensing data from pharmacies currently cover more than 3 million residents in the Netherlands. These data are linked to the National Medical Registry (LMR) that comprises information on diagnoses from in-patient hospital admissions and day care admissions. - The Danish National Health Registers, which consist of several electronic healthcare registries linked for the entire population (approximately 5.5 million inhabitants) [35]. The Ministry of the Interior records information on migrations and dates of birth and death. Information on hospital admissions is kept in the National Hospital Discharge Register, which covers in-patient and out-patient hospital diagnoses, as well as diagnoses from day care admissions and emergency rooms. The Danish Medicines Agency keeps a nationwide registry of all prescription drugs sold at pharmacies throughout the country. Further, there are various disease-specific registries, such as the nation-wide Danish Multiple Sclerosis Registry and the Danish Multiple Sclerosis Treatment Registry [36,37]. We focus on two different exposure-outcome associations: (A) the risk of fractures in patients with multiple sclerosis (B) the risk of fractures with thiazolidinediones The background to these associations will be discussed next

8 CHAPTER 1 INTRODUCTION A. Risk of fractures with multiple sclerosis Multiple sclerosis (MS) is a disorder of the central nervous system and is characterised by acute focal inflammatory demyelination and axonal loss, leading to the chronic multifocal sclerotic plaques [38]. In northern Europe, the disease affects approximately 1 person in every 1,000 people [39] and it is the most common cause of neurological disability in young adults [38]. The first symptoms of the disease usually arise between the age of 20 and 40 years and twice as many women as men are affected. Symptoms of the disease vary greatly, from person to person and from time to time in one person. They may include fatigue, imbalance, muscle weakness, uncoordinated movements, mood swings, numbness, blurred vision and diz ziness. Most patients have periods of relatively good health (remissions) alternating with periods of worsening symptoms (relapses). Approximately 80% of patients present with relapsing-remitting MS and, typically, the illness passes through phases of relapse with full recovery, relapse with persistent deficit, and secondary progression [40]. Treatment options for patients with MS include the use of interferon beta, glatiramer acetate, natalizumab, mitoxantrone, and methylprednisolone (during relapses) [38]. Osteoporosis is a skeletal disease characterised by low bone mass and microarchitectural disruption, which leads to bone fragility and hence susceptibility to fracture [41]. It is a serious public health issue that affects many people: about 50% of women and 20% of men older than 50 years will experience a fracture in their remaining lifetime [42]. Patients with osteoporosis most commonly suffer from fractures of the spine, hip or wrist, but other bones such as the humerus or ribs can also be affected [43]. From a patient s perspective, a fracture and the subsequent loss of mobility and autonomy often represent a major drop in quality of life. Hip fractures have a huge impact on the patient s life, because they cause serious disability and carry an approximate 12-month excess mortality of 20% [44]. For patients with low bone mass, the use of bisphosphonates, strontium ranelate, raloxifene and denosumab have been approved for the prevention of fractures [45]. In patients with fractures, parathyroid hormone and analogues are available as anabolic therapy. Patients with MS may have an increased risk of fractures, because they often have osteoporosis [46,47] and they fall more often than the general population [48-50]. The increased risk of falling may be explained by poorer postural balance, impaired vision, disability or spasticity. Osteoporosis in MS patients is probably the result of immobility, a low level of daily activity, vitamin D deficiency, or the inflammatory nature of the disease [51]. In addition, patients with MS may be at an increased risk of fractures due to exposure to glucocorticoids (GCs) [52]. It has previously been demonstrated that patients who receive several courses of highdose GCs (daily dose >15 mg and cumulative exposure >1 g in prednisolone equivalents) have a substantially increased risk of fracture [53]. Patients with MS who experience a relapse are often prescribed a short-course of methylprednisolone with a dosage that corresponds to g prednisone equivalents. On the other hand, patients with advanced MS may have to use a wheelchair possibly lowering the risk of falling. However, the decreased mobility in a later stage of the disease causes a reduced bone mineral density (BMD) elevating the risk of fracture. Thus, the association between MS and risk of fracture may depend on the stage of the disease. The studies that have evaluated the risk of fracture for patients with MS only included smaller number of patients and rates were not compared with control individuals [54-56]. Further, the contributions of falling, disability and the use of GCs in the association between MS and fracture are unclear. B. Risk of fractures with thiazolidinediones Diabetes mellitus (DM) is a group of metabolic diseases characterised by high blood sugar, either because the pancreas does not produce enough insulin or because the body does not adequately respond to the insulin that is produced. In type 1 diabetes mellitus (T1DM), autoimmune-mediated destruction of pancreatic b-cell islets leads to absolute insulin deficiency [57]. In type 2 diabetes mellitus (T2DM), insufficient insulin is produced, the body becomes resistant to normal or even high levels of insulin, or both. In the year 2010, approximately 285 million adults had DM worldwide (6.4% of all people), and this value is predicted to rise to around 439 