Canine neoplasia in the UK: estimates of incidence rates from a population of insured dogs

Transcription

1 PAPERS Canine neoplasia in the UK: estimates of incidence rates from a population of insured dogs Neoplasia is common in pet dogs but accurate figures for the incidence of tumours in this, as in other species, are sparse. The purpose of this study was to document the occurrence of tumours in a defined population of dogs. From a database of 130,684 insured dogs, claims relating to the investigation or treatment of tumours or tumour-like lesions during a 12-month period were accessed and followed up. A total of 2546 claims were tumour related and were classified according to tumour site and type. Because the demographics of the insured population were skewed towards younger animals, a standard population, as described in the veterinary literature, was used in the calculation of tumour incidence rates. The skin and soft tissues were the most common sites for tumour development, with a standardised incidence rate of 1437 per 100,000 dogs per year, followed by alimentary (210), mammary (205), urogenital (139), lymphoid (134), endocrine (113) and oropharyngeal (112). Canine cutaneous histiocytoma was the most common single tumour type, with a standardised incidence rate of 337 per 100,000 dogs per year, followed by lipoma (318), adenoma (175), soft tissue sarcoma (142), mast cell tumour (129) and lymphosarcoma (114). These data are unique and provide a valuable basis for future research into the aetiology and epidemiology of canine tumours. J. M. DOBSON, S. SAMUEL, H. MILSTEIN, K. ROGERS* AND J. L. N. WOOD* Journal of Small Animal Practice (2002) 43, Department of Clinical Veterinary Medicine, University of Cambridge, Cambridge CB3 0ES *Epidemiology Unit, Animal Health Trust, Newmarket, Suffolk CB8 7UU INTRODUCTION Cancer is an important disease in pet dogs. Recent studies have shown it to be one of the major causes of death in insured dogs in both the UK and Sweden (Bonnett and others 1997, Michell 1999). There is a large amount of clinical information on the diagnosis and therapy of different types of neoplasia and some information on risk factors for different cancers (Reif and others 1995, Richards and others 2000). However, there is little up-to-date information available on the frequency or incidence of different types of cancer in the canine population anywhere in the world. Most previous epidemiological studies of canine cancer have been based on veterinary teaching hospital populations or on surveys of intakes into pathology laboratories (Goldschmidt and Shofer 1992). The Veterinary Medical Data Program, instigated in 1964 by the National Cancer Institute and based at Michigan State University, is one of the largest studies reported to date (Priester and McKay 1980); however, this was a hospital-based study in which participating veterinary schools submitted coded medical information on all cases seen to a central unit. Very few studies have attempted to estimate population-based rates of disease and most of these have been in North America. The major study on the incidence of tumours in dogs and cats was conducted from 1963 to 1967 in the Alameda and Contra Costa counties of California (Dorn and others 1968a,b). This was based on a centralised histological examination of all cases of malignant neoplasia from veterinary practices in the state, using a survey to estimate the population at risk. A defined population of dogs and cats were studied in Oklahoma and those with tumours were entered into the Tulsa Registry of Canine and Feline Neoplasms (MacVean and others 1978). Lengerich and others (1992) used a one-stage random digit telephone survey to establish population statistics and ascertain the prevalence of cancer cases in the principal catchment area of the Purdue Comparative Oncology Program animal tumour registry in Indiana. More recently, a group in Guelph have used electronic clinical records in five practices in Ontario to estimate the incidence of cancer in companion animals (Reid-Smith and others 2000). The use of computerised data to study health and disease in companion animals has been suggested by several authorities (Thrusfield 1983, Willeberg 1986). In this study data from a pet insurance company was used to estimate the incidence of neoplasia in a defined population of insured dogs in the UK. 240 JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE 2002

