Three RCTs (n=10,731), four observational studies (n=67,382) and 214 case reports (51 on pioglitazone and 163 on rosglitazone) were included in the review.
When the RCTs were pooled, an increased in risk of heart failure was found with thiazolidinediones compared to control, odds ratio 2.1 (95% CI: 1.08 to 4.08, p=0.03), resulting in an estimated number needed to harm (NNH) to be approximately 50 over 2.2 years; moderate statistical heterogeneity was found (I2=59%). Similarly, when the observational studies were pooled an increased risk of heart failure was found with thiazolidinediones compared to control, odds ratio 1.55 (95% CI: 1.33 to 1.80, p<0.00001); moderate statistical heterogeneity was found (I2=47%). When observational studies with post-MI (myocardial infarction) diabetic patients were excluded, no evidence of statisitical heterogenity was found (I2=0%).
The median duration for the onset of heart failure was 24 weeks (range one to 260). Heart failure due to thiazolidinediones did not occur as a result of cumulative ingested dose. Patients on low doses (defined as ≤15 mg pioglitazone or ≤4 mg rosiglitazone) were also at risk of developing heart failure. Heart failure was not found to develop earlier in those receiving high doses. The adverse reaction was not limited to the elderly, with 26% of the reported cases occurring in patients less than 60 years old. The authors were unable to identify any susceptibility factors. Death occurred in nine patients.