(A) Graphic demonstration of statistically significant and clinically important treatment benefits. Five trial results (A to E) and their interpretation with reference to zero effect (a risk ratio of 1.0) and a minimal clinically important difference (MCID) of 15% relative risk reduction (corresponding to a risk ratio of 0.85) are shown. Treatment effects (double arrows) are expressed as 95% confidence intervals. (B) Bayesian analysis of clinical importance in TACTICS–TIMI 18: Probability density plot for the difference in outcomes between the 2 management strategies is shown using the Bayesian approach (a noninformative Gaussian prior probability distribution with mean = 0 and SD = 10). The probability for any given threshold (d) can be computed in terms of the area under the probability density curve. The light shaded area to the left of log odds ratio 0 (equivalent to odds ratio of 1) indicates a 98.6% probability of d >0% benefit and the dark shaded area to the left of log odds ratio –0.35 (equivalent to odds ratio of 0.71 or relative risk of 0.75) indicates a 17% probability of d >25% benefit. (C) Posterior probabilities for a range of threshold of benefit are graphically represented. The probability of d >25% risk reduction is shown as an open circle (17%).

Data for the composite end point and the individual component end points are shown for 3 trials: (A) Stent PAMI, (B) HOPE, and (C) TRITON–TIMI 38. Heterogeneity is derived from Cochran's Q chi-square test and is implied at p < 0.10; I² quantifies the impact of heterogeneity. Data for stent PAMI at 1 year of follow-up (23). CI = confidence interval; MI = myocardial infarction; TVR = target vessel revascularization.

Data used are first events (to exclude double counting) (see Table 2). The p values are derived from a global z statistic under the assumption of no correlation among the components of the composite end point (32,33). A sensitivity analysis across myocardial infarction (MI) weighting is also shown (the weights of death and stroke were fixed at 1.0 and 0.5, respectively). Only MI weights 0.4 achieve statistical significance (p 0.05, values below the dotted line).

The rate of the primary end point among asymptomatic patients (with multiple atherothrombotic risk factors) and symptomatic patients (with documented coronary, cerebrovascular, or peripheral arterial disease) is shown as hazard ratio (HR) and 95% confidence interval (CI). The result of subgroup analysis unadjusted for multiple comparisons is shown as an interaction term with 95% CI and p value.

Outcomes at 30 days and at 6 months are stratified by treatment group and cardiac troponin T (TnT) level. The primary end point was death, myocardial infarction (MI), or urgent revascularization. *p = 0.01 at 30 days and p = 0.003 at 6 months for interaction between treatment group and cardiac TnT level with respect to the primary end point. CI = confidence interval; RR = risk ratio.

The primary outcome (composite of death, nonfatal myocardial infarction [MI], nonfatal stroke, and repeat revascularization) at 12 months is stratified by mode of revascularization (percutaneous coronary intervention [PCI] vs. coronary artery bypass graft surgery [CABG]) and the anatomic subset of coronary artery disease: overall cohort, which includes left main (LM) plus 3-vessel disease (3VD), LM cohort subdivided into isolated LM and LM plus 1-vessel disease (1VD), and 3VD (48). Interaction p value unadjusted for multiple comparisons is shown; the dotted line represents the overlap in treatment effect. CI = confidence interval.