Thirty-seven RCTs (3,035 patients) were included in the review. Twenty-one trials (1,201 patients) compared pioglitazone with control, 15 trials (1,784 patients) compared rosiglitazone with control and one evaluated both medications versus control (50 patients). Study quality was variable: all studies had baseline comparability of groups, 21 studies described adequate allocation concealment, 22 reported patient withdrawals and 31 reported analysis methods. Only eight trials were described as double-blinded; the remainder were single-blind or open label.
Thiazolidinediones showed a statistically significant benefit on HbA1c compared to control for both pioglitazone (WMD -0.12%, 95% CI -0.38 to -0.16, Ι²=20%) and rosiglitazone (WMD -0.47%, 95% CI -0.62 to -0.33, Ι²=63%). Subgroup analyses showed that pioglitazone had significant benefits compared to placebo or previous glycaemic control whilst rosiglitazone was also more effective than sulphonylurea.
Twenty-five RCTs also assessed fasting plasma glucose and the pooled estimates showed overall benefits of both drugs. Pioglitazone gave significant reductions compared to placebo or previous glycaemic control, metformin and voglibose. Rosiglitazone gave significant reductions compared to placebo or previous glycaemic control, metformin and sulphonylurea.
Two RCTs assessed incidence of major cardiac events for pioglitazone and two for rosiglitazone. Both drugs showed statistically significant benefits (pioglitazone RR 0.24, 95% CI 0.09 to 0.63, Ι²=36%; rosiglitazone risk ratio 0.41, 95% CI 0.19 to 0.87; Ι² = 0%). Results for other non-glycaemic outcomes were also reported.
There was some evidence of publication bias.