Background: Obesity (body mass index [BMI] ≥30 kg/m²) is reportedly associated with worse prognosis in early breast cancer. But, this association could depend strongly on estrogen receptor (ER) positivity and ovarian activity (or young age). Methods: Through EBCTCG, anonymised information on each individual with early breast cancer in any trial is requested periodically, including BMI (at randomisation), ER status, menopausal status, age, treatment, recurrence and death. The global Steering Committee has requested analyses of the independent effects of BMI on outcome. Results are provided for 80,000 patients (in 70 trials) who had all these data items; most also had data on tumour diameter and nodal status. Mean follow-up was 8 woman-years. Cox regression (stratified for trial and treatment, and adjusted for age) assesses the relevance of BMI to mortality with recurrence (as a surrogate for the relevance of BMI to breast cancer mortality). Few had BMI <20 kg/m²; the standard WHO cut-points define overweight (25-30 kg/m²). Results: In 20,000 women with ER-poor disease there was little association of BMI with breast cancer mortality, and none after adjustment for tumour diameter and nodal status. In 60,000 with ER+ disease, BMI was positively associated with breast cancer mortality in pre/peri- and in post-menopausal women (each 2P<0.00001). But, after adjustment for tumour characteristics the association remained clearly significant only in 20,000 pre/peri-menopausal women with ER+ disease (breast cancer mortality rate ratio comparing BMI ≥30 versus BMI 20-25 kg/m² [RR] = 1.34, 95%CI 1.22-1.47, 2P<0.00001, with a steady trend between BMI <25, 25-30, 30-35 and ≥35 kg/m²); little association remained in 40,000 post-menopausal women with ER+ disease (RR=1.06, 95%CI 0.99-1.14, 2P=0.12; heterogeneity between RRs 2P<0.0001). In analyses of ER+ disease subdivided by age (instead of menopausal status), obesity appeared importantly relevant only to age about 55 years. Findings were not materially altered by excluding the first 5 years of follow-up. Conclusions: In women with early breast cancer, obesity appears strongly independently related to breast cancer mortality only in pre/peri-menopausal ER+ disease.