2 Table 1. Classification of tumours by histological type Broad tissue category Malignancy Specific histology (examples) Epithelial Malignant Squamous cell carcinoma Basal cell carcinoma Transitional cell carcinoma Adenocarcinoma MATERIALS AND METHODS Population at risk and identification of cases of neoplasia All dogs insured with a single UK pet insurance company between June 1, 1997 and May 31, 1998 were included in this study. The total number, age and breed distribution of these animals was estimated by a cross-sectional evaluation of all dogs Adenoma (sebaceous, perianal, mammary) Basal cell tumour Naevus Mesenchymal Malignant Soft tissue sarcoma Fibrosarcoma Haemangiosarcoma Haemangiopericytoma Liposarcoma Neurofibrosarcoma Fibroma (epulis = peripheral odontogenic fibroma) Lipoma Bone Malignant Osteosarcoma Haematopoietic Malignant Lymphosarcoma (including histiocytic) Leukaemia Myeloma Histiocytosis (malignant/systemic) Plasmacytoma Histiocytoma (canine cutaneous) Nervous tissue Malignant Chemodectoma Glioma Astrocytoma Neuroma Other Malignant Mast cell tumour Not specified Melanoma (oral) Melanoma (cutaneous) Sertoli cell tumour Granulosa cell tumour Seminoma Leydig cell tumour Interstitial cell tumour insured on June 1, The charges for insuring the dogs were determined and paid on an annual premium that varied depending on the different policies. The policies all provided cover for non-routine healthcare and paid for all treatment costs, less a fixed amount of around 30 for each incident. Data on all claims for veterinary treatment were held on a computer database Table 2. Classification of tumours by body site Broad site categories Skin and soft tissue Oral/pharyngeal Nasal/respiratory Alimentary Lymphoid Bone CNS Mammary Urogenital Endocrine CNS Central nervous system Examples of sites included Head Trunk Limb Perianal Gingiva Tongue Pharynx Nares Nasal cavity Larynx Trachea Lung Stomach Intestine Rectum/anus Liver/spleen Pancreas Lymph nodes Bone marrow Appendicular skeleton Axial skeleton Skull Spine Brain Spinal cord Eye Mammary Kidney Bladder Urethra Prostate Testicle Ovary Uterus Vagina Adrenal Pituitary Thyroid against the identification of each dog. Each claim was accompanied by veterinary certification which included the diagnosis or nature of the claim. All claims relating to the investigation or treatment of tumours FIG 1. Age structure of the insured population, compared to that of a normal UK dog population (Thrusfield 1989) and the overall agespecific incidence of cancer in this study JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE

3 Table 3. Age-standardised incidence rates of canine neoplasia by site and tumour type Site Tumour type* Standardised incidence rate (/100,000 dogs/year) Lower CI Upper CI Skin and soft tissue Histiocytoma Lipoma Adenoma Mast cell tumour Soft tissue sarcoma Other* Subtotal Alimentary Haemangiosarcoma Carcinoma Adenocarcinoma Haemangioma Adenoma Other* Subtotal Mammary Carcinoma Adenocarcinoma Adenoma Mixed mammary Other* Subtotal Urogenital Seminoma Sertoli cell tumour Other* Subtotal Lymphoid Lymphosarcoma Leukaemia Other* Subtotal Endocrine Cushing s syndrome Carcinoma Other* Subtotal Oral/pharyngeal Epulis Melanoma Other* Subtotal Bone Osteosarcoma Bone sarcoma Chondrosarcoma Other* Subtotal Nasal/respiratory Carcinoma Adenocarcinoma Other* Subtotal CNS Not specified Subtotal Other Chemodectoma Haemangiosarcoma Other* Subtotal * Including not specified Cardiac tumour CI Confidence interval broad and specific levels of classification in each category as shown in Tables 1 and 2. Where histological investigations had not been performed, the tumour type was defined as unknown, with the exception of bone tumours, most of which were diagnosed by radiography as osteosarcoma, and discrete fatty masses, which were diagnosed clinically as lipoma. Statistical analysis A crude annual incidence was estimated from the number of claims classified as neoplastic divided by the number of animals at risk during the year. Estimates were subdivided by site, histological type and malignancy and breed-specific incidence rates were also derived. Crude estimates of cancer incidence were age-standardised using the published age structure of a normal UK dog population (Thrusfield 1989) as an external standard. Standard methods for external standardisation were used to adjust incidence estimates for the effects of age (Rothman 1986). Other than where clearly stated, all incidence rates presented in this paper are age standardised. All computations were performed using SAS (SAS Institute). RESULTS or tumour-like lesions between June 1, 1997 and May 31, 1998 were accessed and examined. Where possible, each tumour was classified from the information provided with the claim by anatomical site and histological type, along with animal signalment. This information was recorded in a spreadsheet. Where the information available was insufficient to classify a claim as neoplastic, or where it was incomplete with respect to histological type or site, the veterinary surgeon was contacted and asked for further information. In cases of repeat claims for treatment of the same tumour in one animal, only one tumour event was recorded. If one animal was diagnosed with more than one tumour type during the period of study, these were recorded as separate incidents. The diagnosis supplied by the veterinary surgeon on the claim form (or on follow-up) was assumed to be correct. Cases where the diagnosis was uncertain (for example, suspected tumour of ) and could not be substantiated were not included in the study, neither were claims for warts or papillomas. Tumour classification For the purpose of further analysis, each tumour was