Validation of an atrial fibrillation risk algorithm in whites and African Americans.

BACKGROUND: We sought to validate a recently published risk algorithm for incident atrial fibrillation (AF) in independent cohorts and other racial groups. METHODS: We evaluated the performance of a Framingham Heart Study (FHS)-derived risk algorithm modified for 5-year incidence of AF in the FHS (n = 4764 participants) and 2 geographically and racially diverse cohorts in the age range 45 to 95 years: AGES (the Age, Gene/Environment Susceptibility-Reykjavik Study) (n = 4238) and CHS (the Cardiovascular Health Study) (n = 5410, of whom 874 [16.2%] were African Americans). The risk algorithm included age, sex, body mass index, systolic blood pressure, electrocardiographic PR interval, hypertension treatment, and heart failure. RESULTS: We found 1359 incident AF events in 100 074 person-years of follow-up. Unadjusted 5-year event rates differed by cohort (AGES, 12.8 cases/1000 person-years; CHS whites, 22.7 cases/1000 person-years; and FHS, 4.5 cases/1000 person-years) and by race (CHS African Americans, 18.4 cases/1000 person-years). The strongest risk factors in all samples were age and heart failure. The relative risks for incident AF associated with risk factors were comparable across cohorts and race groups. After recalibration for baseline incidence and risk factor distribution, the Framingham algorithm, reported in C statistic, performed reasonably well in all samples: AGES, 0.67 (95% confidence interval [CI], 0.64-0.71); CHS whites, 0.68 (95% CI, 0.66-0.70); and CHS African Americans, 0.66 (95% CI, 0.61-0.71). Risk factors combined in the algorithm explained between 47.0% (AGES) and 63.6% (FHS) of the population-attributable risk. CONCLUSIONS: Risk of incident AF in community-dwelling whites and African Americans can be assessed reliably by routinely available and potentially modifiable clinical variables. Seven risk factors accounted for up to 64% of risk.

Duke Scholars

Author Michael J Pencina Biostatistics & Bioinformatics, Division of Biostatistics