Abstract P106: Dietary Patterns and Incidence of Atrial Fibrillation: the Health Professionals Follow-up Study (HPFS)

Introduction: Atrial fibrillation (AF) is a growing public health problem. Overall diet quality has proven informative for studying chronic diseases, but potential relations with AF risk are unknown. Methods: We followed 39,840 US male health professionals free from AF at baseline (age 70.5±9.8 y) for incidence of AF from 2002 to 2010. Diet was assessed every 4 ys by validated questionnaires, with cumulative updating of past diet. We also performed analyses evaluating only more recent diet. A diet pattern was characterized based on biologic considerations using 8 dietary factors that might influence risk factors for AF: higher fruits, vegetables, whole grain, seafood, nuts, and polyunsaturated to saturated fat ratio, and lower glycemic load and trans-fat. Each factor was scored in quintiles (1- 5), in ascending order for more healthful factors and descending order for less healthful factors. These scores were summed to quantify the diet pattern score (range 8-40). Incident AF was identified from initial self-report that was also subsequently confirmed by detailed supplementary questionnaire; these methods were validated against medical records in a subset (positive predictive value=95%). We estimated risk using Cox models adjusting for age, prevalent diseases, lifestyle habits, and health-seeking behaviors (greater multivitamin use; more routine physical, sigmoidoscopy, and rectal exams). Results: During 308,144 person-years follow-up, 879 incident AF cases occurred. Men with higher diet scores were older, less likely to smoke, more likely to be physically active and be on aspirin, and more likely to have health seeking behaviors. The multivariable-adjusted HR (95%CI) for quintiles 2 to5 vs. quintile 1 were 1.06 (0.84, 1.33), 1.04 (0.82, 1.32), 1.15 (0.92, 1.45), and 1.26 (0.98, 1.60), respectively (p trend=0.04). Exploring each of the 8 dietary factors individually, none were significantly associated with incident AF, except seafood which was inversely associated (HR for quintile 5 vs. 1=0.79; 95%CI=0.63, 1.00). Health-seeking behaviors were associated with trend toward AF risk (extreme quartile HR 1.21, 95% CI 0.92-1.59). Stratified by health seeking behavior, the diet score showed positive association with AF only in men with greatest health seeking behaviors (p trend 0.01), but not in the other groups (p>0.8). Sensitivity analyses using only more recent diet, stratifying by aspirin use, or limiting to symptomatic AF cases did not appreciably alter results. Conclusions: Contrary to our hypothesis, a more healthful diet pattern was linked to higher AF risk. This association might partly relate to unmeasured or residual confounding due to greater health seeking behaviors that could increase AF diagnosis. Overall, our novel findings suggest no major benefits of a generally healthy diet on AF incidence, except perhaps for seafood consumption.