$Objective:$ To implement a mendelian randomization (MR) approach to determine whether type 2 diabetes mellitus (T2D), fasting glucose, fasting insulin, and body mass index (BMI) are causally associated with specific ischemic stroke subtypes.

$Methods:$ MR estimates of the association between each possible risk factor and ischemic stroke subtypes were calculated with inverse-variance weighted (conventional) and weighted median approaches, and MR-Egger regression was used to explore pleiotropy. The number of single nucleotide polymorphisms (SNPs) used as instrumental variables was 49 for T2D, 36 for fasting glucose, 18 for fasting insulin, and 77 for BMI. Genome-wide association study data of SNP-stroke associations were derived from METASTROKE and the Stroke Genetics Network (n = 18,476 ischemic stroke cases and 37,296 controls).

$Results:$ Conventional MR analysis showed associations between genetically predicted T2D and large artery stroke (odds ratio [OR] 1.28, 95% confidence interval [CI] 1.16-1.40, p = 3.3 × 10(-7)) and small vessel stroke (OR 1.21, 95% CI 1.10-1.33, p = 8.9 × 10(-5)) but not cardioembolic stroke (OR 1.06, 95% CI 0.97-1.15, p = 0.17). The association of T2D with large artery stroke but not small vessel stroke was consistent in a sensitivity analysis using the weighted median method, and there was no evidence of pleiotropy. Genetically predicted fasting glucose and fasting insulin levels and BMI were not statistically significantly associated with any ischemic stroke subtype.

$Conclusions:$ This study provides support that T2D may be causally associated with large artery stroke.