Pharmacoepidemiology of autoimmune diseases

The current system in place to study safety of medicines after introduction on the market relies on spontaneous reporting. Adverse events occurring long after initiation or cessation of drug use are likely to be missed by this system. In this thesis we explore methods to identify signals of long-term, unexpected adverse events and methods to evaluate these signals. We utilised data from the UK General Practice Research Database (GPRD), a large primary care database. A systematic review investigating the validity of medical diagnoses recorded in this database illustrated that the GPRD is a powerful tool to study morbidity in primary care. However, intimate knowledge of the complexities of the database is needed to ensure the best use is made of the database. Pre-existing hypotheses of drug-induced systemic lupus erythematosus (SLE) were evaluated using the GPRD. Associations between risk of SLE and exposure to hydralazine, minocycline, and carbamazepine were confirmed using both a matched case-control design and the self controlled case series method. Spontaneous reports of drug-induced SLE recorded in the UK Yellow Card database indicated that symptoms of SLE often resolve after withdrawal of the suspected drug. Using the Smile Plot method to generate signals of drug-induced SLE, we were not able to identify known signals of drug-induced lupus. However, we did identify factors strongly associated with treatment of early symptoms of disease. These findings indicated a high specificity of the Smile Plot method. To improve sensitivity, better hierarchical coding systems for drugs are needed to ensure appropriate grouping. Lastly, we utilised the GPRD to provide an example of a systematically performed drug safety study. In a small subset of data, we generated hypotheses of drug-induced Hashimoto's disease. Associations were subsequently evaluated in a larger subset of data. No drugs were clearly associated with risk of Hashimoto's disease.