Atrial fibrillation (AF) is the most common cardiac arrhythmia. We designed an umbrella review to systematically assess the epidemiological credibility of the associations of non-genetic factors with risk of AF. We searched PubMed and EMBASE from inception to December 31, 2018 to identify systematic reviews and meta-analyses of observational studies for the association of non-genetic factors with risk of AF. For each meta-analysis, we used the random-effects model, and we estimated the 95% confidence and prediction intervals. We also assessed between-study heterogeneity, small-study effects and excess significance bias. We identified 34 eligible papers that examined 51 associations of 42 unique non-genetic factors with risk of AF. Eighteen associations remained statistically significant at P value < 1 × 10⁻⁶. Thirty-one associations presented large or very large between-study heterogeneity. Eight associations presented evidence for small-study effects and 13 associations had evidence for excess significance bias. Ten associations, i.e. corrected QT interval, alcohol consumption (highest vs. lowest category, per 1 drink/day increase), body mass index (> 30 units vs. < 30 units, per 5 units increase), waist circumference, body weight, type 2 diabetes mellitus, and smoking (ever vs. never, per 10 cigarettes/day increase) were supported by convincing or highly suggestive evidence in meta-analyses of prospective cohort studies. Type 2 diabetes mellitus, markers of adiposity, alcohol consumption, smoking, and corrected QT interval constitute credible risk factors of AF. Our proposed grading may guide the design of future studies, including Mendelian randomization studies, to assess whether these associations are causal.

心房颤动（AF）是最常见的心律不齐。我们设计了一项总括性综述，以系统地评估非遗传因素与房颤风险之间关系的流行病学可信度。从开始到2018年12月31日，我们搜索了PubMed和EMBASE，以鉴定观察性研究的系统评价和荟萃分析，以了解非遗传因素与房颤风险之间的关系。对于每个荟萃分析，我们使用随机效应模型，并估计95％的置信度和预测间隔。我们还评估了研究之间的异质性，小研究效应和显着性偏倚。我们鉴定了34篇符合条件的论文，这些论文检查了42种独特的非遗传因素与房颤风险的51种关联。18个关联在P值<1×10时仍具有统计学意义^-6。31个关联呈现出很大的或非常大的研究间异质性。八个协会提供了小研究效果的证据，而十三个协会则提供了过度显着性偏倚的证据。十个关联，即校正的QT间隔，酒精消耗（最高/最低类别，每增加1喝酒/天），体重指数（> 30单位对<30单位，每增加5单位），腰围，体重，在前瞻性队列研究的荟萃分析中，令人信服或极富启发性的证据支持2型糖尿病和吸烟（从无到有，每10支香烟/天增加）。2型糖尿病，肥胖，饮酒，吸烟和校正的QT间隔标志物是AF的可靠危险因素。我们建议的分级可能会指导未来研究的设计，