Methods:

This meta-analysis used the databases MEDLINE (from 1948 to November 31, 2011), EMBASE (from 1988 to 2011 week 48), and COCHRANE (from inception until the third quarter of 2011) to identify all available studies reporting cases of atrial fibrillation based on the amount of exercise. References of the retrieved articles and the review articles were also screened for eligible studies. Both prospective cohort studies and nested case-control studies examining the relation of regular physical activity and atrial fibrillation risk were eligible for inclusion. Studies on professional athletes, studies not reporting atrial fibrillation in controls, as well as traditional case-control studies, where it is difficult to establish temporality between exposure and outcome, were excluded. All of the studies used for the meta-analysis divided subjects into four or five groups based on the cumulative physical activity per week, amount and intensity of work-related physical activity, leisure-type physical activity, or amount of exercise per week. For all of the studies included, the extreme groups (i.e., maximum vs. minimal amount of physical activity) were used for the current analyses. Extracted data from the eligible studies were meta-analyzed using the fixed-effects model.