NEW YORK – A majority of experts in the US Food and Drug Administration's Oncologic Drugs Advisory Committee (ODAC) voted on Thursday that the available data did not support the first-line use of three immunotherapies when combined with chemotherapy in gastric and esophageal cancer patients whose tumors have low or no PD-L1 expression.
The panel considered data for two checkpoint inhibitors already approved in these settings for all-comers, Merck's Keytruda (pembrolizumab) and Bristol Myers Squibb's Opdivo (nivolumab), along with data submitted within BeiGene's biologics license application (BLA) seeking approval for Tevimbra (tislelizumab), a potential new option for these cancers.
The agency previously approved Keytruda with chemo and Opdivo with chemo as a first-line treatment for patients with HER2-negative unresectable or metastatic gastric or gastroesophageal adenocarcinoma regardless of PD-L1 expression levels. Keytruda plus chemo and Opdivo plus BMS's other checkpoint inhibitor Yervoy (ipilimumab) with chemo also have PD-L1 expression-agnostic approvals as a first-line treatment for unresectable or metastatic esophageal squamous cell carcinoma. The FDA is expected to decide in December whether to approve Tevimbra as a new gastric or esophageal cancer treatment.
FDA experts and the three pharma sponsors presented data at the meeting on Thursday from several Phase III trials of these checkpoint inhibitors that stratified patients by PD-L1 expression levels. Trials of Opdivo and Keytruda both used PD-L1 combined positive score (CPS) to measure expression, while the two studies of Tevimbra used a different PD-L1 expression scoring method called tumor area positivity (TAP) score. The FDA asked ODAC experts to consider the available data and weigh the efficacy and risks of these drugs in patients with a PD-L1 CPS or TAP score less than 1.
For the gastric cancer indication, 10 experts on the panel voted that the data was not favorable for PD-L1-negative group, two panelists voted yes that the data was favorable, and one panelist abstained. For esophageal cancer indication, 11 panelists voted that the data was not favorable for PD-L1-negative group, one panelist voted yes, and one panelist abstained.
While the ODAC panel voted that the risk-benefit assessment of these drugs was not favorable in the PD-L1-negative gastric and esophageal cancer subgroups, it was not a vote to narrow the indications of these drugs in PD-L1 positive patients.
"There is fairly widespread consensus across the panel that the risk-benefit profile does not favor [checkpoint inhibitor] use in patients that have PD-L1 scores less than 1," Christopher Lieu, acting chairperson of the ODAC, said at the meeting. "Having said that, I think that we've heard from the panel members that there is some discomfort with the small amount of patients that we're trying to make decisions from and understanding that the data is never going to be perfect."
Although it's unclear based on the committee's vote whether the FDA will narrow the indications of these drugs to PD-L1-positive patients, it was clear from the discussion at the meeting that the agency is considering doing so while also thinking about how to maintain access for PD-L1 negative patients who want to take these drugs despite the risks to see if they benefit.
Mixed efficacy results
The panel's discussion on immunotherapies in gastric and esophageal cancer focused on the small overall survival benefit seen in patients with PD-L1 scores less than 1, compared to those with higher PD-L1 expression scores.
Experts from BMS and the FDA presented data from the Phase III CheckMate-649 trial, which led to the 2021 approval of Opdivo with chemo in first-line gastric and gastroesophageal junction (GEJ) cancer. In its presentation, BMS argued that the gastric/GEJ indication should remain as is due to the Opdivo regimen showing some benefit across PD-L1 subgroups, including biomarker-negative patients, in CheckMate-649.
In the PD-L1 CPS less than 1 group, the median overall survival was 13.08 months on Opdivo-chemo compared to 12.48 months on chemo alone, according to BMS's analysis. The PD-L1 CPS greater than or equal to 1 group had a slightly greater benefit in overall survival, 13.96 months in the Opdivo-chemo arm versus 11.33 months in the chemo alone arm. The PD-L1-high group, with a CPS greater than or equal to 10, had the greatest survival benefit over chemo, 15.01 months on Opdivo-chemo versus 10.87 months on chemo alone. According to BMS's own exploratory analysis of overall survival by PD-L1 expression, however, the hazard ratio favored Opdivo in each subgroup, but in the PD-L1 CPS less than 1 group, the hazard ratio was close to 1, suggesting it was not statistically significant.
