Abstract 4147213: Does Well-Controlled, Recent-Onset Diabetes Increase Stroke Hazard in Atrial Fibrillation? A Population-Based Cohort Study

Introduction: Stroke prevention with anticoagulation is an important aspect of atrial fibrillation (AF) management. Anticoagulation is often based on the premise that the risk of stroke increases with each additional CHA2DS2VASc risk factor. Diabetes is accepted as a risk factor for stroke in AF, but the threshold at which stroke risk increases is uncertain. Specifically, it is unknown if AF patients with recent-onset well-controlled diabetes (ROWCD) are at increased risk of stroke in AF compared to AF patients without diabetes. Hypothesis: We hypothesize that patients with AF and diabetes of <5 years’ duration with glycated hemoglobin (HbA1c) ≤7% who do not use insulin (henceforth designated as the ROWCD group) do not have a significantly different adjusted hazard of stroke than AF patients without diabetes. Methods/Approach: Using linked administrative databases in Ontario, Canada, we conducted a population-based retrospective cohort study of patients ≥66 years who had a new AF diagnosis between Apr 1 2013 and Mar 31 2022. Exclusion criteria included insulin use. Cause-specific hazard regression was used to quantify the adjusted hazard ratio (HR) for stroke over 2 years of follow-up in patients with AF plus diabetes relative to AF patients without diabetes (adjusting for baseline stroke risk factors and time-varying anticoagulation status). The analysis was repeated after excluding people with diabetes of ≥5 years duration, HbA1c >7%, or missing HbA1c data (to generate the ROWCD subset). Results/Data: The primary analysis included 233,692 patients with AF (mean age 78.8 years, 50.9% male), of whom 64,972 (27.8%) had diabetes. Diabetes was associated with an adjusted HR of 1.15 (1.08-1.22) for stroke in the full cohort (p< 0.001). After excluding people with diabetes of ≥5 years duration, HbA1c >7% or missing HbA1c, there were 7,415 patients with ROWCD. Mean diabetes duration was 3.0 years, mean HbA1c was 6.2%, with 38.5% metformin use, 10.6% dipeptidyl peptidase-4 inhibitor use, and 5.3% sulfonylurea use. There were 3,503 (2.1%) strokes in the non-diabetes group and 139 (1.9%) in the ROWCD group. ROWCD was not significantly associated with adjusted stroke hazard (HR 0.96, 95% CI 0.81-1.13, p=0.61). Conclusions: These findings suggest that people with AF who have ROWCD may not warrant a point on the CHA2DS2VASc score, as this subset of diabetes is not associated with increased hazard of stroke in AF.