Background: Atrial fibrillation (AF) is the commonest chronic arrhythmia with a prevalence of 9% in octogenarians and accounts for 24% of the stroke risk in this population. Although trials demonstrate reductions in stroke with warfarin, audit data show that it is still underused. However, anti-coagulation in the very elderly is not without risk. Methods: Randomised open labelled prospective study of primary thromboprophylaxis for AF. Patients aged >80 and <90 were randomised to receive dose-adjusted warfarin (INR 2.0-3.0) or aspirin 300 mg. All patients had permanent AF, were ambulant, had Folstein mini mental score >25 and had no contraindications to either treatment. Follow-up was for 1 year with 3 monthly visits. The primary outcome measure was a comparative frequency of combined endpoints comprising death, thromboembolism, serious bleeding and withdrawal from the study. Results: Seventy-five patients (aspirin 39; warfarin 36) were entered (mean age 83.9, 47% male). There were significantly more adverse events with aspirin (13/39; 33%) than warfarin (2/36; 6%), P = 0.002. 10/13 aspirin adverse events were caused by side effects and serious bleeding; there were three deaths (two aspirin, one warfarin). Conclusion: Dose-adjusted warfarin was significantly better tolerated with fewer adverse events than aspirin 300 mg in this elderly population. Although aspirin 75 mg may have been better tolerated, there is no evidence for efficacy in AF at this dose. © 2007 Oxford University Press.