classified according to its histological type and anatomical site, with A total of 3242 claims were initially identified as being potentially tumour related. In 1989 of these claims it was possible to classify the tumour on the basis of the details contained in the claim. The remaining 1253 cases were followed up by contacting the veterinary surgeon and 557 of these cases were subsequently included and classified at least by general categories on the basis of the additional information. Thus, a total of 2546 claims were identified as being related to neoplasia. The population at risk was estimated to be 130,684 dogs, giving a crude overall incidence estimate of 1948 cases per 100,000 dogs per year. The age-specific incidence rate was relatively low in younger animals but increased 242 JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE 2002

4 Table 4. Age-standardised incidence rates of canine neoplasia by specific site General site Specific site Standardised incidence rate (/100,000 dogs/year) Lower CI Upper CI Skin and soft tissue Limb Perianal Multiple Ear Head Chest Other* Subtotal Alimentary Liver/spleen Intestine Anorectal Pancreas Other* Subtotal Mammary Mammary Other* Subtotal Urogenital Testicle Bladder Prostate Other* Subtotal Lymphoid Lymphoma/lymphosarcoma Leukaemia Lip, tongue and not specified oral Multiple Other* Subtotal Endocrine Cushing s syndrome Adrenal Thyroid Other* Subtotal Oral/pharyngeal Gingiva Lip, tongue and not specified oral Pharynx Other* Subtotal Bone Limb Mandible Other* Subtotal Nasal/respiratory Nose Lung Multiple Subtotal Central nervous system Brain Eye Spinal cord Subtotal Other Heart Subtotal * Including not specified In most cases specific site not available CI Confidence interval other body systems had lower rates still, as shown in Table 3. Canine cutaneous histiocytoma (CCH) was the most common tumour, with a standardised incidence rate of 377 per 100,000 dogs per year, followed by lipoma, adenoma, soft tissue sarcoma, mast cell tumour and lymphosarcoma. While histiocytoma, lipoma and mast cell tumour were virtually exclusive to the skin and soft tissues, adenocarcinoma was reported in the mammary gland, alimentary tract and nasal/respiratory tracts. When tumour site was examined in detail (Table 4), some variation in the frequency of tumours within the broad locations was evident; for example, the testicle was the most common site for tumour development within the urogenital system and the liver/spleen within the alimentary system. In terms of tissue of origin, epithelial tumours were slightly more common than mesenchymal tumours (714 versus 610 per 100,000 dogs per year). Overall, it was not possible to distinguish benign from malignant in 30 per cent of the tumours. Of those that were classified, benign tumours were more common than malignant (Table 5) and there was considerable variation in the proportions of benign and malignant tumours at different sites. sharply after the age of six to a peak of 81 per 100,000 dogs per year in dogs aged 10 years, then decreased in animals older than this (Fig 1). Comparison of the age distribution of the insured population with that of a normal population (Thrusfield 1989) showed that young animals were heavily overrepresented (Fig 1). As shown, the incidence of neoplasia increased markedly with age, and so overall population estimates of cancer incidence were likely to be highly biased. Thus, all crude incidence rates were age standardised, using the age structure of the published normal population (Thrusfield 1989) as the external standard. The age-standardised, overall incidence rate for neoplasia in this population was 2671 cases per 100,000 dogs per year. The incidence rates of neoplasia, categorised by histological type within sites, are shown in Table 3. The skin and soft tissues were by far the most common sites for tumour development, with a standardised incidence of 1437 per 100,000 dogs per year. This was seven times more common than tumour development in the mammary gland, which was associated with a standardised incidence of 205 cases per 100,000 dogs per year. The urogenital, lymphoid, endocrine, alimentary and DISCUSSION This study was based on a population of insured dogs which is likely to differ in some respects from the UK dog population as a whole. In the former, there is likely to be a higher proportion of purebred dogs with different levels of neutering. In the population studied, there was an overrepresentation of younger animals (Fig 1). While adjustments were made for the overall age structure of the population in the estimates of incidence given here, it was not possible to adjust for the effects of breed or type. It is clear from other studies (eg, Richards and others 2000), as well as from this dataset (Wood and others 2000), JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE

