Impact of the definition of osteoarthritis and of the timing of its onset on the association between type 2 diabetes mellitus and osteoarthritis : Clinical Practice Research Datalink

Aims: In a previous case-control study in a large primary care database, the Clinical Practice Research Datalink (CPRD), type 2 diabetes mellitus (T2DM) was associated with a decreased rate of total joint replacement (TJR). As this was in contrast to the hypothesis, selection bias due to the used definition of osteoarthritis (OA) or misclassification of the onset of OA were raised as possible explanations. We therefore aimed to explore the effect of the definition of OA, and hypothesized timing of its onset on the association between T2DM and OA. Methods: All patients using a non-insulin anti-hyperglycaemic drug (NIAD) between 1989 and 2012 in the CPRD were included and matched to unexposed patients. Cox proportional hazard models were fitted estimating the risk of TJR or OA in T2DM patients compared to patients without T2DM. These analyses were repeated in sensitivity scenarios and joint-specific analyses. To assess whether misclassification of onset of OA may affect the association, analyses were repeated with addition of a latency period of up to 10 years after start of follow-up. Results: The use of TJR as a proxy for OA (hazard ratio (HR) = 0.74; 95% Confidence Interval (CI) = 0.70–0.78) resulted in a HR that was approximately 0.2 lower than when OA diagnostic codes were used (HR = 0.93; 95% CI = 0.90–0.95). The joint-specific subgroup analyses, sensitivity scenarios, and latency analyses showed similar results. Conclusion: When examining the association between T2DM and OA, the use of TJR as a proxy for OA resulted in a 20% lower estimate than the OA diagnosis