Twenty-two RCTs (10,123 participants, range 46 to 2,159) met the inclusion criteria. Twelve RCTs were placebo-controlled. Fourteen RCTs reported appropriate methods of treatment assignment, 12 reported blinding, but three did not specifically report this was of outcome assessors. Drop-out rates ranged from zero to 8.3% across trial arms.
Overall incidence of non-fatal myocardial infarction was 5.1% with GPI and 8.3% with control (RR 0.66, 95% CI 0.55 to 0.79). Results were similar for abciximab, small molecule GPI and when the analysis was restricted to placebo controlled RCTs.
Incidence of minor bleeding significantly increased with GPI (3.0% versus 1.7%, RR 1.70, 95% CI 1.28 to 2.26; number of RCTs unclear) as did incidence of thrombocytopenia (0.8% versus 0.04%, RR 4.77, 95% CI 1.67 to 13.64; eight RCTs).
There were no significant differences between GPI and control for target vessel revascularisation (10 RCTs), major bleeding (15 RCTs), stroke (five RCTs) and all-cause mortality at either time point (10 RCTs).
There was no significant heterogeneity or evidence of publication bias for any analysis. Results of the meta-regression did not identify any of the variables investigated as effect modifiers.