Select Topic:
Browse by Series:

PD-L1 Testing in Squamous NSCLC

Panelists: Mark A. Socinski, MD, Florida Hospital Cancer Institute; Tracey L. Evans, MD, University of Pennsylvania; David M. Jackman, MD, Dana-Farber Cancer Institute; Edward S. Kim, MD, Levine Cancer Institute; Jared Weiss, MD, UNC Lineberger Comprehensive Cancer Center
Published: Thursday, Feb 23, 2017



Transcript:

Mark Socinski, MD:
Tracey, let me ask you: are we at a time where, in the squamous population, PD-L1 testing should be standard-of-care in all patients?

Tracey L. Evans, MD: Absolutely, and that is a new position for me, and that just came along with presentation.

Mark Socinski, MD: Was it hard for you?

Tracey L. Evans, MD: It wasn’t that difficult, I somehow managed to get over it, but I say that facetiously. It has been very difficult for our institution actually, because it meant purchasing the platform by which we do the test in the way it was done in the KEYNOTE-024 study. We had to do that as a send-out, and only just recently, we were able to get the machinery that we needed to do the test in-house. So, yes, it was not difficult for me, but for my colleagues, it was actually quite difficult.

Mark Socinski, MD: And expensive probably.

Tracey L. Evans, MD: And expensive, right. But when the first checkpoint inhibitor that initially got approved in lung cancer—that was nivolumab in squamous—in that initial randomized study versus docetaxel, PD-L1 level didn’t matter. And we’ve had a lot of debate about whether there was any worth in testing PD-L1. Because even in the studies that came later, the nonsquamous studies, and then the atezolizumab and the pembrolizumab second-line studies, where the level did seem to predict for an enhanced population, even in the 0% patients, they still did as well with better toxicities than docetaxel. I was not a big tester until KEYNOTE-024 came around, and clearly in the high expressing patients who had greater than 50% expression, they’ve got a better option out there than chemotherapy—and its independent histology. It’s true in the squamous, it’s true in the nonsquamous. So, everybody needs to be tested for PD-L1, and we now have that as reflex testing at our institution, which is also wonderful because we had to order it separately before.

Mark Socinski, MD: Okay, so that’s the first step. We all agree that the patient should be tested. The next issue is, how should they be tested? I like to refer to the last 18 months as the “immuno-tsunami.” We’ve had to change the standard-of-care in second-line, now first-line. Each of the immunotherapeutic agents have been developed with a companion or complementary antibody. It’s not feasible to have 4 or more different platforms for testing these. Jared, what’s your perspective on what’s being done to get us to a 1-test platform, if you will?

Jared Weiss, MD: If you look historically, for example, at the development of HER2 in breast cancer, there were originally multiple tests to look at this, and that, of course, wasn’t going to fly in the long run. Our current situation, as you pointed out, is we have 4 different antibodies developed by 2 different companies with 4 different sets of staining conditions and lots of different ways of reading them. Not surprisingly, you had heterogeneity and how it helps you. Some of these are even described differently in different studies of the same drug.

There have been 2 major attempts to reconcile this: the Blueprint Project and the French Harmonization Project. Blueprint looked at 4 of these antibodies: 22C3, 288, SP263, and SP142. What they found is that all of these really stained mostly similarly on tumor cells, with the exception of SP142. That stained fewer tumor cells and fewer immune cells. Overall, the discordance rate, in some fashion, between the assays was about one-third, although that fell dramatically if you left out SP142. And interestingly, they made attempts to say, “Okay, well, is it just that we’re looking at these differently? What if we look at one antibody, score them all the same way—take the rules from one test and look at the other test—can the concordance go up? Are we really biologically looking at the same thing?” And it turns out, it gets worse if you try to look any of these tests in a way that’s different from the way they were developed. That attempt failed.

So, what do you do with this in practice? You look at the FDA label for the drug that you’re considering using and you get the test associated with it, and you get it and you score it as you’re supposed to. For example, we were talking about first-line pembrolizumab. If that’s what you’re considering, you need to get 22C3 and you need to look for at least 50% staining in tumor cells, regardless of intensity, and what’s going on in immune cells.

