Search Videos by Topic or Participant
Browse by Series:

PD-L1 Expression in Bladder Cancer

Panelists: Daniel P. Petrylak, MD, Yale University Cancer Center; Dean F. Bajorin, MD, Memorial Sloan Kettering Cancer Center; Robert Dreicer, MD, MS, Virginia Cancer Center; Arjun V. Balar, MD, Laura and Isaac Perlmutter Cancer Center; Elizabeth R. Plimack, MD, MS, Fox Chase Cancer Center; David I. Quinn, MBBS, PhD, USC Norris Cancer Hospital
Published: Thursday, Mar 23, 2017


Transcript:

Daniel P. Petrylak, MD:
So, David, why is there such a difference in opinion about PD-L1 expression and response to these agents?

David I. Quinn, MBBS, PhD: I think the data are conflicting and we tend to be reductionist in the way we look at things. And we would love a marker where we send it off, it comes back, and says, “Look, this marker links to therapy X. Please follow the dots.” Unfortunately, it’s not as simple as that, and, as Betsy and Dean were discussing earlier, it may be that patients that are in this immunologically challenged state respond both to chemotherapy and to checkpoint inhibition very well. So, I think we have conflicting data. We have sets where we’re looking at presumably the same population of patients where with one of the monoclonal antibody therapies, we get a response that seems to relate to PD-L1 expression. And then, with another one that we would assume is not terribly different in our clinical experience, it’s the reverse, we don’t get an effect. And also, the disease state matters. Whether the patient is eligible for chemotherapy as first line, that seems to be a different set to platinum-pretreated patients. We’re having trouble grappling with what’s a dynamic marker. We know it changes. We just have a snapshot of a particular tumor that’s taken, which may be the day before we want to start therapy or within a week, or it could be from a cystectomy or nephroureterectomy that was done 7 years before, and they’re likely to be very different.

Robert Dreicer, MD, MS: David, I think that last part is very critical. None of the studies that have been reported have really told us what percentage of samples come from what timeline. We’re interpreting things based on, frankly, somewhat incomplete data. So, it’s even harder to speculate about that.

Daniel P. Petrylak, MD: Exactly. In fact, in Arjun’s study in platinum-ineligible patients where he found no correlation between PD-L1 expression and the overall outcome, that perhaps is because that tissue is probably the closest that we would have to the actual treatment. And perhaps you’d like to comment on that because that may be the reason why we didn’t see anything.

Arjun V. Balar, MD: Right. The study you’re referencing is IMvigor-210 cohort 1 and similarly KEYNOTE-052. Both of those studies tested immune checkpoint inhibitor therapy, PD-1 and PD-L1 respectively, in the first-line setting in cisplatin-ineligible patients. And simply, by virtue of the fact that these are first-line patients, you’re going to have younger tissue that went into the biomarker testing. In fact, we looked at that. The median age of the tissue is probably somewhere in the range of about 100 days or so. Whereas, in the second-line studies, it’s much longer, probably on the order of something like 300 days or so. And that certainly can impact the utility of the paraffin-embedded block that the tissue has been sitting in wax for years versus younger tissue, and then, the applicability of doing IHC staining on such old tissue.

Daniel P. Petrylak, MD: In fact, on the phase I trial, there were some patients that had archival tissue from years back that turned out to be negative. Then, we rebiopsied them and we had some positive patients out of that. It’s a very, very difficult issue. Betsy, you want to comment?

Elizabeth R. Plimack, MD, MS: One other maybe simpler explanation for why PD-L1 is less predictive later is that the studies where the discovery set defined the PD-L1 cut point in the test across drugs are all where it performed the best. So, your atezolizumab trial, the frontline KEYNOTE trial, the biomarker looked good, it was developed in that set. And then, when you apply it to subsequent sets, it’s just a law of statistics, right? There’s fallout. And so, maybe that’s a simple explanation for what we’re seeing. For durvalumab, for instance, we’ve only seen the test set. So, I think we really need to see, for instance, how that pans out.

Robert Dreicer, MD, MS: And it’s important to remind people there are multiple PD-L1 assays and none of them have been used in the same studies. They look at different things. We’re a long way away from understanding how to apply that across the disease state.

Elizabeth R. Plimack, MD, MS: But we have to validate a biomarker in order to believe in it.

Dean F. Bajorin, MD: So, we have a dynamic disease over time, we have a dynamic immunological environment over time, and then we have multiple assays with different antibodies with different reads and we expect one answer. I think that that shows we can’t boil it down.

Daniel P. Petrylak, MD: What we really need is an imaging agent or some liquid biopsy that we can monitor on a real-time basis.

Elizabeth R. Plimack, MD, MS: We need something that validates across sets.

Arjun V. Balar, MD: But the challenge is clearly the immune system is not binary. It’s a long continuum, so you’re trying to measure a dynamic thing in a very static environment. There is some correlation with PD-L1 expression. In my opinion, I think that’s the most we’re going to get with PD-L1 expression. We need something better.

