KEYNOTE-355: Pembrolizumab Plus Chemotherapy in Previously Untreated Locally Recurrent Inoperable or Metastatic Triple-Negative Breast Cancer

Co-primary endpoint 1 (OS, CPS ≥10): 23.0 vs 16.1 months — HR 0.73 (95% CI 0.55–0.95; p=0.0185) ✓ Significant Co-primary endpoint 2 (OS, CPS ≥1): 17.6 vs 16.0 months — HR 0.86 (95% CI 0.72–1.04; p=0.1125) ✗ Not significant OS, ITT population: 17.2 vs 15.5 months — HR 0.89 (95% CI 0.76–1.05) — not formally tested Safety signal: Immune-mediated AEs (26.5% vs 6.4%; grade ≥3: 5.3% vs 0%); 2 treatment-related deaths in pembrolizumab arm

📱 Clinic quick reference: Mobile version — save for point-of-care use

Clinical Bottom Line

KEYNOTE-355 demonstrated that adding pembrolizumab to first-line chemotherapy significantly extends overall survival in patients with metastatic triple-negative breast cancer whose tumors express PD-L1 with a combined positive score of 10 or more. This is the patient population that benefits — and the population for which pembrolizumab carries its FDA-approved indication in this setting. Patients with lower PD-L1 expression did not derive a statistically significant survival benefit from the addition of pembrolizumab.

This trial, together with IMpassion130 (atezolizumab plus nab-paclitaxel, which was voluntarily withdrawn from the US market in 2021), established checkpoint inhibitor therapy as a component of first-line treatment for PD-L1–positive metastatic TNBC. In current practice, pembrolizumab plus chemotherapy is the standard first-line immunotherapy-based regimen for this population. The CPS ≥10 threshold is critical for patient selection — the data clearly show that the survival benefit concentrates in this biomarker-defined subgroup and does not extend to the broader CPS ≥1 or unselected populations.

From a safety perspective, the addition of pembrolizumab to chemotherapy introduced immune-mediated adverse events — most commonly thyroid dysfunction — without a dramatic increase in the overall grade ≥3 toxicity burden. Two treatment-related deaths occurred in the pembrolizumab arm. Clinicians should maintain vigilance for immune-mediated toxicities and ensure that patients understand the monitoring requirements before initiating therapy.

Trial Overview

Study Design

• Trial name and registration: KEYNOTE-355, NCT02819518
• Design: Phase 3, randomized, double-blind, placebo-controlled
• Randomization: 2:1 (pembrolizumab–chemotherapy : placebo–chemotherapy), stratified by type of chemotherapy received during the trial (a taxane or gemcitabine–carboplatin), tumor PD-L1 expression (CPS ≥1 or CPS <1), and previous treatment with the same class of neoadjuvant or adjuvant chemotherapy as that received during the trial (yes or no)
• Setting: Previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer; first-line
• Analysis type: Protocol-specified final analysis of overall survival
• Funding: Merck Sharp and Dohme

Co-primary endpoints were progression-free survival (PFS) and overall survival (OS), each tested hierarchically first in the CPS ≥10 subgroup, then CPS ≥1, then the ITT population. PFS was formally tested and reported in the prior interim analysis (Cortes et al., Lancet 2020). This publication reports the final analysis of OS.

Mechanism of Action

Pembrolizumab is a humanized monoclonal antibody that binds to the programmed death-1 (PD-1) receptor on T cells, blocking interaction with its ligands PD-L1 and PD-L2. By disrupting this immune checkpoint, pembrolizumab restores T-cell–mediated antitumor immune responses [2].

Patient Population

• Key eligibility: Adult patients with centrally confirmed, locally recurrent inoperable or metastatic TNBC per ASCO-CAP guidelines; measurable disease per RECIST v1.1; ECOG PS 0 or 1; adequate organ function. Patients who completed treatment for stage I–III breast cancer were eligible if ≥6 months had elapsed between completion of treatment with curative intent and first documented recurrence.
• Key exclusions: Prior investigational agent within 4 weeks; prior anti-PD-1/PD-L1/PD-L2 therapy; active autoimmune disease within 2 years; active CNS metastases; history of noninfectious pneumonitis; active infection warranting systemic therapy.
• Biomarker selection: PD-L1 CPS assessed centrally using PD-L1 IHC 22C3 pharmDx (Agilent Technologies); efficacy analyzed in CPS ≥10, CPS ≥1, and ITT populations.
• Sample size flow: 847 patients randomized (566 pembrolizumab–chemotherapy, 281 placebo–chemotherapy); 843 treated (562 + 281 in safety population). Number screened not reported in this publication.

