BACKGROUND: Immune checkpoint inhibitors are the standard of care for first-line treatment of patients with metastatic renal cell carcinoma, yet optimised treatment of patients whose disease progresses after these therapies is unknown. The aim of this study was to determine whether adding atezolizumab to cabozantinib delayed disease progression and prolonged survival in patients with disease progression on or after previous immune checkpoint inhibitor treatment. METHODS: CONTACT-03 was a multicentre, randomised, open-label, phase 3 trial, done in 135 study sites in 15 countries in Asia, Europe, North America, and South America. Patients aged 18 years or older with locally advanced or metastatic renal cell carcinoma whose disease had progressed with immune checkpoint inhibitors were randomly assigned (1:1) to receive atezolizumab (1200 mg intravenously every 3 weeks) plus cabozantinib (60 mg orally once daily) or cabozantinib alone. Randomisation was done through an interactive voice-response or web-response system in permuted blocks (block size four) and stratified by International Metastatic Renal Cell Carcinoma Database Consortium risk group, line of previous immune checkpoint inhibitor therapy, and renal cell carcinoma histology. The two primary endpoints were progression-free survival per blinded independent central review and overall survival. The primary endpoints were assessed in the intention-to-treat population and safety was assessed in all patients who received at least one dose of study drug. The trial is registered with ClinicalTrials.gov, NCT04338269, and is closed to further accrual. FINDINGS: From July 28, 2020, to Dec 27, 2021, 692 patients were screened for eligibility, 522 of whom were assigned to receive atezolizumab-cabozantinib (263 patients) or cabozantinib (259 patients). 401 (77%) patients were male and 121 (23%) patients were female. At data cutoff (Jan 3, 2023), median follow-up was 15·2 months (IQR 10·7-19·3). 171 (65%) patients receiving atezolizumab-cabozantinib and 166 (64%) patients receiving cabozantinib had disease progression per central review or died. Median progression-free survival was 10·6 months (95% CI 9·8-12·3) with atezolizumab-cabozantinib and 10·8 months (10·0-12·5) with cabozantinib (hazard ratio [HR] for disease progression or death 1·03 [95% CI 0·83-1·28]; p=0·78). 89 (34%) patients in the atezolizumab-cabozantinib group and 87 (34%) in the cabozantinib group died. Median overall survival was 25·7 months (95% CI 21·5-not evaluable) with atezolizumab-cabozantinib and was not evaluable (21·1-not evaluable) with cabozantinib (HR for death 0·94 [95% CI 0·70-1·27]; p=0·69). Serious adverse events occurred in 126 (48%) of 262 patients treated with atezolizumab-cabozantinib and 84 (33%) of 256 patients treated with cabozantinib; adverse events leading to death occurred in 17 (6%) patients in the atezolizumab-cabozantinib group and nine (4%) in the cabozantinib group. INTERPRETATION: The addition of atezolizumab to cabozantinib did not improve clinical outcomes and led to increased toxicity. These results should discourage sequential use of immune checkpoint inhibitors in patients with renal cell carcinoma outside of clinical trials. FUNDING: F Hoffmann-La Roche and Exelixis.