BACKGROUND: For women with early-stage oestrogen receptor (ER)-positive breast cancer, adjuvant tamoxifen reduces 15-year breast cancer mortality by a third. Aromatase inhibitors are more effective than tamoxifen in postmenopausal women but are ineffective in premenopausal women when used without ovarian suppression. We aimed to investigate whether premenopausal women treated with ovarian suppression benefit from aromatase inhibitors. METHODS: We did a meta-analysis of individual patient data from randomised trials comparing aromatase inhibitors (anastrozole, exemestane, or letrozole) versus tamoxifen for 3 or 5 years in premenopausal women with ER-positive breast cancer receiving ovarian suppression (goserelin or triptorelin) or ablation. We collected data on baseline characteristics, dates and sites of any breast cancer recurrence or second primary cancer, and dates and causes of death. Primary outcomes were breast cancer recurrence (distant, locoregional, or contralateral), breast cancer mortality, death without recurrence, and all-cause mortality. As distant recurrence invariably results in death from breast cancer several years after the occurrence, whereas locoregional recurrence and new contralateral breast cancer are not usually fatal, the distant recurrence analysis is shown separately. Standard intention-to-treat log-rank analyses estimated first-event rate ratios (RR) and their confidence intervals (CIs). FINDINGS: We obtained data from all four identified trials (ABCSG XII, SOFT, TEXT, and HOBOE trials), which included 7030 women with ER-positive tumours enrolled between June 17, 1999, and Aug 4, 2015. Median follow-up was 8·0 years (IQR 6·1-9·3). The rate of breast cancer recurrence was lower for women allocated to an aromatase inhibitor than for women assigned to tamoxifen (RR 0·79, 95% CI 0·69-0·90, p=0·0005). The main benefit was seen in years 0-4 (RR 0·68, 99% CI 0·55-0·85; p<0·0001), the period when treatments differed, with a 3·2% (95% CI 1·8-4·5) absolute reduction in 5-year recurrence risk (6·9% vs 10·1%). There was no further benefit, or loss of benefit, in years 5-9 (RR 0·98, 99% CI 0·73-1·33, p=0·89) or beyond year 10. Distant recurrence was reduced with aromatase inhibitor (RR 0·83, 95% CI 0·71-0·97; p=0·018). No significant differences were observed between treatments for breast cancer mortality (RR 1·01, 95% CI 0·82-1·24; p=0·94), death without recurrence (1·30, 0·75-2·25; p=0·34), or all-cause mortality (1·04, 0·86-1·27; p=0·68). There were more bone fractures with aromatase inhibitor than with tamoxifen (227 [6·4%] of 3528 women allocated to an aromatase inhibitor vs 180 [5·1%] of 3502 women allocated to tamoxifen; RR 1·27 [95% CI 1·04-1·54]; p=0·017). Non-breast cancer deaths (30 [0·9%] vs 24 [0·7%]; 1·30 [0·75-2·25]; p=0·36) and endometrial cancer (seven [0·2%] vs 15 [0·3%]; 0·52 [0·22-1·23]; p=0·14) were rare. INTERPRETATION: Using an aromatase inhibitor rather than tamoxifen in premenopausal women receiving ovarian suppression reduces the risk of breast cancer recurrence. Longer follow-up is needed to assess any impact on breast cancer mortality. FUNDING: Cancer Research UK, UK Medical Research Council.