Baseline Characteristics

Baseline characteristics were reported as well balanced between the two groups. Detailed baseline data were presented in Supplementary Table S1, which was not available for extraction. The CPS ≥1 subgroup comprised 636 patients (75.1% of ITT), and the CPS ≥10 subgroup comprised 323 patients (38.1% of ITT); baseline characteristics of these subgroups were described as similar to those of the intention-to-treat population.

Treatment Protocol

Experimental Arm: Pembrolizumab–Chemotherapy (n=566 randomized; safety population n=562)

Pembrolizumab (200 mg every 3 weeks) plus investigator's choice of chemotherapy (nab-paclitaxel, paclitaxel, or gemcitabine–carboplatin).

• Dose and schedule: Pembrolizumab 200 mg IV every 3 weeks; nab-paclitaxel 100 mg/m² days 1, 8, 15 of every 28-day cycle; OR paclitaxel 90 mg/m² days 1, 8, 15 of every 28-day cycle; OR gemcitabine 1000 mg/m² plus carboplatin AUC 2 days 1 and 8 of every 21-day cycle
• Treatment duration: Up to 35 intravenous administrations of pembrolizumab or until confirmed disease progression, unacceptable toxic effects, withdrawal of consent, or physician's decision
• Median treatment duration: 26.4 weeks
• Median relative dose intensity: Not reported in this publication

Control Arm: Placebo–Chemotherapy (n=281 randomized; safety population n=281)

Placebo every 3 weeks plus investigator's choice of chemotherapy (nab-paclitaxel, paclitaxel, or gemcitabine–carboplatin).

• Dose and schedule: Placebo IV every 3 weeks; nab-paclitaxel 100 mg/m² days 1, 8, 15 of every 28-day cycle; OR paclitaxel 90 mg/m² days 1, 8, 15 of every 28-day cycle; OR gemcitabine 1000 mg/m² plus carboplatin AUC 2 days 1 and 8 of every 21-day cycle
• Treatment duration: Up to 35 intravenous administrations of placebo or until confirmed disease progression, unacceptable toxic effects, withdrawal of consent, or physician's decision
• Median treatment duration: 23.1 weeks
• Median relative dose intensity: Not reported in this publication

Efficacy Outcomes

Co-Primary Endpoint 1: Overall Survival — CPS ≥10 Subgroup

Definition: Overall survival among patients whose tumors expressed PD-L1 with a CPS of 10 or more Analysis population: CPS ≥10 subgroup (ITT population with CPS ≥10; N=323: 220 pembrolizumab–chemotherapy, 103 placebo–chemotherapy) Statistical method: Stratified log-rank test; stratified Cox proportional-hazards model with Efron's method of tie handling Alpha allocation: 0.0227 two-sided (for this final analysis) Median follow-up: 44.1 months

Pembrolizumab–chemotherapy: 23.0 months (95% CI, 19.0 to 26.3); 155 of 220 patients (70.5%) died Placebo–chemotherapy: 16.1 months (95% CI, 12.6 to 18.8); 84 of 103 patients (81.6%) died Comparison: HR 0.73 (95% CI, 0.55 to 0.95; two-sided P=0.0185)

According to the prespecified criterion for statistical significance for the final analysis (alpha level of 0.0227), the addition of pembrolizumab to chemotherapy resulted in significantly longer overall survival than chemotherapy alone in the CPS ≥10 subgroup.

The estimated overall survival at 18 months was 58.3% (95% CI, 51.4 to 64.5) in the pembrolizumab–chemotherapy group and 44.7% (95% CI, 34.9 to 53.9) in the placebo–chemotherapy group.

