Harry H. Yoon, MD

KEYNOTE-181’s invited discussant, Harry H. Yoon, MD, Associate Professor of Oncology at the Mayo Clinic, Rochester, Minnesota, said the findings have a “potentially immediate clinical impact” for second-line treatment of esophageal cancer, including esophageal squamous cell carcinoma and Siewert type I gastroesophageal junction adenocarcinoma. The current standard of care for these tumors above the gastroesophageal junction is taxane- or irinotecan-based regimens and ramucirumab plus paclitaxel for tumors at the gastroesophageal junction or stomach. Pembrolizumab is currently approved in the third-line setting for patients with programmed cell death ligand 1 (PD-L1)–positive, combined positive score (CPS) ≥ 1 gastric/gastroesophageal junction adenocarcinoma. The new findings extend its benefit to the second-line setting and to those with squamous cell histology in patients with PD-L1 CPS ≥ 10, he said.

Since KEYNOTE-181 failed to meet its co-primary endpoint of overall survival in the intent-to-treat population, pembrolizumab may not be indicated in unselected patients with esophageal cancer. For the co-primary analysis of patients with squamous cell carcinoma, he noted that the “Kaplan-Meier curves for overall survival do separate,” but the difference missed the prespecified P value by .0018 (P = .0095). This was not likely due to underpowering, he suggested, as this appears to be the largest second-line study in patients with metastatic squamous cell carcinoma.

Co-primary Analysis

“KEYNOTE-181 had three co-primary endpoints, and because of this, the alpha had to be shared among them to reduce the possibility of a type 1 error,” he explained. “One can imagine if there had been only two primary endpoints, this result [in patients with squamous cell carcinoma] might have been positive.”

For the co-primary analysis in the PD-L1 CPS ≥ 10 subgroup, however, “the results are clinically meaningful and statistically significant,” he said, noting that this degree of PD-L1 expression was seen in 35% of the whole population and 42% of the squamous cell carcinoma subgroup.

“The most compelling part is the tail on the curve, showing continued separation,” he noted. “In the pembrolizumab arm, 43% of patients were alive at 12 months. That’s a number needed to treat of just 4 patients to have 1 alive at 12 months. The response rate and progression-free survival also favored pembrolizumab, and it was better tolerated than chemotherapy.”

‘Reasonable Internal Validity’

Dr. Yoon added: “Given that randomization of treatment was not stratified by PD-L1 status, some would ask how is the PD-L1 subgroup analysis different from a secondary analysis, the results of which would only be hypothesis-generating? I think there’s room for disagreement, but my perspective is that there is reasonable internal validity for two main reasons: first, there was prespecified statistical compensation to reduce a false-positive result (such as could occur if baseline factors favored the pembrolizumab arm). Second and more important, there is coherent biologic rationale based on earlier evidence that a higher CPS predicts for better response to pembrolizumab.”

Dr. Yoon described previous pembrolizumab studies in support of this statement and concluded that KEYNOTE-181 validates the hypothesis generated in KEYNOTE-180 and KEYNOTE-61 “that enriching the drug target in the tumor microenvironment, up to CPS ≥ 10, identifies patients in whom pembrolizumab outperforms chemotherapy.”

Clinical Implications

A caveat is that multivariable analysis could help clarify whether the positive results in the CPS ≥ 10 subgroup could be explained by a higher frequency of favorable characteristics in the pembrolizumab arm, and the strength of those results could influence guidelines and implementation in clinical practice.

As for clinical practice, Dr. Yoon said he currently orders PD-L1 and HER2 testing upon diagnosis of metastatic disease in patients with gastroesophageal adenocarcinomas and mismatch repair testing in most solid tumors. With these new findings, it is now “reasonable to consider” ordering PD-L1 testing as well in patients with squamous cell histology of the esophagus, he said.

For treatment in the second-line setting, it is now reasonable to consider pembrolizumab as an option for patients with advanced or metastatic adenocarcinoma or squamous cell carcinoma of the esophagus or Siewert type I adenocarcinoma of the esophagogastric junction who have a PD-L1 CPS ≥ 10, he said. He encouraged guideline committees to consider changes as well, based on the findings. ■

DISCLOSURE: Dr. Yoon has received honoraria and is a consultant/advisor for Astellas, LSK Biopharma, Merck Sharp & Dohme, and BeiGene; has received institutional research funding from Boston Biomedical, Lilly/ImClone, Merck, and Roche/Genentech; and has received reimbursement for travel expenses from Lilly.