million (7.7%) by 2030 [58]. T2DM is the predominant form of the disease and accounts for at least 90% of cases [59]. Cardiovascular morbidity in patients with T2DM is two to four times greater than that of non-diabetic people [60]. Management of T2DM involves the control of blood glucose and lipid concentrations, blood pressure, bodyweight and smoking, as well as regular screening for complications [58]. Treatment options for patients with T2DM include biguanides, sulfonylureas, thiazolidinediones (TZDs) and insulin [61]. TZDs have been demonstrated to improve insulin resistance when applied to patients with T2DM. Recently, an association of rosiglitazone with risk of cardiovascular outcomes [62] has led to withdrawal of the drug in Europe and restricted access in the United States, but pioglitazone is still used in the management of T2DM. Several studies have found that TZD use is associated with decreased BMD and elevated risk of fracture [63,64]. One possible explanation is that TZDs affect the differentiation of mesenchymal stem cells, leading to increased adipogenesis at the expense of the formation of osteoblasts [65,66]. Other explanations include the inhibition of the aromatase, and hence the estrogen biosynthesis in adipose tissue and the affection of the insulin-like growth

9 CHAPTER 1 INTRODUCTION factor (IGF) system [67,68]. Not only the use of TZDs, but also the underlying T2DM has been associated with fracture [69,70]. Even though patients with T2DM have on average a higher body mass index (BMI) as well as a higher BMD compared with the general population [70,71], which would in theory protect them against fractures, previous studies have found an increased risk of fractures in T2DM. Possible mechanisms include direct effects of the high glucose levels on bone turnover [72], increased urine calcium loss [73] and changes in vitamin D metabolism [74]. Complications of diabetes, such as neuropathy and angiopathy, may also contribute to an increased risk of fracture [75-77]. Possible mechanisms are a decreased mobility, estrogen deficiency, and high bone turnover due to secondary hyperparathyroidism. If the TZD effect on osteoblastogenesis predominates, the increase in fracture risk should be equal in men and women. If other mechanisms, such as the effect on the aromatase, predominate, women may be at higher risk than men. At the start of our project (in 2008), there was no consensus whether TZDs affected specific types of bones, whether women had greater fracture risks than men and whether the underlying disease confounded the associations between TZD use and fracture risk. Objective of the thesis The overall aim of this thesis was to explore the value of using data from multiple countries when estimating the risk of fractures for patients with MS and the risk of fractures for patients exposed to TZDs. Our specific aims were: - to estimate fracture risks for patients with MS, in relation to falling, disability and exposure to GCs - to estimate fracture risks for patients exposed to TZDs, stratified to fracture type, sex, and proxies of disease severity - to explore the advantages and disadvantages of combining electronic healthcare data from different countries - to describe and determine the impact on the results of differences in study populations, duration of follow-up and recording of outcomes and risk factors between the different electronic healthcare registries - to investigate the differences between a meta-analysis that combines aggregate summary results and one that combines individual patient data. Outline of the thesis Figure 1 schematically displays the registries in which the two associations were studied and the corresponding chapters. Figure 1. General Practice Research Database (UK) PHARMO Record Linkage System (NL) Danish National Health Registers (DK) General Practice PHARMO Record Danish National Research Database (UK) Chapter 2.1 & 2.2 Linkage System (NL) Chapter 2.3 Health Registers (DK) Chapter 2.4 & 2.5 General Practice Research Database (UK) General Practice Research Database (UK) Multiple sclerosis & Multiple Risk of fracture sclerosis & Risk of fracture Chapter 2.1 & 2.2 Chapter 2.3 Chapter 2.4 & 2.5 Chapter 4.1 Chapter 4.1 Thiazolidinediones & Thiazolidinediones Risk of fracture & Risk of fracture PHARMO Record Linkage System (NL) Danish National Health Registers (DK) PHARMO Record Danish National Linkage System (NL) Chapter 3.1 Health Registers (DK) Chapter 3.2 Chapter 3.1 Chapter 3.2 Chapter 4.2 Chapter 4.2 Chapter 2 covers five studies on the risk of fractures in patients with MS. In the first study, the GPRD was utilised to estimate relative and absolute risks of fracture for patients with MS. This was followed by a GPRD study estimating the individual long-term risks of osteoporotic and hip fracture in MS patients. The next study concerned a population-based cohort study in PHARMO RLS, in which relative risks of fracture among patients with MS were estimated. The last two studies in Chapter 2 were based on data from the Danish National Health Registers. Chapter 2.4 analysed incidence rates of fractures for patients with MS stratified by fracture type, sex, and age. In Chapter 2.5, the risk of fracture was investigated for incident