Co-Primary Endpoint 2: Overall Survival — CPS ≥1 Subgroup

Definition: Overall survival among patients whose tumors expressed PD-L1 with a CPS of 1 or more Analysis population: CPS ≥1 subgroup (ITT population with CPS ≥1; N=636: 425 pembrolizumab–chemotherapy, 211 placebo–chemotherapy) Statistical method: Stratified log-rank test; stratified Cox proportional-hazards model with Efron's method of tie handling Alpha allocation: 0.0344 two-sided (since CPS ≥10 subgroup met significance, alpha was recycled) Median follow-up: 44.1 months

Pembrolizumab–chemotherapy: 17.6 months (95% CI, 15.5 to 19.5); 336 of 425 patients (79.1%) died Placebo–chemotherapy: 16.0 months (95% CI, 12.8 to 17.4); 177 of 211 patients (83.9%) died Comparison: HR 0.86 (95% CI, 0.72 to 1.04; two-sided P=0.1125)

No significant between-group difference in overall survival was observed according to the prespecified criterion (alpha level of 0.0344).

The estimated overall survival at 18 months was 48.4% (95% CI, 43.5 to 53.0) in the pembrolizumab–chemotherapy group and 41.4% (95% CI, 34.7 to 48.0) in the placebo–chemotherapy group.

Overall Survival — Intention-to-Treat Population

Analysis population: ITT population (N=847: 566 pembrolizumab–chemotherapy, 281 placebo–chemotherapy) Median follow-up: 44.1 months

Pembrolizumab–chemotherapy: 17.2 months (95% CI, 15.3 to 19.0); 460 of 566 patients (81.3%) died Placebo–chemotherapy: 15.5 months (95% CI, 13.9 to 17.2); 238 of 281 patients (84.7%) died Comparison: HR 0.89 (95% CI, 0.76 to 1.05)

Significance was not tested because of the prespecified multiplicity strategy: OS in the ITT population could only be formally tested if the CPS ≥1 subgroup had met the significance boundary, which it did not.

The estimated overall survival at 18 months was 47.8% (95% CI, 43.6 to 51.9) in the pembrolizumab–chemotherapy group and 41.8% (95% CI, 36.0 to 47.5) in the placebo–chemotherapy group.

Multiplicity Control

Familywise type I error was controlled at a two-sided alpha of 0.05, split among PFS (0.01), OS (0.036), and ORR (0.004). The group sequential method with graphical approach of Maurer and Bretz was used for alpha allocation across interim and final analyses and for recycling between endpoints and populations. OS was tested hierarchically: CPS ≥10 first, then CPS ≥1 (only if CPS ≥10 met significance), then ITT (only if CPS ≥1 met significance). Because the CPS ≥10 subgroup met significance at the final analysis, alpha was recycled to the CPS ≥1 subgroup (alpha 0.0344). Because CPS ≥1 did not meet its boundary, OS in the ITT population was not formally tested.

Key Secondary Endpoints

Progression-Free Survival (Updated)

PFS was a co-primary endpoint formally tested at the prior interim analysis. The following are updated descriptive results at the final analysis.

CPS ≥10 subgroup: Median PFS was 9.7 months in the pembrolizumab–chemotherapy group vs 5.6 months in the placebo–chemotherapy group (HR 0.66; 95% CI, 0.50 to 0.88).

CPS ≥1 subgroup: Median PFS was 7.6 months vs 5.6 months (HR 0.75; 95% CI, 0.62 to 0.91).

ITT population: Median PFS was 7.5 months vs 5.6 months (HR 0.82; 95% CI, 0.70 to 0.98).

Objective Response Rate (Updated)

CPS ≥10 subgroup: A confirmed objective response occurred in 116 of 220 patients (52.7%) in the pembrolizumab–chemotherapy group and in 42 of 103 (40.8%) in the placebo–chemotherapy group.

CPS ≥1 subgroup: ORR was 44.9% (191/425) vs 38.9% (82/211).

ITT population: ORR was 40.8% (231/566) vs 37.0% (104/281).

Duration of Response (Updated)

CPS ≥10 subgroup: Median duration of response was 12.8 months in the pembrolizumab–chemotherapy group vs 7.3 months in the placebo–chemotherapy group.

Exploratory Endpoints: Overall Survival by Additional CPS Cutoffs

The trial reported OS by additional PD-L1 CPS thresholds to characterize the relationship between PD-L1 expression and benefit from pembrolizumab. Note that Figure 2 analyses used the unstratified Cox model, which accounts for the slight difference in HR for the CPS ≥10 subgroup compared with the stratified primary analysis.

