Association between osteoarthritis and increased risk of cardiovascular diseases : investigation of the role of NSAIDs as an underlying mechanism

Osteoarthritis (OA) has been reported as an independent risk factor for cardiovascular diseases (CVD). There is no cure for OA and non-steroidal anti-inflammatory drugs (NSAIDs) are the mainstay of OA treatment. NSAIDs are known to be associated with various cardiovascular adverse effects, but their direct impact on CVD risk among OA patients is not well studied. There is a need to understand better the underlying mechanism of the increased risk of CVD among OA patients and to what extent NSAIDs play a role. This thesis conducted three studies using health administrative data (HAD) from British Columbia (BC), Canada and the Canadian Community Health Survey (CCHS) focusing on the role of NSAIDs in the OA-CVD association. The objective of study 1 was to quantify the role of NSAIDs in the increased risk of CVD among OA patients. This longitudinal study performed a mediation analysis using a marginal structural model and showed that a substantial proportion of total CVD risk among OA patients was attributable to NSAID use. The objective of study 2 was to evaluate the overall cardiovascular safety of various NSAIDs that are used in treating OA patients. This retrospective cohort study used time-dependent Cox regression analysis to estimate CVD risk associated with NSAID use overall and four unique groups of NSAIDs, i.e., coxibs, naproxen, ibuprofen and other conventional NSAIDs. This study showed that exposure to NSAIDs substantially increased CVD risk compared to unexposed person-time. It also showed that, relative to unexposed person-time coxibs and naproxen may increase CVD risk more than conventional NSAIDs including ibuprofen. The objective of study 3 was to identify a valid approach in imputing body mass index (BMI), an important confounding variable in the OA-CVD relationship for which information is usually not available in HAD. Multiple imputation was compared with proportion-based imputation (PBI) approach using plasmode simulated dataset created from CCHS data. This study showed that multiple imputation was superior to PBI in imputing BMI category using information from an external dataset. As a collective work, this thesis provides a better understanding of OA-CVD association that hopefully will improve clinical management of OA.Pharmaceutical Sciences, Faculty ofGraduat