Background: Cusatuzumab, a high-affinity anti-CD70 antibody, has shown preliminary activity as a treatment for acute myeloid leukaemia when combined with azacitidine. We aimed to determine the optimum dose for future trials of cusatuzumab in combination with azacitidine in patients with previously untreated acute myeloid leukaemia who are not eligible for intensive chemotherapy. Methods: In this randomised, phase 2, open-label, dose-optimisation study we enrolled adult patients aged 18 years or older with newly diagnosed acute myeloid leukaemia not eligible for intensive chemotherapy, and with Eastern Cooperative Oncology Group scores of 0–2, from 40 hospitals and centres across seven countries. In part one of the trial, participants were randomly allocated 1:1 to 10 mg/kg or 20 mg/kg intravenous cusatuzumab on days 3 and 17, combined with subcutaneous or intravenous azacitidine 75 mg/m2 on days 1–7 in 28-day cycles. The primary efficacy outcome was the rate of complete remission in the intention-to-treat group. The two dose cohorts were evaluated independently without between-cohort statistical comparison. Safety analyses were performed in all patients who received one dose of study drug. Part two of the trial was planned to be a single-arm expansion to evaluate cusatuzumab plus azacitidine at the cusatuzumab dose level selected in part one (primary hypothesis ≥35% rate of complete remission vs null hypothesis of 20%); however, changes in the acute myeloid leukaemia treatment landscape during this trial made it unlikely that enrolment to part two of the study would be clinically feasible, so the study stopped at the end of part one. The trial was registered at ClinicalTrials.gov, NCT04023526. Findings: 103 patients were enrolled between Aug 30, 2019, and Feb 25, 2020, and randomly assigned to either cusatuzumab 10 mg/kg (n=51) or 20 mg/kg (n=52). Median follow-up was 7·2 months (IQR 10·7 months). 57 of 103 (55%) patients were male and 46 (45%) patients were female, 78 (76%) were White, one (1%) was Asian, and 24 (23%) did not report their race. In the 10 mg/kg group, complete remission rate was 12% (six of 51 patients; 95% CI 6–23) and in the 20 mg/kg group was 27% (14 of 52; 17–40). Grade 3 or worse treatment-emergent adverse events (TEAEs) were similar between the cusatuzumab 10 mg/kg (n=51) and 20 mg/kg (n=51) cohorts and included thrombocytopenia (24 patients [47%] vs 29 [57%]), anaemia (24 [47%] vs 17 [33%]), and neutropenia (20 [39%] in both cohorts). Serious TEAEs were also similar in the two cohorts (44 [86%] vs 40 [78%]). Treatment-related TEAEs leading to death were reported in both groups (three patients [6%] in the 10 mg/kg group vs one patient [2%] in the 20 mg/kg group); the reported causes of death were pneumonia (n=2) and septic shock (n=2). Interpretation: Although part one of this study was not designed to formally compare the two dose cohorts for efficacy, the totality of clinical data for cusatuzumab studies performed to date indicate that cusatuzumab 20 mg/kg plus azacitidine represents the optimal dose for further studies. A phase 1b study investigating the triple combination of cusatuzumab with venetoclax and azacitidine is underway (NCT04150887). Funding: Janssen Research & Development and argenx.