Rosiglitazone Associated with Increased Cardiovascular Risks

Rosiglitazone was approved by the U.S. Food and Drug Administration in 1999 to treat type 2 diabetes. Previous data have suggested an associated risk of cardiovascular complications. A new systematic review and meta-analysis of randomized, controlled trials evaluated this risk.

Individual patient-level data were obtained from GlaxoSmithKline’s (GSK), and summary-level data were collected from GSK’s Study Register platforms, MEDLINE, PubMed, Embase, Web of Science, Cochrane Central Registry of Controlled Trials, Scopus, and; databases were searched from inception through January 2019. Eligible studies were phase II to IV clinical trials analyzing rosiglitazone against any control in adult patients for at least 24 weeks.

A search of yielded 33 eligible trials encompassing 21,156 total patients with available patient-level data; 103 trials encompassing 23,683 patients that did not have patient-level data were used for the meta-analyses for myocardial infarction, and 103 trials encompassing 22,772 patients without patient-level data were included in the meta-analyses for cardiovascular-related mortality. When evaluating the 29 trials that had patient-level data and were used in previous meta-analyses with GSK’s summary level data, the use of patient-level data identified more myocardial infarction events compared with summary-level data in 26 trials, as well as fewer cardiovascular-related deaths in five trials.

When excluding trials without patient-level data and using a constant continuity correction of 0.5 and a random effects model accounting for trials that had no events in one group, rosiglitazone had a 33% greater risk than controls of a composite event; odds ratios (ORs) for specific events were: myocardial infarction, 1.17; heart failure, 1.54; cardiovascular-related death, 1.15; and non-cardiovascular related death, 1.18. ORs were reduced in analyses including trials without patient-level data for myocardial infarction (1.09) and cardiovascular-related death (1.12). Outcomes remained largely similar when trials with zero events in both arms were used and when either continuity correction was used.