Time-varying incidence of ischaemic stroke in untreated secondary atrial fibrillation: a 14-centre competing risk analysis

Background The 2024 ESC guidelines recommend "Evaluation and dynamic reassessment" for atrial fibrillation (Class I, Level of Evidence C). In secondary AF triggered by acute illness, establishing the optimal timing for reassessment is hindered by critical literature gaps. Historical observational data frequently suffer from immortal time bias and fail to properly account for the high competing risk of acute mortality. These limitations obscure the true natural history and early time-varying incidence of ischaemic stroke in untreated patients. Purpose We sought to map the natural history and precise time-varying incidence of ischaemic stroke using a competing risk framework, establishing an evidence-based timeline to inform clinical reassessment. Methods We analysed a 14-centre registry spanning community, academic, and quaternary settings (2014–2025). From an initial 20,537 patients with an AF diagnosis, we excluded 1,462 with prior AF. To minimise immortal time bias, we excluded patients with same-day mortality (n=93) and in-hospital stroke prior to treatment assignment (n=105). We evaluated a natural history cohort of 12,905 patients managed without anticoagulation. We applied Fine-Gray subdistribution hazard models, explicitly treating death without ischaemic stroke as a competing risk. Results The untreated cohort (n=12,905) had a mean age of 72.0 ± 12.9 years. The cohort was 41.8% female (n=5,391) and 89.0% White (n=11,484). Mean admission blood pressure was 130.9 ± 28.5 / 77.8 ± 19.8 mmHg, with a mean BMI of 29.7 ± 8.2. Mean ejection fraction was 54.8 ± 16.1%. The mean duration to first AF diagnosis was 0.2 ± 1.9 days, with 0.4 ± 1.0 days in recurrent AF. Prior ischaemic stroke was present in 13.4% (n=1,732), and the mean length of stay was 10.2 ± 14.1 days. Acute clinical triggers included sepsis (n=4,547), kidney failure (n=7,623), respiratory failure (n=6,438), metabolic encephalopathy (n=1,990), liver injury (n=322), and alcohol toxicity (n=1,730). The 15-day cumulative incidence of ischaemic stroke reached 16% (95% CI: 15%–16%). Between days 15 and 30, the incidence rate attenuated, reaching 18% (95% CI: 17%–19%) at 30 days. Concurrently, all-cause mortality reached 15% (95% CI: 14%–15%) by day 30, driving a 30% (95% CI: 29%–30%) composite risk of ischaemic stroke or death. Conclusion(s) The natural history of untreated secondary AF is characterised by an early-peaking ischaemic stroke hazard concentrated predominantly within the first 15 days, occurring concurrently with substantial competing mortality. By resolving previous methodological limitations through rigorous competing risk analysis, these findings map the precise temporal kinetics of the disease. Recognising that the highest thromboembolic hazard occurs within the first two weeks provides an evidence-based timeline that can inform the optimal scheduling of clinical evaluation and dynamic reassessment in acute care settings.