OBJECTIVE: To investigate if repetitive transcranial magnetic stimulation (rTMS) was as effective as electroconvulsive therapy (ECT) in treating major depressive episodes and to perform a cost-effectiveness analysis. DESIGN: A single-blind pragmatic multicentre randomised controlled trial (RCT) with 6 months of follow-up to test equivalence of rTMS with ECT. SETTING: The South London and Maudsley NHS Trust and Pembury Hospital in the Invicta Mental Health Trust in Kent. PARTICIPANTS: Right-handed adult patients referred for ECT for treatment of a major depressive episode (DSM-IV) were assessed. During the 2.5-year trial period, 260 patients were referred for ECT, of whom 46 entered the trial. The main reason for not entering the trial was not consenting to ECT while being formally treated under the UK Mental Health Act 1983. INTERVENTIONS: Patients were randomised to receive a 15-day course of rTMS of the left dorsolateral prefrontal cortex (n = 24) or a course of ECT (n = 22). MAIN OUTCOME MEASURES: Patients were assessed before randomisation, at end of treatment and at the 6-month follow-up. Primary outcome measures were the 17-item Hamilton Rating Scale for Depression (HRSD) and proportion of remitters (defined as HRSD score <or=8) at the end-of-treatment time point. Secondary outcomes included self-ratings for mood on the Beck Depression Inventory-II (BDI-II) and visual analogue mood scales (VAMS), the Brief Psychiatric Rating Scale (BPRS), plus subjective and objective side-effects. Low scores on the BDI-II, VAMS and BPRS are positive in terms of health. The results were analysed on an intention-to-treat basis. Cost data were collected using the Client Service Receipt Inventory and the Short Form with 36 Items was used to obtain quality of life measures. Health economic outcomes were cost of treatments, costs incurred during the 6-month follow-up period and gains in quality-adjusted life-years (QALYs). RESULTS: One patient was lost to follow-up at end of treatment and another eight at 6 months. The end-of-treatment HRSD scores were lower for ECT, with 13 (59%) achieving remission compared with four (17%) in the rTMS group. However, HRSD scores did not differ between groups at 6 months. BDI-II, VAMS and BPRS scores were lower for ECT at end of treatment and remained lower after 6 months. Improvement in subjective reports of side-effects following ECT correlated with antidepressant response. There was no difference between the two groups before or after treatment on global measures of cognition. Although individual treatment session costs were lower for rTMS than ECT, the cost for a course of rTMS was not significantly different from that for a course of ECT as more rTMS sessions were given per course. Service costs were not different between the groups in the subsequent 6 months but informal care costs were significantly higher for the rTMS group and contributed substantially to the total cost for this group during the 6-month follow-up period. There also was no difference in gain in QALYs for ECT and rTMS patients. Analysis of cost-effectiveness acceptability curves demonstrated that rTMS has very low probability of being more cost-effective than ECT. CONCLUSIONS: ECT is a more effective and potentially cost-effective antidepressant treatment than 3 weeks of rTMS as administered in this study. Optimal treatment parameters for rTMS need to be established for treating depression. More research is required to refine further the administration of ECT in order to reduce associated cognitive side-effects while maintaining its effectiveness. There is a need for large-scale, adequately powered RCTs comparing different forms of ECT. The next generation of randomised trials of rTMS should also seek to compare treatment variables such as stimulus intensity, number of stimuli administered and duration of treatment, with a view to quantifying an effect size for antidepressant action.