BACKGROUND: Immunotherapy in combination with chemotherapy has shown promising efficacy across many different tumour types. We report the prespecified second interim overall survival analysis of the phase 3 IMpassion130 study assessing the efficacy and safety of atezolizumab plus nab-paclitaxel in patients with unresectable, locally advanced or metastatic triple-negative breast cancer. METHODS: In this randomised, placebo-controlled, double-blind, phase 3 trial, done in 246 academic centres and community oncology practices in 41 countries, patients aged 18 years or older, with previously untreated, histologically documented, locally advanced or metastatic triple-negative breast cancer, and Eastern Cooperative Oncology Group performance status of 0 or 1 were eligible. Patients were randomly assigned (1:1) using a permuted block method (block size of four) and an interactive voice-web response system. Randomisation was stratified by previous taxane use, liver metastases, and PD-L1 expression on tumour-infiltrating immune cells. Patients received atezolizumab 840 mg or matching placebo intravenously on day 1 and day 15 of every 28-day cycle and nab-paclitaxel 100 mg/m of body surface area intravenously on days 1, 8, and 15 until progression or unacceptable toxicity. Investigators, patients, and the funder were masked to treatment assignment. Coprimary endpoints were investigator-assessed progression-free survival per Response Evaluation Criteria in Solid Tumors version 1.1 and overall survival, assessed in the intention-to-treat population and in patients with PD-L1 immune cell-positive tumours (tumours with ≥1% PD-L1 expression). The final progression-free survival results were previously reported at the first interim overall survival analysis. The prespecified statistical testing hierarchy meant that overall survival in the subgroup of PD-L1 immune cell-positive patients could only be formally tested if overall survival was significantly different between the treatment groups in the intention-to-treat population. This study is registered with ClinicalTrials.gov, NCT02425891. FINDINGS: Between June 23, 2015, and May 24, 2017, 902 patients were enrolled, of whom 451 were randomly assigned to receive atezolizumab plus nab-paclitaxel and 451 were assigned to receive placebo plus nab-paclitaxel (the intention-to-treat population). Six patients from each group did not receive treatment. At the second interim analysis (data cutoff Jan 2, 2019), median follow-up was 18·5 months (IQR 9·6-22·8) in the atezolizumab group and 17·5 months (8·4-22·4) in the placebo group. Median overall survival in the intention-to-treat patients was 21·0 months (95% CI 19·0-22·6) with atezolizumab and 18·7 months (16·9-20·3) with placebo (stratified hazard ratio [HR] 0·86, 95% CI 0·72-1·02, p=0·078). In the exploratory overall survival analysis in patients with PD-L1 immune cell-positive tumours, median overall survival was 25·0 months (95% CI 19·6-30·7) with atezolizumab versus 18·0 months (13·6-20·1) with placebo (stratified HR 0·71, 0·54-0·94]). As of Sept 3, 2018 (the date up to which updated safety data were available), the most common grade 3-4 adverse events were neutropenia (38 [8%] of 453 patients in the atezolizumab group vs 36 [8%] of 437 patients in the placebo group), peripheral neuropathy (25 [6%] vs 12 [3%]), decreased neutrophil count (22 [5%] vs 16 [4%]), and fatigue (17 [4%] vs 15 [3%]). Treatment-related deaths occurred in two (<1%) patients in the atezolizumab group (autoimmune hepatitis related to atezolizumab [n=1] and septic shock related to nab-paclitaxel [n=1]) and one (<1%) patient in the placebo group (hepatic failure). No new treatment-related deaths have been reported since the primary clinical data cutoff date (April 17, 2018). INTERPRETATION: Consistent with the first interim analysis, this second interim overall survival analysis of IMpassion130 indicates no significant difference in overall survival between the treatment groups in the intention-to-treat population but suggests a clinically meaningful overall survival benefit with atezolizumab plus nab-paclitaxel in patients with PD-L1 immune cell-positive disease. However, this positive result could not be formally tested due to the prespecified statistical testing hierarchy. For patients with PD-L1 immune cell-positive metastatic triple-negative breast cancer, atezolizumab plus nab-paclitaxel is an important therapeutic option in a disease with high unmet need. FUNDING: F Hoffmann-La Roche and Genentech.