Daniel P. Petrylak, MD: Exactly.

David I. Quinn, MBBS, PhD: I think the PD-L1 story, we’re going to hear from the lung cancer people that they’re trying to compare at least 4 of the different antibodies out there. And they do seem to perform differently in different settings. So, we’ll need to move through that and I would suggest that with our current therapeutics, PD-L1 in tumor tissue is not going to be the entire answer, but there may be linkage in one therapeutic, which is really good. That may make a difference to that particular therapy and the biomarker that goes with it.

Dean F. Bajorin, MD: That’s a really good point because in the lung tissue, there’s harmonization on PD-L1 expression on the tumor, but it’s less so on the infiltrating immune cells. It really becomes highly complex.

David I. Quinn, MBBS, PhD: When you do these assessments, you get more questions than you answer because I think the heterogeneity of what’s in those lymphocytes, even if you’re just looking at PD-L1, is fascinating. I don’t know what it means, but we need to try and work it out.

Daniel P. Petrylak, MD: I think as David McConkey said in his talk the other day, we need to get past PD-L1 and start looking at other things, and I think that’s important. But, David, you actually mentioned a case recently from California of a patient who was trying to get one of the checkpoint inhibitors for us, and the insurance company denied it based upon PD-L1 stain. Maybe you can elaborate on that because it’s important that we don’t eliminate patients based upon their staining.

David I. Quinn, MBBS, PhD: In our Los Angeles County Hospital network, almost 2 years ago, we were faced with a problem. We have a capitated network with very limited funding. If we gave everybody checkpoint inhibitors who possibly qualified, we’d probably close the hospital in April of the year and have to get to December. We do get help for what’s predominantly an indigent population from our pharmaceutical partners, but the pharmacy was trying to rationalize use. And, at the time, we had early data, not validated, to suggest and that PD-L1 would be a good marker in renal cell, in bladder, lung, and certainly in melanoma, where I think it is useful. And so, they said we need to have PD-L1 positivity for us to give the checkpoint inhibitor. Now, we’re due to review this and sit down after the 2-year period. I think there will be some changes, but I think they just had to have a response to rationalizing therapy in a dynamic milieu.

As it happens, we are able to get most of our indigent patients who would benefit the opportunity of getting one or other checkpoint inhibitor, and we just have to work at it. And we’re talking about policy, which has aged rapidly, and we’ll hopefully catch up with the information. But it’s just one way of trying to limit and rationalize the therapeutic.

Transcript Edited for Clarity
Slider Left
Slider Right


Transcript:

Daniel P. Petrylak, MD:
So, David, why is there such a difference in opinion about PD-L1 expression and response to these agents?

David I. Quinn, MBBS, PhD: I think the data are conflicting and we tend to be reductionist in the way we look at things. And we would love a marker where we send it off, it comes back, and says, “Look, this marker links to therapy X. Please follow the dots.” Unfortunately, it’s not as simple as that, and, as Betsy and Dean were discussing earlier, it may be that patients that are in this immunologically challenged state respond both to chemotherapy and to checkpoint inhibition very well. So, I think we have conflicting data. We have sets where we’re looking at presumably the same population of patients where with one of the monoclonal antibody therapies, we get a response that seems to relate to PD-L1 expression. And then, with another one that we would assume is not terribly different in our clinical experience, it’s the reverse, we don’t get an effect. And also, the disease state matters. Whether the patient is eligible for chemotherapy as first line, that seems to be a different set to platinum-pretreated patients. We’re having trouble grappling with what’s a dynamic marker. We know it changes. We just have a snapshot of a particular tumor that’s taken, which may be the day before we want to start therapy or within a week, or it could be from a cystectomy or nephroureterectomy that was done 7 years before, and they’re likely to be very different.

Robert Dreicer, MD, MS: David, I think that last part is very critical. None of the studies that have been reported have really told us what percentage of samples come from what timeline. We’re interpreting things based on, frankly, somewhat incomplete data. So, it’s even harder to speculate about that.

Daniel P. Petrylak, MD: Exactly. In fact, in Arjun’s study in platinum-ineligible patients where he found no correlation between PD-L1 expression and the overall outcome, that perhaps is because that tissue is probably the closest that we would have to the actual treatment. And perhaps you’d like to comment on that because that may be the reason why we didn’t see anything.

Arjun V. Balar, MD: Right. The study you’re referencing is IMvigor-210 cohort 1 and similarly KEYNOTE-052. Both of those studies tested immune checkpoint inhibitor therapy, PD-1 and PD-L1 respectively, in the first-line setting in cisplatin-ineligible patients. And simply, by virtue of the fact that these are first-line patients, you’re going to have younger tissue that went into the biomarker testing. In fact, we looked at that. The median age of the tissue is probably somewhere in the range of about 100 days or so. Whereas, in the second-line studies, it’s much longer, probably on the order of something like 300 days or so. And that certainly can impact the utility of the paraffin-embedded block that the tissue has been sitting in wax for years versus younger tissue, and then, the applicability of doing IHC staining on such old tissue.

