ABSTRACT NUMBER: ESOC2026A262 STROKE RISK WITH SGLT2I, GLP-1RA, AND DPP-4I IN TYPE 2 DIABETES: REAL-WORLD LONGITUDINAL ESTIMATES IN THE CONTEXT OF TRIAL AND OBSERVATIONAL EVIDENCE

Abstract Background and aims The impact of SGLT2 inhibitors (SGLT2i), GLP1 receptor agonists (GLP1RA), and DPP4 inhibitors (DPP4i) on stroke risk in type 2 diabetes remains uncertain. RCTs show stroke reduction with GLP1RA but neutral findings for SGLT2i and DPP4i, while observational studies often suggest broader benefits. This study estimated the effects of these drug classes on stroke risk using nationwide longitudinal data and causal inference methods and compared findings with meta-analyses of RCTs and observational studies. Methods This nationwide longitudinal study included all individuals with type 2 diabetes first registered in the Swedish National Diabetes Register from 2010-2019. Treatments, comorbidities, and laboratory values were updated over time. Marginal structural models with inverse probability of treatment weighting were used to account for time-varying exposures, confounders, and treatment switching. Hazard ratios were obtained from weighted Cox regression. Estimates were pooled across five multiply imputed datasets and compared with systematically identified meta-analyses. Results We included 267,708 individuals [58.9% males, baseline mean(SD) age 62.8(11.6), mean(SD) follow-up 1640 (1039) days] contributing 3,329,254 clinical registrations. During follow-up, 11,494 (4.29%) ischemic and 13,130 (4.90%) total strokes occurred. GLP1RA use was associated with a lower risk of ischemic (HR 0.80, 95% CI 0.67-0.97) and total stroke (HR 0.82, 95% CI 0.69-0.98). SGLT2i and DPP4i showed no clear associations. Results aligned with RCT meta-analyses. Conclusions GLP1RA were associated with reduced stroke risk, while SGLT2i and DPP4i showed neutral effects. By applying longitudinal causal inference methods and comparing directly with meta-analyses, this work helps bridge discrepancies between trial and real-world evidence. Conflict of interest AM: Nothing to disclose, TA: Nothing to disclose, DB: Nothing to disclose, BE: Nothing to disclose, MvE: Nothing to disclose, AV: Nothing to disclose, KSS: Nothing to disclose.