Merck also argued that the overall survival benefits were consistent across PD-L1 subgroups. But in the PD-L1 CPS less than 1 subgroup in the Phase III KEYNOTE-859 trial, the benefit was similarly less pronounced, with a median overall survival on Keytruda of 12.7 months compared 12.2 months on chemo, with a hazard ratio near 1. The PD-L1-high group, again, demonstrated the best survival outcomes over chemo, with a median overall survival on Keytruda-chemo of 15.7 months compared to 11.8 months on chemo alone.
While the results were positive for the overall study population regardless of PD-L1 status, the FDA noted that clinical guidelines from the American Society of Clinical Oncology and the National Comprehensive Cancer Network deviated from the FDA's decision to approve Opdivo and Keytruda in the all-comer setting. ASCO and NCCN recommended oncologists prescribe Opdivo and chemo only for gastric or GEJ cancer patients whose tumors had a PD-L1 CPS greater than or equal to 5 and Keytruda with chemo for patients whose tumors had PD-L1 CPS greater than or equal to 10.
For Tevimbra, the Phase III RATIONALE-305 results also showed an increasing benefit with higher PD-L1 expression. The median overall survival in the group with PD-L1 TAP scores less than 1 receiving Tevimbra-chemo was 15.4 months, compared to 13.8 months for those on chemo alone. However, this subgroup was small in this trial, including only 69 patients, and the hazard ratio for overall survival was nearly 1.
The pharma sponsors noted that safety and tolerability was consistent across PD-L1 subgroups. But an FDA expert at the meeting noted that there is additional toxicity from adding a checkpoint inhibitor to chemo. Across all the trials, there was a 3 percent to 11 percent increase in grade 3 and 4 treatment-emergent adverse events among those also given a checkpoint inhibitor, Vaibhav Kumar, a clinical reviewer in the division of oncology 3 within the FDA's Center for Drug Evaluation and Research, said in his presentation at the ODAC meeting.
The panel also discussed results from three Phase III trials, in which esophageal cancer patients received Keytruda, Opdivo, or Tevimbra in combination with chemo as a first-line treatment. The trials similarly showed greater benefit for subgroups with higher PD-L1 expression. In subgroups with PD-L1 scores less than 1, each of the trials showed an overall survival hazard ration of 1 or above, suggesting that there was no benefit over chemo.
In the KEYNOTE-590 trial, evaluating Keytruda and chemo versus chemo alone in esophageal cancer, the median overall survival in those with PD-L1 scores less than 1 was the same in both arms. In this subgroup, esophageal cancer patients on Opdivo plus chemo had a numerically lower median overall survival compared to those on just chemo in the CheckMate-648 study, 10.2 months versus 12.5 months, respectively. BeiGene's Tevimbra plus chemo also performed worse than chemo alone in the PD-L1 TAP less than 1 subgroup in the RATIONALE-306 trial, with a median survival of 11.8 months versus 16.1 months, respectively.
Lieu noted that the CheckMate-649, KEYNOTE-859, and RATIONALE-305 results demonstrated that "PD-L1 is predictive of response" to these immunotherapies. For esophageal cancer trials, he noted there was a "lack of overall survival benefit" in the PD-L1 less than 1 subgroup, but also acknowledged that sample sizes in these studies were very small.
"We see significant activity at PD-L1 greater than 10, we see modest benefit for [PD-L1] between 1 and 10, and I see no evidence of benefit in PD-L1 scores less than 1," Lieu said.
PD-L1 limitations
Despite ODAC panelists largely agreeing that the benefit of these immunotherapy-chemo regimens is greater in PD-L1-positive gastric and esophageal cancer patients based on the available data, the pharma sponsors and several public commenters raised concerns about limiting options for patients who lack access to PD-L1 testing and due to inconclusive PD-L1 testing results.
The pharma companies and ODAC panelists also worried that the measurement of PD-L1 expression can be inconsistent due to lack of standardization between tests and scoring methods and due to inter-reporter variability.
Both Merck and BMS performed analyses using real-world data from Flatiron Health showing gaps in PD-L1 testing for patients with gastric, GEJ, and esophageal cancer. BMS's analysis of the Flatiron data showed that about 40 percent of all patients did not receive PD-L1 testing prior to treatment.
