Summary Background Early improvements in disease-free survival have been noted when an aromatase inhibitor is given either instead of or sequentially after tamoxifen in postmenopausal women with oestrogen-receptor-positive early breast cancer. However, little information exists on the long-term effects of aromatase inhibitors after treatment, and whether these early improvements lead to real gains in survival. Methods 4724 postmenopausal patients with unilateral invasive, oestrogen-receptor-positive or oestrogen-receptor-unknown breast cancer who were disease-free on 2–3 years of tamoxifen, were randomly assigned to switch to exemestane (n=2352) or to continue tamoxifen (n=2372) for the remainder of a 5-year endocrine treatment period. The primary endpoint was disease-free survival; overall survival was a secondary endpoint. Efficacy analyses were intention-to-treat. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN11883920. Results After a median follow-up of 55·7 months (range 0–89·7), 809 events contributing to the analysis of disease-free survival had been reported (354 exemestane, 455 tamoxifen); unadjusted hazard ratio 0·76 (95% CI 0·66–0·88, p=0·0001) in favour of exemestane, absolute benefit 3·3% (95% CI 1·6–4·9) by end of treatment (ie, 2·5 years after randomisation). 222 deaths occurred in the exemestane group compared with 261 deaths in the tamoxifen group; unadjusted hazard ratio 0·85 (95% CI 0·71–1·02, p=0·08), 0·83 (0·69–1·00, p=0·05) when 122 patients with oestrogen-receptor-negative disease were excluded. Conclusions Our results suggest that early improvements in disease-free survival noted in patients who switch to exemestane after 2–3 years on tamoxifen persist after treatment, and translate into a modest improvement in overall survival.