Carl Heneghan, Alison Ward, Rafael Perera, Clare Bankhead, Alice Fuller, Richard Stevens, Kairen Bradford, Sally Tyndel, Pablo Alonso-Coello, Jack Ansell, Rebecca Beyth, Artur Bernardo, Thomas Decker Christensen, Manon Cromheecke, Robert G Edson, David Fitzmaurice, Alain PA Gadisseur, Josep M Garcia-Alamino, Chris Gardiner, Michael Hasenkam, Alan Jacobson, Scott Kaatz, Farhad Kamali, Tayyaba Irfan Khan, Eve Knight, Heinrich Kortke, Marcel Levi, David Bruce Matchar, Barbara Menendez-Jandula, Ivo Rakovac, Christian Schaefer, Andrea Siebenhofer, Juan Carlos Souto, Rubina Sunderji, Kenneth Gin, Karen Shalansky, Heinz Völler, Otto Wagner, Armin Zittermann

• Background Uptake of self-testing and self-management of oral coagulation has remained inconsistent, despite good evidence of their effectiveness. To clarify the value of self-monitoring of oral anticoagulation, we did a meta-analysis of individual patient data addressing several important gaps in the evidence, including an estimate of the effect on time to death, first major haemorrhage, and thromboembolism. Methods We searched Ovid versions of Embase (1980-2009) and Medline (1966-2009), limiting searches to randomised trials with a maximally sensitive strategy. We approached all authors of included trials and requested individual patient data: primary outcomes were time to death, first major haemorrhage, and first thromboembolic event. We did prespecified subgroup analyses according to age, type of control-group care (anticoagulation-clinic care vs primary care), self-testing alone versus self-management, and sex. We analysed patients with mechanical heart valves or atrial fibrillation separately. We used a random-effect modelBackground Uptake of self-testing and self-management of oral coagulation has remained inconsistent, despite good evidence of their effectiveness. To clarify the value of self-monitoring of oral anticoagulation, we did a meta-analysis of individual patient data addressing several important gaps in the evidence, including an estimate of the effect on time to death, first major haemorrhage, and thromboembolism. Methods We searched Ovid versions of Embase (1980-2009) and Medline (1966-2009), limiting searches to randomised trials with a maximally sensitive strategy. We approached all authors of included trials and requested individual patient data: primary outcomes were time to death, first major haemorrhage, and first thromboembolic event. We did prespecified subgroup analyses according to age, type of control-group care (anticoagulation-clinic care vs primary care), self-testing alone versus self-management, and sex. We analysed patients with mechanical heart valves or atrial fibrillation separately. We used a random-effect model method to calculate pooled hazard ratios and did tests for interaction and heterogeneity, and calculated a time-specific number needed to treat. Findings Of 1357 abstracts, we included 11 trials with data for 6417 participants and 12 800 person-years of follow-up. We reported a significant reduction in thromboembolic events in the self-monitoring group (hazard ratio 0.51; 95% CI 0.31-0.85) but not for major haemorrhagic events (0.88, 0.74-1.06) or death (0.82, 0.62-1.09). Participants younger than 55 years showed a striking reduction in thrombotic events (hazard ratio 0.33, 95% CI 0.17-0.66), as did participants with mechanical heart valve (0.52, 0.35-0.77). Analysis of major outcomes in the very elderly (age >= 85 years, n=99) showed no significant adverse effects of the intervention for all outcomes. Interpretation Our analysis showed that self-monitoring and self-management of oral coagulation is a safe option for suitable patients of all ages. Patients should also be offered the option to self-manage their disease with suitable health-care support as back-up.…