Rosiglitazone may be linked with a raised risk of cardiovasculardisease, a study has found –but the FDA has not required the maker totake any specific action for now due to limitations of the study andconflicting evidence from other research.
The meta-analysis was published early online by the New England Journal of Medicine.
Thiazolidinediones(glitazones) effectively reduce blood glucose levels, but there islimited evidence on their long-term effect on adverse clinical outcomesof diabetes. The authors note that most deaths in diabetes result fromcardiovascular (CV) causes, so used data from existing trials to assessrosiglitazone’s effects on CV outcomes.
The researchersscreened available clinical trials comparing rosiglitazone to controlsto identify those meeting eligibility criteria: having a randomisedcomparator group, similar treatment duration and more than 24 weeks’drug exposure. They selected only studies reporting CV deaths ormyocardial infarctions (MIs), with patients receiving rosiglitazonegrouped and compared with those receiving the control.
Outcomemeasures were CV death or MI, taken from trial data. It was impossibleto determine time to event or whether the same patient had bothoutcomes.
Of 116 trials screened, 48 met initialeligibility criteria. Six were excluded because they reported no MI orCV deaths. Three main study groups were identified (rosiglitazone vscontrol respectively):
• Five trials submitted to the FDA for initial registration (n=1,967 vs n=793).
• Thirty-five studies identified from the GSK clinical trial registry (n=9,502 vs n=5,961).
• Two large published trials: DREAM (n=2,635 vs n= 2,634) and ADOPT (n=1,456 vs n=2,895).
Totalpopulation for analysis was 27,843 (n=15,560 vs n=12,283). Mean patientage was 57, with an excess of men (60.7% vs 53.3%), with relativelypoor diabetic control (mean HbA1c 8.2%, both groups). There were 158 MIand 61 CV deaths overall: 86 vs 72 MI, and 39 vs 22 CV deaths. Summaryodds ratio for MI in the rosiglitazone group, 1.43, was statisticallysignificant at conventional levels (95% CI 1.03–1.98; p=0.03). Summaryodds ratio for CV death was not statistically significant but showed atrend to increase in the rosiglitazone group (OR 1.64; 95% CI0.98–2.74; p=0.06).
The authors conclude that treatment with rosiglitazone may increase risk of MI and of CV death.
Theynote their study’s limitations, including lack of access to originaldata; use of pooled data mostly from short-term trials not designed tocollect these data, and in which events were not independentlyadjudicated; and relatively few events overall.
Nevertheless,the authors suggest further evaluation is needed to clarify the risksand benefits of rosiglitazone. They note that this and related drugswere licensed on the basis of a surrogate outcome – effects on bloodglucose control – not on clinical outcomes. The RECORD trial, due toreport in 2009, may provide further data.
New Engl J Med, published early online 21 May 2007; doi:10.1056/NEJMoa072761