Allopurinol and Cardiovascular Events: Time-Related Biases in Observational Studies.

Abstract

Several observational studies reported that allopurinol, an effective treatment for gout, was associated with important reductions in cardiovascular (CV) events, with calls for large, randomized trials, although some results were conflicting. The present study was undertaken to assess the extent of time-related biases in these observational studies. We searched the literature for all observational studies reporting on allopurinol and CV events, focusing on 2 time-related biases. Time-related confounding bias results from studies using cohorts of patients all exposed to allopurinol, with comparisons based on episodes of allopurinol discontinuation, where confounding factors are not updated over follow-up time. Immortal time bias arises from the exposure misclassification of periods of cohort follow-up during which the outcome under study cannot occur. We identified 12 studies, of which 8 were affected by time-related confounding bias or immortal time bias, while the remaining 4 studies avoided these biases. The studies affected by time-related confounding bias resulted in significant reductions in the incidence of CV events with allopurinol use (pooled hazard ratio [HR] 0.88 [95% confidence interval (95% CI) 0.85-0.92]), as did the studies affected by immortal time bias (pooled HR 0.79 [95% CI 0.72-0.87]). The 4 studies that avoided these biases resulted in a pooled HR of 1.07 (95% CI 0.91-1.25). Observational studies reporting significantly reduced incidence of CV events with allopurinol use were affected by time-related biases. Overall, studies that avoided these biases did not find a protective effect. The ALL-HEART randomized trial will provide important and accurate evidence on the potential effectiveness of allopurinol on CV outcomes.