Pooling data from two large-scale randomized controlled trials across 10,712 participants followed for up to three years, this analysis applied the GenericML framework — using Best Linear Predictor, Group Average Treatment Effects, and baseline profile classification — to test whether certain patient subgroups benefit more from intensive blood pressure control than others. Intensive BP control produced only a modest average cardiovascular risk reduction (BLP β1 = −0.0136), and critically, no statistically significant evidence of treatment effect heterogeneity emerged. Even participants clustered into metabolically adverse high-risk profiles showed no meaningfully different treatment response compared to the lowest-benefit group.
This finding carries substantial implications for precision cardiovascular medicine. A dominant assumption in hypertension management — that machine learning can identify which patients benefit most from aggressive BP targets — receives a pointed challenge here. The landmark SPRINT and ACCORD trials already showed divergent population-level results; this reanalysis suggests those differences may not translate into actionable individual-level predictions. For clinicians and patients, the message is sobering: identifying high-risk phenotypes does not automatically reveal who gains the most from treatment intensification. Importantly, this is a preprint not yet peer-reviewed, and the pooling of heterogeneous trial populations may mask subgroup signals. The work is methodologically rigorous but confirmatory in tone — reinforcing caution rather than overturning guidelines. Adults on intensive BP regimens should not alter management based on this finding alone, pending peer review.
