Elizabeth Smyth, MD, Offers Insight on KEYNOTE-859 Findings

Get Permission

The invited discussant of KEYNOTE-859,1 Elizabeth Smyth, MD, consultant in gastrointestinal oncology at Cambridge University Hospitals NHS Foundation Trust in the United Kingdom, called the findings “practice-changing” but cautioned that better patient selection is needed to optimally apply them.

Dr. Smyth put the KEYNOTE-859 results into context with the “large pantheon” of global phase III trials showing “very consistently” that treatment with immune checkpoint inhibitors plus chemotherapy is most effective at improving overall survival in patients with higher levels of PD-L1 expression. KEYNOTE-859 now provides evidence for pembrolizumab in advanced gastric cancer, where there was no previous indication, she said.

KEYNOTE-859 showed that pembrolizumab plus fluorouracil and platinum chemotherapy may provide “variably” meaningful improvements in overall survival that are often dependent on PD-L1 expression, and these findings “are likely to support” pembrolizumab as a new treatment option for locally advanced or metastatic HER2-negative gastric or gastroesophageal junction adenocarcinoma. These treatment decisions will vary according to regional regulatory approval, funder appraisals, and clinician/patient discussion of benefit according to PD-L1 status, Dr. Smyth added.

Elizabeth Smyth, MD

Elizabeth Smyth, MD

Results Consistent, Interpretation Less So

Although a fairly equal number of randomized trials of checkpoint inhibitors in gastric cancer have been positive or negative, they have altogether yielded these “lessons”: The majority of gastric cancers are not sensitive to anti–PD-1 monotherapy, anti–PD-1 monotherapy is not superior to chemotherapy for nonselected or even weakly PD-L1–positive populations, and the use of chemotherapy plus anti–PD-1 improves response rates and overall survival, with the most benefit in patients with PD-L1–high tumors.

“Results of the trials have been consistent, but their interpretation by regulatory authorities has not,” Dr. Smyth commented. In the United Kingdom, the recommendation is for PD-L1 status to guide the use of checkpoint inhibitors; outside of Europe, regulatory authorities have permitted their use in all patients, though they have recognized the evidence is strongest for their use in PD-L1–high tumors.

KEYNOTE-859 adds considerably to the evidence, as this was a large, well-powered clinical trial with a global reach in patients with gastric or gastroesophageal junction cancers treated with standard-of-care oxaliplatin. With an overall survival benefit seen (hazard ratio [HR] = 0.78), she said that “one might question whether a difference of less than 6 weeks is clinically meaningful.” However, Dr. Smyth unquestionably acknowledged that “there are benefits at the tail of the curve for patients treated with pembrolizumab: The 2-year survival is improved by 9.3%, which is a 50% relative gain. Of note, this comes in at just below the 10% improvement in 2-year survival, which is needed for an ESMO Magnitude of Clinical Benefit Scale score of 4.”

Identifying Patients Who Might Benefit

Although these changes are “real and substantial,” only one-third of patients were alive at 2 years, Dr. Smyth noted. “How can we correctly identify the patients who are deriving this benefit?” The greatest difference vs chemotherapy alone was seen in patients selected according to biomarkers of immune sensitivity—PD-L1 and microsatellite instability–high (MSI-H)—so, she asked, how confidently can PD-L1 status guide treatment choice?

“We could choose to focus on just patients who have PD-L1 CPS ≥ 10 tumors (HR = 0.64), who definitely should be receiving immune checkpoint inhibitors. If we include the remaining two-thirds of patients with a CPS less than 10, there is much less benefit (HR = 0.86), although a few patients with scores close to the cutoff would miss out on effective treatment. Or we could focus on the one in five patients who have PD-L1–negative tumors, who clearly do not benefit (HR = 0.92) and should not receive pembrolizumab. If we treat patients with any PD-L1 score (HR = 0.73), we are treating many patients who won’t benefit, because this group contains patients with PD-L1–low and –high expression,” she pointed out.

The message from the PD-L1 subsets, Dr. Smyth said, is that to make informed treatment decisions, one needs to know whether the patient has no chance of benefit (ie, PD-L1–negative), has a high chance of benefit (ie, a PD-L1 CPS ≥ 10), or has an intermediate chance of benefit (ie, PD-L1–positive but with a CPS < 10). However, for this last subset, information is not available from the current data set. Some hints about the benefit of checkpoint inhibitors in intermediate-expressing tumors, however, might be extrapolated from CheckMate 649. Using special software, the researchers found that the addition of nivolumab added no significant benefit in progression-free or overall survival among patients with PD-L1 CPS 1–4 tumors.2

Limitations of PD-L1 as a Biomarker

There are significant drawbacks to using PD-L1 as a biomarker in gastric and gastroesophageal cancers, mainly factors related to spatial and temporal heterogeneity, Dr. Smyth continued. In these “chaotic” tumors, PD-L1 expression can vary within the tumor itself; concordance is low between primary and metastatic tumors and across time periods; there are doubts about the interchangeability of the two most common PD-L1 assays; and there is concern about intra- and interobserver reproducibility of test results.

Dr. Smyth emphasized that a more accurate means of patient selection is needed for the use of checkpoint inhibitors in gastric cancer. Meanwhile, reliance in the clinic is mostly on the “arbitrary” and “potentially flawed” categories of PD-L1 expression. More personalization is needed to avoid toxicities that many patients needlessly experience and to control rising health-care costs as indications for checkpoint inhibitors grow. Cross-validation studies are important for giving oncologists, patients, regulatory authorities, and funders “confidence in this biomarker,” she said.

DISCLOSURE: Dr. Smyth reported financial relationships with Amal Therapeutics, Aptitude Health, Amgen, Astellas, AstraZeneca, BeiGene, BMS, Celgene, Daiichi Sankyo, Elsevier, Everest Clinical Research, First Word Group, Five Prime Therapeutics, Gritstone Oncology, Imedex, Merck, My Personal Therapeutics, Novartis, Pfizer4, Roche, Sai-Med, Servier, Touch Oncology, Turning Point Therapeutics, and Zymeworks.


1. Rha SY, Wyrwicz LS, Yanez Weber PE, et al: Pembrolizumab plus chemotherapy as first-line therapy for advanced HER2-negative gastric or gastroesophageal junction cancer: Phase III KEYNOTE-859 study. ESMO Virtual Plenary. Abstract VP1-2023. Presented February 16, 2023. 

2. Zhao JJ, Yap DWT, Chan YH, et al: Low programmed death-ligand 1-expressing subgroup outcomes of first-line immune checkpoint inhibitors in gastric or esophageal adenocarcinoma. J Clin Oncol 40:392-402, 2022.


Related Articles

KEYNOTE-859 Interim Analysis: Overall Survival Benefit for First-Line Pembrolizumab in Advanced Gastric Cancer

Results of the interim analysis of KEYNOTE-859 are in, and they confirm the overall survival benefit of first-line immunotherapy plus chemotherapy in advanced gastric cancer.1 Pembrolizumab plus a fluoropyrimidine- and platinum-containing doublet provided a statistically significant improvement in...