Objective To investigate the efficacy and safety of warfarin in the prevention of cerebral infarction in nonvalvular atrial fibrillation (NVAF). Methods One hundred and thirty-six NVAF patients were randomized into warfarin group [receiving adjusted-dose warfarin,international normalized ratio(INR)was 2.0 - 3.0], aspirin group( receiving aspirin 100 mg/d) and control group (treated without anticoagulants )by random digits table. Followed up 18 months, and the main end point events and adverse effect of the three groups were compared. Results In 136 cases,4 cases lost,and 77 cases(58.3%) were male. The mean dose of warfarin was(2.5 ± 1.0) mg. During the follow-up period, main end point events occurred in 12 cases,with 1 case (2.50%, 1/40) in warfarin group, 4 cases(9.52%, 4/42 ) in aspirin group and 7 cases ( 14.00%, 7/50)in control group. There was no significant difference in main end point events among the three groups ( x2 =2.084,P =0.353). But in the patients with 3 or above risk factors,there was significant difference in the survival curve among the three groups ( x2 = 6.404, P = 0.041 ). The incidence rate of bleeding was higher in warfarin group than that in aspirin group,but there was no significant difference [5.00%(2/40) vs. 2.38%(1/42),P ＞ 0.05]. Conclusions Warfarin can improve survival rate especially in the patients with 3 or above risk factors,and the complication of bleeding occurs mostly when INR ＞ 3.0.Under closed monitoring (INR 2.0-3.0),adjusted-dose warfarin is safety and efficacy.