5 Table 5. Relative proportions of malignant and benign neoplasms Malignant Not Incidence rate (%) (%) specified (/100,000 (%) dogs/year) Skin and soft tissue Alimentary Mammary Urogenital Lymphoid Endocrine Oral/pharyngeal Bone Nasal/respiratory Central nervous system All tumour sites that different breeds are variously predisposed to different tumour types and so the effects of breed on the overall estimates of incidence in this study were not clear. However, the authors are unaware of other studies which provide sufficient demographic information to allow direct standardisation and comparison with other studies of cancer epidemiology. The major limitation of this study lies in potential misclassification of cases/ tumours due to the method of data collection. In this study no attempt was made to review or validate the diagnosis entered on the claim form by the veterinary surgeon in charge of the case. It is possible that some tumours were incorrectly diagnosed, either because they were not submitted for histological diagnosis or because of errors in sample collection and/or histological interpretation. A relatively high rate of histological diagnosis would have been achieved in this study as the insurance policies covered histology fees. However, histological examination was not performed in every case. For example, tumours of bone were often diagnosed by radiography without biopsy, Cushing s disease (hyperadrenocorticism) would have been diagnosed by endocrinological testing and evidently many mammary tumours were not sent for histological examination (hence the high incidence of not specified malignancy at the latter two sites). The fact that claims for suspected tumours were not included unless the tumour diagnosis could be confirmed through follow-up may have resulted in an under-estimation of incidence as insufficient information was available on a number of animals to allow their inclusion as cases. Care was taken to only consider types of neoplasia that were likely to have resulted in an insurance claim. It was for this reason that claims for wart removal or papilloma were excluded from the study. For the reasons outlined above, care should be taken in the interpretation of the results and the 95 per cent confidence limits around every incidence figure should be taken into account when considering the likely true incidence. Histological classification of tumours is not always straightforward and while standard systems of classification have been published (WHO 1974, 1976), some of these are outdated and few are adhered to universally by different clinicians or histopathologists. Therefore, a broad system of tumour classification based on tissue of origin, malignancy and actual diagnosis was used in this study. Similar methods were employed by Misdorp (1990); however, the present study classifies the tissue origin of mast cell tumours and melanoma as other and includes the histiocytic tumours (CCH and forms of histiocytosis) within the haematopoietic category. The majority of tumours (74 per cent) were classified at the precise level, although some tumour types were grouped in order to allow interpretation of the data. For example, adnexal tumour was chosen to cover a variety of different tumour types arising from hair follicles, sweat and sebaceous glands associated with the skin (such as trichoepitheliomas, pilomatricomas and sebaceous epitheliomas) and soft tissue sarcoma was chosen to include all malignant mesenchymal tumours (eg, fibrosarcomas) as is conventional in human oncology (Enzinger and Weiss 1995). It was impossible to distinguish accurately between benign and malignant neoplasms in every case. In some instances, this was due to the lack of a precise histological diagnosis (eg, mammary tumours and Cushing s disease, as above). In others, such as cutaneous melanomas, mast cell tumours and Sertoli cell tumours, it was not possible to predict the behaviour of the tumour from the histological diagnosis alone because factors such as tumour grade and location were needed. In cases where such information was not available, tumours were classified as not specified at this level. Variation in rates of detection and reporting of tumours at different sites might influence the findings of this type of study. The results presented showed that the skin and soft tissues were by far the most common site for the development of tumours in this dog population. While undoubtedly important, there may be some bias in reporting because tumours arising in the skin and soft tissues are more readily detectable than tumours arising at deeper sites, which may only be apparent at (infrequently performed) postmortem examinations. It is difficult to make direct comparisons between this and other epidemiological studies because of differences in the population at risk and in methods of tumour classification. The total incidence of neoplasia in the Californian dog population studied in the 1960s (Dorn and others 1968b) was 1134 cases per 100,000 dogs per year, of which were malignant. More recently, Reid-Smith and others (2000) reported an estimated incidence rate of 4817 tumours per 100,000 dog years (including 852 malignant) in southern Ontario. The figures from the present study (2671 total tumours and malignant tumours per 100,000 dogs per year) are intermediate between these two studies. There are many factors, other than an actual increase in the incidence of neoplasia in dogs over the past 30 years, that might account for some of these differences. Only 13 per cent of the dogs in the Californian study were nine years of age or older, whereas in the population to which this study was standardised (Thrusfield 1989), the proportion was nearly 20 per cent. The proportion of animals in these 244 JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE 2002

