Real-World Association Between Elevated Lipoprotein(a) and Cerebrovascular and Cardiovascular Outcomes: A Federated EHR Cohort Study

BackgroundElevated lipoprotein(a) [Lp(a)] is causally implicated in atherosclerotic cardiovascular disease, but its prospective association with incident ischemic stroke in real-world clinical populations remains incompletely characterized. Observational data are complicated by testing-indication bias and aggressive preventive management in identified high-Lp(a) patients. MethodsWe conducted a retrospective cohort study using the TriNetX US Collaborative Network (66-67 healthcare organizations). Adults with Lp(a) measurements (LOINC 10835-7; January 2015-December 2025) were categorized as C-HIGH (Lp(a) [&ge;]50 mg/dL) or C-LOW (<50 mg/dL) and balanced by 1:1 propensity score matching on cardiovascular diagnoses, medications, tobacco history, and four laboratory variables. Pre-specified analyses included an alternative cutoff ([&ge;]30 mg/dL), dose-response evaluation across three strata, and a 2-year landmark analysis. ResultsAfter expanded propensity score matching, the primary analysis included 97,882 matched adults (48,941 per arm). Incident ischemic stroke/TIA occurred in 2.45% vs 2.69% (Cox HR 0.956, 95% CI 0.878-1.041; log-rank p=0.299). The Lp(a) [&ge;]30 mg/dL sensitivity analysis (121,076 matched) yielded HR 0.947 (0.878-1.023; p=0.168). Dose-response analysis showed no significant association at any stratum. A 2-year landmark analysis confirmed a null late-period effect (HR 1.071, 0.905-1.268; p=0.427). A post-hoc composite (stroke/TIA, cardiac arrest, heart failure) was null (HR 0.979; p=0.512). A broader MACE-like composite including MI yielded a nominally significant Cox HR of 1.058 (1.004- 1.116; p=0.034), with entirely null crude estimates, attributable to differential follow-up time. ConclusionsElevated Lp(a) was not associated with incident ischemic stroke or TIA across multiple thresholds and follow-up windows in this large federated cohort. These real-world findings are consistent with testing-indication bias and treatment attenuation in clinically identified high-Lp(a) populations.