Transcript:Benjamin P. Levy, MD: Marina, I want to go back. So, you’re not testing for PD-L1. Are there other biomarkers that you think may be useful either now or in the future? We’re getting a lot of data coming out at ASCO looking at some novel things.

Marina Garassino, MD: Yes. There are data on several things. The first one is mutational burden, which can be interesting, but you need in homogenous tumors. So, maybe this is not the mutational load at all, the important thing, but maybe you need something different. Then, there are data on TNF (tumor necrosis factor), there are data on cytokines. We know that maybe KRAS-mutated patients can benefit more from these kind of drugs, but, again, they are subgroup analyses and not preplanned. I think that the story is quite complicated and we need more time.

Benjamin P. Levy, MD: Govindan.

Ramaswamy Govindan, MD: I agree with Marina. I want to add a few more things. One is it’s not just the mutation burden, it’s the quality of those mutant peptides. After all, the immune system has to see something that’s foreign for it to react. These are called neoantigens. These altered peptides are able to present effectively to the immune cells. And so, there are no computer-based algorithms that can predict, based on the mutation profile, what could effectively present for the class 1 reaction. These new antigens seem to predict, a little bit better, the response to outcomes. And also, it is these new antigens present in the trunk, or present in all the clones, that seem to predict better. It gets so complicated. But, out of that, there is a nice silver lining. There was an excellent presentation at ASCO this year from the Hopkins group looking at the mismatch repair deficiency. In fact, a paper was presented on colorectal cancer in New England Journal of Medicine last year. In colorectal cancer, clearly those with mismatch repair deficiency seemed to respond. And, also, there’s a cohort on the non-colorectal cancer with mismatch repair deficiency. That was the focus of a presentation at this meeting, and it was an impressive 50 plus percent response rate.

And so, these are all various cancers you wouldn’t think of as responding to checkpoint inhibitors: gastric cancer, pancreatic cancer, even endometrial cancer. The appeal of that is that’s a test that is really available. We can do this based on immunohistochemistry or PCR, and there is evidence of mismatch repair deficiency in lung cancer quite a bit as well. So, that’s all very interesting. And then, the tumor that I know of that has the highest response rate to checkpoint inhibitors is not something we commonly talk about: Merkel cell cancer. The reason is that these have viral expressed antigens, and they have the highest response to PD-L1 inhibitors. A lot of this virus is getting integrated on a genome, and so it may be another thing that…And then, there were excellent papers yesterday looking at the T-cell receptor clonal rearrangement, both in the tumor and in the peripheral blood, that can select patients who would respond better to these things. It’s really a fascinating field.

This biomarker field is not going to be simple because you need two parts: something that elicits the immune response in the first place and something that holds it back. Because if that weren’t happening, patients won’t present with clinically evident cancer in the first place. So, we’ve got to look at both of these things. We’ve been looking at the one part of that out of a plethora of checkpoints that can slow down the immune system. We’re looking at one of that, and that’s why we are in this pickle looking at PD-L1.

Benjamin P. Levy, MD: It is a pickle. Sarah, Yale’s done a lot of work looking at PD-L1 and actually reporting on some discordant data based on the PD-L1 platform you’re using. What’s your perspective on this?

Sarah B. Goldberg, MD, MPH: I agree with what’s been said. It’s a complicated story with PD-L1 and with other biomarkers. But, I think what we’ve seen in trials, and from personal experience treating patients, is when you have a patient where you test for PD-L1 and it’s negative, they can clearly still respond. It seems in several trials now the response rate is lower. So, there’s less of a chance of response, but they can respond. And, I think that’s really the question when I think about whether I am going to test someone for PD-L1: will I still want to treat them anyway even if it’s negative? Because maybe I just don’t want to know. And if I’m going to treat them anyway because the other options are not great, there’s still this chance of long-term, really significant benefit even in the EGFR patients. I might want to treat anyway, so I won’t test. If it would change my mind, if I might choose something else instead, I consider testing. But, even with a mutational burden and other biomarkers that are being looked at, it’s really an interesting field, but I don’t think we have enough information now to say if we test then, we’ll know who to treat, because we still may see benefit in the negative patients. It may be a different story in the first-line setting where we have other active treatment options, but at least in second-line setting and beyond, I don’t tend to test.

Benjamin P. Levy, MD: I agree. I think the negative predictive value of this PD-L1 testing is poor. We still have patients who are negative that respond. We’re not routinely testing, but I think there may be a role. If you look at the gradations of PD-1 testing, they do seem to respond, they do seem to correlate with the response. Tucked away in the appendix of the KEYNOTE-001, depending on the cutoff level, there was higher responses based on the staining. Mark?

Mark G. Kris, MD: One of the nice things about being an alleged expert in this and that is you get to visit other hospitals, and I have to tell you I’ve learned the most about this at your hospital from Lieping Chen. The first thing I’d like to do is step back a bit. There’s very relevant biology here. PD-L1 agents cannot work without a tumor antigen that can be recognized by the immune system, without T cells, and without PD-L1 cloaking. All three of those things have to be there or this stuff will not work. It’s not possible, no magic, it will not work. So, why are these tests not perfect? Well, the tests actually are little pictures of this. The neoantigen is all about the antigen, but our ability to define what’s the real critical antigen, that’s going to turn on the T cell, is not good. And PD-L1 is the same thing. We don’t have a robust way of doing that, and, of course, we don’t understand at all how to get the right cytotoxic T cell there. Each of these things teases at that, but none of them define it. But, I don’t want people to misinterpret that PD-L1 expression doesn’t matter. It’s absolutely critical or this stuff will not work, and you have to have T cells there. They say, “Oh, we didn’t find any T cells in the specimen.” Well, you had to miss it, because this stuff will not work if you don’t have the T cell. So, that said, I don’t think there’s anything today that we can use for this. All of these things, they enrich, but they don’t predict.

Ramaswamy Govindan, MD: If I can add one more thing. This is also to make matters more complicated. This is a very dynamic marker. So, the PD-L1-negative tumor can become PD-L1-positive, and also it can be present in one area and not in other areas. All these things add to that.

Mark G. Kris, MD: The other thing I saw up there, they have all these beautiful pictures of this tumorous tissue. And the PD-L1 is like a little edge and everything else is negative. If you biopsy, you were going to biopsy that huge thing in the middle, and you would have told me this is the PD-L1.

Chandra P. Belani, MD: But, the PD-L1 expression does enrich patients for response.

Mark G. Kris, MD: It does.

Chandra P. Belani, MD: And if it is so dynamic that we miss it, that’s why those negative patients are responding. We are missing it in certain situations. And then, of course, you need the neoantigen or the neoepitope for the answer.

Benjamin P. Levy, MD: I think there’s a lot of work to be done. It’s very exciting in ASCO to see some of the work that’s coming down looking at this. What a complicated story, and I think it will take some time to tease out what’s going on. But, certainly, there’s a lot of good work going into trying to define those patients most likely to respond, but also, just as important, trying to look at why patients aren’t responding. And I think that’s where combination strategies come in handy.

Transcript Edited for Clarity