There was another attempt to bring this all together, to harmonize this. The French Study looked at 3 of these antibodies: 288, 22C3, and SP2163. At a 50% level, they were 95% concordant. But then they took it a step further and they said, “Look, in the real world, people are often using laboratory-derived tests.” They looked at 27 real lab-derived tests that people were looking to use and it was just about half that were discordant. Again, reinforcing the point that you really need to go to the standardized test for whatever drug you’re looking to use, and use it as indicted. I would say for those of us in the research world, at some point, we probably want to start torturing PD-L1 until it confesses something and consider better biomarkers.

Mark Socinski, MD: Well, I don’t disagree with that. However, I also think that when you look across all the studies and all the antibodies we have, PD-L1, in one sense, has been informative. As difficult as the IHC is to do and get a standardized test, it does tell us something.

Tracey L. Evans, MD: To an extent.

Jared Weiss, MD: Positive predictive value, you can push it to give you some positive predictive value. But as Tracey explained very well a few minutes ago, negative predictive value, particularly in later lines, has been a little more problematic.

Mark Socinski, MD: Right. And so, at each of our institutions, is the PD-L1 testing done in-house?

Jared Weiss, MD: We have not taken the leap to purchasing the machine, and 22C3 remains a send-out.

David M. Jackman, MD: It’s in-house.

Tracey L. Evans, MD: In-house recently.

Edward S. Kim, MD: We were in-house in October, so that’s when we implemented it.

Mark Socinski, MD: We just started using the SP263 antibody internally, so that’s what we do, and we’ve convinced our pathologist to do it reflexively at the time of diagnosis. I think that that helps in that. I think that’s where we have to be. I think Ed and I had the pleasure of learning a lot about PD-L1 testing a week or so ago, and I thought that was quite informative. I came away with that saying that the 3 antibodies—and get SP142 out of there—have very similar performance characteristics, and I think that that’s what it is.

But I know at the community level, most people are sending them out. Community oncologists have to send them someplace. And I know on one of the lab slips that we have, I don’t remember which third-party company it was, but it actually said, “Are you planning to use the Merck drug or the BMS (Bristol-Myers Squibb) drug?” And they would use a different test based on that sort of thing. So, I know that that’s there.

Transcript Edited for Clarity

Slider Left
Slider Right


Transcript:

Mark Socinski, MD:
Tracey, let me ask you: are we at a time where, in the squamous population, PD-L1 testing should be standard-of-care in all patients?

Tracey L. Evans, MD: Absolutely, and that is a new position for me, and that just came along with presentation.

Mark Socinski, MD: Was it hard for you?

Tracey L. Evans, MD: It wasn’t that difficult, I somehow managed to get over it, but I say that facetiously. It has been very difficult for our institution actually, because it meant purchasing the platform by which we do the test in the way it was done in the KEYNOTE-024 study. We had to do that as a send-out, and only just recently, we were able to get the machinery that we needed to do the test in-house. So, yes, it was not difficult for me, but for my colleagues, it was actually quite difficult.

Mark Socinski, MD: And expensive probably.

Tracey L. Evans, MD: And expensive, right. But when the first checkpoint inhibitor that initially got approved in lung cancer—that was nivolumab in squamous—in that initial randomized study versus docetaxel, PD-L1 level didn’t matter. And we’ve had a lot of debate about whether there was any worth in testing PD-L1. Because even in the studies that came later, the nonsquamous studies, and then the atezolizumab and the pembrolizumab second-line studies, where the level did seem to predict for an enhanced population, even in the 0% patients, they still did as well with better toxicities than docetaxel. I was not a big tester until KEYNOTE-024 came around, and clearly in the high expressing patients who had greater than 50% expression, they’ve got a better option out there than chemotherapy—and its independent histology. It’s true in the squamous, it’s true in the nonsquamous. So, everybody needs to be tested for PD-L1, and we now have that as reflex testing at our institution, which is also wonderful because we had to order it separately before.

Mark Socinski, MD: Okay, so that’s the first step. We all agree that the patient should be tested. The next issue is, how should they be tested? I like to refer to the last 18 months as the “immuno-tsunami.” We’ve had to change the standard-of-care in second-line, now first-line. Each of the immunotherapeutic agents have been developed with a companion or complementary antibody. It’s not feasible to have 4 or more different platforms for testing these. Jared, what’s your perspective on what’s being done to get us to a 1-test platform, if you will?

