Summary Background Tamoxifen preserves bone in postmenopausal women, but non-steroidal aromatase inhibitors accelerate bone loss and increase fracture risk. We aimed to study the effect on bone health in a subgroup of women included in the Intergroup Exemestane Study (IES), a large randomised trial that compared the switch to the steroidal aromatase inhibitor exemestane with continuation of tamoxifen in the adjuvant treatment of postmenopausal breast cancer. Methods Results were analysed from 206 evaluable patients from the IES, in which postmenopausal women with histologically confirmed and completely resected unilateral breast cancer (that was oestrogen-receptor positive or of unknown status), who were disease-free after 2–3 years of treatment with tamoxifen were randomised to continue oral tamoxifen 20 mg/day or switch to oral exemestane 25 mg/day to complete a total of 5 years of adjuvant endocrine therapy. The primary endpoint was change in bone-mineral density (BMD) assessed by dual energy X-ray absorptiometry. Changes in biochemical markers of bone turnover were also analysed in this substudy, and the incidence of fractures in the entire study reported. The IES is registered on the Current Controlled Trials website http://www.controlled-trials.com/ISRCTN11883920. Findings Within 6 months of switching to exemestane, BMD was lowered by 0·051 g/cm3 (2·7%; 95% CI 2·0–3·4; p<0·0001) at the lumbar spine and 0·025 g/cm3 (1·4%; 0·8–1·9; p<0·0001) at the hip compared with baseline. BMD decreases were only 1·0% (0·4–1·7; p=0·002) and 0·8% (0·3–1·4; p=0·003) in year 2 at the lumbar spine and hip, respectively. No patient with BMD in the normal range at trial entry developed osteoporosis. Bone resorption and formation markers increased at all time points in women receiving exemestane (p<0·001). With a median follow-up in all IES participants (n=4274) of 58 months, 162 (7%) and 115 (5%) patients in the exemestane and tamoxifen groups, respectively, had fractures (odds ratio 1·45 1·13–1·87; p=0·003). Interpretation These results indicate that the increase in survival shown previously with the IES switch strategy is achieved at the expense of some detriment to skeletal health, so the risk-benefit ratio to women needs to be individually assessed.