Insights From: Ezra Cohen, MD, Moores Cancer Center; UC San Diego Health; Jared Weiss, MD, UNC Lineberger Comprehensive Cancer Center
Jared Weiss, MD: We would like to have a great biomarker for patient selection with recurrent metastatic head and neck cancer. The ideal would be a situation like ER/PR in breast cancer or the EGFR mutation in lung cancer, where you have a good biomarker. If it’s positive, you know your patient should get the treatment. If it’s negative, you know that they should get something else.
There have been a lot of candidate biomarkers to try to fulfill this role in head and neck cancer. The first and most obvious was PD-L1—the actual target of the treatment and the interaction between PD-1 and PD-L1. It’s true that in some of the studies, we do see a positive predictive value for PD-L1, meaning that the patients with the highest level of PD-L1 expression seem to benefit more than patients with the lowest level or no expression.
The problem is, in the clinical context that we’re looking at, a negative predictive value—who’s not going to benefit—is the more useful biomarker. In this context, in my opinion, PD-L1 fails us in head and neck cancer in its current form. And so, we look to other possible candidate biomarkers. The most obvious second choice is HPV. We know that the HPV-driven head and neck cancers are biologically distinct from the smoking-driven cancers, with a lot of major biologic differences.
In the spectrum of immunotherapy, we know that their phenotype is rather different, and the preclinical arguments for response are very different. This can be measured in a lot of different ways. The most common are the FISH test for HPV or P16 IHC. While some of the individual smaller studies have shown some differences there, it’s not clear that there is any population for which chemotherapy would be better than the PD-1 or PD-L1 inhibitors. And so, again, HPV fails us as a biomarker for excluding patients, at least in the immunotherapy versus chemotherapy question. It doesn’t operate the way we would have an ideal biomarker operate. In my opinion, we should be fairly agnostic to HPV status in our current selection of patients for immunotherapy.
Where might we go from here? There are a number of candidates. Perhaps the furthest along is mutational burden. What’s nice about this biomarker is that there are commercially available assays that give it to you already. If you order next-generation sequencing, some of the commercially available platforms will give you a measure of mutational load. And, it does seem to have some predictive capacity.
My favorite candidate biomarker, personally—reserving the right to be very wrong—is the interferon-gamma signatures. If you look at the KEYNOTE-012 data with pembrolizumab, we published those data showing hints of true negative predictive value for that assay. To me, as a clinician, that’s really what I thirst for—a test that tells me who won’t benefit, so I know to move on to something else. That’s not commercially available. You can’t use it in standard clinical practice now. But, that’s my personal hope for a true biomarker in the future.
Ezra Cohen, MD: When we think about using these agents, anti–PD-1 antibodies, and when we think about where the field is going and where we need to go, we begin to think about biomarkers and we think about better patient selection. One of the obvious biomarkers in squamous cell carcinoma of the head and neck is HPV. We know that there are 2 etiologies for this disease now—the virus HPV and mostly tobacco-driven, or HPV-negative.
We recognize that the HPV-positive patients tend to have a more immune infiltrated tumor with both T cells and other types of immune cells—mostly macrophages—and we hypothesized, at the beginning, that those patients would have a higher response rate when it came to immunotherapy. It turns out, interestingly enough, that both HPV-positive and HPV-negative patients respond to anti–PD-1 therapy with about the same numbers. So, the response rate is very similar. If you look at the studies, as the data emerge, it does look like the HPV-positive patients may have a higher response rate. It does look like, in randomized trials, the HPV-positive patients may have greater efficacy—slightly lower hazard ratios in the survival curves. But, HPV-negative patients clearly respond and clearly can benefit from these agents.
HPV per se, right now, is not a good predictive biomarker when it comes to using anti–PD-1 therapy. One might ask, “Does it matter how HPV integrates or how HPV is residing in the cell?” And although the data are very early, the answer, so far, appears to be, “No.” It is actually integrated into the DNA. Or, it is present episomally, and does not seem to affect the efficacy or the activity of anti–PD-1 therapy.