Refining Predictive Risk Models for Stroke in Atrial Fibrillation: A Scoping Review and Meta-analysis for Aotearoa New Zealand, Māori and Pacific Peoples

Abstract

AIM: The predictive risk model CHA2DS2 VASc helps clinicians assess the risk of stroke in patients with atrial fibrillation (AF). Originally developed and validated in predominantly European populations, it may not accurately reflect the stroke risk for diverse ethnic groups; in Aotearoa New Zealand, Māori and Pacific peoples with AF are at higher stroke risk. As part of global efforts to address health inequities, there is growing interest in adapting predictive models to suit local- and ethnic-specific risks better. Our objectives were to determine: 1) if stroke risk from AF varies by ethnic background/race, 2) stroke rates in non-anticoagulated AF cohorts, and 3) model performance of CHA2DS2 VASc across different geographical regions. Finally, we provide an overview of methodological considerations for risk model development.

METHODS: We searched English language peer-reviewed studies reporting stroke rates in unselected cohorts with AF, published between 1995 and 2024. For stroke risk, we included cohorts with over 5,000 non-anticoagulated patients. The sources of evidence were PubMed, Scopus and EMBASE.

RESULTS: Twenty-seven studies were eligible for inclusion. We found significantly elevated stroke risk in African Americans and Hispanics with AF compared with whites (odds ratio [OR] 1.44 [95% confidence interval (CI) 1.25-1.66] and OR 1.11 [95% CI 1.05-1.18] respectively). In Māori and Pacific peoples with AF, the risk of stroke was higher than in New Zealand Europeans, but this difference was not significant (OR 1.28 [95% CI 0.89-1.82], p=0.18 and OR 1.29 [95% CI 0.93-1.52], p=0.17 respectively). Stroke risk (0.6/100-6.8/100 person-years) and CHA2DS2 VASc performance (c-statistics 0.55-0.8) varied substantially between studies.

CONCLUSION: We support the local refinement of risk prediction models in line with cardiology society recommendations.