These exploratory analyses are striking in their consistency: the survival benefit of adding pembrolizumab to chemotherapy is concentrated in tumors with CPS ≥10. At CPS 1–9, the HR was 1.09 (95% CI, 0.85 to 1.40), suggesting no benefit — and potentially a trend toward harm — in this intermediate expression group. Among patients with CPS <1, the HR was 0.97, indicating no benefit.

Safety

Safety Population

Pembrolizumab–chemotherapy: n=562; Placebo–chemotherapy: n=281.

Safety Summary

Treatment-related adverse events of grade 3, 4, or 5 occurred in 68.1% of the pembrolizumab–chemotherapy group and 66.9% of the placebo–chemotherapy group. Discontinuation due to AE, dose reduction, and dose interruption rates were not reported in this publication. Serious AE rates were not reported in this publication.

Grade ≥3 Adverse Events of Clinical Significance

The source reports grade 3, 4, or 5 combined (not separately).

Adverse Events of Special Interest

⚠️ Immune-Mediated Adverse Events: Critical Safety Signal

Immune-mediated adverse events occurred at substantially higher rates with pembrolizumab–chemotherapy compared with placebo–chemotherapy.

• Any immune-mediated AE: 149 (26.5%) vs 18 (6.4%)
• Grade ≥3 immune-mediated AE: 30 (5.3%) vs 0
• Fatal immune-mediated AE: No patients died due to immune-mediated adverse events

The grade ≥3 incidence of 5.3% (vs 0% in the control arm) meets the threshold for a critical safety signal. Individual immune-mediated AEs:

Clinical implication: Thyroid function should be monitored at baseline and periodically during treatment. Pneumonitis, though uncommon (2.5%), was exclusive to the pembrolizumab arm and was grade ≥3 in 6 patients (1.1%). Patients should be counseled about respiratory symptoms. Severe skin reactions were all grade ≥3 in the pembrolizumab arm (10 patients, 1.8%).

Deaths

• Total deaths (ITT): 460 of 566 (81.3%) in the pembrolizumab–chemotherapy group and 238 of 281 (84.7%) in the placebo–chemotherapy group
• Treatment-related deaths: 2 (0.4%) in the pembrolizumab–chemotherapy group (1 from acute kidney injury and 1 from pneumonia); 0 in the placebo–chemotherapy group
• Deaths due to disease progression: Not reported separately in this publication
• Deaths due to immune-mediated AEs: 0

Subgroup Analyses

The publication reported overall survival by PD-L1 CPS cutoffs using unstratified Cox models (Figure 2). These analyses are the central biomarker-driven subgroup data for this trial.

Note: The CPS ≥10 HR of 0.71 (unstratified) differs from the primary analysis HR of 0.73 (stratified). This discrepancy is explicitly explained in the figure caption: the overall population used stratified Cox regression; the subgroup analyses used the unstratified Cox model.

Prespecified subgroup analyses by additional clinical and demographic variables were referenced in the paper (Supplementary Figure S4) but were not available for extraction.

These subgroup analyses were not powered for formal statistical comparisons. Results should be interpreted as hypothesis-generating. However, the consistent pattern across CPS cutoffs — benefit concentrated at CPS ≥10, no benefit at CPS <10 — is the central clinical finding of this analysis.

Key Comparator Trials

Cross-trial comparisons are limited by differences in patient populations, trial designs, and endpoints. These data are presented for context, not for direct statistical comparison.

Contextual Notes

KEYNOTE-355 and IMpassion130 are the two pivotal first-line checkpoint inhibitor trials in metastatic TNBC, but they differ in important ways. IMpassion130 used atezolizumab (anti-PD-L1) with nab-paclitaxel only, while KEYNOTE-355 allowed three chemotherapy backbones. The biomarker platforms differ: IMpassion130 used the SP142 assay measuring PD-L1 on immune cells (IC+), while KEYNOTE-355 used the 22C3 assay measuring combined positive score (CPS). These assays identify overlapping but non-identical patient populations, making direct cross-trial biomarker comparisons unreliable.

