The Lancet August 28, 2021

Pembrolizumab plus chemotherapy versus chemotherapy alone for first-line treatment of advanced oesophageal cancer (KEYNOTE-590): a randomised, placebo-controlled, phase 3 study

Jong-Mu Sun, Lei Shen, Manish A Shah, et al.

Bottom Line

The addition of first-line pembrolizumab to standard chemotherapy significantly improved both overall survival and progression-free survival in patients with advanced esophageal and gastroesophageal junction cancers compared to chemotherapy alone, particularly in those with PD-L1 CPS ≥ 10.

Key Findings

1. At a median follow-up of 22.6 months, median overall survival (OS) in patients with esophageal squamous cell carcinoma (ESCC) and a PD-L1 combined positive score (CPS) ≥ 10 was 13.9 months in the pembrolizumab group vs 8.8 months in the placebo group (Hazard Ratio [HR] 0.57; 95% CI, 0.43-0.75; p<0.0001).
2. In the overall intent-to-treat (ITT) population, median OS was 12.4 months with pembrolizumab plus chemotherapy versus 9.8 months with placebo plus chemotherapy (HR 0.73; 95% CI, 0.62-0.86; p<0.0001).
3. Median progression-free survival (PFS) in the ITT population was 6.3 months in the pembrolizumab arm compared to 5.8 months in the placebo arm (HR 0.65; 95% CI, 0.55-0.76; p<0.0001).
4. In the subgroup of all patients with PD-L1 CPS ≥ 10, median OS was 13.5 months versus 9.4 months in favor of the pembrolizumab combination (HR 0.62; 95% CI, 0.49-0.78; p<0.0001).
5. The objective response rate (ORR) in all randomized patients was 45.0% for the pembrolizumab group versus 29.3% for the placebo group.
6. Grade 3 to 5 treatment-related adverse events occurred in 72% of patients receiving pembrolizumab plus chemotherapy and in 68% of patients receiving placebo plus chemotherapy.

Study Design

Design
Phase 3 RCT
Double-Blind
Sample
749
Patients
Duration
22.6 mo
Median
Setting
Multicenter, 26 countries
Population Adult patients with locally advanced unresectable or metastatic adenocarcinoma or squamous cell carcinoma of the esophagus, or Siewert type 1 gastroesophageal junction adenocarcinoma, who had an ECOG performance status of 0 or 1 and no prior systemic therapy for advanced disease.
Intervention Pembrolizumab 200 mg IV every 3 weeks (up to 35 cycles) plus 5-fluorouracil (800 mg/m^2/day on days 1-5) and cisplatin (80 mg/m^2 on day 1) every 3 weeks (up to 6 cycles).
Comparator Placebo IV every 3 weeks plus 5-fluorouracil and cisplatin at the same doses and scheduling as the intervention arm.
Outcome Overall survival (OS) in patients with ESCC and PD-L1 CPS ≥ 10; OS and progression-free survival (PFS) in patients with ESCC, patients with PD-L1 CPS ≥ 10, and all randomized patients (ITT).

Study Limitations

The chemotherapy backbone was restricted to cisplatin and 5-fluorouracil, which is standard globally but does not account for the increasing preference for oxaliplatin-based regimens (like FOLFOX) in Western practice due to more favorable toxicity profiles.
The majority of enrolled patients (73%) had squamous cell carcinoma and over half the population was Asian, potentially limiting the statistical power and generalizability of the findings for patients with esophageal adenocarcinoma, which is the predominant histology in Western nations.
Using a PD-L1 CPS threshold of 10 adds a layer of biomarker testing complexity, and the absolute magnitude of benefit in patients with CPS < 10 or those with adenocarcinoma was less pronounced than in the ESCC CPS ≥ 10 subgroup.

Clinical Significance

KEYNOTE-590 initiated a paradigm shift in gastrointestinal oncology by establishing pembrolizumab combined with platinum/fluoropyrimidine chemotherapy as a new standard-of-care first-line treatment for advanced esophageal and Siewert type 1 gastroesophageal junction cancers. It directly led to global regulatory approvals (including the US FDA in March 2021) for this combination, transcending historical limitations of chemotherapy-alone survival outcomes.