10 CHAPTER 1 INTRODUCTION MS patients from Denmark. In Chapter 3, fracture risks were examined for patients exposed to TZDs. The first study describes how PHARMO RLS was utilised to estimate fracture risks for current users of TZDs and other antidiabetic drugs. In Chapter 3.2, the same study question was investigated using data from the Danish National Health Registers. Chapter 4 focuses on the combination of data from the different electronic healthcare registries. The first study dealt with the association between MS and fracture risk. Results from the GPRD and Danish studies were compared and combined. Chapter 4.2 shows risks of fracture in relation to TZD exposure for patients from PHARMO RLS, the Danish registries and the GPRD, also comparing patient characteristics and combining study results. This thesis concludes with a general discussion of the results. REFERENCES 1. Alvarez-Requejo A, Porta M. Pharmacoepidemiology in practice. Current status and future trends. Drug Saf Jul;13(1): Strom BL. What is pharmacoepidemiology? In: Strom BL, Kimmel SE, editors. Textbook of Pharmacoepidemiology. Philadelphia, PA: John Wiley & Sons, Ltd; pp Sutton AJ, Higgins JP. Recent developments in meta-analysis. Stat Med Feb 28;27(5): Green J, Czanner G, Reeves G, Watson J, Wise L, Beral V. Oral bisphosphonates and risk of cancer of oesophagus, stomach, and colorectum: case-control analysis within a UK primary care cohort. BMJ Sep 1;341:c Cardwell CR, Abnet CC, Cantwell MM, Murray LJ. Exposure to oral bisphosphonates and risk of esophageal cancer. JAMA Aug 11;304(6): Patsopoulos NA, Analatos AA, Ioannidis JP. Relative citation impact of various study designs in the health sciences. JAMA May 18;293(19): Meier CR, Schlienger RG, Kraenzlin ME, Schlegel B, Jick H. HMG-CoA reductase inhibitors and the risk of fractures. JAMA Jun 28;283(24): US Food and Drug Administration FDA (2010). Possible fracture risk with high dose, long-term use of proton pump inhibitors. 11. van Staa TP, Wegman S, de Vries F, Leufkens B, Cooper C. Use of statins and risk of fractures. JAMA Apr 11;285(14): de Vries F, de Vries C, Cooper C, Leufkens B, PostmarketDrugSafetyInformationforPa- van Staa TP. Reanalysis of two studies with tientsandproviders/ucm htm. Ac- contrasting results on the association be- cessed 28 May tween statin use and fracture risk: the Gen- 6. Farmer RD, Lawrenson RA, Todd JC, Williams TJ, MacRae K. Oral contraceptives and ve- eral Practice Research Database. Int J Epidemiol Oct;35(5): nous thromboembolic disease. Analyses of 13. Parmar MK, Torri V, Stewart L. Extracting the UK General Practice Research Database summary statistics to perform meta-analyses and the UK Mediplus database. Hum Reprod of the published literature for survival end- Update Nov-Dec;5(6): points. Stat Med Dec 30;17(24): De Vries F, van Staa TP, Leufkens HG. Proton 34. pump inhibitors, fracture risk and selection 14. Stewart LA, Parmar MK. Meta-analysis of bias: three studies, same database, two an- the literature or of individual patient data: swers. Osteoporos Int 2011 May;22(5):1641- is there a difference? Lancet Feb 2. 13;341(8842):

11 CHAPTER 1 INTRODUCTION 15. Broeze KA, Opmeer BC, van der Veen F, (accessed 3 July 2012) 31. Kieler H, Artama M, Engeland A, Ericsson O, 43. Rachner TD, Khosla S, Hofbauer LC. Osteopo- Bossuyt PM, Bhattacharya S, Mol BW. Individual patient data meta-analysis: a promising approach for evidence synthesis in reproductive medicine. Hum Reprod Update Nov-Dec;16(6): Lyman GH, Kuderer NM. The strengths and limitations of meta-analyses based on aggregate data. BMC Med Res Methodol Apr 25;5: (accessed 3 July 2012) 25. Coloma PM, Schuemie MJ, Trifirò G, Gini R, Herings R, Hippisley-Cox J, Mazzaglia G, Giaquinto C, Corrao G, Pedersen L, van der Lei J, Sturkenboom M; EU-ADR Consortium. Combining electronic healthcare databases in Europe to allow for large-scale drug safety monitoring: the EU-ADR Project. Pharma- Furu K, Gissler M, Nielsen RB, Nørgaard M, Stephansson O, Valdimarsdottir U, Zoega H, Haglund B. Selective serotonin reuptake inhibitors during pregnancy and risk of persistent pulmonary hypertension in the newborn: population based cohort study from the five Nordic countries. BMJ Jan 12;344:d (accessed 3 July 2012) rosis: now and the future. Lancet Apr 9;377(9773): Center JR, Nguyen TV, Schneider D, Sambrook PN, Eis man JA. Mortality after all major types of osteoporotic frac ture in men and women: an observational study. Lancet 1999; 353: TA160publicinfo.pdf (accessed 5 March Signorello LB, McLaughlin JK, Lipworth L, Fri- coepidemiol Drug Saf Jan;20(1): (accessed 3 July 2012) is S, Sørensen HT, Blot WJ. Confounding by indication in epidemiologic studies of commonly used analgesics. Am J Ther May- Jun;9(3): Egger M, Schneider M, Davey Smith G. Spurious precision? Meta-analysis of observational studies. BMJ Jan 10;316(7125): (accessed 3 July 2012) 20. Stang PE, Ryan PB, Racoosin JA, Overhage JM, Hartzema AG, Reich C, Welebob E, Scarnecchia T, Woodcock J. Advancing the science for active surveillance: rationale and design for the Observational Medical Outcomes Partnership. Ann Intern Med Nov 2;153(9): (accessed 3 July 2012) 22. Platt R, Carnahan RM, Brown JS, Chrischilles E, Curtis LH, Hennessy S, Nelson JC, Racoosin JA, Robb M, Schneeweiss S, Toh S, Weiner MG. The U.S. Food and Drug Administration s Mini-Sentinel program: status and direction. Pharmacoepidemiol Drug Saf Jan;21 Suppl 1: (accessed 3 July 2012) html (accessed 3 July 2012) (accessed 3 July 2012) 29. Mokhles MM, Trifirò G, Dieleman JP, Haag MD, van Soest EM, Verhamme KM, Mazzaglia G, Herings R, Luise C, Ross D, Brusselle G, Colao A, Haverkamp W, Schade R, Camp G, Zanettini R, Sturkenboom MC. The risk of new onset heart failure associated with dopamine agonist use in Parkinson s disease. Pharmacol Res Mar;65(3): Trifirò G, Mokhles MM, Dieleman JP, van Soest EM, Verhamme K, Mazzaglia G, Herings R, de Luise C, Ross D, Brusselle G, Colao A, Haverkamp W, Schade R, van Camp G, Zanettini R, Sturkenboom MC. Risk of cardiac valve regurgitation with dopamine agonist use in Parkinson s disease and hyperprolactinaemia: a multi-country, nested case-control study.drug Saf Feb 1;35(2): ) (accessed 3 July 2012) (accessed 3 July 2012) 36. Brønnum-Hansen H, Koch-Henriksen N, Stenager E. The Danish multiple sclerosis registry. Scand J Pub Health 2011; 39: Koch-Henriksen N, Sorensen PS. The Danish Multiple Sclerosis Group. The Danish National Project of interferon-beta treatment in relapsing-remitting multiple sclerosis. Mult Scler 2000; 6: Rejdak K, Jackson S, Giovannoni G. Multiple sclerosis: a practical