Level 2: Mid-level evidence

Two meta-analyses, each involving 42 randomized trials, suggest that rosiglitazone (Avandia) is associated with increased cardiovascular risk.

One published meta-analysis included 27,843 patients with diabetes or prediabetes (N Engl J Med. 2007;356:2457-2471). A systematic search was not described, but 116 phase 2, 3, and 4 trials were screened for inclusion. Forty-eight trials met inclusion criteria of randomization to treatment with rosiglitazone vs. treatment without rosiglitazone, similar duration of treatment in all groups, and drug exposure >24 weeks. Six trials that did not report any MIs or cardiovascular deaths (CVDs) were excluded. A systematic evaluation of trial quality was not reported.

Comparing rosiglitazone vs. control, the rate of MI was 0.6% vs. 0.62% in an analysis not adjusting for differences in study populations, 0.43% vs. 0.36% in an analysis of all small trials combined (N=16,385) (P=.15), 0.57% vs. 0.34% in the Diabetes Reduction Assessment with Ramipril and Rosiglitazone Medication (DREAM) trial (N=5,269) (P=.22), and 1.85% vs. 1.44% in the A Diabetes Outcome Progression Trial (ADOPT) (N=4,351) (P=.27). A combined meta-analysis reported significant results (P=.03) equivalent to NNH 375. The rate of CVD was 0.38% vs. 0.24% in an analysis not adjusting for differences in study populations, 0.38% vs. 0.19% in an analysis of all small trials combined (N=10,257) (P=.02), 0.51% vs. 0.38% in the DREAM trial (P=.67), and 0.14% vs. 0.18% in the ADOPT trial (P=.78). A combined meta-analysis did not reach statistical significance (P=.06).

In analyses comparing rosiglitazone with specific alternative drugs (metformin [Glucophage], sulfonylurea, insulin) or placebo, there were no significant differences in rates of MI or CVD.Another meta-analysis (unpublished) was reported by the FDA (FDA Press Release 2007 May 21; full-text available online free of charge at: www.fda.gov/bbs/topics/NEWS, accessed November 12, 2007, and FDA MedWatch 2007 May 21; full-text available online without charge at: www.fda.gov/cder/drug, accessed November 12, 2007). This analysis was limited to 14,237 patients with diabetes and compared rosiglitazone (with or without insulin, sulfonylureas, and/or metformin) with placebo (with or without other antidiabetic drugs). Thirty-eight trials lasted six months. The overall incidence of myocardial ischemia was 1.99% with rosiglitazone vs. 1.51% with controls (NNH 208). However, there was no significant association between rosiglitazone and myocardial ischemic events in an observational study with 33,363 patients, a randomized trial (ADOPT) with median four-year follow-up in 4,351 patients, and partial data from another randomized trial (DREAM) with median three-year follow-up in nearly 5,300 patients.