Historical Context

Historically, patients with locally advanced unresectable or metastatic esophageal cancer faced a grim prognosis, with standard first-line chemotherapy yielding median overall survival times of only 9 to 10 months. Prior to KEYNOTE-590, immunotherapy had shown meaningful activity in the second-line setting (KEYNOTE-181) for ESCC with PD-L1 expression. KEYNOTE-590 successfully moved PD-1 blockade into the first-line setting, cementing concurrent chemoimmunotherapy as the benchmark strategy for treating this aggressive malignancy.

Guided Discussion

High-yield insights from every perspective

Med Student
Medical Student

What is the mechanism of action of pembrolizumab, and why is the Combined Positive Score (CPS) rather than the Tumor Proportion Score (TPS) used to assess PD-L1 expression in gastrointestinal and esophageal cancers like those in KEYNOTE-590?

Key Response

Pembrolizumab is an IgG4 monoclonal antibody that blocks the PD-1 receptor on T-cells, preventing the inhibitory signals from PD-L1/PD-L2. In gastrointestinal malignancies, immune infiltrates (macrophages, lymphocytes) within the tumor microenvironment heavily express PD-L1 and play a massive role in immune evasion. CPS accounts for PD-L1 expression on both tumor cells and these tumor-infiltrating immune cells, making it a much more accurate predictive biomarker for immune checkpoint inhibitor efficacy than TPS, which only scores tumor cells.

Resident
Resident

A 65-year-old male presents with newly diagnosed metastatic esophageal squamous cell carcinoma with a PD-L1 CPS of 5. Based on the findings of KEYNOTE-590, how would you counsel him regarding the addition of pembrolizumab to his cisplatin/5-FU backbone?

Key Response

While the FDA approved pembrolizumab plus chemotherapy for all patients with advanced esophageal/GEJ cancer based on KEYNOTE-590, the trial showed that the magnitude of overall survival and progression-free survival benefit was heavily driven by patients with a PD-L1 CPS >= 10. Residents must understand this nuance: while technically indicated for a CPS of 5, the patient should be counseled that their expected benefit is smaller than the heavily publicized headline results, requiring a careful shared decision-making discussion regarding the added risks of immune-related adverse events.

Fellow
Fellow

KEYNOTE-590 utilized a cisplatin and 5-FU backbone, but contemporary clinical practice often favors FOLFOX due to a more favorable toxicity profile and outpatient administration. How does the choice of chemotherapy backbone (oxaliplatin vs. cisplatin) potentially modulate the tumor microenvironment and the efficacy of concurrent PD-1 blockade in advanced esophageal cancer?

Key Response

Fellows must critically extrapolate trial data to real-world practice. Oxaliplatin is known to induce immunogenic cell death (ICD) more potently than cisplatin by triggering the exposure of damage-associated molecular patterns (DAMPs) like calreticulin, which enhances dendritic cell activation and subsequent T-cell priming. Therefore, substituting FOLFOX for cisplatin/5-FU is not only clinically convenient but may theoretically synergize better with pembrolizumab, a rationale that justifies off-protocol backbone switching in modern practice.

Attending
Attending

Given that the KEYNOTE-590 trial demonstrated a statistically significant overall survival benefit in the all-comer population, but subset analyses suggest the effect in the CPS < 10 subgroup is marginal or absent, how do you justify the financial toxicity and risk of life-altering irAEs when treating a CPS 0 patient in your clinical practice?

Key Response

Attendings must balance statistically positive trial designs against clinically meaningful benefit. The hierarchical testing of KEYNOTE-590 allowed the 'all-comer' endpoint to be met because the massive benefit in the CPS >= 10 subgroup diluted the lack of benefit in the CPS < 10 group. An insightful attending uses this to teach that regulatory approval does not mandate clinical use; exposing a CPS 0 patient to the financial burden and toxicity of pembrolizumab without clear biologic rationale represents poor value-based care and highlights the flaws of all-comer trial designs.

