BackgroundWe evaluated a new chemoimmunotherapy combination based on the anti-PD1 monoclonal antibody pembrolizumab and the pyrimidine antimetabolite gemcitabine in HER2- advanced breast cancer (ABC) patients previously treated in the advanced setting, in order to explore a potential synergism that could eventually obtain long term benefit in these patients.MethodsHER2-negative ABC patients received 21-day cycles of pembrolizumab 200 mg (day 1) and gemcitabine (days 1 and 8). A run-in-phase (6 + 6 design) was planned with two dose levels (DL) of gemcitabine (1,250 mg/m2 [DL0]; 1,000 mg/m2 [DL1]) to determine the recommended phase II dose (RP2D). The primary objective was objective response rate (ORR). Tumor infiltrating lymphocytes (TILs) density and PD-L1 expression in tumors and myeloid-derived suppressor cells (MDSCs) levels in peripheral blood were analyzed.ResultsFourteen patients were treated with DL0, resulting in RP2D. Thirty-six patients were evaluated during the first stage of Simon’s design. Recruitment was stopped as statistical assumptions were not met. The median age was 52; 21 (58%) patients had triple-negative disease, 28 (78%) visceral involvement, and 27 (75%) ≥ 2 metastatic locations. Progression disease was observed in 29 patients. ORR was 15% (95% CI, 5–32). Eight patients were treated ≥ 6 months before progression. Fourteen patients reported grade ≥ 3 treatment-related adverse events. Due to the small sample size, we did not find any clear association between immune tumor biomarkers and treatment efficacy that could identify a subgroup with higher probability of response or better survival. However, patients that experienced a clinical benefit showed decreased MDSCs levels in peripheral blood along the treatment.ConclusionPembrolizumab 200 mg and gemcitabine 1,250 mg/m2 were considered as RP2D. The objective of ORR was not met; however, 22% patients were on treatment for ≥ 6 months. ABC patients that could benefit of chemoimmunotherapy strategies must be carefully selected by robust and validated biomarkers. In our heavily pretreated population, TILs, PD-L1 expression and MDSCs levels could not identify a subgroup of patients for whom the combination of gemcitabine and pembrolizumab would induce long term benefit.Trial registrationClinicalTrials.gov and EudraCT (NCT03025880 and 2016–001,779-54, respectively). Registration dates: 20/01/2017 and 18/11/2016, respectively.