Pioglitazone and bladder cancer: two studies, same database, two answers

Abstract

Various studies have used the UK Clinical Practice Research Datalink (CPRD) to evaluate the same side effects of drugs, often with opposite conclusions. The CPRD was formerly known as the General Practice Research Database. Examples include third generation pills and venous thromboembolism, proton pump inhibitors and fracture [1], oral bisphosphonates and gastrointestinal cancers [2] and now, pioglitazone and bladder cancer [3, 4]. Only rarely have CPRD data been re-analyzed to find out why results vary so much. The judge requested re-analysis during the third generation pill lawsuit, but the results were not published as a scientific paper. The other examples are our re-analysis of statins and fracture risk [5] and our paper on confounding by respiratory disease severity [6]. We were quite surprised to read the recent CPRD study on pioglitazone and bladder cancer without a proper re-analysis or even a discussion [3] of the findings of Azoulay and colleagues [4]. They used almost the same study population but a different study design. A re-analysis of CPRD is often unrewarding. It requires more resources than conducting an original study. Major clinical journals are less likely to publish a third or fourth CPRD study on the same topic and the paper is less likely to be cited. However, to remove the confusion, both groups of researchers could re-analyze each other's data and this journal and the BMJ could reward these efforts with two new publications. A similar letter about bisphosphonates and oesophageal cancer was published in the BMJ in 2010 [2] and the journal has recently published a re-analysis of the data [7].