Abstract Body: Objective
The contribution of atrial fibrillation (AF) to global stroke burden is not well described. This study synthesized available evidence using the Burden of Proof (BoP) methodology and a Meta-regression-Bayesian, Regularized, trimmed (MR-BRT) model, to provide a robust stroke risk estimate for the general population while accounting for multiple potential biases.

Methods
We conducted a systematic review and Bayesian meta-analysis of population-based observational studies of AF and stroke, sourced from PubMed and EMBASE (1970-2024). To adequately capture potential global burden, all (paroxysmal, persistent/sustained, chronic and, unspecified) types of AF were considered as exposure and the study outcome was any types of incident (ischemic, hemorrhagic, or total) stroke. Relative risk (RR) assessments - hazard ratio, risk ratio, and odds ratio were extracted from published observational studies. Binary covariates were generated to identify potential systematic bias, including the representativeness, precision in exposure and outcome status determination, variations in definitions, and adjustment was made for confounders - age, sex, education, lifestyle factors, study design, and comorbidities. We used the MR-BRT model to estimate a conservative mean AF-stroke dichotomous risk function and the overall pooled mean RR, adjusting for study heterogeneity and publication bias.

Results
Of 6,447 articles reviewed, 49 studies were included, with participants aged 18 to 99 and sample sizes ranging from 250 to 4.3 million. Individuals with AF had a 99% higher risk of stroke (RR 1.99, 95% uncertainty interval [UI]: 1.91, 2.08), with an increasing range of 92% to 108%, indicating a significant association despite study heterogeneity (γ). The most conservative estimate was 1.92 (95% UI: 1.90, 1.99). No publication bias was found based on the funnel plot analysis.

Conclusion
This study quantifies the risk of all strokes in the general population due to any type of AF. Differences in age, comorbidities, and medication use (especially anticoagulants) across populations and the level of confounder adjustment may influence the observed risk. The BoP framework highlights the importance of synthesizing high-quality evidence to guide targeted interventions and health policy decisions aimed at reducing stroke globally. These findings support more aggressive AF management, particularly in lower-resource settings, as part of stroke prevention strategies to reduce stroke incidence.