Background: In the single-arm, phase 2 DESTINY-Breast01 trial, trastuzumab deruxtecan showed robust activity in patients with HER2-positive metastatic breast cancer who were refractory or resistant to trastuzumab emtansine; a population with scarce effective treatments. In DESTINY-Breast02, we aimed to compare the efficacy and safety of trastuzumab deruxtecan with treatment of physician's choice in this patient population. Methods: This randomised, open-label, multicentre, phase 3 trial was conducted at 227 sites (hospitals, university hospitals, clinics, community centres, and private oncology centres) in North America, Europe, Asia, Australia, Brazil, Israel, and Türkiye. Eligible patients were aged 18 years or older, had unresectable or HER2-positive metastatic breast cancer, previously received trastuzumab emtansine, disease progression, an Eastern Cooperative Oncology Group performance status of 0 or 1, and adequate renal and hepatic function. Patients were randomly assigned (2:1) to receive trastuzumab deruxtecan (intravenously at 5·4 mg/kg once every 3 weeks) or treatment of physician's choice using block randomisation. Treatment of physician's choice was either capecitabine (1250 mg/m2; orally twice per day on days 1–14) plus trastuzumab (8 mg/kg intravenously on day 1 then 6 mg/kg once per day) or capecitabine (1000 mg/m2) plus lapatinib (1250 mg orally once per day on days 1–21), with a 21-day schedule. The primary endpoint was progression-free survival based on blinded independent central review in the full analysis set. This study is registered with ClinicalTrials.gov, NCT03523585. Findings: Between Sept 6, 2018, and Dec 31, 2020, 608 patients were randomly assigned to receive trastuzumab deruxtecan (n=406; two did not receive treatment) or treatment of physician's choice (n=202; seven did not receive treatment). 608 (100%) patients were included in the full analysis set. The median age was 54·2 years (IQR 45·5–63·4) in the trastuzumab deruxtecan group and 54·7 years (48·0–63·0) in the treatment of physician's choice group. 384 (63%) patients were White, 603 (99%) were female, and five (<1%) were male. The median follow-up was 21·5 months (IQR 15·2–28·4) in the trastuzumab deruxtecan group and 18·6 months (8·8–26·0) in the treatment of physician's choice group. Median progression-free survival by blinded independent central review was 17·8 months (95% CI 14·3–20·8) in the trastuzumab deruxtecan group versus 6·9 months (5·5–8·4) in the treatment of physician's choice group (HR 0·36 [0·28–0·45]; p<0·0001). The most common treatment-emergent adverse events were nausea (293 [73%] of 404 with trastuzumab deruxtecan vs 73 [37%] of 195 with treatment of physician's choice), vomiting (152 [38%] vs 25 [13%]), alopecia (150 [37%] vs eight [4%]), fatigue (147 [36%] vs 52 [27%]), diarrhoea (109 [27%] vs 105 [54%]), and palmar–plantar erythrodysaesthesia (seven [2%] vs 100 [51%]). Grade 3 or higher treatment-emergent adverse events occurred in 213 (53%) patients receiving trastuzumab deruxtecan versus 86 (44%) receiving treatment of physician's choice; whereas drug-related interstitial lung disease occurred in 42 (10%; including two grade 5 death events) versus one (<1%). Interpretation: DESTINY-Breast02 shows the favourable benefit–risk profile of trastuzumab deruxtecan in patients with HER2 positive metastatic breast cancer, as previously reported in DESTINY-Breast01, and is the first randomised study to show that one antibody-drug conjugate can overcome resistance to a previous one. Funding: Daiichi Sankyo and AstraZeneca.