overview for clinicians. Br Med Bull. 2010;95: Pugliatti M, Sotgiu S, Rosati G. The worldwide prevalence of multiple sclerosis. Clin Neurol Neurosurg Jul;104(3): Compston A, Coles A. Multiple sclerosis. Lancet Apr 6;359(9313): Sambrook P, Cooper C. Osteoporosis. Lancet Jun 17;367(9527): van Staa TP, Dennison EM, Leufkens HG, Cooper C. Epidemiology of fractures in England and Wales. Bone Dec;29(6): Hearn AP, Silber E. Osteoporosis in multiple sclerosis. Mult Scler 2010; 16: Gibson JC, Summers GD. Bone health in multiple sclerosis. Osteoporos Int 2011 Dec;22(12): Nilsagård Y, Lundholm C, Denison E, Gunnarsson LG. Predicting accidental falls in people with multiple sclerosis - a longitudinal study. Clin Rehabil 2009; 23: Peterson EW, Cho CC, von Koch L, Finlayson ML. Injuri ous falls among middle aged and older adults with multiple sclerosis. Arch Phys Med Rehabil 2008; 89: Finlayson ML, Peterson EW, Cho CC. Risk factors for falling among people aged 45 to 90 years with multiple scle rosis. Arch Phys Med Rehabil 2006; 87: Schett G, David JP. The multiple faces of autoimmune-mediated bone loss. Nat Rev Endocrinol 2010; 6:

12 CHAPTER 1 INTRODUCTION 52. Dovio A, Perazzolo L, Osella G, Ventura M, 60. Stumvoll M, Goldstein BJ, van Haeften TW. 68. Lecka-Czernik B, Ackert-Bicknell C, Adamo 75. Nisbeth U, Lindh E, Ljunghall S, Backman Termine A, Milano E, Bertolotto A, Ange- Type 2 diabetes: principles of pathogen- ML, Marmolejos V, Churchill GA, Shockley U, Fellström B. Increased fracture rate in li A. Immediate fall of bone formation and esis and therapy. Lancet Apr 9-15; KR, Reid IR, Grey A, Rosen CJ. Activation of diabetes mellitus and females after renal transient increase of bone resorption in the 365(9467): peroxisome proliferator-activated recep- transplantation. Transplantation 1999; 67(9): course of high-dose, short-term glucocorticoid therapy in young patients with multiple sclerosis. J Clin Endocrinol Metab 2004; 89: De Vries F, Bracke M, Leufkens HG, Lammers JW, Cooper C, Van Staa TP. Fracture risk with intermittent highdose oral glucocorticoid therapy. Arthritis Rheum 2007; 56: Stenager E, Jensen K. Fractures in multiple sclerosis. Acta Neurol Belg 1991; 91: Troiano RA, Jotkowitz A, Cook SD, Bansil S, Zito G. Rate and types of fractures in corticosteroid-treated multiple sclerosis patients. Neurology 1992; 42: Logan WC Jr, Sloane R, Lyles KW, Goldstein B, Hoenig HM. Incidence of fractures in a cohort of veterans with chronic multiple sclerosis or traumatic spinal cord injury. Arch Phys Med Rehabil 2008; 89: Zimmet P, Alberti KG, Shaw J. Global and societal implications of the diabetes epidemic. Nature Dec 13;414(6865): Nolan CJ, Damm P, Prentki M. Type 2 diabetes across generations: from pathophysiology to prevention and management. Lancet Jul 9;378(9786): Gonzalez EL, Johansson S, Wallander MA, Rodriguez LA. Trends in the prevalence and incidence of diabetes in the UK: J Epidemiol Community Health 2009; 63: 61. Tahrani AA, Bailey CJ, Del Prato S, Barnett AH. Management of type 2 diabetes: new and future developments in treatment. Lancet Jul 9;378(9786): Schernthaner G, Chilton RJ. Cardiovascular risk and thiazolidinediones--what do meta-analyses really tell us? Diabetes Obes Metab Dec;12(12): Loke YK, Singh S, Furberg CD. Long-term use of thiazolidinediones and fractures in type 2 diabetes: a meta-analysis. CMAJ 2009; 180: Betteridge DJ. Thiazolidinediones and fracture risk in patients with Type 2 diabetes. Diabet Med Jul;28(7): Grey A. Skeletal consequences of thiazolidinedione therapy Osteoporos Int 19(2): Benvenuti S, Cellai I, Luciani P, Deledda C, Baglioni S, Giuliani C, Saccardi R, Mazzanti B, Dal Pozzo S, Mannucci E, Peri A, Serio M. Rosiglitazone stimulates adipogenesis and decreases osteoblastogenesis in human mesenchymal stem cells J Endocrinol Invest 30(9):RC26 RC Rubin GL, Zhao Y, Kalus AM, Simpson ER. Peroxisome proliferator-activated receptor gamma ligands inhibit estrogen biosynthesis in human breast adipose tissue: possible implications for breast cancer therapy. Cancer Res Mar 15;60(6): tor gamma (PPARgamma) by rosiglitazone suppresses components of the insulin-like growth factor regulatory system in vitro and in vivo. Endocrinology Feb;148(2): Janghorbani M, Van Dam RM, Willett WC, Hu FB. Systematic review of type 1 and type 2 diabetes mellitus and risk of fracture Am J Epidemiol 166(5): Vestergaard P. Discrepancies in bone mineral density and fracture risk in patients with type 1 and type 2 diabetes a meta analysis Osteoporos Int 18(4): Nguyen NT, Nguyen XM, Lane J, Wang P. Relationship between obesity and diabetes in a US adult population: findings from the national health and nutrition examination survey, Obes Surg 2011; 21(3): Williams JP, Blair HC, McDonald JM, McKenna MA, Jordan SE, Williford J, Hardy RW. Regulation of osteoclastic bone resorption by glucose. Biochem Biophys Res Commun 1997; 235(3): McNair P, Madsbad S, Christensen MS, Christiansen C, Faber OK, Binder C, Transbøl I. Bone mineral loss in insulin treated diabetes mellitus: studies on pathogenesis. Acta Endocrinol (Copenh) 1979; 90(3): Bouillon R. Diabetic bone disease. Calcif Tissue Int 1991; 49(3): Vogt MT, Cauley JA, Kuller LH, Nevitt MC. Bone mineral density and blood flow to the lower extremities: the study of osteoporotic fractures. J Bone Miner Res 1997; 12(2): Hadjidakis D, Diamantopoulos E, Kokkinakis E, Sfakianakis M, Merakos G,Raptis SA. Bone mineral density in type II diabetics with macroangiopathy. J Bone Miner Res 1997; 12(Suppl 1): S

13 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS CHAPTER 2

14 2.1 THE RISK OF FRACTURE IN PATIENTS WITH MULTIPLE SCLEROSIS: THE UK GENERAL PRACTICE RESEARCH DATABASE Marloes T. Bazelier Tjeerd-Pieter van Staa Bernard M.J. Uitdehaag Cyrus Cooper Hubert G.M. Leufkens Peter Vestergaard Joan Bentzen Frank de Vries Easy reading is damn hard writing. Nathaniel Hawthorne ( ) Published in: J Bone Miner Res Sep; 26(9):

