Patient-Level Data Strengthen Rosiglitazone-CV Link: Meta-analysis

This speaks to the “importance of transparency and data-sharing when it comes to clinical trial evidence,” the lead author says.

Patient-Level Data Strengthen Rosiglitazone-CV Link: Meta-analysis

Individual patient-level data (IPD) obtained from GlaxoSmithKline confirm the cardiovascular hazard of rosiglitazone (Avandia), with investigators reporting that the antidiabetic drug is in fact connected with an increase in CVD events overall, particularly an increased risk of heart failure.

When the meta-analysis was diluted by including clinical trials without granular, patient-level information, the risk of CVD events was attenuated, Joshua D. Wallach, PhD (Yale School of Public Health, New Haven, CT), and colleagues report in a paper published online this week in the BMJ.

“It was really an exercise about the focus and importance of transparency and data-sharing when it comes to clinical trial evidence,” Wallach told TCTMD. The researchers sought to learn how using IPD might reveal information about CV risk factors, he said. “What brings rosiglitazone back into [this] is that it’s been 20 years [since its approval] and the uncertainty has been there for a while,” Wallach explained, making the medication an ideal candidate for this sort of approach that can shore up the evidence base.

First approved in 1999 by the US Food and Drug Administration for the treatment of type 2 diabetes, rosiglitazone was initially rejected by the European Medicines Agency before being approved in 2000. Annual sales topped $3 billion by 2006, before being felled by a 2007 meta-analysis suggesting a heightened MI risk with its use.

European regulators took rosiglitazone off the market in 2010. The FDA, on the other hand, updated the product label and limited the drug’s availability through its Risk Evaluation and Mitigation Strategy program, but the agency subsequently withdrew these measures in 2013 after a readjudication of the controversial RECORD trial showed no signals of CV harm for rosiglitazone versus other diabetes medications.

Its US market share never rebounded, however. As such, the latest meta-analysis doesn’t speak so much to clinical practice as it does to the benefits of transparency in healthcare data, Wallach observed. In the long rosiglitazone saga, “no other study has really used the raw data from trials and then also combined it with everything you can learn from publications and clinical trial registries,” he said.

Crunching the Numbers

For the new meta-analysis, the patient-level details were obtained via (CSDR). The consortium of clinical study sponsors and funders seeks “to drive scientific innovation and improve medical care by facilitating access to patient-level data from clinical studies,” its website notes.

From CSDR, Wallach and colleagues identified 33 trials with IPD for 21,156 participants. Their meta-analysis of these studies found that patients treated with rosiglitazone had a higher overall risk of cardiovascular events (composite outcome of acute myocardial infarction, heart failure, cardiovascular-related death, and noncardiovascular-related death). Among all the individual components, however, only heart failure risk was significantly higher in the rosiglitazone group.

Rosiglitazone Versus Control: Meta-analysis Using IPD



95% CI

Composite Outcome



Acute MI



Heart Failure



CV-Related Death



Non-CV-Related Death



The researchers then, using sources like PubMed, Embase, and others, included 103 trials without IPD into their calculations for both acute MI and cardiovascular-related death. The addition of these data attenuated what were already nonsignificant associations, resulting in an odds ratio of 1.09 for acute MI and 1.12 for cardiovascular-related death.

Additionally, absolute numbers of outcomes differed based on the data source: of the 29 trials with IPD that were included in previous meta-analyses using GlaxoSmithKline’s summary-level data, 26 showed more MIs and five had fewer cardiovascular-related deaths compared with the less-granular data.

The results, Wallach said, demonstrate how data-sharing platforms can help update prior insights and why this is important when evaluating drug safety. On the backdrop of growing interest in the sharing of data, this study shows how existing resources can be used, he pointed out. CSDR was launched by GlaxoSmithKline, but its offerings now also include trials from Bayer, Novartis, and Roche.

At Wallach’s institution, there’s the Yale Open Data Access (YODA) project, partnered with Johnson & Johnson. “Instead of [researchers] going to the pharmaceutical company and asking for trial data, the YODA team evaluates proposals for the data and acts as the decision-maker” for whether the project is worth pursuing, he said, adding that there are similar structures elsewhere, such as the pairing of Duke and Bristol-Myers Squibb.

“These platforms ensure that all shared data are deidentified, and they also require requestors to prespecify their research questions and methods,” the investigators explain in their paper. “Furthermore, they employ a ‘trusted intermediary’ approach, with independent review committees screening detailed proposals and making data-sharing decisions.”


  • This project was conducted as part of the Collaboration for Research Integrity and Transparency at Yale, funded by the Laura and John Arnold Foundation.
  • Wallach reports no relevant conflicts of interest.

We Recommend