An FDA committee’s September 2024 vote to limit the use of Merck’s Keytruda and BMS’ Opdivo in stomach and esophageal cancers based on PD-L1 expression levels reflects an emerging trend that leverages ever-maturing datasets.
In the last decade, cancer treatment has pivoted to a precision medicine approach, with immunotherapies providing the backbone for many regimens. With this shift, patient biomarker data has become a crucial aspect of clinical trials, and recently, the FDA has been reassessing labels based on maturing datasets after approval.
Often, therapies for diseases with high mortality rates such as cancer are greenlit under the FDA’s accelerated approval pathway before all the data are evident. This has been the case for PD-1 inhibitors, including Merck’s Keytruda (pembrolizumab) and Bristol Myers Squibb’s Opdivo, both of which are approved for treating several types of cancer.
“When [PD-1 inhibitors] first came out, we were like ‘everyone gets pembro,’ no matter the PD-L1 expression level because there was literally nothing else,” Harpreet Singh, former FDA division director of oncology and current chief medical officer at Precision Medicine, told BioSpace.
Now, treatment options have expanded, and the FDA is setting limits on when certain inhibitors should be used. In September 2024, the FDA’s Oncology Drugs Advisory Committee (ODAC) voted 10-2 against the use of Keytruda and Opdivo as first-line treatments for advanced HER2-negative stomach cancer in patients with PD-L1 expression scores less than 1, and 11-1 against the use of Keytruda and Opdivo in first-line unresectable or metastatic esophageal squamous cell carcinoma with low or no PD-L1 expression. The therapies are currently approved for use in all patients regardless of PD-L1 expression.
Despite the overwhelmingly negative ODAC votes in September, the FDA has yet to release a verdict, and Singh said “it’s not clear whether they will.” But there is precedent.
In 2022, the FDA limited the use of PARP inhibitors from Clovis Oncology and GSK to ovarian cancer patient populations with certain BRCA mutations. Prior to that, in 2018 the agency incorporated PD-L1 status into the labels of Keytruda and Tecentriq for existing frontline approvals for patients with urothelial cancers. The change came after a study revealed that patients with PD-L1 low status had decreased overall survival in the single-agent immunotherapy arms compared to chemotherapy.
“FDA is a data-driven scientific agency,” Singh said. “As the data grows and evolves, they’re adjusting their behavior to go where the data takes them.”
Following the Data
At the crux of the September ODAC votes was the risk-benefit profile of the PD-1 inhibitors in patients with low levels of the biomarker. Ultimately, the panelists concluded that if the PD-1 inhibitors do not appear to have substantial benefit for patients with PD-L1 scores lower than 1, their use is more likely to harm than to help.
PD-L1 is a measure of how involved the immune system is in a cancer, explained Elad Sharon, a medical oncologist at the Dana-Farber Cancer Institute. If an immune response is evident by the presence of PD-L1 on the cell surfaces, then PD-1 inhibition has a chance to work and benefit the patient, he told BioSpace. But if the immune system doesn’t interact with the tumor microenvironment to begin with, this type of inhibition is unlikely to provide any benefit.
“If you’re not providing benefit to a patient, then you probably are only providing harm,” Sharon said, noting that immuno-oncology agents thwart a normal regulatory pathway of the immune system. “You essentially take that tool out of the toolbox for the immune system, and that’s how you end up with these toxicities like inflammation or other areas of concern.”
These instances of regulatory reassessment follow the collection of further evidence, Singh said. The initial approval of Keytruda in 2014 was based on single-arm trials, which showed sufficient evidence of benefit to warrant approval, she explained. However, as a broader range of treatments became available, more biomarker data began to emerge. As these data matured for gastric and esophageal tumors, the hazard ratio uncovered wasn’t so favorable for patients without PD-L1 expression.
The FDA is the only regulatory body in the world that receives a company’s raw data, Singh explained. In the case of Keytruda and Opdivo in stomach and esophageal cancer, the agency utilized the subject-level datasets from each company to aggregate and interrogate the massive amounts of data, including biomarker status.
“They’re looking for trends across therapeutic classes . . . to see critical patterns that companies can’t identify alone because they don’t have each other’s subject level data,” Singh said.
The question comes down to the FDA’s role in these situations, she continued. Should the agency aggregate and publish the data to inform providers or take it a step further and restrict the label? A Merck spokesperson told BioSpace that the company “cannot speculate on the future actions of the FDA” and is “committed to continuing discussions” with the agency.
It’s a balance between leading with the data and patient access, Singh said, indicating her preference to publish the data and allow practitioners to ultimately make the call for their patients. However, if the label is restricted, she clarified that very few patients will be affected as gastroesophageal junction tumors are rarely negative for this biomarker.
Biomarker-Driven Drug Development
Perhaps the biggest lesson to be learned from the increased biomarker scrutiny after approval is for drug developers, Sharon said. He recommended the FDA require pre-specified analyses of relevant biomarkers in early-stage clinical trials.
“Identifying appropriate patient populations and developing statistically valid conclusions from clinical trials is an essential role that sponsors have to play,” he said.
Early-stage trials are crucial for determining if a drug works the way it’s expected to. Often, therapeutics have a compensatory mechanism or accessory pathway that can be affected, making it unwise to begin a trial with the assumption that a drug will be ineffective for a particular population, Sharon explained. Ideally, companies should track biomarkers but not exclude patients. These data can then be used to target the drug to the population expected to see the greatest benefit as well as develop an accompanying biomarker test for selection purposes.
“You can’t really separate the two,” Sharon said. Selectivity is only as good as the tests available, he explained, and “every time you end up with a new diagnostic technology, you have the ability to retest your assumptions and see if you can more precisely treat patients according to their specific needs.”
Ultimately, it’s the responsibility of both the FDA and the sponsors to determine which patients will actually benefit from a treatment, Sharon concluded—and what is really needed is more rigorous data.
“We have the technology to lead us to the right patient with the right therapy at the right time more now than ever,” he said. “We just have to make sure that when we’re developing new agents, we rigorously test and co-develop these biomarkers so that we can achieve that significant level of precision and benefit for our patients.”
