The total of 94 RCTs (n= 26,418) comprised 84 published trials and 10 unpublished trials. Data from 93 trials was included in the meta analysis; data from the PROACTIVE trial (n=5 238) was not included in some of the meta-analysis, because the number of events was much larger than for the other studies.
Randomisation and blinding were described as adequate in 31 (33%) and 21 (22%) of the 94 trials. Withdrawals and dropouts were reported in 66 trials (70%).
All-cause mortality: There was no significant differences between pioglitazone and other treatments in the analysis of all trials, and for the PROACTIVE trial. In the meta-analysis where PROACTIVE was excluded and trials with at least one reported event were included, there was a significant reduction in mortality observed with pioglitazone treatment (odds ratio 0.30, 95% CI 0.14 to 0.63). Statistically significant reductions of all-cause mortality were observed for the analyses of Type 2 diabetic patients (relative risk 0.41, 95% CI 0.23 to 0.72; 68 trials) and long-term (≥24 weeks) compared with short-term studies (≤24 weeks). No significant reductions were observed when pioglitazone was compared with different treatments or placebo in analyses that included trials with zero events.
Non-fatal coronary events: In analyses of all trials, of trials with at least one event and of Type 2 diabetics, there were no statistically significant differences between pioglitazone and other treatments (relative risk 0.82, 95% CI: 0.55 to 1.23 for the last comparison; figures for the other comparisons were not provided). In the PROACTIVE trial, use of pioglitazone was associated with a significant reduction in non-fatal coronary events.
Non-fatal heart failure: A statistically significant increase in risk of heart failure with the use of pioglitazone was observed in the PROACTIVE trial and in trials that reported non-fatal heart failure that required hospitalisation (relative risk 1.32, 95% CI: 0.88 to 1.98). A small observed increase in the incidence of non-fatal heart failure was observed for all trials, but was not statistically significant (odds ratio 0.38, 95% CI: 0.90 to 2.12).