Abstract In a worldwide collaboration, information was sought and centrally checked on mortality and recurrence for each woman in any randomised trial that began before 1985 of any aspect of systemic adjuvant therapy for early breast cancer. Checked data were available for 75 000 women (about 90% of those ever randomised), of whom 32% had died and another 10% had experienced recurrence. The parts now reviewed include 30 000 women in tamoxifen trials, 3000 in ovarian ablation trials, 11 000 in polychemotherapy trials, 15 000 in other chemotherapy comparisons, and 6000 in immunotherapy trials. Highly significant reductions in the annual rates both of recurrence and of death are produced by tamoxifen (25% SD 2 recurrence and 17% SD 2 mortality: 2p<0·00001), by ablation below age 50 (26% SD 6 recurrence and 25% SD 7 mortality: 2p=0·0004), and by polychemotherapy (28% SD 3 recurrence and 16% SD 3 mortality: 2p<0·00001), but not by ablation at older ages or by immunotherapy. (Tamoxifen also reduced the risk of development of contralateral breast cancer by 39% SD 9: 1p<0·00001.) For tamoxifen and for polychemotherapy the avoidance of recurrence is chiefly during years 0-4 (this difference being maintained but not increased afterwards), but the avoidance of mortality is highly significant both during and after years 0-4, so the cumulative differences in survival produced by these relatively brief treatments (median: 2 years tamoxifen, 1 year polychemotherapy) are larger at 10 than at 5 years. There is little information beyond year 10 (except for ovarian ablation, which produces separately significant mortality reductions both during and after years 0-9). Both direct and indirect randomised comparisons show long-term polychemotherapy (eg, 12 months) to be no better than shorter (eg, 6 months) regimens, but do show polychemotherapy to be significantly better than single-agent chemotherapy. Indirect randomised comparisons do not reveal significant differences between different forms of polychemotherapy, or differences between different tamoxifen doses, but do show that long-term tamoxifen (eg, 2 years, or even 5 years) is significantly more effective than shorter tamoxifen regimens. In old age (70+) tamoxifen is of demonstrated efficacy, but chemotherapy has not been evaluated. Between ages 50 and 69 direct comparisons show that chemotherapy plus tamoxifen is better (1p<0·00001) than chemotherapy alone both for recurrence and for mortality, and better (1 p<0·00001) than tamoxifen alone for recurrence. In women aged under 50 chemotherapy and ovarian ablation appear, by an indirect comparison, to be of comparable efficacy, and the combination may be still better. The 30-40% proportional risk reductions that can be produced by combined chemo-endocrine therapy in middle age are similar for node-positive and for node-negative patients, but the absolute improvement in 10-year survival is about twice as great for the former (at least 12 deaths avoided per 100 women treated) as for the latter.