A post-hoc analysis of the ASPREE randomized controlled trial examined 1,270 U.S. participants aged 65+ who self-identified as non-Hispanic Black (n=897) or Hispanic (n=373). Using a Random Survival Forest X-learner to estimate individualized treatment effects, daily 100-mg aspirin was associated with a 35% lower hazard of losing disability-free survival (HR 0.65; 95% CI 0.45–0.93) overall. Among the highest predicted-benefit tertile, the effect was striking — a 64% hazard reduction (HR 0.36) and an 11.1 percentage-point absolute risk reduction over five years — while no benefit appeared in the lowest tertile.
The original ASPREE trial's null result for aspirin across the full older-adult population dampened enthusiasm for prophylactic aspirin use, leading major guidelines to discourage its routine use in adults over 60. This subgroup signal — now interrogated with machine-learning heterogeneity methods — reopens a narrower but clinically meaningful question: can we identify who actually benefits? That is a genuinely important precision-medicine framing, particularly given that minoritized populations are historically underrepresented in cardiovascular trials.
However, caution is essential. This is a post-hoc, hypothesis-generating analysis with a modest analytic cohort, and machine-learning benefit scores derived from observational splits within a trial carry real overfitting risk. The authors themselves call for external validation before clinical adoption. Critically, this is a preprint posted to medRxiv and has not yet undergone peer review — findings and effect sizes could change materially. Incremental but directionally important.