15 CHAPTER 2 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS ABSTRACT INTRODUCTION Background Patients with multiple sclerosis (MS) may be at an increased risk of fracture owing to a greater risk of falling and decreased bone mineral density when compared with the general population. This study was designed to estimate the relative and absolute risk of fracture in patients with MS. Methods We conducted a population-based cohort study using data from the UK General Practice Research Database linked to the National Hospital Registry ( ). Incident MS patients (n=5565) were matched 1:6 by year of birth, sex, and practice with patients without MS (controls). Cox proportional-hazards models were used to derive adjusted hazard ratios (HRs) for fracture associated with MS. Time-dependent adjustments were made for age, comorbidity, and drug use. Absolute 5- and 10-year risks of fracture were estimated for MS patients as a function of age. Results Compared with controls, MS patients had an almost threefold increased risk of hip fracture [HR=2.79, 95% confidence interval (CI) ] and a risk of osteoporotic fracture that was increased 1.4-fold (HR=1.35, 95% CI ). Risk was greater in patients who had been prescribed oral/intravenous glucocorticoids (GCs; HR=1.85, 95% CI ) or antidepressants (HR=1.79, 95% CI ) in the previous 6 months. Absolute fracture risks were low in younger MS patients but became substantial when patients were older than 60 years of age. Multiple sclerosis (MS) is a neurodegenerative disease characterized by the gradual accumulation of focal plaques of demyelination, particularly in the periventricular areas of the brain. According to the World Health Organization, MS affects more than 1.3 million people worldwide, including 630,000 Europeans, 520,000 Americans, and 66,000 Eastern Mediterranean patients [1]. Patients with MS may be at risk of fracture [2] owing to an increased risk of falling [3 5] and low bone mineral density (BMD) [6 10]. An increased risk of falling may result from imbalance, disability, or spasticity [3,5,11]. Osteoporosis occurs more frequently among patients with MS, probably as a result of immobility [7,10], a low level of daily activity [6,12], vitamin D deficiency [10], and the use of glucocorticoids (GCs) [6,7,13]. Furthermore, it has been shown that patients with rheumatoid arthritis (RA), chronic obstructive pulmonary disease (COPD), or inflammatory bowel disease whose cumulative exposure to GCs was greater than 1 g (in prednisone equivalents) had a substantially increased risk of osteoporotic fracture [14]. MS patients who suffer from a relapse are often prescribed oral or intravenous methylprednisolone (MPH) with a dosage that corresponds to 1.5 to 3.3 g of prednisone equivalents. A recent study showed that in patients with RA, the underlying disease process also may contribute to the fracture risk [15]. In patients with MS, there may be a similar association between disease activity and risk of fracture: Inflammation, which is highest during relapses, may be associated with a reduction in BMD in MS patients [8]. However, in MS patients, individual risk profiles for fracture have not been determined. The aim of this study was to determine the relative risk of fracture in patients with MS compared with population-based controls and to calculate the absolute 5- and 10-year risks of fracture in MS patients. 2.1 Conclusion It is concluded that MS is associated with an increased risk of fracture. Fracture risk assessment may be indicated in patients with MS, especially those prescribed GCs or antidepressants. METHODS Data source Information for this study was obtained from the General Practice Research Database (GPRD). The GPRD comprises prospectively collected computerized medical records for over 10 million patients under the care of general practitioners (GPs) in the United Kingdom (UK). It has been the source for numerous epidemiologic studies, and the accuracy and completeness of these data have been well validated and documented [16]. Previous studies of GPRD data have shown a high level of data validity with respect to the reporting of fractures [17] GPRD data have been linked to the national Hospital Episode Statistics (HES) in England for approximately 45% of all practices [18]. In this study, data were linked from April 1997 until March

16 CHAPTER 2 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS Study population The case population consisted of all patients aged 18 years or older with at least one recorded diagnosis of MS during the period of GPRD or HES data collection. For this study, GPRD data collection started in 1987 and ended in August 2009 and HES data collection started in April 1997 and ended in March 2008 (thereby defining the total study period as 1987 to 2009). Patients with a history of MS before the start of data collection were excluded, thereby restricting the case population to incident MS patients. Patients with MS were stratified to probable and possible cases at baseline (date of first diagnosis/ index date). A probable case was defined as someone with a diagnosis of MS in the GPRD and a diagnosis of MS in the HES or a diagnosis of MS either in the GPRD or the HES and at least one of the following: (1) an MRI within 6 months of the first diagnosis, (2) two or more subsequent MS diagnoses (GPRD), and/or (3) a prescription for either corticotropin, glatiramer, interferon b1a or b1b, or natalizumab at any time during follow-up. All other patients with a diagnosis of MS in the GPRD or the HES were considered possible cases. The case definition of MS has been developed by a senior neurologist (BU) and a senior pharmacoepidemiologist who has 8 years experience with analysis of GPRD (FdV). We defined patients free of osteoporosis treatment at baseline as those who had not been prescribed bisphosphonates, raloxifene, strontium ranelate, and/or parathyroid hormone (PTH) before the start of follow-up. Each MS patient was matched by year of birth, sex, and practice to six control individuals (patients without a diagnosis of MS at any time during their period of registration). The index date of MS diagnosis was the date of the first record of MS after the start of valid data collection. Control patients were assigned the same index date as their matched case. Each patient then was followed from his or her index date to the end of data collection, the date of transfer of the patient out of the practice area, or the patient s death, whichever came first. Study outcomes All patients (cases and controls) were followed up for the occurrence of fractures. The types of fracture were classified according to the International Classification of Diseases (ICD-10) categories. These included skull (S02), neck (S12), ribs (S22), pelvis (S32), shoulder (S42), forearm (S52), hand (S62), hip/femur (S72), ankle (S82), foot (S92), or unspecified fractures (T02, T08, T10, T12). A clinical osteoporotic fracture was defined as a fracture of the radius/ ulna, vertebrae, femur, hip, humerus, pelvis, or ribs. The total follow-up period was divided into 30-day intervals. The presence of risk factors was assessed by reviewing the computerized medical records for any evidence of risk factors before the start of an interval. Potential confounders that were determined at baseline included sex, body mass index (BMI), smoking status, alcohol use, history of falling, and history of fracture. Potential confounders that were determined for a timedependent analysis included age, a history of chronic diseases (congestive heart failure, RA, cerebrovascular disease [19], inflammatory bowel disease, dementia, depression, epilepsy, and COPD[20]), as well as evidence of fatigue or visual disturbances 6 months before the start of a 30-day interval. A prescription for orally or intravenously administered GCs [14,21], statins [22], antiarrythmics, antidiabetics, antidepressants [23], antipsychotics [24], hypnotics/anxiolytics, asthma medication, anticonvulsants, hormone-replacement therapy, vitamin D, levothyroxine, baclofen, or opioids (potencies equivalent to tramadol or higher) in the previous 6 months also were considered as potential confounders. The prescription of a medication was used as a proxy for use of that drug. In the United Kingdom, exacerbations of MS are treated with a short course of orally or intravenously administered MPH [25]. Prescriptions can be issued by the GP, or the patient may be referred to an MS clinic for intravenous MPH on 3 consecutive days. Information on MPH exposure during MS relapses was retrieved from anonymous free text, as recorded by the GP or by the consultant neurologist, in a discharge summary or letter to the GP. We searched all free-text records of each MS patient with the following search terms: methylpredniso, methyl-predniso, methyl predniso, solumedrol, solu medrol, solu-medrol, ivmp, medrone, ivmp, booster, bolus, pulse ther, and steroid pulse. Records then were made anonymous, and the date of administration, MPH dose, and route of administration were noted. Average daily dose and cumulative dose of oral/ intravenous GCs were determined for the 6-month period before the start of an interval. The average daily dose was defined as the total dose of GCs (in prednisolone equivalents) that was prescribed in the previous 6 months divided by the number of days. In the analysis to derive absolute risk, additional specific risk factors were considered, including vertigo, dizziness, imbalance, disturbance of sensation, spasticity, sexual dysfunction, paroxysmal symptoms, cognitive dysfunction, vitamin D deficiency, and proxy indicators of increased disability (i.e., home visits by the GP, nursing care, and patient receiving residential care/living in a care home or using a wheelchair or walking aid) 6 months before the start of an interval. Furthermore, the use of nonsteroidal anti-inflammatory drugs (NSAIDs), meprobamate, tizanidine, dantrolene, modafinil, methylphenidate, or amantadine in the previous 6 months was included in this analysis. To create a cohort of patients unexposed to osteoporosis treatment at baseline, medication records of all participants of the study (MS cases and controls) were searched for prescriptions of bisphosphonates, raloxifene, strontium ranelate, and PTH

17 CHAPTER 2 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS Statistical analysis We compared baseline characteristics between cases and controls using chi-square testing. Two main analyses were conducted using Cox proportional-hazards models. The first analysis compared the fracture risk in MS patients with that in control patients to yield an estimate of the relative risk [as a hazard ratio (HR)] of fracture in MS. In that analysis, the calculations were adjusted for all potential confounders that changed the HR more than 1% in an age/sexadjusted analysis. Analyses were stratified to duration of disease, drug use in the previous 6 months (including daily and cumulative exposure to GCs), falling 3 months to 1 year before, and a history of fatigue and the disability proxy indicator in the previous 6 months. In addition to the full cohort analysis, we calculated HRs in an osteoporosis-treatment-naive cohort, where we restricted the analysis to MS patients and controls free of osteoporosis treatment at baseline. In the second analysis, we calculated the long-term risk of osteoporotic and hip fracture using the full cohort. The Cox model allows calculation of an individual s probability of fracture (i.e., survivor function) for each set of patient characteristics. For the analysis of long-term risk, we fitted the regression model with the general and specific risk factors, which were determined at baseline, using forward selection. All characteristics, except age, were included as categorical variables in the regression models. For the