Checkpoint Inhibitor Biomarkers for Lung Cancer

Transcript:Mark A. Socinski, MD: I’m beginning to see that in many of the second opinions that I’m doing in my practice in Pittsburgh, we get the outside laboratory report and there’s a result for EGFR, ALK, and usually ROS1, and then there’s PD-L1. Often, I don’t know what antibodies they’re using, how they’re doing it, what it means, and all this stuff. Where are we with biomarkers? Let me ask Roy. Is it prime time now or is this something that we still have a long way to sort out?

Roy S. Herbst, MD, PhD: I think it’s somewhere in the middle. It’s clear that biomarkers can help a little bit, talking just now about PD-L1. You know that PD-L1 binds to PD-1, so if you can measure PD-L1 on the tumor, that should be a positive predictive marker. Again, it’s not the only marker, and that is one of the problems. The other issues are the heterogeneity within tumors and the fact that PD-L1 expression is induced by gamma interferon, so it’s going to differ based on the state of the tumor.

Some of the biopsies that we looked at were done within a few weeks before treatment; some are a year or more before. So, there are all different variables. Then you have all the different antibodies that are being used; antibodies that are staining tumor cells or stroma and have different epitopes. As if all that wasn’t enough to make it a little bit confused, you then have the fact that it’s not like the EGFR mutation where you either have it or you don’t. Now, is it 10%, 20%, 30%? We have all different assays, all different scales.

That said, in the nivolumab trials, some biomarker work was done in the squamous setting, and it didn’t seem to make much of a difference. In the nonsquamous tumors, there were some data that I actually discussed at ASCO, and it’s very difficult to determine what to do there. Those patients didn’t all have the PD-1 biomarker; only about 80% did. It wasn’t tested prospectively, it was done afterward, so the patients were not stratified by the biomarker.

For the 50% of patients who had less than 1% PD-L1 positivity on the tumor cell, the hazard ratio was not much different than 1.0. There was a bit of a tail to the curve, and the toxicity was less for nivolumab versus the docetaxel. So, one can make the case, even in that setting, that one should use nivolumab because there’s a chance of a little bit of a tail to the Kaplan-Meier curve perhaps, and then certainly you don’t want to keep the other 50% of patients from getting the drug.

There are other biomarkers being developed for pembrolizumab. There’s also another strategy for atezolizumab. We’ve worked on that and looked very carefully at measuring not just the tumor cell compartment, but also the immune compartment. It makes sense that some of the T-cells and macrophages also express PD-L1 because this is all working to adapt to the immune response. It’s a bit complicated!

From the investigator point of view, are we going to have a central lab running four or five different PD-L1 assays? I agree that you get a report and you don’t know what it means. It’s a work in progress, and then of course there are other markers that are being looked at, as well.

Naiyer Rizvi, MD: I think it’s an important point in terms of this assay because, as you both said, what’s being done commercially right now has not been validated in terms of clinical outcomes. But, undoubtedly, the biomarkers that were developed for the nivolumab and pembrolizumab clinical trials will be available in the near future, and then it becomes a more useful assay where we have some clinical data and a validated assay to accompany decision making.

If you have a patient who has nonsquamous histology and the tumor is PD-L1-negative and the hazard ratio is 1 compared with chemotherapy, I think that one has to make a decision about whether nivolumab makes sense or not.

Geoffrey R. Oxnard, MD: That’s a patient I put on a clinical trial maybe, instead of on a drug that has relatively less activity. So PD-L1 negativity could be enriched for clinical trials perhaps.

Naiyer Rizvi, MD: Or maybe docetaxel and Cyramza would be better than docetaxel alone.

Mark A. Socinski, MD: Good question.

Geoffrey R. Oxnard, MD: My hope is that it’s going to be like the early EGFR days. The drug erlotinib becomes available and it’s the battle of the biomarkers. Is it mutations, is it IHC, or is it FISH? And we’d sort it out because the drug is available and everyone could fight over it.

My hope is that someone is going to sort out the better biomarker and clarify this for the rest of us. But, in the meantime, we’re pushing around so-so biomarkers waiting for the good ones.