Merck's clinical expert, Yelena Janjigian, chief attending of gastrointestinal oncology at Memorial Sloan Kettering Cancer Center, cited several limitations to PD-L1 testing in her presentation at the meeting. She noted, for example, that PD-L1 testing may not be possible because it's hard to obtain samples from patients or because of tumor heterogeneity. She also noted test variability and differences in pathologists' interpretation of PD-L1 CPS as further challenges.
The committee also discussed how choosing a cutoff for PD-L1 positivity may be difficult. Ravi Madan, an ODAC voting member and senior clinician and head of the Prostate Cancer Clinical Research Section at the National Cancer Institute, pointed out that the trials the committee reviewed were not designed to answer this question about benefit at different PD-L1 expression cutoffs, which makes choosing a threshold more difficult.
Jeffrey Meyerhardt, another ODAC voting member and chief clinical research officer at Dana-Farber Cancer Institute, added that any biomarker cutoffs can be difficult to determine because there may be outlier patients who benefit from the drug. "The reality is, for good or bad, we do this all the time in oncology," Meyerhardt said. "We don't give gemcitabine for gastric cancer, for example. Is there a gastric cancer patient who potentially could benefit from gemcitabine? I'm sure there is. There's probably multiple, but we still have to use some data to be able to decide who's potentially going to benefit or not."
Overall, the panel did not come to a conclusion about what the PD-L1 cutoff for treatment eligibility should be in these settings or how it should be chosen if the indications are limited to PD-L1 expressors.
"Limiting immunotherapy using a somewhat unreliable biomarker is a little bit concerning," Lieu said. "I would love to see patients have the opportunities to receive these drugs between PD-L1 scores of 1 and 10. But … that requires some discussion between the patient and their provider in terms of the risks, because we're asking our patients to undertake greater grade 3 and 4 risks for unclear benefit, particularly at lower scores."
Two members of the panel reviewing the gastric cancer data, James Hillard, the patient representative, and Randy Hawkins, the consumer representative and an assistant professor of internal medicine at Charles R. Drew University of Medicine & Science, both voted that the risk-benefit assessment was favorable in all-comers since there were some responders among the PD-L1-negative groups and expressed concerns over missing these patients by narrowing the indication.
The patient representative on the esophageal cancer panel, Dana Deighton, also voted that the data were favorable in all-comers, noting that her vote reflected her experience with patients having limited treatment options and struggling to find any treatment that would work.
Richard Pazdur, the director of the FDA's Oncology Center of Excellence, urged the pharma sponsors to work on standardizing PD-L1 testing methods across clinical trials and pushed for their commitment to doing so at the meeting. The representatives for Merck, BMS, and BeiGene all signaled that they welcomed efforts to collaborate with others and harmonize PD-L1 testing.
"This has been not a great experience, having all of these different tests here," Pazdur said. "I want to emphasize we made a concerted effort at FDA in trying to harmonize these tests with external organizations. As we move forward, and [are] looking at new biomarkers, we really have to develop platforms across the commercial concerns of the companies."
For PD-L1 testing, there have been some efforts to standardize tests and reporting across different methods. One project, called Blueprint, examined the interchangeability of PD-L1 assays and provided data on how comparable the results were. In 2017, the International Association for the Study of Lung Cancer (IASLC) released the Atlas of PD-L1 Immunohistochemistry Testing in Lung Cancer to provide clinicians and patients clarity on which PD-L1 assays may best suit their needs.
There have been other efforts to harmonize biomarker testing methods beyond PD-L1. Friends of Cancer Research convened a consortium to harmonize tumor mutational burden (TMB) testing, which resulted in a calibration tool for developers of next-generation sequencing assays to correct the variability in how TMB is calculated across different gene panels and analysis pipelines. The same group also launched an effort in 2020 to harmonize the use of homologous recombination deficiency (HRD) as a biomarker.
Pazdur also sought commitment from pharma sponsors to establish expanded access or compassionate use programs to provide their immunotherapies to PD-L1-negative gastric cancer patients, if the indication was narrowed and they still wished to provide their drugs to this group. The companies noted they had expanded access programs, but each said they would consider a program specifically for PD-L1 negative patients for these indications.
"I do want to address the concerns of patients. We realize the issues here with the biopsy [and testing]," Pazdur said. "If we do restrict [the indication] and if somebody wants the drug, it probably would not be paid for. So, we want to make sure that there might be other avenues that patients may have access to this drug."