We obtained results from all randomized trials evaluating the benefits of adjunctive Gp IIb-IIIa inhibitors among STEMI patients undergoing primary angioplasty. The literature was scanned by formal searches of electronic databases (MEDLINE and CENTRAL) from January 1990 to September 2008. Furthermore, oral presentations and/or expert slide presentations were included [searched on the TCT (www.tctmd.com), EuroPCR (www.europcr.com), ACC (www.acc.org), AHA (www.aha.org), and ESC (www.escardio.org) websites from January 2002 to September 2008].The following key words were used: randomized trial, myocardial infarction, reperfusion, primary angioplasty, Gp IIb-IIIa inhibitors, abciximab, tirofiban, and eptifibatide.

Inclusion criteria were: (i) randomized treatment allocation; (ii) availability of complete clinical data. Exclusion criteria: (i) follow-up data in less than 90% of patients; (ii) ongoing studies or irretrievable data. No language restrictions were enforced.

Data were independently abstracted by two investigators. In case of incomplete or unclear data, authors, where possible, were contacted. Disagreements were resolved by consensus. Data were managed according to the intention-to-treat principle.

Primary and secondary endpoints were mortality and re-infarction at 30 days, respectively. Safety endpoint was the rate of major bleeding complications.

Statistical analysis was performed using the Review Manager 4.27, SPSS 15.0 statistical package and Comprehensive Meta-analysis Version 2. Odds ratio (OR) and 95% confidence intervals (95% CI) were used as summary statistics. The pooled OR was calculated by using a fixed-effect model with the Mantel?Haenszel method. The DerSimonian and Laird random effect model was additionally applied to calculate pooled OR in case of significant heterogeneity across studies. Between-study heterogeneity was analysed by means of I² = [(Q ? df)/Q]?100%, where Q is the ?² statistic, and df is its degrees of freedom. The potential publication bias was examined by constructing a ?funnel plot?, in which the standard error (SE) of the ln OR was plotted against the OR (30 day mortality).

The relationship between benefits in mortality and risk profile in each study (study level variable) was evaluated by a weighted random-effects meta-regression analysis regressing the log OR against the control group event rate expressed as odds using the inverse of the variance of the log OR as weight.²⁰

We additionally performed a weighted random-effects meta-regression analysis regressing the log OR against the average log event rate observed in experimental and control group combined, using the inverse of the variance of the log OR as weight²⁰ and a weighted random-effects meta-regression analysis regressing the log odds in the experimental group against the log odds in the control group, using the inverse of the variance of the log odds as weight.²¹

Results are reported as beta coefficients and two-sided P-values. Similar analysis was conducted for 30 day re-infarction.

The study was performed in compliance with the Quality of Reporting of Meta-Analyses (QUORUM) guidelines.²²

Of the 859 potentially relevant articles initially screened, a total of 17 trials were initially identified^{3?16,18,19,23,24} (Figure?1). One trial⁴ was excluded because part of the ISAR trial.⁵ Thus, a total of 16 randomized trials^{3,5?16,18,19,23,24} were finally included in the meta-analysis., involving 10 085 patients (5094 or 50.5% in the Gp IIb-IIIa inhibitors group and 4991 or 49.5% in the control group). Characteristics of the included trials are shown in Table?1. High-dose bolus (600 mg) of clopidogrel was administrated in only three trials,^19,23,24 whereas in the Horizons trial¹⁸ the dosage of bolus of clopidogrel was left to the discretion of the treating physicians. In the study by Zorman et al.,¹⁰ patients were randomized to early vs. late administration (in the cath lab) or no abciximab. In the Horizons trial,¹⁸ the control group received Bivalirudin as anticoagulation therapy instead of unfractionated heparin.

As shown in Figure?2, overall Gp IIb-IIIa inhibitors did not reduce 30 day mortality [2.8 vs. 2.9%, OR (95% CI) = 0.96 (0.76?1.22), P = 0.75, P_het = 0.48, NNT = 1000]. A trend in benefits in mortality was observed after the exclusion of the Horizons trial [2.6 vs. 3.3%, OR (95% CI)= 0.78 (0.58, 1.04), P = 0.09, P_het = 0.84; NNT = 143]. As shown in Figure?3, no publication biased was observed.

