The risks of stroke or systemic embolism and major bleeding are considered similar between paroxysmal and sustained atrial fibrillation (AF), and warfarin has demonstrated superior efficacy to aspirin, irrespective of the AF type. However, with the advent of novel oral anticoagulants (NOACs) and antiplatelet agents, the optimal antithrombotic prophylaxis for paroxysmal AF remains unclear.We searched Medline, Embase, CENTRAL, and China Biology Medicine up to October week 1, 2015. Randomized controlled trials of AF patients assigned to NOACs, warfarin, or antiplatelets, with reports of outcomes stratified by the AF type, were included. A fixed-effects model was used if no statistically significant heterogeneity was indicated; otherwise, a random-effects model was used.Six studies of 69,990 nonvalvular AF patients with ≥1 risk factor for stroke were included. Postantithrombotic treatment, paroxysmal AF patients showed lower risks of stroke (risk ratio [RR], 0.72; 95% confidence interval [CI], 0.59-0.87), stroke or systemic embolism (RR, 0.74; 95% CI, 0.63-0.86), and all-cause mortality (RR, 0.75; 95% CI, 0.67-0.83), while the major bleeding risk was comparable (RR, 0.96; 95% CI, 0.85-1.08). We were unable to detect the superiority of anticoagulation over antiplatelets for paroxysmal AF (RR, 0.72; 95% CI, 0.43-1.23), while it was more effective than antiplatelets for sustained AF (RR, 0.42; 95% CI, 0.33-0.54). NOACs showed superior efficacy over warfarin and trended to show reduced major bleeding irrespective of the AF type.The AF type is a predictor for thromboembolism, and might be helpful in stroke risk stratification model in combination with other risk factors. With the appearance of novel anticoagulant and antiplatelet agents, the best antithrombotic choice for paroxysmal AF needs further exploration.