Objective To evaluate the efficacy and safety of subcutaneous injection of fondaparinux during perioperative period of radiofrequency catheter ablation of atrial aibrillation (AF), figure out whether it will affect the targeted time of INR after using warfarin, and compare with dalteparin and enoxaparin for ecchymosis effects at the injection sites. Methods Ninety-four AF patients undergoing catheter ablation were randomly divided into 3 groups, including 24 patients in enoxaparin group, 20 patients in dalteparin group, and 50 patients in fondaparinux group. Enoxaparin, dalteparin or fondaparinux was subcutaneously injected for 5?10 days before catheter ablation, and continued for at least 3?5 days after ablation before INR achieving the target of 1.8?2.0 under the use of oral warfarin. Results There was no statistical difference in baseline characteristics among three groups (P＞0.05). The thromboembolic events were 0/24 patients in enoxaparin group, 0/20 in dalteparin group and 1/50 in fondaparinux group, and there was no statistical difference in the incidence of thromboembolic events among three groups. There was no serious bleeding events in all three groups. The time for achieving the targeted INR after ablation had no statistical difference among three groups (P＞0.05). The incidence and the long diameter of ecchymosis in injection sites were significantly lower/less in fondaparinux group than in enoxaparin and dalteparin groups on the seventh day after injection of fondaparinux or LMWH (non-ecchymosis ratio: fondaparinux, 0.66; enoxaparin, 0.21; dalteparin, 0.15; P＜0.05; long diameter of ecchymosis: fondaparinux, 0.7cm; enoxaparin, 2.3cm; dalteparin, 3.3cm; P＜0.05). Conclusion Compared with enoxaparin and dalteparin, fondaparinux in perioperative anticoagulation therapy of catheter ablation of AF neither increases the major bleeding and thromboembolic events, nor affects the targeted time of INR after using warfarin. Fondaparinux could reduce the incidence and scope of ecchymosis in subcutaneous injection sites.