As shown in Figure?4A, we observed a relationship between patient's risk profile and mortality benefits from adjunctive Gp IIb-IIIa inhibitors (beta = ?14.12, P = 0.008; without Horizons: beta = ?10.31, P = 0.08). The relationship was confirmed when benefits were plotted against average events between experimental and control groups (beta = 0.68, P = 0.07; without Horizons: beta = ?0.47, P = 0.23) (Figure?4B). No significant relationship was observed between the odds of the experimental and control groups (beta = 0.23, P = 0.21; without Horizons: beta = 0.40, P = 0.06) (Figure?4C). In particular, the deviation of the regression line, corresponding to the pooled treatment effect, from the line of identity (Figure?4C) suggests heterogeneity and further supports the relationship between risk profile and mortality benefits.

As shown in Figure?5, overall Gp IIb-IIIa inhibitors did not reduce 30 day re-infarction [1.5 vs. 1.9%, OR (95% CI) = 0.82 (0.59?1.13), P = 0.22, P_het = 0.54; NNT = 250]. Similar findings were observed after the exclusion of the Horizons trial [1.3 vs. 1.9%, OR (95% CI) = 0.70 (0.46?1.08), P = 0.11, P_het = 0.5; NNT = 167].

As shown in Figure?6A, no significant relationship was observed between the risk of re-infarction and benefits from adjunctive Gp IIb-IIIa inhibitors (beta = ?17.88, P = 0.25; without Horizons: beta = ?10.31, P = 0.08). These data were confirmed when the benefits were plotted against average events between experimental and control groups (beta = ?0.10, P = 0.79; without Horizons: beta = ?0.47, P = 0.23) (Figure?6B). A significant relationship was observed between the odds of the experimental and control groups (beta = 0.55, P = 0.03; without Horizons: beta = 0.48, P = 0.12) (Figure?6C), with a minor deviation of the regression line from the line of identity when compared with that observed for mortality.

As shown in Figure?7, Gp IIb-IIIa inhibitors were associated with higher risk of major bleeding complications [3.9 vs. 2.6%, OR (95% CI) = 1.50 (1.19?1.89), P = 0.0005, P_het = 0.85]. Similar results were observed after exclusion of the Horizons trial [3.3 vs. 2.3%, OR (95% CI) = 1.41 (1.04, 1.93), P = 0.03, p_het = 0.8].

The use of a risk score based on individual patient's data would have overcome the limitation due to the use of mortality in control group as marker of risk profile.¹⁹ In fact, the observed proportions of events in the control groups of the trials, as a measure of the underlying risk, may produce misleading results. This arises through a bias due to regression to the mean, especially observed in meta-analyses which include some small trials or in which the variability in the true underlying risks across trials is small. In order to minimize this limitation and to confirm initial findings, we have performed additional analyses thought to be more appropriate such as the use of average proportion of events in the control and treated groups as the measure of underlying risk or to plot the proportion of events in the treated group against that in the control group.²⁰

Even though high-dose clopidogrel bolus may obtain a quicker and better platelet aggregation inhibition, the peak effect is still observed after 3?4 h,²⁶ whereas Gp IIb-III inhibitors act very rapidly. In fact, benefits from Gp IIb-IIIa inhibitors were observed in the On-TIME 2 trial²³ despite early administration of high-dose 600 mg clopidogrel, whereas in the BRAVE trial¹⁹ clopidogrel was administrated just before the procedure and thus less probably affecting the results of the study. However, due to paucity of data, no definite conclusion can be drawn on this issue. In addition, we analysed short-term outcomes, whereas some Gp IIb-IIIa inhibitor studies have demonstrated increased survival benefit with longer follow-up.²⁵ No uniform definition of major bleeding complications was adopted in trials included in the current meta-analysis. Finally, we could not exclude a clinical heterogeneity between studies, due to different inclusion and exclusion criteria, and study drugs.

This meta-analysis shows a significant relationship between benefits in mortality from Gp IIb-IIIa inhibitors and patient's risk profile. Thus, Gp IIb-IIIa inhibitors should be strongly considered among high-risk patients.