The Strengths and Pitfalls of PD-L1 as a Biomarker for Immunotherapy

Meg Barbor, MPH

July 2018 - Immunotherapy


PD-1 and PD-L1 expression on tumor cells is decidedly unique compared with other biomarkers currently used in oncology, and its use has been demonstrated in numerous clinical scenarios. But although the benefits associated with PD-1/PD-L1 targeted therapies have been unprecedented, the risks should not be underestimated.

At the 2018 Hematology/Oncology Pharmacy Association conference, 2 experts discussed the use of PD-L1 as a biomarker in clinical practice.

PD-L1 Testing in the First- and Second-Line Settings

In patients with non–small-cell lung cancer (NSCLC) associated with at least 50% PD-L1 expression, significant responses were seen with front-line pembrolizumab (Keytruda) compared with investigator’s choice of chemotherapy in the KEYNOTE-024 study. Alternatively, the CheckMate-026 study of nivolumab (Opdivo) enrolled all patients with some level of PD-L1 expression but found no advantage to using nivolumab in the front-line setting compared with chemotherapy.

“This supports the argument for PD-L1 testing in the front-line setting for advanced or metastatic NSCLC, especially when considering monotherapy,” said Josiah Land, PharmD, BCOP, Clinical Pharmacy Specialist, Thoracic Medical Oncology, Memorial Sloan Kettering Cancer Center, New York City.

In the CheckMate-017 (squamous) and CheckMate-057 (nonsquamous) studies of nivolumab versus docetaxel (Taxotere) in patients with previously treated advanced or metastatic NSCLC, durable responses were seen with nivolumab across all histologies and PD-L1 expression levels. But PD-L1 as a predictive biomarker in these trials was not useful at predicting durable response. Patients with the highest expression of PD-L1 had durable responses, but the patients who did not express PD-L1 had superior responses, as well.

“What’s interesting about this trial, and what muddies the water outside of the front-line indication, is that PD-L1 didn’t really help us predict a quality of response. So it’s hard to use PD-L1 as a predictive biomarker in the second-line setting, because the responses weren’t necessarily consistent among all levels of expression,” he said.

In the OAK clinical trial, patients with NSCLC who received atezolizu­mab (Tecentriq) had more responses and improved survival compared with those receiving docetaxel regardless of PD-L1 expression, again complicating the argument for PD-L1 as a predictive biomarker in this setting.

In CheckMate-025, PD-L1 expression was a poor prognostic indicator of response to nivolumab versus everoli­mus (Afinitor) in patients with advanced renal-cell carcinoma, “thus murking the waters outside of front-line NSCLC treatment,” Dr Land told attendees.

And in CheckMate-040, patients with hepatocellular carcinoma had similar response rates with nivolumab regardless of PD-L1 expression. Similar responses have been seen in colorectal cancer clinical trials.

“It sounds like I’m droning on about this, but it’s hard to really nail this down as a predictive biomarker, because we’re seeing similar outcomes whether the patients do or do not express PD-L1,” Dr Land said.

According to Dr Land, although the data do support the importance of PD-L1 testing in the front-line treatment of NSCLC, confirming PD-L1 as a predictive biomarker in previously treated patients with NSCLC or with other solid tumors is more complicated.

How Is PD-L1 Different?

A new era of oncology requires a new type of thinking, suggested Jacob K. Kettle, PharmD, BCOP, Clinical Oncology Specialist, University of Missouri Health Care, Columbia. PD-L1 as a biomarker is different from other known biomarkers that are inherent or overwhelmingly predictive, he said.

“After about 2 decades of this line of thinking, we’ve got this pond full of biomarkers—CD20, EGFR, BRAF, ALK, HER2—that have been cornerstones of therapy,” Dr Kettle said.

“I think the temptation was to put PD-L1 in the same bucket, but PD-L1 is dynamic, inducible, and disease-state dependent,” he said.

PD-L1 is not a cornerstone biomarker but rather an enriching biomarker, Dr Kettle noted, as was displayed by the enrichment of responses in treated patients with advanced NSCLC and high PD-L1 expression in KEYNOTE 010, which compared pembrolizumab with docetaxel.

Raising the PD-L1 threshold often leads to better response rates, but lowering it means more patients are receiving treatment. These are both valuable goals but are diametrically opposed.

“We can’t satisfy both, and I think that creates complex clinical decision-making and more profound philosophical issues,” Dr Kettle said. For example, patients with low PD-L1 expression in the OAK study displayed fairly meager response rates, whereas approximately 33% of patients with high PD-L1 expression responded.

“There’s a huge difference between those 2, very clearly showing there’s a big difference between how patients are going to react to these drugs,” Dr Kettle said. “If you’re a patient and have an option that is barely a 1 in 10 chance, that’s a lot different from offering treatment that has a 1 in 3 chance,” he added.

“Just Because You Can, Doesn’t Mean You Should”

For a biomarker to be clinically relevant, it should identify responders as well as nonresponders.

“I think we need to recalibrate expectations. None of us want to leave a useful therapy on the table, but we need to remember that there are definitely consequences to using treatment, especially when patients aren’t likely to benefit,” Dr Kettle said. “Therapy that doesn’t work is worse than doing nothing.”

There is also a growing understanding of the concept of hyperprogression, whereby some patients experience rapid tumor growth after initiation of PD-L1 inhibitor therapy and a shorter overall survival.

“I’m not trying to discourage PD-L1 therapy; it has radically changed how we practice and what we can offer patients. It offers potentially durable responses and a better tolerability profile. But we need to fight the urge to just instinctively and inclusively reach for these drugs because they’re on-label. Maybe that’s not the right move all the time,” Dr Kettle said.

There is also the financial impact, he added. The cost of finding a single response with nivolumab in PD-L1–negative NSCLC can be close to $500,000.

“It’s a tough question, but the reality is that there is a societal impact to this,” he said.

Enriching Biomarkers

The application of these types of biomarkers is nuanced but should help to align the choice of therapy with a patient’s values. PD-L1 expression is a useful tool, but because it is part of a dynamic and complex microenvironment, more research is needed to optimize its application in clinical decision-making.

“There are some patients who say, ‘I don’t care what the risks are; if there’s a chance I can get a little more time, I want whatever’s on the table,’ and if that’s the case, then you know that’s a great patient to pursue therapy, even though there’s a low chance of response,” Dr Kettle said.

Combining biomarkers has shown promise in predicting responses to immunotherapy, and in the CheckMate-026 trial of front-line therapy with single-agent nivolumab in NSCLC, pairing high tumor mutational burden biomarker and high PD-L1 expression overwhelmingly predicted responses in patients.

“That is extraordinarily encouraging to me moving forward, that hopefully we find this perfect balance in the future,” Dr Kettle said, adding that these conclusions are cursory, and that the data can change rapidly.

“But if our mission is greatness, then we have to be humble enough to say, ‘We need to do things differently.’ We need to keep pushing the boundaries and confront those uncertainties. Addressing them—as opposed to ignoring them—is how we’re going to get to where we want to go,” he said.