Skip to main content
main-content

Medicine Matters oncology

Hello and welcome. We are going to be talking about the use of PD-1, PD-L1 checkpoint inhibitors in the perioperative setting of non-small-cell lung cancer. I'm Professor Heather Wakelee, Chief of the Division of Medical Oncology at Stanford University, Deputy Director of the Stanford Cancer Institute, and President of the International Association for the Study of Lung Cancer. And I am joined today by Professor Nasser Altorki. Dr. Altorki, do you want to introduce yourself?


Yes. So I am Nasser Altorki. I'm the Chief of Thoracic Surgery at Weill Cornell Medicine in New York. And I'm also the leader of the experimental therapeutics program at the Meyer Cancer Center at the same place. And I'm happy to be here with Dr. Wakelee to discuss such a timely topic.


Great. So we're going to be having the medical oncologists and the surgical oncologist perspective as we talk through an amazing amount of data that's been coming out recently regarding the use of checkpoint inhibitors in early-stage lung cancer. And we know a lot about the use of checkpoint inhibitors in the metastatic non-small-cell lung cancer setting. And in that world, we know that PD-L1 levels are really important, such that if a patient has high PD-L1 with a greater than 50% level, we often will think about just giving single agent checkpoint inhibitors. And if it's lower, we're going to be looking at that usually in combination with chemotherapy. And of course, that's in the setting where there is no driver mutation, which is a whole separate topic. And we'll kind of try to cover that a little bit today, as well. I think it's a bit more complicated when we're talking about the perioperative setting. But a question for you as a surgeon, when you have a newly diagnosed patient, when do you think about looking at PD-L1 level and probably also the driver mutations, as well?


So as you know before the last couple of years, we didn't really look at those things. Because we looked at them from we wanted to keep a running database of what the mutational profile is and all that, but really had no immediate use for them in dictating the management of the patient. But in the last 3 years it's just been amazing. First, the ADAURA Trial. So now we actually have to get EGFR on those patients, not necessarily on the diagnostic biopsy but at the very least on the resection material. And the data on IMpower010, it just changed everything. And I can see that on KEYNOTE-091, which is the [pembrolizumab] study, it's almost going to seal the deal on all that. And now we actually had reached out to our pathologists saying that we would like that to be a reflex testing on all our-- The question really is, we know we need to get EGFR at any time point during the continuum of patient care before or after resection, question is PD-L1, when do we get?


Right. And I think one of the things that can be confusing too is there are several different PD-L1 tests. And you mentioned a couple of the big trials. So we've got now three randomized phase III trials with the neoadjuvant, CheckMate 816 with neoadjuvant nivolumab added to chemo, and then the IMpower010 with single agent adjuvant atezolizumab, and the KEYNOTE-091 with single agent adjuvant pembrolizumab. These are, of course, many neoadjuvant trials that were not randomized with positive data. And they're all using slightly different PD-L1 assays, which makes it a little bit confusing. You and I were both very involved in the IMpower010 study where we started using that SP142 assay, which looks at both the tumor and immune cells for PD-L1. And then, before the trial was analyzed decided instead of using that assay to switch to the SP263, which only looks at the tumor.


And what's the PD-L1 level on the tumor? And I think it's important for the audience to understand that a little bit further, where IASLC had done the Blueprint Project, and there were others looking at all these different PD-L1 assays, the 22C3, 263, others, and really everything clustered together with very, very close alignment except the SP142, probably because of that looking at the immune cells versus just the tumor. And so that's why with the IMpower trial we looked at the SP263, which we know correlates really, really closely with the 22C3 assay used for pembrolizumab without issues for nivolumab. And so I'm wondering, do you have an institutional standard? Or do you use different PD-L1 assays when the medical oncologists are looking at the different checkpoint inhibitors?


Well, we were lucky to have one of our Weill pathologists to have the part of the Blueprint project. And we actually use SP263. Have used it all along. And it's good to see that all the other three antibodies line up well with respect to the expression of PD-L1 in the tumor cells. And that may be a better guide to treatment decisions. I'm not sure from a research point of view that it's the best way to look at PD-L1, because obviously there's a lot of PD-L1 in the tumor microenvironment. But that's a research discussion. But in terms of guiding therapy, I think that's emerging as the best way to do it. So we use SP263.


Yeah. That's what we tend to use as well. But then if I have a patient in metastatic setting, or I'm thinking about pembrolizumab, if I have a 263 assay that's high, I kind of look at that as equivalent to 22C3 result. So I tend to believe that Blueprint data is essentially.


In the grading of what percent?


That's a great question. I think for metastatic, I use the 50% and higher, that is when I'm more comfortable using single agent. When we talk about the perioperative setting, it gets kind of confusing. And so maybe we can review the data little bit. Maybe we can first talk about nivolumab and the CheckMate 816.


