IJCS | Volume 32, Nº4, July/August 2019

385 Garlipp et al. Physical activity and atrial fibrillation Int J Cardiovasc Sci. 2019;32(4):384-390 Original Article studies report a J or U curve between the correlation of intensity of physical activity and AF. 4-10 On the other hand, some studies show a linear correlation between the amount of physical activity and the development of AF. 11-14 Considering the regular practice of physical activity as a cardiovascular benefit and literature inconsistency about its potential arrhythmogenic effect, the objective of this study was to analyze the effects of physical activity on the incidence of AF. Methods Data source and search In order to ensure a review of studies on the effects of physical exercise on the development of atrial fibrillation, we searched the following databases: PubMed, BVS Health and Cochrane. The following descriptors were used: “atrial fibrillation”, “exercise”, “physical activity” and “exercise therapy”. The search was limited to articles published in English whose full texts were reviewed. We used prospective, retrospective, cross- sectional and cohort studies. Data collection took place between September and October 2017 and data analysis was performed between October and December 2017. References in all articles included were examined for other relevant publications. Participants were patients without AFwho underwent physical exercise and were followed up. Figure 1 shows the flowchart of the meta-analysis selection process according to the Jadad quality scale. Criteria for data inclusion and extraction The inclusion criteria were as follows: (1) study design: all cohort studies, prospective, cross-sectional, observational and randomized clinical trials with patients who performed physical exercises and the development of AF. The studies were excluded from the analysis if: (1) they included patients with previous atrial fibrillation, (2) they included athletes and/or patients submitted to vigorous physical exercise (3) they could not extract concrete data frompublished results, such as comments, letters, cases, abstracts, reviews, experimental studies and animal studies, (4) the results were not clearly reported. Screening, selection, data extraction and risk of bias evaluation were performed independently and duplicated by two researchers and, ultimately, the potential of disagreement and disagreement was resolved by the corresponding author. Quality assessment The quality assessments were evaluated by the composite scale of Jadad, a numerical score of zero meaning the weakest to seven meaning the strongest. The scale contains the following points: (1) generation of random sequence (0-2), (2) concealment of allocation (0-2), (3) blinding double (0-2), (4) description of withdrawals and drop-outs (0-1). The total score of 4-7 indicates high quality. Methods of data synthesis and risk of bias in individual studies All statistical analyses were provided using Review Manager 5.3 to provide the mean difference (MD) and the relative risk (RR) ratiowith 95% confidence intervals (95% CI). The statisticalmethod of heterogeneity indexwas used to assess heterogeneity. The level of significance was 5%. Results Search on the chosen databases resulted in 731 articles. After reviewing the titles, abstracts and articles repeated, a total of 11 studies were included in this systematic review. Study and characteristics of patients In the 11 studies included in the meta-analysis, there were 276,323 participants and, in the studies, the number of participants ranged from 2,014 to 81,317, with ages varying between 12 and 90 years. The characteristics of the studies and the patients are presented in Table 1. Four studieswere prospective (Bapat et al., 6 Grundvold et al., 7 Mokhayeri et al., 8 Morseth et al., 9 andMozaffarian et al., 10 ) with mean follow-up periods of 7.7, 35, 11, 20 and 12 years, respectively; a cohort study (Williams et al., 14 ) withmean follow-up period of 6.2 years; a Post-Hoc analysis of a prospective study (Everett et al., 13 ), a cross- sectional study (Myrstad et al., 11 ) and an observational study (Qureshi et al., 12 ). In addition to these, one was a Post-Hoc analysis of a RandomizedClinical Trial (Aizer et al., 4 ) with follow-up period of 12 years, and a prospective observational study (Azarbal et al., 5 ) with follow-up period of 11.5 years. From the analyzed studies, no data were identified regarding patient hospitalization. Study of the effect of exercise on the population analyzed Figure 2 shows the results of the meta-analysis considering the selected studies.