Reassessing the Neutrality of ISCHEMIA: Aggregate-Data Reanalysis Reveals Instability of Composite Endpoints and Hidden Efficacy of Invasive Therapy

The ISCHEMIA trial reported no reduction in cardiovascular death or myocardial infarction (MI) with an initial invasive strategy. However, this neutrality depends on the design of the primary composite endpoint, which combined spontaneous type-1 and type-3 MI (plaque rupture or sudden death) with procedural type-4a/5 MI defined by permissive biomarker thresholds and minimal prognostic significance. In addition, ≈20% crossover in the conservative arm diluted the intention-to-treat (ITT) contrast.

Using only published aggregate data, we conducted design-based reanalyses. We separated spontaneous from procedural MI and repeated analyses after excluding patients censored by procedural events (“restricted” analysis). We corrected for misattribution by reallocating the 20 first procedural MIs reported in the conservative arm—events that necessarily occurred after revascularization—to the invasive arm. We estimated the causal effect of early revascularization among compliers using an instrumental-variable Wald estimator. We derived the threshold weight w\ ^* w^\ ^* w\ ^* at which procedural MI would neutralize the composite, calculated fragility, and estimated numbers needed to treat/harm (NNT/NNH). Finally, we reconstructed parametric pseudo–Kaplan–Meier curves by calibrating Weibull models to the published 4-year risks.

Across 5,179 patients, spontaneous MI were halved by invasive therapy (2.9% vs 5.8%; relative risk 0.51), confirmed in restricted analysis (RR 0.52). Procedural MI were more frequent in the invasive arm (2.7% vs 0.8%; RR 3.45). After reallocation of the 20 misattributed events, the all-MI composite reversed (RR 1.08). Instrumental-variable analysis indicated that early revascularization among compliers reduced spontaneous MI by −4.8 percentage points (NNT ≈21) while increasing procedural MI by +3.2 points (NNH ≈31). The threshold was w ^* =0.81: unless a procedural MI is valued at ≥81% of a spontaneous MI, invasive therapy is favorable. Fragility analysis showed that only 20 events (0.4% of the cohort) sufficed to reverse the composite. Weibull curves illustrated early procedural hazard and late spontaneous benefit.

Neutrality in ISCHEMIA is thus an artifact of endpoint construction and misattribution. Invasive therapy consistently reduces clinically meaningful spontaneous MI, while procedural events are prognostically trivial.