Sex-Specific Exposure-Effect Relationship Between Physical Activity and Incident Atrial Fibrillation in the General Population: A Dose-Response Meta-Analysis of 16 Prospective Studies

Abstract

Background: Since evidence regarding the relationship between physical activity (PA) and atrial fibrillation (AF) incidence is inconsistent among studies, we performed a dose-response meta-analysis to comprehensively evaluate the exposure-effect association between PA and incident AF and the potential sex difference in the general population. Methods: The PubMed and Embase databases were searched for eligible studies published up to July 2020 (PROSPERO: CRD42018091692). The non-linear or linear exposure-effect relationship between PA and AF was examined using the robust error meta-regression method. Results: A total of 16 prospective studies involving 1,449,017 individuals and 39,884 AF cases were included. We observed an inverse non-linear association between PA level and incident AF (I ² = 0%, p _{non-linearity} < 0.001). In the linear model, a 5 metabolic equivalent of task (MET)-h/week increase in PA was associated with a decreased risk of AF [risk ratio (RR) = 0.992, 95% confidence interval (CI): 0.988-0.996, I ² = 0%]. In the sex-stratified analysis, we observed an inverse non-linear relationship between PA level and AF risk in females (I ² = 90%, p _{non-linearity} < 0.0001) but not in males (I ² = 0%, p _{non-linearity} = 0.40). In the linear model, a 5 MET-h/week increase in PA was associated with a reduced risk of AF in females (RR = 0.982, 95% CI: 0.975-0.989, I ² = 71%) but not in males (RR = 0.998, 95% CI: 0.994-1.002, I ² = 0%), with a significant interaction observed between the two groups (p _interaction < 0.0001). Conclusion: There was an inverse non-linear relationship between PA level and incident AF in the general population. The beneficial effect of PA in reducing AF risk might be predominantly observed in females.

Keywords: atrial fibrillation; dose-response; meta-analysis; physical activity; risk factor.

Figure 1

Flow chart of study selection for our dose–response analysis.

Figure 2

Dose–response analysis of physical activity and atrial fibrillation in the total population. (A,B) Non-linear and linear models of PA and AF in the range of 0–80 MET-h/week. (C,D) Non-linear and linear models of PA and AF at the high level of PA. The bold and dashed lines represent the estimated RR and 95% CI, respectively. The non-linear models were fit by using a restricted cubic spline. PA, physical activity; AF, atrial fibrillation; RR, risk ratio; CI, confidence interval.

Figure 3

Dose–response analysis of physical activity and atrial fibrillation in the sex-stratified analysis. (A,B) Non-linear and linear models of PA and AF in females. (C,D) Non-linear and linear models of PA and AF in males. The bold and dashed lines represent the estimated RR and 95% CI, respectively. The non-linear models were fit by using a restricted cubic spline. PA, physical activity; AF, atrial fibrillation; RR, risk ratio; CI, confidence interval.