Notably, atezolizumab's indication in metastatic TNBC was voluntarily withdrawn from the US market in 2021 after the confirmatory IMpassion131 trial failed to show benefit for atezolizumab combined with paclitaxel [5]. This left pembrolizumab as the only checkpoint inhibitor with an active approval in this setting.

The OS result for atezolizumab in the PD-L1 IC+ population of IMpassion130 (HR 0.67) was descriptive only, as the hierarchical testing plan required the ITT PFS result to be significant first — which it was — but the ITT OS was not significant, precluding formal testing of OS in the PD-L1+ subgroup [4]. The magnitude of the point estimate in the biomarker-positive populations appears similar between the two trials, though the different assays and statistical frameworks prevent meaningful comparison.

Grey Zones and Unanswered Questions

• CPS 1–9 population: The HR of 1.09 (95% CI, 0.85 to 1.40) in the CPS 1–9 subgroup raises concern that patients in this intermediate PD-L1 range may not benefit — and could potentially be harmed — by the addition of pembrolizumab. This subgroup was not powered for formal comparison, but the point estimate and its clinical implications warrant caution when PD-L1 CPS falls between 1 and 9.

• Optimal chemotherapy backbone: Three chemotherapy partners were permitted (nab-paclitaxel, paclitaxel, gemcitabine–carboplatin), and chemotherapy type was a stratification factor. However, this publication does not report OS outcomes by chemotherapy backbone. Whether the benefit of pembrolizumab is consistent across all three chemotherapy partners — particularly given the failure of atezolizumab with paclitaxel in IMpassion131 — remains an open question.

• Prior neoadjuvant immunotherapy: KEYNOTE-355 enrolled patients in the first-line metastatic setting, but the trial enrolled before neoadjuvant pembrolizumab became standard in early-stage TNBC (via KEYNOTE-522). As neoadjuvant pembrolizumab use increases, a growing proportion of patients presenting with metastatic TNBC will have received prior pembrolizumab. KEYNOTE-355 excluded patients with prior anti-PD-1/PD-L1 therapy, so the trial cannot inform retreatment decisions.

• Subsequent therapy data: The publication does not report subsequent therapy rates or types. In the era of sacituzumab govitecan and other later-line options, understanding the post-progression treatment landscape and its potential confounding effect on OS is important.

• Brain metastases: Patients with active CNS metastases were excluded. The efficacy of pembrolizumab plus chemotherapy in patients with stable, treated brain metastases — a common clinical scenario in metastatic TNBC — is not defined by this trial.

Clinical Implications

Where This Fits in the Treatment Sequence

Pembrolizumab plus chemotherapy (nab-paclitaxel, paclitaxel, or gemcitabine–carboplatin) is the standard first-line treatment for patients with locally recurrent inoperable or metastatic TNBC whose tumors express PD-L1 with a CPS ≥10, as assessed by the 22C3 pharmDx assay. NCCN guidelines recommend this combination as a Category 1, preferred first-line option for this population [3].

For patients whose tumors have CPS <10, chemotherapy alone remains the standard first-line approach, and the KEYNOTE-355 data do not support the addition of pembrolizumab.

Practical Considerations

• PD-L1 testing: CPS must be assessed using the PD-L1 IHC 22C3 pharmDx assay. Results from other PD-L1 assays (SP142, SP263, 28-8) are not interchangeable and should not be used for treatment selection in this setting.
• Chemotherapy backbone selection: The choice among nab-paclitaxel, paclitaxel, and gemcitabine–carboplatin should be individualized based on prior therapy history, toxicity profile, patient preference, and logistical considerations (e.g., infusion schedules, steroid premedication requirements for taxanes).
• Pembrolizumab duration: Treatment continued up to 35 administrations (approximately 2 years) or until progression. Median treatment duration was 26.4 weeks in the experimental arm.
• Immune-mediated AE monitoring: Thyroid function tests at baseline and regularly during treatment. Monitor for pneumonitis symptoms (cough, dyspnea). Liver function monitoring given the 6.0% rate of grade ≥3 ALT elevation.

Unanswered Questions

The two most practice-relevant open questions are: (1) whether the benefit of pembrolizumab is consistent across all three permitted chemotherapy backbones, and (2) how to manage patients who progress after neoadjuvant pembrolizumab (KEYNOTE-522) and subsequently develop metastatic disease.

