Upper limb problems are common after breast cancer treatment. To investigate the clinical effectiveness and cost-effectiveness of a structured exercise programme compared with usual care on upper limb function, health-related outcomes and costs in women undergoing breast cancer surgery. This was a two-arm, pragmatic, randomised controlled trial with embedded qualitative research, process evaluation and parallel economic analysis; the unit of randomisation was the individual (allocated ratio 1 : 1). Breast cancer centres, secondary care. Women aged ≥ 18 years who had been diagnosed with breast cancer and were at higher risk of developing shoulder problems. Women were screened to identify their risk status. All participants received usual-care information leaflets. Those randomised to exercise were referred to physiotherapy for an early, structured exercise programme (three to six face-to-face appointments that included strengthening, physical activity and behavioural change strategies). The primary outcome was upper limb function at 12 months as assessed using the Disabilities of Arm, Hand and Shoulder questionnaire. Secondary outcomes were function (Disabilities of Arm, Hand and Shoulder questionnaire subscales), pain, complications (e.g. wound-related complications, lymphoedema), health-related quality of life (e.g. EuroQol-5 Dimensions, five-level version; Short Form questionnaire-12 items), physical activity and health service resource use. The economic evaluation was expressed in terms of incremental cost per quality-adjusted life-year and incremental net monetary benefit gained from an NHS and Personal Social Services perspective. Participants and physiotherapists were not blinded to group assignment, but data collectors were blinded. Between 2016 and 2017, we randomised 392 participants from 17 breast cancer centres across England: 196 (50%) to the usual-care group and 196 (50%) to the exercise group. Ten participants (10/392; 3%) were withdrawn at randomisation and 32 (8%) did not provide complete baseline data. A total of 175 participants (89%) from each treatment group provided baseline data. Participants' mean age was 58.1 years (standard deviation 12.1 years; range 28-88 years). Most participants had undergone axillary node clearance surgery (327/392; 83%) and 317 (81%) had received radiotherapy. Uptake of the exercise treatment was high, with 181 out of 196 (92%) participants attending at least one physiotherapy appointment. Compliance with exercise was good: 143 out of 196 (73%) participants completed three or more physiotherapy sessions. At 12 months, 274 out of 392 (70%) participants returned questionnaires. Improvement in arm function was greater in the exercise group [mean Disabilities of Arm, Hand and Shoulder questionnaire score of 16.3 (standard deviation 17.6)] than in the usual-care group [mean Disabilities of Arm, Hand and Shoulder questionnaire score of 23.7 (standard deviation 22.9)] at 12 months for intention-to-treat (adjusted mean difference Disabilities of Arm, Hand and Shoulder questionnaire score of -7.81, 95% confidence interval -12.44 to -3.17; p = 0.001) and complier-average causal effect analyses (adjusted mean difference -8.74, 95% confidence interval -13.71 to -3.77; p ≤ 0.001). At 12 months, pain scores were lower and physical health-related quality of life was higher in the exercise group than in the usual-care group (Short Form questionnaire-12 items, mean difference 4.39, 95% confidence interval 1.74 to 7.04; p = 0.001). We found no differences in the rate of adverse events or lymphoedema over 12 months. The qualitative findings suggested that women found the exercise programme beneficial and enjoyable. Exercise accrued lower costs (-£387, 95% CI -£2491 to £1718) and generated more quality-adjusted life years (0.029, 95% CI 0.001 to 0.056) than usual care over 12 months. The cost-effectiveness analysis indicated that exercise was more cost-effective and that the results were robust to sensitivity analyses. Exercise was relatively cheap to implement (£129 per participant) and associated with lower health-care costs than usual care and improved health-related quality of life. Benefits may accrue beyond the end of the trial. Postal follow-up was lower than estimated; however, the study was adequately powered. No serious adverse events directly related to the intervention were reported. This trial provided robust evidence that referral for early, supported exercise after breast cancer surgery improved shoulder function in those at risk of shoulder problems and was associated with lower health-care costs than usual care and improved health-related quality of life. Future work should focus on the implementation of exercise programmes in clinical practice for those at highest risk of shoulder problems. This trial is registered as ISRCTN35358984. 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. 26, No. 15. See the NIHR Journals Library website for further project information.