variables of age, sex, and each of the risk factors, we also investigated possible statistical interactions with MS, although none was added subsequently to the model. The 5- and 10-year risks of osteoporotic and hip fracture then were estimated, conditional on patient survival. We evaluated risks for MS patients in different sex and age categories using their individual risk profiles. In addition, we estimated risks with GC or antidepressant use, adding the corresponding risk factor to the model. Various methods were used to test the fitting of the Cox models, including a test of the proportional-hazards assumption. We also compared the observed 5-year probability of fracture (based on the Kaplan-Meier estimate) with the probability predicted by the Cox model. To assess the internal validity of the model further, the C-statistic was calculated, and we performed a 10-fold cross-validation. We applied the shrinkage factor that we found to the b coefficients of the model, and we adjusted the C-statistic for overestimation. A sensitivity analysis was conducted to assess the impact of defining the index date as the first MS diagnosis 1 year after start of follow-up. In another sensitivity analysis, we excluded all probable MS patients who were classified as probable based on their MRI. All data management and statistical analyses were conducted using SAS Version 9.1/9.2 software (SAS Institute, Inc., Cary, NC, USA). RESULTS The study population included 5565 patients with MS and 33,360 population-based controls. The mean age at index date (first diagnosis of MS) was 44.8 years, and 70.0% of the MS patients were female. The mean duration of follow-up after the index date was 5.7 years for the MS patients and 6.0 years for the control individuals. Before the index date, 32.2% of the MS patients had been prescribed at least one antidepressant, 10.0% at least one anticonvulsant, and 13.2% at least one oral/intravenous GC. In control patients, these proportions were 21.2%, 2.5%, and 6.4%, respectively. Fractures were recorded during follow-up in 394 MS patients (7.1%) and 1742 control individuals (5.2%). Of all MS patients, 3386 were classified as probable and 2179 as possible cases at baseline. Further descriptive details of the participants are given in Table 1. Table 2 shows that patients with MS had a 1.2-fold increased risk of any fracture: adjusted HR=1.23 [95% confidence interval (CI) ]. The risk of osteoporotic fracture was increased 1.4-fold (HR=1.35, 95% CI ), and the risk of hip fracture was increased almost 3-fold (HR=2.79, 95% CI ). The risk of vertebral or radius/ulna fracture was not increased. When we compared probable MS patients with possible MS patients, we found that the risk of osteoporotic fracture was higher in probable patients: HR=1.46 (95% CI ) in probable and HR=1.14 (95% CI ) in possible MS patients, although the difference was not significant (p=.152). There was a significant different risk of hip fracture between probable MS patients (HR=3.75, 95% CI ) and possible MS patients (HR=1.64, 95% CI ). Because the risks of fracture in probable patients were comparable with the risks in all MS patients, we performed the subsequent analyses in the full MS cohort, thereby increasing the number of patients. The risks in the osteoporosistreatment-naive cohort (5494 MS patients, 32,669 controls) were similar to those in all MS patients. Table 3 shows that the risk of osteoporotic fracture with any GC use in the previous 6 months was increased (HR=1.85, 95% CI ). The risk was doubled in patients who had recently been exposed to daily dosages of 7.5mg of prednisolone equivalents or more compared with control patients (HR=2.35, 95% CI ). In addition, we found a similar increase in risk of osteoporotic fracture when we evaluated exposure to cumulative doses of 1 g or more of prednisolone equivalents in the previous year (HR=2.35, 95% CI ; data not shown). The fracture risk with recent GC use was higher in patients who had been prescribed a short course in the previous year (HR=2.67, 95% CI ), the risk being greater with orally administered GCs (HR=4.65, 95% CI ) than with intravenously administered GCs (HR=1.39, 95% CI ; data not shown). In patients who had been treated recently with antidepressants, the risk of osteoporotic fracture was almost doubled

18 CHAPTER 2 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS Table 1. Baseline characteristics of MS patients and controls Characteristic MS patients (%) Controls (%) n=5,565 n=33,360 Female 3897 (70.0) (70.0) Age at index date Mean By category (12.0) 4018 (12.0) (25.9) 8663 (26.0) (29.1) 9700 (29.1) (19.2) 6391 (19.2) (13.8) 4588 (13.8) Mean duration of disease (years) Smoking Never 2135 (38.4) (45.5) a Current 1543 (27.7) 7173 (21.5) a Ex-smoker 815 (14.6) 4287 (12.9) a Unknown 1072 (19.3) 6723 (20.2) BMI < (8.0) 1957 (5.9) a (34.6) (34.1) (23.7) 8803 (26.4) a > (16.0) 5476 (16.4) Unknown 990 (17.8) 5740 (17.2) Disease history Fracture 829 (14.9) 4498 (13.5) a Falling 362 (6.5) 995 (3.0) a Fatigue 441 (7.9) 1752 (5.3) a Asthma 582 (10.5) 3478 (10.4) COPD 56 (1.0) 310 (0.9) Congestive heart failure 33 (0.6) 157 (0.5) Diabetes mellitus 157 (2.8) 854 (2.6) Rheumatoid arthritis 34 (0.6) 234 (0.7) Cerebrovascular incident 158 (2.8) 390 (1.2) a Epilepsy 132 (2.4) 448 (1.3) a History of drug use Statins 277 (5.0) 1278 (3.8) a Antiarrythmics 32 (0.6) 163 (0.5) Antidiabetics 130 (2.3) 684 (2.1) Antidepressants 1794 (32.2) 7066 (21.2) a Antipsychotics 226 (4.1) 953 (2.9) a Anxiolytics/hypnotics 1187 (21.3) 5048 (15.1) a Anticonvulsants 558 (10.0) 823 (2.5) a Opioids 386 (6.9) 1217 (3.6) a Oral / i.v. glucocorticoids 737 (13.2) 2132 (6.4) a Abbreviations: CI, confidence interval; BMI, body mass index; COPD, chronic obstructive pulmonary disease; i.v., intravenous. Table 2. Risk of fracture in MS patients compared with controls, by type of fracture Full cohort analysis Osteoporosis-treatment-naïve