The primary analysis of IMvigor130 showed a significant progression-free survival benefit with first-line atezolizumab plus platinum-based chemotherapy (group A) versus placebo plus platinum-based chemotherapy (group C) in patients with locally advanced or metastatic urothelial cancer. However, this finding did not translate into significant overall survival benefit for group A versus group C at the final analysis, precluding formal statistical testing of outcomes with atezolizumab monotherapy (group B) versus group C. Here we report the final overall survival results for group B versus group C; this report is descriptive and should be considered exploratory due to the study's statistical design. In this global, partially blinded, randomised, controlled, phase 3 study, patients (aged ≥18 years) who had locally advanced or metastatic urothelial cancer previously untreated in the metastatic setting and Eastern Cooperative Oncology Group performance status of 0-2 were enrolled at 221 hospitals and oncology centres in 35 countries. Patients were randomly assigned (1:1:1), using a permuted block method (block size of six) and an interactive voice and web response system, stratified by PD-L1 status, Bajorin score, and investigator's choice of platinum-based chemotherapy, to receive either atezolizumab plus platinum-based chemotherapy (group A), atezolizumab alone (group B), or placebo plus platinum-based chemotherapy (group C). Sponsors, investigators, and patients were masked to assignment to atezolizumab or placebo in group A and group C; atezolizumab monotherapy in group B was open label. For groups B and C, atezolizumab (1200 mg) or placebo was administered intravenously every 3 weeks. Chemotherapy involved 21-day cycles of gemcitabine (1000 mg/m body surface area on day 1 and day 8 of each cycle) plus the investigator's choice of carboplatin (area under the curve 4·5 mg/mL per min or 5 mg/mL per min) or cisplatin (70 mg/m body surface area), administered intravenously. Co-primary endpoints were progression-free survival and overall survival in group A versus group C, and overall survival in group B versus group C, tested hierarchically, in the intention-to-treat (ITT) population, and then the populations with high PD-L1 tumour expression (immune cell IC expression score of IC2/3) if the results from group A versus group C were significant. Here, we report the co-primary endpoint of overall survival for group B versus group C in the ITT and IC2/3 populations. The ITT population for this analysis comprised concurrently enrolled patients in groups B and C who were randomly assigned to treatment. For the safety analysis, all patients enrolled in group B and group C who received any study treatment were included. The trial is registered with ClinicalTrials.gov, NCT02807636, and is active but no longer recruiting. Between July 15, 2016, and July 20, 2018, 1213 patients were enrolled and randomly assigned to treatment, of whom 362 patients were assigned to group B and 400 to group C, of whom 360 and 359, respectively, were enrolled concurrently (ITT population). 543 (76%) of 719 patients were male, 176 (24%) were female, and 534 (74%) were White. As of data cutoff (Aug 31, 2022), after a median follow-up of 13·4 months (IQR 6·2-30·8), median overall survival was 15·2 months (95% CI 13·1-17·7; 271 deaths) in group B and 13·3 months (11·9-15·6; 275 deaths) in group C (stratified hazard ratio 0·98 95% CI 0·82-1·16). The most common grade 3-4 treatment-related adverse events were anaemia (two 1% in patients who received atezolizumab monotherapy vs 133 34% in those who received placebo plus chemotherapy), neutropenia (one <1% vs 115 30%), decreased neutrophil count (0 vs 95 24%), and decreased platelet count (one <1% vs 92 24%). Serious adverse events occurred in 163 (46%) patients versus 196 (50%). Treatment-related deaths occurred in three (1%; n=1 each, pneumonia, interstitial lung disease, large intestinal obstruction) patients who received atezolizumab monotherapy and four (1%; n=1 each, diarrhoea, febrile neutropenia, unexplained death, toxic hepatitis) who received placebo plus chemotherapy. The final analysis from IMvigor130 did not show a significant improvement in overall survival with first-line atezolizumab monotherapy compared with platinum-based chemotherapy in the intention-to-treat population. The safety profile of atezolizumab monotherapy remained acceptable after extended follow-up, with no new safety signals. F Hoffmann-La Roche.