Broad 6-month composite outcomes after acute ischemic stroke can mask opposing early benefit and harm pathways following antithrombotic treatments. Using the corrected public-use dataset from the International Stroke Trial (IST), the authors performed a secondary analysis to quantify how recorded early events mediate differences in 6-month death-or-dependency.
The analysis included 19,285 randomized IST participants for whom a 6-month outcome classification could be assigned. One hundred fifty participants with missing 6-month classification were excluded from the analysis. The dataset and analysis were presented as a secondary, post hoc exploration of early pathways linking randomized antithrombotic allocation to later disability-inclusive outcomes.
The mediator of interest was the first recorded selected event occurring within 14 days of randomization or earlier death/discharge. Events classified as the first selected event were:
This first-event framing was used to probe whether early harms (for example bleeding) or early ischemic prevention (for example fewer recurrent strokes) could explain later differences in death-or-dependency.
The investigators estimated standardized risks, risk differences per 1000 patients, and risk ratios using robust Poisson regression and g-computation. To decompose total effects into components mediated by the first recorded event versus direct effects, they estimated interventional direct and indirect effects using multinomial mediator and outcome models and marginalized factorial co-allocation. These methods produce interpretable contrasts attributable to pathways that operate through the early-event mediator versus those acting by other mechanisms.
Allocation to aspirin was associated with a small reduction in the composite 6-month outcome of death-or-dependency. Reported estimates were a risk difference of −12.1 per 1000 patients and a risk ratio of 0.981. When decomposed, the interventional direct component accounted for −9.8 per 1000 and the first-event–mediated component accounted for −2.3 per 1000, indicating most of the modest benefit was not explained solely by the recorded first events within 14 days.
Low-dose heparin had an approximately null total effect on 6-month death-or-dependency (3.0 per 1000). This null composite effect reflected opposing components: an unfavorable interventional direct component of 6.4 per 1000 and a favorable first-event–mediated component of −3.4 per 1000. In other words, early-event mediation tended to offset an unfavorable direct effect.
Higher-dose heparin also yielded a near-null composite effect on the disability-inclusive endpoint (−1.9 per 1000). However, a sensitivity analysis focused on mortality revealed an unfavorable signal for higher-dose heparin, reported as 15.6 per 1000 in the mortality sensitivity analysis. The main composite estimates thus conceal potentially divergent effects on death versus dependency components.
In this historical megatrial dataset, small or near-null effects on a 6-month death-or-dependency outcome were compatible with opposing early pathways: prevention of early ischemic events versus increased early bleeding hazards. Recorded first events within 14 days explained only part of the contrasts between randomized treatments and later outcomes; much of the treatment–outcome contrast was attributable to direct components not captured by the selected first-event mediator.
The authors caution that these estimates are methodological and interpretive, derived from a historical trial dataset, and are not contemporary prescribing guidance.
The report is a secondary analysis of a historical randomized trial and therefore subject to the limits of the original data collection, event ascertainment, and the predefined mediator list. The authors note that first recorded early events explained only part of outcome contrasts, implying unmeasured early or later pathways likely contributed. The paper frames results as methodological insight into how early harms and benefits can offset when aggregated into a single later outcome, rather than as definitive guidance for current antithrombotic prescribing in acute ischemic stroke.
These findings illustrate the need for careful mediation-aware analyses when interpreting composite long-term outcomes after acute stroke and remind clinicians that historical trial estimates may not directly translate into current practice.