Background: New options are needed for patients with heavily pretreated breast cancer. Etirinotecan pegol is a long-acting topoisomerase-I inhibitor that prolongs exposure to, but reduces the toxicity of, SN38 (the active metabolite of irinotecan). We assessed whether etirinotecan pegol is superior to currently available treatments for patients with previously treated, locally recurrent or metastatic breast cancer. Methods: In this open-label, multicentre, randomised phase 3 study (BEACON; BrEAst Cancer Outcomes with NKTR-102), conducted at 135 sites in 11 countries, patients with locally recurrent or metastatic breast cancer previously treated with an anthracycline, a taxane, and capecitabine (and two to five previous regimens for advanced disease) were randomly assigned (1:1) centrally via an interactive response system to etirinotecan pegol (145 mg/m2 as a 90-min intravenous infusion every 3 weeks) or single-drug treatment of physician's choice. Patients with stable brain metastases and an Eastern Cooperative Oncology Group performance status of 0-1 were eligible. Randomisation was stratified with a permuted block scheme by region, previous eribulin, and receptor status. After randomisation, patients and investigators were aware of treatment assignments. The primary endpoint was overall survival in the intention-to-treat population. This study is registered with ClinicalTrials.gov, number NCT01492101. Findings: Between Dec 19, 2011, and Aug 20, 2013, 852 patients were randomly assigned; 429 to etirinotecan pegol and 423 to treatment of physician's choice. There was no significant difference in overall survival between groups (median 12·4 months [95% CI 11·0-13·6] for the etirinotecan pegol group vs 10·3 months [9·0-11·3] for the treatment of physician's choice group; hazard ratio 0·87 [95% CI 0·75-1·02]; p=0·084). The safety population includes the 831 patients who received at least one dose of assigned treatment (425 assigned to etirinotecan pegol and 406 to treatment of physician's choice). Serious adverse events were recorded for 128 (30%) patients treated with etirinotecan pegol and 129 (32%) treated with treatment of physician's choice. Fewer patients in the etirinotecan pegol group had grade 3 or worse toxicity than those in the treatment of physician's choice group (204 [48%] vs 256 [63%]; p<0·0001). The most common grade 3 or worse adverse events were diarrhoea (41 [10%] in the experimental group vs five [1%] in the control group), neutropenia (41 [10%] vs 125 [31%]), and peripheral neuropathy (two [<1%] vs 15 [4%]). Three patients in the etirinotecan pegol group died of treatment-related adverse events (pneumonia, myelodysplastic syndrome, and acute renal failure) and two in the treatment of physician's choice group (neutropenic sepsis and septic shock). Interpretation: This trial did not demonstrate an improvement in overall survival for etirinotecan pegol compared to treatment of physician's choice in patients with heavily pre-treated advanced breast cancer. The toxicity profile noted in the etirinotecan pegol group differed from that in the control group. In view of the frequency of cross-resistance and overlapping toxicities noted with many available drugs and the need for effective drugs in highly refractory disease, etirinotecan pegol may warrant further research in some subgroups of patients. Funding: Nektar Therapeutics.