Serum Uric Acid as a Sex-Dependent Risk Marker of Post-Stroke Epilepsy After Acute Ischemic Stroke: Complementary...

Observational studies have consistently reported associations between serum uric acid (SUA) and post-stroke epilepsy (PSE) in acute ischemic stroke (AIS), including a recently published large cohort study identifying a U-shaped relationship. However, whether these associations reflect causality and whether SUA provides incremental predictive value beyond established clinical factors remains unknown.

we conducted a two-sample Mendelian randomization (MR) analysis to test whether lifelong genetically predicted serum urate shares a genetic causal architecture with general epilepsy susceptibility, using 299 independent genetic instruments. In parallel, we performed a prespecified secondary analysis of a multicenter AIS cohort (n = 21,459) using multivariable logistic regression with restricted cubic splines (RCS) to characterize the observational SUA-PSE relationship. Formal sex interaction testing (Wald tests) and sex-stratified spline analyses were performed. Predictive utility was evaluated through discrimination (AUC), calibration, decision curve analysis (DCA), and internal validation with 1000 bootstrap resamples. The two analytical components address related but distinct biological questions: observational analysis tests the admission-SUA-to-1-year-PSE pathway, while MR tests whether lifelong genetically predicted urate shares a causal architecture with general epilepsy susceptibility.

MR analysis found no evidence supporting a shared lifelong genetic causal architecture between serum urate and general epilepsy susceptibility across all four methods (IVW: OR = 1.043, 95% CI 0.926-1.174, p = 0.487; MR-Egger p = 0.582; weighted median p = 0.589; weighted mode p = 0.996), with no pleiotropy (Egger intercept p = 0.865) or heterogeneity (Cochran Q p = 0.206). In the observational cohort (n = 21,459; 936 PSE events (4.36%)), RCS analyses confirmed a significant nonlinear SUA-PSE association (P for nonlinearity < 0.001). Importantly, a formal sex-by-SUA interaction revealed strikingly divergent patterns (overall p < 0.001; nonlinear component p = 0.0002): Women in the highest SUA tertile had markedly elevated PSE risk (adjusted OR = 2.33, 95% CI 1.71-3.19), a finding confirmed with sex-specific tertile cut-points (OR = 1.50, p = 0.005). The apparent protective association in men using overall tertiles (OR = 0.36, 95% CI 0.28-0.46) was attenuated and non-significant with sex-specific cut-points (OR = 0.84, p = 0.16), indicating sensitivity to stratification method. Adding SUA to a base clinical model produced a modest but statistically significant increment in discrimination (AUC: 0.8498 to 0.8607, ΔAUC = 0.011, DeLong p = 0.001) with consistent positive net benefit on DCA.

Genetically predicted lifelong SUA does not share a causal genetic architecture with general epilepsy susceptibility; this null does not exclude acute post-stroke, context-specific mechanisms, which operate on biological timescales outside the scope of Mendelian randomization. The sex-dependent observational association (female high-SUA OR = 1.50 with sex-specific tertiles; sex × SUA continuous interaction p = 3.99 × 10^-9) generates the hypothesis that sex-stratified SUA monitoring merits prospective investigation, pending external validation.