analysis MS patients: n=5,565; Controls: n=33,360 MS patients: n=5,494; Controls: n=32,669 Age-sex adj. HR Fully adj. HR (95% CI) (95% CI) Rate /1000 person-years Fully adj. HR (95% CI) Fracture, n= Age-sex adj. HR (95% CI) Rate /1000 person-years Fracture, n= No MS MS Any fracture ( ) 1.23 ( ) a ( ) 1.22 ( ) a Osteoporotic j ( ) 1.35 ( ) b ( ) 1.31 ( ) b Hip ( ) 2.79 ( ) c ( ) 3.05 ( ) g Vertebral ( ) 0.94 ( ) d ( ) 0.93 ( ) h Radius/ulna ( ) 1.16 ( ) e ( ) 1.12 ( ) e Other ( ) 1.27 ( ) f ( ) 1.19 ( ) f Abbreviations: adj, adjusted; HR, hazard ratio; CI, confidence interval. a) Adjusted for (i) and the use of opioids in the previous 6 months, history of cerebrovascular disease, epilepsy b) Adjusted for (i) and the use of opioids in the previous 6 months, history of cerebrovascular disease, epilepsy, BMI c) Adjusted for (i) and the use of opioids in the previous 6 months, history of fatigue in the previous 6 months, BMI d) Adjusted for age, sex, the use of oral / intravenous glucocorticoids, antidepressants, opioids in the previous 6 months, history of fracture at index date e) Adjusted for (i) and history of epilepsy, history of visual disturbance in the previous 6 months f) Adjusted for (i) and the use of opioids in the previous 6 months, history of cerebrovascular disease, epilepsy g) Adjusted for age, sex, the use of oral / intravenous glucocorticoids, antidepressants, anticonvulsants, opioids in the previous six months, history of falling at index date, history of fracture at index date, history of smoking, BMI h) Adjusted for age, sex, the use of oral / intravenous glucocorticoids, antidepressants in the previous 6 months, history of fracture at index date i) age, sex, the use of oral / intravenous glucocorticoids, antidepressants, hypnotics/anxiolytics, anticonvulsants in the previous 6 months, history of falling at index date, history of fracture at index date, history of smoking j) The numbers in the subcategories of osteoporotic fracture do not add up precisely, because a patient can sustain >1 fracture -and therefore different types of fracturesat the same date 2.1 a) Statistically significant difference (p<0.05) between MS patients and controls, based on chi-square test

19 CHAPTER 2 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS Table 3. Risk of osteoporotic fracture in MS patients compared with all controls, by history of drug use & disease indicators Full cohort analysis Osteoporosis-treatment-naïve analysis MS patients: n=5,565; Controls: n=33,360 MS patients: n=5,494; Controls: n=32,669 (%) Fully adj. HR (95% CI) a (%) Fully adj. HR (95% CI) a Fracture, Fracture, n= n= No MS MS Osteoporotic fracture ( ) ( ) Duration of disease <1year ( ) ( ) 1-5 years ( ) ( ) >5years ( ) ( ) History of drug use in previous 6 months Antidepressants Yes ( ) b ( ) No ( ) ( ) Antipsychotics Yes ( ) ( ) No ( ) ( ) Hypnotics / anxiolytics Yes ( ) ( ) No ( ) ( ) Anticonvulsants Yes ( ) ( ) No ( ) ( ) Oral / i.v. glucocorticoids No use ( ) ( ) Any use ( ) ( ) By average daily dose (mg prednisolone equivalents) < 7.5 mg ( ) ( ) >= 7.5 mg ( ) ( ) History of falling (3 months 1 year before) Yes ( ) ( ) No ( ) ( ) History of MS indicators 6 months before Fatigue Yes ( ) ( ) No ( ) ( ) Disability proxy Yes ( ) ( ) No ( ) ( ) Abbreviations: adj, adjusted; HR, hazard ratio; CI, confidence interval; i.v., intravenous. a) Adjusted for age, sex, the use of oral / intravenous glucocorticoids, antidepressants, hypnotics/anxiolytics, anticonvulsants, opioids in the previous 6 months, history of falling at index date, history of fracture at index date, history of cerebrovascular disease, epilepsy, history of smoking, BMI b) Statistically significant difference (p<0.05) between MS patients with a history of medication use, and MS patients unexposed to the same class of medication, based on Wald test

20 CHAPTER 2 RISK OF FRACTURE WITH MULTIPLE SCLEROSIS Table 4. 5-year risk of fracture in MS patients at the 5th, 50th, and 95th percentiles of risk profiles Osteoporotic fracture Hip fracture 5th 50th 95th 5th 50th 95th Women Men Figure and 10-year risks of osteoporotic and hip fracture (%) in MS patients in relation to age. This also was apparent for MS patients with a record of disability in the previous 6 months. For patients with a record of falling 1 year to 3 months earlier, the HR was 2.23 (95% CI ). No association between the duration of disease (from the first MS diagnosis) and risk of fracture was apparent. In the osteoporosis-treatment-naive analysis, we found similar risks as in the full cohort analysis. Figure 1 displays the 5- and 10-year risks of osteoporotic and hip fractures (percentages) in MS patients as a function of their age. In addition, separate curves were added for the use of oral/ intravenous GCs or antidepressants in the previous 6 months. The distribution of the 5-year fracture risks in MS patients by sex and age categories is shown in Table 4. For example, among women aged 70 to 79 years, the median 5-year risk of osteoporotic fracture was 9.0%. However, there was considerable variation in the risk of osteoporotic fracture in this age range because the risk was 6.2% for the women in the 5th percentile of the risk profile and 22.3% for the women in the 95th percentile of the risk profile. The C-statistic was moderate (0.69) for the prediction of osteoporotic fracture and excellent (0.89) for the prediction of hip fracture. Compared with our primary analysis (Table 2), a sensitivity analysis yielded similar results when we used a lead-in time of 1 year after the start of data collection. Defining the index date as the first MS diagnosis after 1 year of data collection reduced the number of MS patients to n=4339. The fully adjusted HR for any fracture was 1.22 (95% CI ); for osteoporotic fracture, it was 1.40 (95% CI ); and for hip fracture, the HR was 2.75 (95% CI ). Exclusion of all probable MS patients who were classified as probable 36 37