6 older age groups strongly influences the overall estimate of incidence; in the present study, around one-third of all tumours occurred in animals over nine years. The results of the Canadian study were not standardised for age and, as this was a veterinary practice-based study, it is likely that the age structure of the population of dogs was quite different. Neither of these studies provide sufficient age-specific data to allow direct standardisation and comparison with the results of this study. The breed distribution was not reported in either the Canadian or Californian studies, but this can also strongly influence cancer incidence. It is also likely that the rate of tumour diagnosis has increased over the past 30 years as a result of advances in veterinary practice and higher expectations of pet owners. The major proportional trends, in terms of tumour sites, are quite similar to those reported by Dorn and others (1968b). The skin and soft (connective) tissues were the predominant site in both studies and the proportions of tumours at other sites were not dissimilar. One notable difference is in the incidence of malignant melanoma of the skin, a major category in the Californian dogs but an infrequent diagnosis in the UK population. While it is tempting to suggest an environmental cause relating to UV light exposure, some breeds are more at risk of cutaneous melanoma than others (Goldschmidt and Shofer 1992) and the breed composition of the Californian dog population was not described. At the tumour level, the incidence of the major tumour types for example, mast cell tumours, lymphosarcomas and osteosarcomas is approximately as one would expect based on estimates from previous studies (Dorn and others 1968b, Priester and McKay 1980). The relative importance of soft tissue sarcomas in the dog becomes more apparent when they are classed as a group as opposed to being listed under specific diagnoses. However, it should be noted that haemangiosarcoma was the predominant sarcoma of the spleen and other alimentary sites. The high incidence of CCH in this study was notable (377 cases per 100,000 dogs per year). CCH is a common tumour in young dogs. Its incidence was previously reported to be 117 per 100,000 dog per year (Taylor and others 1969), considerably less than in the population of dogs in the present study. Age and breed are important risk factors for CCH (Goldschmidt and Shofer 1992). In both studies, around 43 per cent of the at-risk population was three years of age or less, but the breeds may have been quite different. Also, because this tumour tends to occur in younger animals and initially appears as a rapidly growing mass, owners are probably more likely to seek veterinary advice, and veterinary surgeons are more likely to act than in cases in older animals with slower growing masses, irrespective of insurance status. Comparison of the incidence of human cancer in the UK with CCH in dogs reveals some striking similarities and differences. Breast cancer is the most common malignancy in women and the mammary gland was the third most common tumour site in dogs in this study. However, carcinomas of the lung, large bowel and prostate, the most common human tumours excluding tumours of the breast, did not feature highly in the canine population. Indeed, as a group, carcinomas are not as predominant in this canine population as they are in humans. As the aetiology and pathogenesis of canine tumours is likely to be similar to that in humans, comparative studies may be of considerable value to medical research (Reif and others 1992, 1995, Knapp and Waters 1997). Although risk factors are known to exist for many specific cancer types, these were not evaluated in this study. Rather, the overall, mean estimates of incidence were given instead. Evaluation of predisposing factors was not the purpose of this study, but in an another study (Wood and others 2000), breed and age-specific risks have been evaluated for the more common types of neoplasia. The effects of age and breed vary markedly between different types and it is important that more care is taken to evaluate both factors when considering the risk for individual dogs. Conclusions This study has produced estimates of the incidence of different cancer types seen in a population of UK dogs, which provide a useful benchmark in the evaluation of any possible temporal trends. The results presented, when considered along with those from more detailed analyses of the effects of age and breed (Wood and others 2000), emphasise the importance of