Causal Inference of Blood Pressure Reduction and Coronary Heart Disease Risk in the Framingham Study

Standard cardiovascular risk calculators, including the Framingham Risk Score and the ACC/AHA Pooled Cohort Equations, estimate the conditional probability P(CHD | SysBP = s) rather than the interventional quantity P(CHD | do(SysBP = s)). When confounding is present, this distinction has direct clinical consequences: observational estimates may systematically overstate the absolute benefit of antihypertensive treatment. We applied Pearl's do-calculus to the Framingham Heart Study Offspring Cohort (n = 4,240; primary analysis on 3,776 complete cases; 574 ten-year coronary heart disease (CHD) events). A structurally corrected directed acyclic graph (DAG) was specified, incorporating four biologically motivated corrections, and subjected to conditional independence testing. The average causal effect (ACE) of a 20 mmHg systolic blood pressure (SysBP) reduction was estimated by g-computation with 1,500-iteration bootstrap confidence intervals, corroborated by sex-stratified propensity score matching (PSM) and inverse probability weighting (IPW). Conditional average treatment effects (CATE) were estimated using R-Learner and T-Learner metalearners with gradient-boosted nuisance models. G-computation yielded an ACE of 3.40% absolute risk reduction (95% bootstrap CI: 2.64%–4.14%), compared with a naive observational estimate of 4.14%, a relative overestimation of approximately 21.8% (+35.7% on the log-odds scale). The E-value lower bound was 2.18. Statistically significant heterogeneity in treatment effect was detected across age strata (Kruskal–Wallis p < 0.001) and diabetes status (Mann–Whitney p < 0.001); however, diabetic subgroup estimates were unstable and underpowered (n = 109), and no reliable subgroup inference can be drawn without replication. These findings suggest that observational cardiovascular risk tools may overestimate the absolute benefit of blood pressure reduction, with implications for clinical risk stratification and prescribing thresholds.