LAUSR

Atrial Fibrillation and Cancer— Validation in the Real World To the Editor We read with interest the article by Conen et al1 published in the April issue of JAMA Cardiology. In an analysis of the Women’s Health Study, the authors found an association between incident atrial fibrillation (AF) and cancer. While the study cohort was large, the incidences of AF and cancer were low, placing the findings at risk of a type II error. In addition, this association was not generalizable to men. Therefore, we sought to validate these findings in a larger realworld cohort among both sexes. Using the Explorys platform,2 an aggregated electronic database spanning inpatient and outpatient records from 26 major integrated health care systems across the United States, we identified 11 207 890 women and 9 003 530 men with characteristics similar to the population in the study by Conen et al1 (ie, 45 years or older and free of major cardiovascular disease and cancer at baseline). These cohorts were examined for newonset AF or cancer between June 2011 and May 2016 and analyzed using logistic regression. There were 388 270 and 833 520 patients with incident AF and cancer, respectively. The crude incidence of AF was 0.31% per year for women and 0.42% per year for men, similar to the 0.22% reported by Conen et al.1 Among those with new-onset AF, the age-adjusted odds ratio (age-OR) for incident cancer was 1.66 (95% CI, 1.62-1.71) for women and 1.66 (95% CI, 1.62-1.70) for men, comparable with the age-adjusted hazard ratio of 1.58 (95% CI, 1.34-1.87) reported by Conen et al.1 The risk of incident cancer following new-onset AF was highest during the first year (women: age-OR, 2.58; 95% CI, 2.50-2.66; men: age-OR, 2.58; 95% CI, 2.51-2.65) and returned to baseline risk by the following year, in accordance with the findings by Conen et al.1 For patients with new cancer, we observed an annual incidence of 0.78% per year in women and 0.85% per year in men, consistent with the 0.77% per year rate reported by Conen et al.1 Among those with new cancer, the risk of incident AF was highest during the first year (women: age-OR, 4.46; 95% CI, 4.35-4.58; men: age-OR, 4.37; 95% CI, 4.27-4.47), which was comparable with the age-adjusted hazard ratio of 4.67 (95% CI, 2.85-7.64) during the first 3 months reported by Conen et al1. Finally, we confirmed that the risk of incident AF in those with new cancer was persistently elevated beyond 1 year (women: age-OR, 1.30; 95% CI, 1.25-1.36; men: age-OR, 1.22; 95% CI, 1.17-1.27), a finding not statistically significant in the study by Conen et al.1 Thus, we validate the association between AF and cancer among women and men in a real-world cohort more than 500fold larger than that used by Conen et al.1 Further studies are needed to better delineate and validate these findings in a prospective manner.