Post-Progression Options

Subsequent therapy data were not reported in this publication. In current clinical practice, patients who progress after first-line pembrolizumab plus chemotherapy for metastatic TNBC are typically considered for sacituzumab govitecan (based on ASCENT), single-agent chemotherapy, or clinical trials. For patients with germline BRCA1/2 mutations, PARP inhibitors (olaparib, talazoparib) remain an option based on OlympiAD and EMBRACA. This external context is provided for reference and is not derived from the JSON dataset.

Regulatory and Guideline Status

Regulatory

• FDA: Pembrolizumab was approved in combination with chemotherapy for the treatment of patients with locally recurrent unresectable or metastatic TNBC whose tumors express PD-L1 (CPS ≥10) as determined by an FDA-approved test. The KEYNOTE-355 OS results supported this approval. The companion diagnostic is the PD-L1 IHC 22C3 pharmDx assay [2].
• EMA: Pembrolizumab in combination with chemotherapy is approved for the first-line treatment of locally recurrent unresectable or metastatic TNBC in adults whose tumors express PD-L1 with CPS ≥10.

⚠️ Regulatory status verified as of March 2026. Confirm current approval status and labeled indications before clinical use at FDA.gov.

Guidelines

• NCCN: Pembrolizumab plus chemotherapy (nab-paclitaxel, paclitaxel, or gemcitabine–carboplatin) is listed as a Category 1, preferred first-line option for locally recurrent unresectable or metastatic TNBC with PD-L1 CPS ≥10 [3].

Companion Diagnostics

The PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies) is the FDA-approved companion diagnostic for identifying patients eligible for pembrolizumab in this indication. The CPS ≥10 threshold is required for treatment selection.

About the Author

Andrew Stevenson is the founder and systems architect of kill-cancer.com, a clinical intelligence platform delivering structured, source-traced oncology trial analysis to practicing clinicians. He holds 17 Google technical certifications in data systems, automation, and applied AI — the engineering foundation behind the extraction and verification pipeline that produces every article on this platform. Every number traces directly to its source publication. Zero calculation. Zero editorializing. Zero hallucination.

Five siblings lost to cancer built the urgency. The engineering builds the trust.

📧 andrew@kill-cancer.com 🌐 kill-cancer.com 💬 kill-cancer.com/forum

This article is intended for healthcare professionals only. It is not medical advice and should not be used as a substitute for professional clinical judgment. Treatment decisions should be made in consultation with qualified healthcare providers based on individual patient circumstances.

All trial data presented in this article are sourced directly from the published clinical trial report and its supplementary materials. Numbers are reproduced exactly as reported; no calculations, derivations, or estimates have been performed.

Trial results are presented as reported in the source publication. Updated data, label changes, or guideline revisions published after the source article may alter clinical applicability.

Comparator trial data presented in Section 8 are sourced from their respective published reports and are provided for contextual purposes only. Cross-trial comparisons have inherent limitations and should not be interpreted as direct statistical comparisons.

References

1. Cortes J, Rugo HS, Cescon DW, et al. Pembrolizumab plus chemotherapy in advanced triple-negative breast cancer. N Engl J Med. 2022;387:217-226. doi:10.1056/NEJMoa2202809
2. Pembrolizumab (KEYTRUDA) prescribing information. U.S. Food and Drug Administration. Accessed March 2026.
3. NCCN Clinical Practice Guidelines in Oncology: Breast Cancer. Version 2.2026. Accessed March 2026.
4. Emens LA, Adams S, Barrios CH, et al. First-line atezolizumab plus nab-paclitaxel for unresectable, locally advanced, or metastatic triple-negative breast cancer: IMpassion130 final overall survival analysis. Ann Oncol. 2021;32(8):983-993. doi:10.1016/j.annonc.2021.05.355
5. Miles D, Gligorov J, André F, et al. Primary results from IMpassion131, a double-blind, placebo-controlled, randomised phase III trial of first-line paclitaxel with or without atezolizumab for unresectable locally advanced/metastatic triple-negative breast cancer. Ann Oncol. 2021;32(8):994-1004. doi:10.1016/j.annonc.2021.05.801