Background There is limited evidence about the effectiveness of rehabilitation in meeting the longer-term needs of stroke patients and their carers. Objective To determine the clinical effectiveness and cost-effectiveness of an extended stroke rehabilitation service (EXTRAS). Design A pragmatic, observer-blind, parallel-group, multicentre randomised controlled trial with embedded health economic and process evaluations. Participants were randomised (1 : 1) to receive EXTRAS or usual care. Setting Nineteen NHS study centres. Participants Patients with a new stroke who received early supported discharge and their informal carers. Interventions Five EXTRAS reviews provided by an early supported discharge team member between 1 and 18 months post early supported discharge, usually over the telephone. Reviewers assessed rehabilitation needs, with goal-setting and action-planning. Control treatment was usual care post early supported discharge. Main outcome measures The primary outcome was performance in extended activities of daily living (Nottingham Extended Activities of Daily Living Scale) at 24 months post randomisation. Secondary outcomes at 12 and 24 months included patient mood (Hospital Anxiety and Depression Scale), health status (Oxford Handicap Scale), experience of services and adverse events. For carers, secondary outcomes included carers’ strain (Caregiver Strain Index) and experience of services. Cost-effectiveness was estimated using resource utilisation costs (adaptation of the Client Service Receipt Inventory) and quality-adjusted life-years. Results A total of 573 patients (EXTRAS, n = 285; usual care, n = 288) with 194 carers (EXTRAS, n = 103; usual care, n = 91) were randomised. Mean 24-month Nottingham Extended Activities of Daily Living Scale scores were 40.0 (standard deviation 18.1) for EXTRAS (n = 219) and 37.2 (standard deviation 18.5) for usual care (n = 231), giving an adjusted mean difference of 1.8 (95% confidence interval –0.7 to 4.2). The mean intervention group Hospital Anxiety and Depression Scale scores were not significantly different at 12 and 24 months. The intervention did not improve patient health status or carer strain. EXTRAS patients and carers reported greater satisfaction with some aspects of care. The mean cost of resource utilisation was lower in the intervention group: –£311 (95% confidence interval –£3292 to £2787), with a 68% chance of EXTRAS being cost-saving. EXTRAS was associated with 0.07 (95% confidence interval 0.01 to 0.12) additional quality-adjusted life-years. At current conventional thresholds of willingness to pay for a quality-adjusted life-year, there is a 90% chance that EXTRAS is cost-effective. Conclusions EXTRAS did not improve stroke survivors’ performance in extended activities of daily living but did improve their overall satisfaction with services. Given the impact on costs and quality-adjusted life-years, there is a high chance that EXTRAS could be considered cost-effective. Future work Further research is required to identify whether or not community-based interventions can improve performance of extended activities of daily living, and to understand the improvements in health-related quality of life and costs seen by provision of intermittent longer-term specialist review. Trial registration Current Controlled Trials ISRCTN45203373. Funding This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 24, No. 24. See the NIHR Journals Library website for further project information.