Scholarly Review

Critical appraisal through the lens of expert reviewers and guideline development

PhD
PhD

KEYNOTE-590 employed a complex hierarchical statistical testing strategy evaluating OS and PFS across nested populations (ESCC CPS >=10, ESCC, all CPS >=10, all patients). What are the methodological vulnerabilities of this step-down approach when the treatment effect is highly heterogeneous between distinctly different biologies (squamous cell carcinoma vs. adenocarcinoma)?

Key Response

Hierarchical testing strictly preserves the overall Type I error rate (alpha) but is vulnerable to masking qualitative interactions. If the treatment effect is massive in one specific subpopulation (e.g., ESCC with CPS >= 10) and the sample size is large, it can drag the pooled 'all patient' hazard ratio across the significance threshold. PhD-level critique focuses on how pooling distinct histological biologies (squamous vs. adeno) under a single statistical umbrella limits generalizability and can lead to overstating the efficacy for less responsive subgroups hidden within the broader 'all-comer' statistical victory.

Journal Editor
Journal Editor

As a peer reviewer evaluating the KEYNOTE-590 manuscript, how would you scrutinize the lack of protocol-mandated crossover to immunotherapy upon progression in the control arm, and how does this impact the validity of the overall survival endpoints compared to modern real-world treatment sequencing?

Key Response

A critical editor must assess whether the experimental arm looks superior merely because the control arm was denied standard-of-care subsequent therapies. In modern practice, a patient progressing on first-line chemotherapy would likely receive a checkpoint inhibitor in the second line (e.g., nivolumab or pembrolizumab based on ATTRACTION-3 or KEYNOTE-181). If the trial's control arm had a low rate of subsequent immunotherapy use, the reported overall survival benefit of early pembrolizumab may be artificially inflated compared to the real-world strategy of sequential therapy.

Guideline Committee
Guideline Committee

NCCN guidelines incorporated KEYNOTE-590 by assigning a Category 1 recommendation for pembrolizumab plus chemotherapy in first-line esophageal cancer across all PD-L1 expression levels, whereas ESMO guidelines restrict this recommendation to patients with a CPS >= 10. How should an international guideline committee reconcile these divergent recommendations based on the same trial data?

Key Response

Guideline committees must weigh permissive regulatory labeling against evidence-based value scales. NCCN often aligns closely with broad FDA labels, aiming to maximize access. Conversely, ESMO utilizes the Magnitude of Clinical Benefit Scale (MCBS), reserving strong recommendations for subgroups demonstrating unambiguous, clinically meaningful survival gains (CPS >= 10). A unified committee would need to transparently define whether their guidelines prioritize broad treatment accessibility or strict biomarker-driven stewardship to prevent overtreatment in the CPS < 10 population.

Clinical Landscape

Noteworthy Related Trials

2019

ATTRACTION-3 Trial

n = 419 · Lancet Oncol

Tested

Nivolumab 240mg every 2 weeks

Population

Advanced esophageal squamous cell carcinoma refractory to previous chemotherapy

Comparator

Chemotherapy (paclitaxel or docetaxel)

Endpoint

Overall Survival

Key result: Nivolumab demonstrated a significant improvement in overall survival over chemotherapy with a favorable safety profile in the second-line setting.
2020

KEYNOTE-181 Trial

n = 628 · J Clin Oncol

Tested

Pembrolizumab 200mg every 3 weeks

Population

Advanced esophageal cancer progressing after first-line therapy

Comparator

Investigator's choice of chemotherapy

Endpoint

Overall Survival

Key result: Pembrolizumab significantly improved overall survival compared to chemotherapy in patients with a PD-L1 CPS >= 10.
2022

CheckMate 648 Trial

n = 970 · NEJM

Tested

Nivolumab + chemotherapy or Nivolumab + Ipilimumab

Population

Previously untreated advanced esophageal squamous-cell carcinoma

Comparator

Chemotherapy alone (fluorouracil + cisplatin)

Endpoint

Overall Survival and Progression-Free Survival

Key result: Both nivolumab-based regimens significantly prolonged overall survival compared to chemotherapy alone in patients with advanced ESCC.

Tailored to your role

Want this tailored to you?

Add your specialty or training stage to get role-specific takeaways and more questions.

Personalize this analysis