OBJECTIVES: New oral anticoagulants (NOACs) have been launched for treating patients with atrial fibrillation (AF) in recent years. We are interested in estimating an ICER (Incremental cost effectiveness ratio) of NOAC versus warfarin. METHODS: We found fourteen randomized trials involving dabigatran, rivaroxaban, apixaban, edoxaban, and aspirin as compared to warfarin. Using a network meta-analysis (NMA) we estimated the efficacy and safety of NOACs versus warfarin. Efficacy and safety measure were defined by stroke or systemic embolism and intracranial hemorrhage. We estimated an incremental drug cost induced by NOACs per reducing such a single event. We assumed that an event would reduce 10 QALYs per patient. RESULTS: Stroke or systemic embolism was said to occur 1.5% per year in warfarin users. Intracranial hemorrhage occurred 0.8% per year. NMA showed that the NOACs would reduce the rate of stroke or systemic embolism by 12% to 55% depending upon different NOAC and the rate of intracranial hemorrhage by 16% to 64%. Since there are a million AF patients in Japan, 23,000 events are supposed to occur within a year with warfarin and 9,630 to 19,920 events with NOACs. Yearly drug costs of warfarin and NOACs in Japan were 65 and 1,900 US dollar per patient. So, incremental drug cost with NOACs was 1,835 million dollar per year, whereas NOACs reduced events by 3,080 to 13,370 per year. Therefore, an incremental cost with NOAC was estimated 137,200 to 595,800 dollar per event. Assuming an event reduces 10 QALYs compared to non-event, an ICER was estimated 13,720 to 59,580 dollar per QALY. Japanese GDP was said to be 40,000 dollar per person. Probability of ICER less than GDP was 0.64 under the assumption of normality. CONCLUSIONS: NOAC for AF might be cost effective in Japan since the probability of ICER less than GDP was as large as 0.64.