BACKGROUND: Shared decision-making (SDM) between professionals and patients is increasingly advocated from ethical principles. Some data are accruing about the effects of such approaches on health or other patient-based outcomes. These effects often vary substantially between studies. OBJECTIVE: Our aim was to evaluate the effects of training GPs in SDM, and the use of simple risk communication aids in general practice, on patient-based outcomes. METHODS: A cluster randomized trial with crossover was carried out with the participation of 20 recently qualified GPs in urban and rural general practices in Gwent, South Wales. A total of 747 patients with known atrial fibrillation, prostatism, menorrhagia or menopausal symptoms were invited to a consultation to review their condition or treatments. After baseline, participating doctors were randomized to receive training in (i) SDM skills; or (ii) the use of simple risk communication aids, using simulated patients. The alternative training was then provided for the final study phase. Patients were randomly allocated to a consultation during baseline or intervention 1 (SDM or risk communication aids) or intervention 2 phases. A randomly selected half of the consultations took place in 'research clinics' to evaluate the effects of more time for consultations, compared with usual surgery time. Patient-based outcomes were assessed at exit from consultation and 1 month follow-up. These were: COMRADE instrument (principal measures; subscales of risk communication and confidence in decision), and a range of secondary measures (anxiety, patient enablement, intention to adhere to chosen treatment, satisfaction with decision, support in decision making and SF-12 health status measure). Multilevel modelling was carried out with outcome score as the dependent variable, and follow-up point (i.e. exit or 1 month later for each patient), patient and doctor levels of explanatory variables. RESULTS: No statistically significant changes in patient-based outcomes due to the training interventions were found: COMRADE risk communication score increased 0.7 [95% confidence interval (CI) -0.92 to 2.32] after risk communication training and 0.9 (95% CI -0.89 to 2.35) after SDM training; and COMRADE satisfaction with communication score increased by 1.0 (95% CI -1.1 to 3.1) after risk communication, and decreased by 0.6 (95% CI 2.7 to -1.5) after SDM training. Patients' confidence in the decision (2.1 increase, 95% CI 0.7-3.5, P < 0.01) and expectation to adhere to chosen treatments (0.7 increase, 95% CI 0.04-1.36, P < 0.05) were significantly greater among patients seen in the research clinics (when more time was available) compared with usual surgery time. Most outcomes deteriorated between exit and 1 month later. There was no interaction between intervention effects. CONCLUSION: Patients can be more involved in treatment decisions, and risks and benefits of treatment options can be explained in more detail, without adversely affecting patient-based outcomes. SDM and risk communication may be advocated from values and ethical principles even without evidence of health gain or improvement in patient-based outcomes, but the resources required to enhance these professional skills must also be taken into consideration. These data also indicate the benefits of extra consultation time.