METHODS: In this randomised, placebo-controlled, double-blind, phase 3 trial, done in 209 sites in 29 countries, we randomly assigned patients 2:1 with untreated locally recurrent inoperable or metastatic triple-negative breast cancer using a block method (block size of six) and an interactive voice-response system with integrated web-response to pembrolizumab (200 mg) every 3 weeks plus chemotherapy (nab-paclitaxel; paclitaxel; or gemcitabine plus carboplatin) or placebo plus chemotherapy. Randomisation was stratified by type of on-study chemotherapy (taxane or gemcitabine-carboplatin), PD-L1 expression at baseline (combined positive score [CPS] ≥1 or <1), and previous treatment with the same class of chemotherapy in the neoadjuvant or adjuvant setting (yes or no). Eligibility criteria included age at least 18 years, centrally confirmed triple-negative breast cancer; at least one measurable lesion; provision of a newly obtained tumour sample for determination of triple-negative breast cancer status and PD-L1 status by immunohistochemistry at a central laboratory; an Eastern Cooperative Oncology Group performance status score 0 or 1; and adequate organ function. The sponsor, investigators, other study site staff (except for the unmasked pharmacist), and patients were masked to pembrolizumab versus saline placebo administration. In addition, the sponsor, the investigators, other study site staff, and patients were masked to patient-level tumour PD-L1 biomarker results. Dual primary efficacy endpoints were progression-free survival and overall survival assessed in the PD-L1 CPS of 10 or more, CPS of 1 or more, and intention-to-treat populations. The definitive assessment of progression-free survival was done at this interim analysis; follow-up to assess overall survival is continuing. For progression-free survival, a hierarchical testing strategy was used, such that testing was done first in patients with CPS of 10 or more (prespecified statistical criterion was α=0·00411 at this interim analysis), then in patients with CPS of 1 or more (α=0·00111 at this interim analysis, with partial alpha from progression-free survival in patients with CPS of 10 or more passed over), and finally in the intention-to-treat population (α=0·00111 at this interim analysis). This study is registered with ClinicalTrials.gov, NCT02819518, and is ongoing.

FINDINGS: Between Jan 9, 2017, and June 12, 2018, of 1372 patients screened, 847 were randomly assigned to treatment, with 566 patients in the pembrolizumab-chemotherapy group and 281 patients in the placebo-chemotherapy group. At the second interim analysis (data cutoff, Dec 11, 2019), median follow-up was 25·9 months (IQR 22·8-29·9) in the pembrolizumab-chemotherapy group and 26·3 months (22·7-29·7) in the placebo-chemotherapy group. Among patients with CPS of 10 or more, median progression-free survival was 9·7 months with pembrolizumab-chemotherapy and 5·6 months with placebo-chemotherapy (hazard ratio [HR] for progression or death, 0·65, 95% CI 0·49-0·86; one-sided p=0·0012 [primary objective met]). Median progression-free survival was 7·6 and 5·6 months (HR, 0·74, 0·61-0·90; one-sided p=0·0014 [not significant]) among patients with CPS of 1 or more and 7·5 and 5·6 months (HR, 0·82, 0·69-0·97 [not tested]) among the intention-to-treat population. The pembrolizumab treatment effect increased with PD-L1 enrichment. Grade 3-5 treatment-related adverse event rates were 68% in the pembrolizumab-chemotherapy group and 67% in the placebo-chemotherapy group, including death in <1% in the pembrolizumab-chemotherapy group and 0% in the placebo-chemotherapy group.

INTERPRETATION: Pembrolizumab-chemotherapy showed a significant and clinically meaningful improvement in progression-free survival versus placebo-chemotherapy among patients with metastatic triple-negative breast cancer with CPS of 10 or more. These findings suggest a role for the addition of pembrolizumab to standard chemotherapy for the first-line treatment of metastatic triple-negative breast cancer.