RECORD LINKAGE TO CONDUCT AN EPIDEMIOLOGIC-STUDY ON THE ASSOCIATION OF RHEUMATOID-ARTHRITIS AND LYMPHOMA IN THE PROVINCE OF SASKATCHEWAN, CANADA

The objective of this effort was to assess the utility of the large au tomated database in Saskatchewan as a resource for pharmacoepidemiolog ic studies. To this end a study was undertaken to test the hypothesis that rheumatoid arthritis (RA) increases the risk of cancer, especiall y lymphoma. This was done by performing a retrospective cohort study b ased on record linkage data from Saskatchewan Health. From hospital di scharge diagnoses in the hospital file an exposed group (RA) and two c omparison groups matched to the RA group by age and sex were identifie d: (1) the RA group consisted of people with a discharge diagnosis of rheumatoid arthritis; (2) the osteoarthritis (OA) group consisted of p eople with OA discharge diagnoses; and (3) a comparison (CN) group con sisted of hospitalized people with no discharge diagnoses of arthritis . Drug exposures were determined by linkage with the Prescription Drug File, cancer outcomes were determined by linkage with the Cancer Foun dation file, and length of eligibility in the health plan and demograp hics information were determined by linkage with the registration file . The data were checked for quality of linkages across files and consi stency with study definitions. Of 13,333 identified subjects, 2.8% wer e excluded because of apparent incorrect assignment to study group or age group or because of ineligibility in health plan during the study period. In order to decrease the possibility of misclassification of e xposure (rheumatoid arthritis), hospital discharge diagnoses were used to exclude subjects with any inflammatory rheumatic diseases (IRD) fr om the CN (7.8%) and OA (8.3%) groups and subjects with IRD other than rheumatoid arthritis (4.6%) from the RA group. To decrease selection bias, those who had cancer within 1 year of enrollment (to exclude tho se in hospital because of symptoms of undiagnosed cancer) were exclude d. Because RA subjects hospitalized by a rheumatologist were most like ly to have valid rheumatoid arthritis diagnoses, each analysis was run twice: once with the entire RA group (N = 1210) and once with those i n the RA group who were rheumatologist-hospitalized (N = 646). Logisti c regression of incidence was used to control for age, sex, and use of individual disease-modifying anti-rheumatoid drugs (DMARDs). For the rheumatologist-hospitalized RA group compared to the CN group, a signi ficant 4-fold greater risk for lymphoma/myeloma was detected when DMAR D use was not controlled for, and a 3.4-fold increase in risk was dete cted even when use of individual DMARDs was controlled for. The Saskat chewan Health database proved to be a powerful, cost-efficient resourc e for constructing an epidemiologic study. By using a number of analys is strategies to test and maximize validity, as well as abstraction of hospital charts to check validity, it was possible to perform a usefu l cohort study of a rare outcome and obtain results consistent with pr evious research using more traditional approaches.