But before that the assessment of positivity is made by the pathologists. So the question is, how much of the inter-observer variability there is. And can you get somebody who will tell you, well, this is 40%, then the other guy says, no way, it's 60% or 50%, which has important implications in the metastatic settings, I suppose.


I think so. And that also gets at how many cells are being looked at. Because this is looking at the percentage of cells that are positive. And so I definitely look at it as an assay with a lot of wiggle room. So I don't I don't think 45% versus 55% are different. I would hope that 0 is meaningful. But of course, if you look at enough cells, and you see maybe 1%, and we make a lot of distinction between the 0% and 1% in many of these tests. But that's a really important aspect of it. It's different, I think, in lung cancer we're so used to thinking about the driver mutations, where it truly is it's there or it's not there. It's like a key aspect of the tumor. And with PD-L1, it's this variable that it's a gray. It's a scale. And so that's a really important point.


And we also worry, and I'm sure you do too, about sampling errors.


Yes.


I mean forget the fact that it's dynamic and changes with therapy, but the heterogeneity is well known in PD-L1 expression. So that's another troubling thing.


Right. And I think that's so relevant when we're talking about neoadjuvant treatment, where you've only just got a small sample with a biopsy. But perhaps it doesn't matter as much when we look at the CheckMate 816 data where that PD-L1 level wasn't as important. And do you want to talk through that data a little bit?


Yeah. First of all, it's an amazing trial. I just heard the follow up of it yesterday. And it just completely blew me away. So as you know, it's a randomized study of patients with stage IB, II and IIIA were randomized to either neoadjuvant chemotherapy, which many of us consider as a standard, at least, for patients with stage IIIA, versus chemo-nivo And there was a significant difference in complete path response rate, 24% in the combination arm versus 2%. And a couple of days ago they presented the event-free survival, which showed almost a 15 point difference at 2 years and a difference in the median survival, which was highly statistically significant. So that's the big headline news of this.


And obviously, it's been approved as a standard of care. So there is no going back. But then, the catch is that we show the unplanned, the key variable analysis. And all favor the nivo group. However, then you start breaking it up. And you look at a response based on PD-L1 expression 1% to 49%, more than 50%, and the per stage benefit. And then things start to not look so clear. The IBs and IIs don't look so much better like the stage IIIs. The 50% look a lot different from the 1% to 49%. But I try to take a deep breath and remind myself well that these are the nuances of clinical trials.


Yeah.


Yeah, so the big update at the AACR in 2022-- and we've seen that. And so to me, it's interesting. I mean, the study was two thirds stage IIIA. And I think it's important that people remember that it's mostly a IIIA patient population. And just as you said, even though the results are sort of for all-comers when you start looking at the nuances, seems like more benefit later stage. Makes sense to me, right?


You've got more opportunity for benefit-- and then more benefit with higher PD-L1. Even without PD-L1 expression, you still have benefit, which kind of makes sense when you think about metastatic disease, where in the metastatic world, if you have chemo plus checkpoint inhibitor, it works regardless of PD-L1 expression. But it works more when there's more PD-L1 expression for the most part, right?


So it follows that theme to me. And so we've been working on trials in the neoadjuvant setting. So we've had a lot of patients enrolled from that standpoint. But I haven't had anyone show up since the neoadjuvant nivolumab was approved. So I haven't had to face that question.


But I think for a patient with the stage IIIA, where our surgeons are saying resectable, but it sure would be nice if it shrunk a little first, I would think that chemo plus nivolumab would be a very reasonable option. And more so if they have the higher PD-L1 where I think that benefit is more likely.


There's been some questions though that have been raised. And it didn't really pan out in CheckMate 816. But I've heard a lot of surgeons talking about concerns that if you do neoadjuvant checkpoint inhibitor, it makes the surgery more difficult. Have you found that at all? I mean, what is your perspective as a surgeon?


So from my point of view, it depends where the bulk of the nodal disease is because that's really where most of the action is. None of us go and cut the tumor out of the lung. So we're always operating in lung parenchyma that's reasonably normal.


But if you have heavy nodal involvement in the hilum for example, and you have a good immune response against the tumor, that generates a fibroblastic response within the substance of the nodes which makes it really hard to dissect.


Having said that, we've done a lot of these cases in the different settings. As you know, we had had this trial with low dose radiation plus anti-PD-L1 and have done almost 60% of these cases minimally invasive. So yeah, a little bit more difficult but certainly doable and certainly did not change the complication rate or the length of hospital stay. And in the end, it's worth it for the patient, I think.


Well, let's talk a bit more about adjuvant too, because I think we've got the metastatic disease setting where higher PD-L1, more benefit. When you add with chemo, less relevant but still in that direction. With the CheckMate 816, with neoadjuvant chemo plus nivolumab, the principles all fit together. And then we have these two adjuvant trials.


