No / Objective To compare the incidence of treatment-related complications, including lymphoedema, after two programmes of shoulder mobilisation in women with invasive breast cancer when surgical treatment included axillary lymph node dissection. Design Randomised controlled trial. Setting Two secondary care National Health Service trusts. Participants One hundred and sixteen women (mean age 57 years, standard deviation 13.1 years) recruited from November 2003 to March 2006 (58 intervention group, 58 control group). Seven patients (6%) did not complete the study. Intervention Arm exercises and shoulder movement restricted to below shoulder level for the first 7 days after surgery. Controls commenced an exercise programme that incorporated exercises above shoulder level within 48 hours. Outcome measures All outcomes were recorded at baseline (pre-operatively) and at 1 year. The primary outcome was incidence of lymphoedema, defined by a limb volume difference of 200 ml or more compared with the contralateral arm. This outcome was measured using volume displacement. Secondary outcome measures included volume differences between the two limbs measured by actual volume displacement difference, wound drainage volumes, range of shoulder movement (manual goniometer), grip strength (hand-held dynamometer) and health-related quality of life (Shoulder Disability Questionnaire, Functional Assessment of Cancer Therapy – Breast). Results All statistical tests were two-sided. Data were analysed using intention-to-treat principles. The incidence of lymphoedema (200 ml or more) increased significantly in women who had undertaken a programme of early full shoulder mobilisation. Twenty-two women (19%) developed lymphoedema (200 ml or more) in their first postoperative year. There were significantly more women with lymphoedema in the early full shoulder mobilisation group (n = 16) compared with the delayed full shoulder mobilisation group (n = 6). The relative risk of developing lymphoedema after early mobilisation was 2.7 (95% confidence interval 1.1 to 6.3; P = 0.031). Limb volume differences were significantly higher in the early mobilisation group. This was apparent in differences in limb volume displacement (P = 0.004) and percentage difference between the two limbs (P = 0.007). There were no statistically significant differences in shoulder movement, grip strength or self-evaluated outcomes between the two groups at 1 year. Conclusion A programme of exercise that delays full shoulder mobilisation for 1 week is recommended after axillary node dissection for invasive breast cancer.