BACKGROUND: Near-patient testing has made self-monitoring of anticoagulation with warfarin feasible, and several trials have suggested that such monitoring might be equal to or better than standard monitoring. We did a systematic review and meta-analysis of all randomised controlled trials that assessed the effects of self-monitoring or self-management (self-testing and self-dosage) of anticoagulation compared with standard monitoring. METHODS: We searched the Cochrane Register of Controlled Trials, MEDLINE, EMBASE to April 2005, and contacted manufacturers and authors of relevant studies. Outcomes analysed were: major haemorrhage, thromboembolic events, death, tests in range, minor haemorrhage, frequency of testing, and feasibility of self-monitoring. FINDINGS: We identified 14 randomised trials of self-monitoring: pooled estimates showed significant reductions in thromboembolic events (odds ratio 0.45, 95% CI 0.30-0.68), all-cause mortality (0.61, 0.38-0.98), and major haemorrhage (0.65, 0.42-0.99). Trials of combined self-monitoring and self-adjusted therapy showed significant reductions in thromboembolic events (0.27, 0.12-0.59) and death (0.37, 0.16-0.85), but not major haemorrhage (0.93, 0.42-2.05). No difference was noted in minor haemorrhage. 11 trials reported improvements in the mean proportion of international normalisation ratios in range. INTERPRETATION: Self-management improves the quality of oral anticoagulation. Patients capable of self-monitoring and self-adjusting therapy have fewer thromboembolic events and lower mortality than those who self-monitor alone. However, self-monitoring is not feasible for all patients, and requires identification and education of suitable candidates.