Daniel P. Petrylak, MD: In fact, on the phase I trial, there were some patients that had archival tissue from years back that turned out to be negative. Then, we rebiopsied them and we had some positive patients out of that. It’s a very, very difficult issue. Betsy, you want to comment?

Elizabeth R. Plimack, MD, MS: One other maybe simpler explanation for why PD-L1 is less predictive later is that the studies where the discovery set defined the PD-L1 cut point in the test across drugs are all where it performed the best. So, your atezolizumab trial, the frontline KEYNOTE trial, the biomarker looked good, it was developed in that set. And then, when you apply it to subsequent sets, it’s just a law of statistics, right? There’s fallout. And so, maybe that’s a simple explanation for what we’re seeing. For durvalumab, for instance, we’ve only seen the test set. So, I think we really need to see, for instance, how that pans out.

Robert Dreicer, MD, MS: And it’s important to remind people there are multiple PD-L1 assays and none of them have been used in the same studies. They look at different things. We’re a long way away from understanding how to apply that across the disease state.

Elizabeth R. Plimack, MD, MS: But we have to validate a biomarker in order to believe in it.

Dean F. Bajorin, MD: So, we have a dynamic disease over time, we have a dynamic immunological environment over time, and then we have multiple assays with different antibodies with different reads and we expect one answer. I think that that shows we can’t boil it down.

Daniel P. Petrylak, MD: What we really need is an imaging agent or some liquid biopsy that we can monitor on a real-time basis.

Elizabeth R. Plimack, MD, MS: We need something that validates across sets.

Arjun V. Balar, MD: But the challenge is clearly the immune system is not binary. It’s a long continuum, so you’re trying to measure a dynamic thing in a very static environment. There is some correlation with PD-L1 expression. In my opinion, I think that’s the most we’re going to get with PD-L1 expression. We need something better.

Daniel P. Petrylak, MD: Exactly.

David I. Quinn, MBBS, PhD: I think the PD-L1 story, we’re going to hear from the lung cancer people that they’re trying to compare at least 4 of the different antibodies out there. And they do seem to perform differently in different settings. So, we’ll need to move through that and I would suggest that with our current therapeutics, PD-L1 in tumor tissue is not going to be the entire answer, but there may be linkage in one therapeutic, which is really good. That may make a difference to that particular therapy and the biomarker that goes with it.

Dean F. Bajorin, MD: That’s a really good point because in the lung tissue, there’s harmonization on PD-L1 expression on the tumor, but it’s less so on the infiltrating immune cells. It really becomes highly complex.

David I. Quinn, MBBS, PhD: When you do these assessments, you get more questions than you answer because I think the heterogeneity of what’s in those lymphocytes, even if you’re just looking at PD-L1, is fascinating. I don’t know what it means, but we need to try and work it out.

Daniel P. Petrylak, MD: I think as David McConkey said in his talk the other day, we need to get past PD-L1 and start looking at other things, and I think that’s important. But, David, you actually mentioned a case recently from California of a patient who was trying to get one of the checkpoint inhibitors for us, and the insurance company denied it based upon PD-L1 stain. Maybe you can elaborate on that because it’s important that we don’t eliminate patients based upon their staining.

David I. Quinn, MBBS, PhD: In our Los Angeles County Hospital network, almost 2 years ago, we were faced with a problem. We have a capitated network with very limited funding. If we gave everybody checkpoint inhibitors who possibly qualified, we’d probably close the hospital in April of the year and have to get to December. We do get help for what’s predominantly an indigent population from our pharmaceutical partners, but the pharmacy was trying to rationalize use. And, at the time, we had early data, not validated, to suggest and that PD-L1 would be a good marker in renal cell, in bladder, lung, and certainly in melanoma, where I think it is useful. And so, they said we need to have PD-L1 positivity for us to give the checkpoint inhibitor. Now, we’re due to review this and sit down after the 2-year period. I think there will be some changes, but I think they just had to have a response to rationalizing therapy in a dynamic milieu.

As it happens, we are able to get most of our indigent patients who would benefit the opportunity of getting one or other checkpoint inhibitor, and we just have to work at it. And we’re talking about policy, which has aged rapidly, and we’ll hopefully catch up with the information. But it’s just one way of trying to limit and rationalize the therapeutic.

Transcript Edited for Clarity
View Conference Coverage
Online CME Activities
TitleExpiration DateCME Credits
Cancer Summaries and Commentaries™: Update from Chicago: Advances in the Treatment of Breast CancerJul 31, 20181.0
Community Practice Connections™: The Next Generation in Renal Cell Carcinoma Treatment: An Oncology Nursing Essentials WorkshopJul 31, 20181.5
Publication Bottom Border
Border Publication
x