Prevention of incident atrial fibrillation in type 2 diabetes: nationwide target trial emulation of SGLT2 inhibitors versus GLP-1 receptor agonists

Background Atrial fibrillation (AF) is a frequent cardiovascular complication of type 2 diabetes mellitus (T2DM). We evaluated the comparative 10-year effectiveness of SGLT2 inhibitors (SGLT2i) versus GLP-1 receptor agonists (GLP-1 RA) for incident AF using a nationwide target trial emulation. Methods We emulated a target trial using electronic health records from a national health maintenance organisation (more than 4.8 million members, 2015 to 2025). Eligible adults had T2DM, no prior AF, valvular heart disease, or hypertrophic cardiomyopathy, and initiated an SGLT2i (dapagliflozin or empagliflozin) or a GLP-1 RA (dulaglutide, semaglutide, liraglutide, exenatide, or lixisenatide) with no exposure to the other class in the prior 6 months. A directed acyclic graph, constructed by two cardiologists, prespecified confounders: age, sex, body mass index, HbA1c, estimated glomerular filtration rate, diabetes duration, year of drug purchase, and comorbidities (heart failure, hypertension, coronary artery disease, cerebrovascular disease, chronic kidney disease, chronic obstructive pulmonary disease, obstructive sleep apnoea, malignancy, rheumatologic disease). Exact 1:1 matching on all prespecified covariates produced 24,309 patients per arm. For the absolute risk context, we assembled a matched conventional-therapy cohort (33,936 patients) excluding SGLT2i and GLP-1 RA, using a synthetic index date defined as the sum of the median time to treatment initiation in the active groups and each patient’s diabetes diagnosis date. Incident AF was identified from inpatient and outpatient records. Ten-year risks were estimated using Kaplan-Meier; risk differences (RD) and risk ratios (RR) were reported with 95% confidence intervals (CI). Results The 10-year AF risk in the matched conventional-therapy cohort was 9.97% (95% CI 9.58 to 10.36). Compared with reference therapy, AF risk was lower with SGLT2i (8.52%, 95% CI 7.77 to 9.25; RD -1.46%, 95% CI -2.29 to -0.62; RR 0.854, 95% CI 0.776 to 0.940) and with GLP-1 RA (8.29%, 95% CI 7.60 to 8.97; RD -1.68%, 95% CI -2.47 to -0.90; RR 0.831, 95% CI 0.759 to 0.911). Direct comparison showed no significant difference between drug classes (RD 0.23%, 95% CI -0.78 to 1.24; RR 1.028, 95% CI 0.911 to 1.159). Conclusions Initiation of either SGLT2i or GLP-1 RA was associated with a significant reduction in 10-year incident AF compared with matched conventional therapy. AF risk did not differ between SGLT2i and GLP-1 RA, supporting class selection based on other clinical priorities.