Novel Predictive Biomarkers for Non-Small Cell Lung Cancer - Episode 6
Greg Riedlinger, MD, PhD: There are a number of challenges in using PD-L1 [programmed-cell death ligand 1] as a predictive biomarker in non—small cell lung cancer testing. Initially, there were many different antibodies and scoring systems. There has been a need for standardization using a single antibody and a single scoring system. At Rutgers CINJ [Cancer Institute of New Jersey], we’re generally using the 22C3 antibody and then using either the TPS [tumor proportion score] or the CPS [combined positive score] so that all our results are standardized in that manner. But there are also potential issues in that different biopsies from different sites may have different scores. Whether it’s a primary tumor versus a metastasis could potentially lead to different scores.
Gregory J. Riely, MD, PhD: Much has been made about the need for standardization of PD-L1 assays. There are competing PD-L1 assays that are on the market. I think that there have been 2 efforts that have made this a little bit less critical today. Number 1 was a big effort by the International Association for the Study of Lung Cancer to explore how the PD-L1 assays compared. What they found is that most of the assays are actually relatively similar and lead to the same interpretation for the same patient’s tumor. So, I think many of the assays are very much the same. But the biggest thing that has led to less of a need for standardization is that there’s really only 1 FDA-approved frontline therapy that requires a PD-L1 assay. If we use the PD-L1 assay associated with the pembrolizumab label, then we’re going to identify patients for first-line therapy most appropriately.
Jared Weiss, MD: In thinking about the limitations of PD-L1 as a biomarker in lung cancer, we need to remember what a good biomarker is. We have good biomarkers, for example, EGFR and ALK. When a good biomarker is positive, your targeted therapy works better than chemotherapy. When the biomarker is negative, it doesn’t work. In simple language as a clinician, a good biomarker tells me what to do. If it’s positive, I give whatever drug in question is. If it’s negative, I don’t. That’s really simple. That’s the case with EGFR. If biomarker is positive, then targeted therapy works better than chemotherapy with a better toxicity profile and better patient convenience. If the biomarker is negative, it’s no better than a diarrhea-inducing placebo.
This isn’t PD-L1. PD-L1 has some positive predictive value, but its negative predictive value is awful. If you hearken back to the second-line studies of nivolumab and pembrolizumab comparing them with docetaxel, what you see in many of those forest plots is targeted immunotherapy beating docetaxel, even in biomarker-negative populations, never mind biomarker-low populations. And so, this is not a biomarker that historically, or continuing to this day, cleanly tells you the right thing to do. It’s not worthless either. There is some positive predictive value, which I think is of more clinical relevance as we move into frontline decision making. I think that’s the case because frontline chemotherapy, which is to say platinum doublets, help people more than docetaxel does in the second line.
Why is this so? Where do these limitations come from? There are many reasons. I think one of the key ones is heterogeneity. There’s heterogeneity in PD-L1 expression within a tumor, and there could be heterogeneity between the primary site and the metastases. When you put a biopsy needle into someone and then study that material, you learn about what is going on at the exact place you put the needle and nothing else. When there’s heterogeneity, as there is with the PD-L1 biomarker, this can become very clinically relevant.
Gregory J. Riely, MD, PhD: The important thing that we need to understand about PD-L1 status is how it changes, and even though we know a lot about PD-L1, we have relatively few data to tell us how PD-L1 changes from the time of initial diagnosis. After therapy, does PD-L1 expression change? We have a disappointing amount of data to describe that. What we do know is that most of the trials we have that describe the value of PD-L1 use initial diagnostic material. The archival material from initial diagnosis is almost always the thing used to determine a patient’s PD-L1 status on clinical trials. But I think we really need to understand more about how that PD-L1 level changes over time, and many efforts are underway to identify those changes.
Transcript Edited for Clarity