Jared Weiss, MD: If you look historically, for example, at the development of HER2 in breast cancer, there were originally multiple tests to look at this, and that, of course, wasn’t going to fly in the long run. Our current situation, as you pointed out, is we have 4 different antibodies developed by 2 different companies with 4 different sets of staining conditions and lots of different ways of reading them. Not surprisingly, you had heterogeneity and how it helps you. Some of these are even described differently in different studies of the same drug.

There have been 2 major attempts to reconcile this: the Blueprint Project and the French Harmonization Project. Blueprint looked at 4 of these antibodies: 22C3, 288, SP263, and SP142. What they found is that all of these really stained mostly similarly on tumor cells, with the exception of SP142. That stained fewer tumor cells and fewer immune cells. Overall, the discordance rate, in some fashion, between the assays was about one-third, although that fell dramatically if you left out SP142. And interestingly, they made attempts to say, “Okay, well, is it just that we’re looking at these differently? What if we look at one antibody, score them all the same way—take the rules from one test and look at the other test—can the concordance go up? Are we really biologically looking at the same thing?” And it turns out, it gets worse if you try to look any of these tests in a way that’s different from the way they were developed. That attempt failed.

So, what do you do with this in practice? You look at the FDA label for the drug that you’re considering using and you get the test associated with it, and you get it and you score it as you’re supposed to. For example, we were talking about first-line pembrolizumab. If that’s what you’re considering, you need to get 22C3 and you need to look for at least 50% staining in tumor cells, regardless of intensity, and what’s going on in immune cells.

There was another attempt to bring this all together, to harmonize this. The French Study looked at 3 of these antibodies: 288, 22C3, and SP2163. At a 50% level, they were 95% concordant. But then they took it a step further and they said, “Look, in the real world, people are often using laboratory-derived tests.” They looked at 27 real lab-derived tests that people were looking to use and it was just about half that were discordant. Again, reinforcing the point that you really need to go to the standardized test for whatever drug you’re looking to use, and use it as indicted. I would say for those of us in the research world, at some point, we probably want to start torturing PD-L1 until it confesses something and consider better biomarkers.

Mark Socinski, MD: Well, I don’t disagree with that. However, I also think that when you look across all the studies and all the antibodies we have, PD-L1, in one sense, has been informative. As difficult as the IHC is to do and get a standardized test, it does tell us something.

Tracey L. Evans, MD: To an extent.

Jared Weiss, MD: Positive predictive value, you can push it to give you some positive predictive value. But as Tracey explained very well a few minutes ago, negative predictive value, particularly in later lines, has been a little more problematic.

Mark Socinski, MD: Right. And so, at each of our institutions, is the PD-L1 testing done in-house?

Jared Weiss, MD: We have not taken the leap to purchasing the machine, and 22C3 remains a send-out.

David M. Jackman, MD: It’s in-house.

Tracey L. Evans, MD: In-house recently.

Edward S. Kim, MD: We were in-house in October, so that’s when we implemented it.

Mark Socinski, MD: We just started using the SP263 antibody internally, so that’s what we do, and we’ve convinced our pathologist to do it reflexively at the time of diagnosis. I think that that helps in that. I think that’s where we have to be. I think Ed and I had the pleasure of learning a lot about PD-L1 testing a week or so ago, and I thought that was quite informative. I came away with that saying that the 3 antibodies—and get SP142 out of there—have very similar performance characteristics, and I think that that’s what it is.

But I know at the community level, most people are sending them out. Community oncologists have to send them someplace. And I know on one of the lab slips that we have, I don’t remember which third-party company it was, but it actually said, “Are you planning to use the Merck drug or the BMS (Bristol-Myers Squibb) drug?” And they would use a different test based on that sort of thing. So, I know that that’s there.

Transcript Edited for Clarity
View Conference Coverage
Online CME Activities
TitleExpiration DateCME Credits
Community Practice Connections™: ALK-Positive NSCLC: Emerging Strategies to Inform Sequencing, Optimize Outcomes, and Address Unmet Clinical Needs Along the Disease ContinuumAug 29, 20181.5
Community Practice Connections™: Oncogenic Tumor Board in Advanced NSCLC: Leveraging Actionable Mutations Along the Disease Continuum to Optimize Patient OutcomesAug 30, 20182.0
Publication Bottom Border
Border Publication
x