Association between plasma homocysteine levels and post-stroke cognitive impairment: a pooled analysis of 564 patients from two independent cohorts

Introduction Post-stroke cognitive impairment (PSCI) affects one-third of stroke survivors, but early biomarkers are lacking. We tested whether plasma homocysteine (Hcy) independently predicts PSCI. Methods Two acute ischemic stroke cohorts were merged (564 patients). PSCI was defined as a Montreal Cognitive Assessment score ≤ 23 at 3–6 months. Associations were tested with multivariable logistic regression, restricted cubic spline, sensitivity, subgroup, and mediation analyses. Results PSCI occurred in 36.9%. Hcy was higher in PSCI (21.99 vs. 15.87 μmol/L, p < 0.001) and associated after adjustment (OR = 1.019 per μmol/L, 95% CI 1.002–1.037, p = 0.030); in Cohort B, where age, sex, education and infarct laterality were recorded, the association remained significant after full adjustment (OR = 1.034, 95% CI 1.011–1.057, p = 0.004). The association was graded, with the steepest relative increase at lower Hcy levels and a plateau at higher levels (restricted cubic spline; non-linearity p = 0.054, borderline), and was not statistically significant at Hcy ≥ 20 or ≥ 25 μmol/L. Combining Hcy with Lp-PLA2, NIHSS, and β2-MG improved discrimination (AUC = 0.748, an apparent estimate that was not externally validated). In an exploratory counterfactual mediation analysis the proportion mediated was small (≈ 4% via NIHSS and ≈ 2% via Lp-PLA2). Discussion Elevated Hcy is associated with PSCI in this pooled analysis, and the association persisted in Cohort B after adjustment for age, sex, education and infarct laterality. The relationship was graded rather than sharply thresholded. As an observational finding, and given the absence of folate, vitamin B12 and renal-function data and the neutral results of B-vitamin trials such as VITATOPS, Hcy is best regarded as a risk marker rather than a proven causal target; dedicated trials are needed before any clinical recommendation.