Distant recurrence in women with oestrogen receptor-positive early breast cancer persists at a constant rate for more than 20 years after diagnosis, with little equivalent data for oestrogen receptor-negative breast cancer. Using the database of the Early Breast Cancer Trialists' Collaborative Group (EBCTCG) we investigated rates of distant breast-cancer recurrence in oestrogen receptor-positive and oestrogen receptor-negative tumours and trends in outcomes over time. In this pooled analysis of randomised controlled trial data, patients in the EBCTCG database of more than 650 000 women in trials of treatment for early-stage breast cancer were screened for eligibility. Women were eligible if they were enrolled between 1990 and 2009 and newly diagnosed with oestrogen receptor-positive breast cancer and scheduled for at least 5 years of endocrine therapy, or oestrogen receptor-negative disease, and if they were younger than 75 years at diagnosis, had a tumour diameter of 50 mm or less, and fewer than ten positive axillary lymph nodes, and no evidence of distant metastases at entry. Trial of neoadjuvant therapy, or those in which adjuvant therapy was unclear, and women with oestrogen receptor-negative, progesterone receptor-positive disease, or those for whom outcome or baseline data were missing were excluded. The primary outcome was time to first distant recurrence as defined by each trial, ignoring any locoregional recurrence or contralateral breast cancer. 10-year risks of distant recurrence by period of diagnosis were compared using Cox regression adjusted for patient and tumour characteristics, trial, and assigned treatment. Of the 652 258 women with early breast cancer in the EBCTCG database on Jan 17, 2023, patient-level data were available from 151 randomised trials that included 155 746 women. Rates of distant tumour recurrence improved similarly in women with oestrogen receptor-positive and oestrogen receptor-negative tumours. 80·5% of the improvement for oestrogen receptor-positive disease and 89·8% of the improvement for eostrogen receptor-negative disease was explained by changes in patient and tumour characteristics and improved treatments, but remained significant (p<0·0001). More recently diagnosed patients were more likely to have node-negative disease. 10-year distant recurrence risks during 1990-99 versus 2000-09 were as follows: for node-negative disease, 10·1% versus 7·3% for oestrogen receptor-positive disease and 18·3% versus 11·9% for oestrogen receptor-negative disease; for disease with one to three positive nodes, 19·9% versus 14·7% for oestrogen receptor-positive disease and 31·9% versus 22·1% for oestrogen receptor-negative disease; and for disease with four to nine positive nodes, 39·6% versus 28·5% for oestrogen receptor-positive disease and 47·8% versus 36·5% for oestrogen receptor-negative disease. After adjustment for therapy, rates were reduced by 25% (oestrogen receptor-positive disease) and 19% (oestrogen receptor-negative disease) after 2000 versus the 1990s, with similar improvements observed in oestrogen receptor-positive disease beyond 5 years. Most of the improvement in trial outcomes is explained by a greater proportion of women with lower-risk disease entering trials and improved adjuvant treatment. After adjustment, women diagnosed since 2000 have about a fifth lower rate of distant recurrence than the 1990s. Long-term risks of distant recurrence for oestrogen receptor-positive disease remain, but are about a tenth lower now than in our previous report. Cancer Research UK, UK Medical Research Council.