There are plenty of patients who don’t benefit and that will create the need. Clinicians will say, “My gosh, if the drug’s not working all the time, then I want to be able to know who it’s working in, just like with erlotinib or crizotinib.” I think that will push the market to find a biomarker.

Mary Jo Fidler, MD: I thought there were interesting data presented at ASCO looking at rebiopsying patients with driver mutations, looking at the PD-L1 expression in the tumor-infiltrating lymphocytes before and after progression on a targeted agent. At least in that group, the PD-L1 expression was relatively low and didn’t seem to change. It wasn’t that their tumor progressed on an ALK inhibitor or EGFR inhibitor and then their PD-L1 status changed; it still remained low.

For patients who come in with a driver mutation, I’m trying to focus them in on trying to find a trial for a second- or third-generation targeted drug.

Naiyer Rizvi, MD: Right, so driver meaning, EGFR and ALK?

Mark A. Socinski, MD: But the data there have been relatively disappointing as far as immunotherapeutics, right?

Naiyer Rizvi, MD: Right, and it’s totally what you’d expect. The response rate in neversmokers is clearly less than in patients who have more damaged tumors; they’re less immunogenic. Justin Gainor presented the data on EGFR and ALK, showing low levels of PD-L1 expression and T-cell infiltration within those tumors that are mutated. I think in scenarios like that; the PD-L1 assay is actually more important because the positivity rate is about 10% in patients who are EGFR- or ALK-positive or neversmokers. You really need to select for those patients more, and I think in those cases, where the response rates are lower with nivolumab and there are less immunogenic tumors, the assay could be very useful.

Mark A. Socinski, MD: I want to know where I can get a mutational load test done.

Geoffrey R. Oxnard, MD: Call the Broad Institute; they do whole exome sequencing.

Naiyer Rizvi, MD: I was on the phone with today.

Mark A. Socinski, MD: Tell us about your data. They were interesting.

Naiyer Rizvi, MD: So the data that Mark is speaking of is that most of the responses to PD-1 inhibitors have been in tumor types that are genetically complex, damaged tumors that have a high mutation load. And when we refer to mutation load, we don’t mean whether they have a driver mutation; instead, we mean the total number of mutations within the exome, so expressed mutations. If you look at head and neck tumors, bladder tumors, or melanoma, obviously they’re all damaged tumors for very different reasons. It could be viral, it could be from UV radiation, or it could be from smoking carcinogens.

We looked at patients that were treated with pembrolizumab, looked at responders versus nonresponders, and did whole exome sequencing to evaluate the number of non-synonymous mutations. Those are the mutations that actually change an amino acid. Conversely, you can have a silent mutation that doesn’t change the amino acid and won’t change the protein’s immunogenicity.

If you sequenced the whole exome and all of the mutations, it didn’t really matter, but if you looked at nonsynonymous mutations, the ones where the protein was changed, we found a highly significant difference in terms of those more mutated tumors that responded much better to pembrolizumab than those that were less mutated. I think we will evolve that way—it’s just the next level.

It’s getting cheaper to do these tests. We have next-generation sequencing to determine the driver mutations for targeted therapy. I have to think that doing whole exome sequencing will become routine in the future. We’ll then get a number of nonsynonymous mutations and use that number; maybe not in isolation, but in combination with PD-L1 expression, to select those patients for immunotherapy. I think it’s coming soon.

Roy S. Herbst, MD, PhD: So, to do this test, are all next-generation platforms the same as far as how they call mutations and how they score them?

Naiyer Rizvi, MD: No, I think the bioinformatics does play into it, especially if you’re doing this analysis off of FFPE tissue. There are a lot of false-positives and errors within the results. I think you have to have a very strong bioinformatics platform to do this analysis. But it was the same case with targeted sequencing a few years ago. I think it’s just a matter of creating the algorithms and making them more widely applicable.

[Editor's note: Since the filming of this exchange, pembrolizumab was approved with a companion diagnostic for PD-L1 for patients with NSCLC. Additionally, nivolumab's NSCLC approval was expanded to include nonsquamous histology, with a complementary diagnostic for PD-L1]

Transcript Edited for Clarity