considering both of these factors when evaluating the risk of different cancer types (Egenvall and others 2000). Animals at high risk for one type of cancer may not be at high risk for others. Molecular genetic evaluation of breed differences relevant for individual cancer risk may shed great light on the aetiology of different cancers. Acknowledgements This work was supported by the BSAVA and the Kennel Club Charitable Trust. The authors are most grateful to PetProtect for providing access to its database and for the considerable assistance received from its staff during the course of the study. References BONNETT, B. N., EGENVALL, A., OLSON, P. & HEDHAMMAR, A. (1997) Mortality in insured Swedish dogs: rates and causes of death in various breeds. Veterinary Record 141, DORN, C. R., TAYLOR, D. O. N., FRYE, F. L. & HIBBARD, H. H. (1968a) Survey of animal neoplasms in Alameda and Contra Costa counties, California 1. Methodology and description of cases. Journal of the National Cancer Institute 40, DORN, C. R., TAYLOR, D. O. N., SCHNEIDER, R., HIBBARD, H. H. & KLAUBER, M. R. (1968b) Survey of animal neoplasms in Alameda and Contra Costa counties, California II. Cancer morbidity in dogs and cats from Alameda County. Journal of the National Cancer Institute 40, EGENVALL, A., BONNETT, B. N., SHOUKRI, M., OLSON, P., HEDHAMMAR, A. & DOHOO, I. (2000) Age pattern of mortality in eight breeds of insured dogs in Sweden. Preventive Veterinary Medicine 46, 1-14 ENZINGER, F. M. & WEISS, S. W. (1995) General considerations. In: Soft Tissue Tumours, 3rd edn. Mosby, St Louis. pp 1-16 GOLDSCHMIDT, M. H. & SHOFER, F. S. (1992) Skin Tumours of the Dog and Cat. Pergamon Press, Oxford KNAPP, D. W. & WATERS, D. J. (1997) Naturally occuring cancer in pet dogs. Important models for developing improved cancer therapy in humans. Molecular Medidine Today 3, 8-11 JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE

7 LENGERICH, E. J., TECLAW, R. F., MENDLEIN, J. M., MARIOLIS, P. & GARBE, P. L. (1992) Pet populations in the catchment area of the Purdue Comparative Oncology Program. Journal of the American Veterinary Medical Association 200, MACVEAN, D. W., MONLUX, A. W., ANDERSON, P. S., SILBERG, S. L. & ROSZELA, J. F. (1978) Frequency of canine and feline tumors in a defined population. Veterinary Pathology 15, MICHELL, A. R. (1999) Longevity of British breeds of dog and its relationships with sex, size, cardiovascular variables and disease. Veterinary Record 145, MISDORP, W. (1990) General considerations. In: Tumours of Domestic Animals, 3rd edn. Ed J. E. Moulton. Berkeley, University of California Press. pp 1-22 PRIESTER, W. A. & MCKAY, F. W. (1980) The occurrence of tumours in domestic animals. National Cancer Institute Monograph 54, US Department of Health and Human Services, NIH Publication No REID-SMITH, R. J., BONNETT, B. N., MARTIN, S. W., KRUTH, S. A., ABRAMS-OGG, A. & HAZLETT, M. J. (2000) The incidence of neoplasia in the canine and feline patient populations of private veterinary practices in southern Ontario. Proceedings of the 9th Symposium of the International Society for Veterinary Epidemiology and Economics. August 2000, Colorado. pp REIF, J. S., DUNN, K. & OGILVIE, G. K. (1992) Passive smoking and canine lung cancer risk. American Journal of Epidemiology 135, REIF, J. S., LOWER, K. S. & OGILVIE, G. K. (1995) Residential exposure to magnetic fields and the risk of canine lymphoma. American Journal of Epidemiology 141, RICHARDS, H. G., MCNEIL, P. E., THOMPSON, H. & REID, S. W. J. (2000) An epidemiological analysis of a canine biopsies database compiled by a diagnostic histopathology service. Proceedings of the Society for Veterinary Epidemiology and Preventive Medicine. pp ROTHMAN, K. (1986) Modern Epidemiology. Little Brown, Boston TAYLOR, D. O. N., DORN, C. R. & LUIS, O. H. (1969) Morphologic and biologic characteristics of the canine cutaneous histiocytoma. Cancer Research 29, THRUSFIELD, M. V. (1983) Application of computer technology to the collection, analysis and use of veterinary data. Veterinary Record 112, THRUSFIELD, M. V. (1989) Demographic characteristics of the canine and feline populations of the UK in Journal of Small Animal Practice 30, WILLEBERG, P. (1986) Epidemiologic use of routinely collected veterinary data: risks and benefits. Proceedings of the 4th International Symposium on Veterinary Epidemiology and Economics, Singapore, November 18-22, pp WHO (1974) International Histological Classification of Tumours of Domestic Animals: Part 1. Bulletin of the World Health Organization 50 (No 1,2) WHO (1976) International Histological Classification of Tumours of Domestic Animals: Part 2. Bulletin of the World Health Organization 53 (No 2,3) WOOD, J. L. N., DOBSON, J. & ROGERS, K. (2000) An epidemiological study of cancer in British dogs based on insurance company data. Proceedings of the 9 th Symposium of the International Society for Veterinary Epidemiology and Economics. August 2000, Colorado. pp JOURNAL OF SMALL ANIMAL PRACTICE VOL 43 JUNE 2002 British Small Animal Veterinary Association. All rights reserved