Early-life developmental factors may influence the risk of cerebrovascular disease in adulthood, but the causal relationship between birth weight and different subtypes of ischemic stroke remains unclear. This study used Mendelian randomization (MR) to investigate the causal association between birth weight and the risk of ischemic stroke (IS) and its major clinical subtypes in adulthood.

A two-stage analytical design combining univariable and multivariable MR (UVMR and MVMR) was adopted. Birth weight data were obtained from the Early Growth Genetics (EGG) Consortium, with a sample size of 143677. Stroke outcome data were obtained from the GIGASTROKE consortium, with a total sample size of approximately 1.29 million, including overall IS and 3 major clinical subtypes: cardioembolism (CE), large artery atherosclerosis (LAA), and small vessel disease (SVD). To ensure sufficient instrumental variables in the multivariable analysis, the screening threshold for birth weight instruments was set at P<1×10^-5. Adult body mass index (BMI) and systolic blood pressure (SBP) were included as covariates in the MVMR model to adjust for adult body size and blood pressure. Causal inference was mainly performed using the random-effects inverse variance weighted (IVW) method. F-statistics, MR-Egger intercept test, and MR-Plelotropy RESidual Sum and Outlier (MR-PRESSO) test were used to assess instrument strength, horizontal pleiotropy, and potential outlier loci. The significance threshold was set at P<0.0125 after Bonferroni correction. Fetal-only genetic effect estimates derived from structural equation modeling were further used in a supplementary sensitivity analysis to preliminarily assess the potential influence of maternal genetic effects on the results.

In the univariable analysis, no significant causal association was observed between birth weight and IS or any of its subtypes. The multivariable analysis included 19 to 21 independent core single-nucleotide polymorphisms (SNPs). The univariable F-statistics were all >10, whereas the conditional F-statistic for birth weight was <10, suggesting a potential risk of weak instruments in the conditional model. After adjusting for adult BMI and SBP, MVMR analysis showed that genetically predicted higher birth weight was associated with a reduced risk of LAA (OR=0.736, 95%CI: 0.587 to 0.922, P=0.0076). No significant associations were observed with IS (OR=0.943, 95%CI: 0.869 to 1.023, P=0.1566), CE (OR=1.018, 95%CI: 0.857 to 1.210, P=0.8369), or SVD (OR=0.850, 95%CI: 0.688 to 1.050, P=0.1323). Supplementary sensitivity analysis using fetal-only genetic effect estimates within the multivariable framework showed a direction consistent with the main analysis for the association between birth weight and reduced LAA risk. However, because only a limited number of core SNPs were available and the association reached nominal significance (P=0.033), this result was used only to support the directional consistency of the main analysis.

After adjusting for adult BMI and SBP, genetically predicted higher birth weight is associated with a reduced risk of LAA in adulthood, whereas no significant causal associations are observed with the risk of overall IS, CE, or SVD in adulthood.