And one sort of follows that theme that makes sense where higher PD-L1, more level, and one is completely opposite and very confusing, right? So with the IMpower010, of course, you and I know that trial exceedingly well. So in the un-randomized, just over 1000 patients, all of whom have had complete resection, they've all had adjuvant chemotherapy and then randomized to get atezolizumab or best supportive care.


The trial did allow for patients who had EGFR mutations in their tumor or ALK translocations. The CheckMate 816 with neoadjuvant nivolumab excluded those patients. So here we are, IMpower, randomized 1000 patients. And the study had about 40% stage IIIA versus over two thirds in the neoadjuvant study. Majority, stage II, maybe 10% stage I.


And this trial showed a highly significant, disease-free survival benefit, hazard ratio of 0.66 in the group that was stage II to IIIA, so excluding the IBs, and with tumors with PD-L1 expression. Now, what's different-- and this was, again, using that SP263 assay. So in this trial, the patients who had no PD-L1 expression had no benefit. So it was only those that had some PD-L1 expression. And that, again, makes sense.


We're giving the checkpoint inhibitor by itself after we finished all the chemo as opposed to in combination. So no PD-L1 expression, no benefit. PD-L1 expression 1% to 49%, maybe benefit, greater than 50% PD-L1 expression, big benefit, disease-free survival hazard ratio of 0.43, so a really, really good number there. And the FDA approved atezolizumab in the adjuvant setting for that stage II to IIIA population if there was some PD-L1 expression.


And a lot of people are talking about preferring it when there's high PD-L1 and not as sure. I think it's important to also highlight that the disease-free survival hazard ratio was 0.79 when you looked at all-comers regardless of PD-L1 expression with stage II to IIIA. But then there's this clear separation where that benefit is really in the high patients-- high PD-L1 expressing tumors. And we'll come back to the EGFR because I think that gets very complicated.


But then when we look at the KEYNOTE-091, which was just presented in an ESMO plenary recently, there the disease-free survival hazard ratio is 0.76 for all-comers. And when you look at, well, what were the stages of the disease? Pretty similar to the IMpower study. When you look at how many patients had PD-L1 expression, just over 50% in both of them, so super similar. That hazard ratio, 0.76 versus 0.79 for disease-free survival, pretty similar.


But then when you look at the greater than 50% PD-L1 expressing tumors in patients with that, in the keynote study, they didn't have any benefit. And so it doesn't make any sense, right? Here, IMpower010--


And it was a co-primary endpoint.


Yeah.


It was powered for that.


Right. So I'm totally confused because here we've got all the other trials, more PD-L1, more benefit. KEYNOTE, it didn't matter, even though in the metastatic setting, that PD-L1 level is so important with pembrolizumab. So I'm still scratching my head. I don't know if you have thoughts about that.


Totally. My thoughts are I'm confused too. I mean, if this was a key secondary endpoint, I'd say, well, maybe they didn't have enough patients. I mean, to be fair, they haven't published the results yet. And we don't know a lot of the details. But that is just like a knock on the head because that's where you expect them. This is consistent.


We're expected to sort of disregard all the data, not only from IMpower010 but also from the metastatic setting. I mean, it's an article of faith that if you have 50% or more PD-L1, you're going to get benefit from immunotherapy. So I am really looking forward to the final data presentation and the details of the manuscript. I think they used the 22--


Yeah, the 22C3 assay, which should be the same as the 263, so--


Yeah, but is it? I mean, you know, now--


In theory it was similar, yeah.


Now we're back to the drawing board. But yes, the benefit in patients with less than 1% is another thing-- that don't express PD-L1. I don't recall. Did they get approval or not, based on the data?


It's not approved yet. But I mean, we just heard this data, like, last month-- so very early in 2022. So I mean, the expectation will be that there will be. But we haven't seen that approval yet.


Because if they get approved, they will be approved for the less than 1% as well because that's where the data are strongest.


But is that really the best thing, right? I mean, to me, it doesn't make sense because-- so if a disease-free survival hazard ratio is 0.76, you're still not helping the majority of people. And so being able to figure out who we're helping and who we're not is so important because in all of the trials with the checkpoint inhibitors, there are patients who develop immune toxicities. And those immune adverse events can be significant. And they can be forever toxicities.


So I think we need to be able to figure out who we're helping and who we're not. They did look at by stage. And with the CheckMate 816 and the IMpower study. In both cases, the higher stage, more benefit. That's especially striking in the IMpower study.


But that's the other part that didn't make sense about the KEYNOTE-091 is the stage IIIA patients were not the ones that were most likely to benefit. And that also has me kind of baffled because it just makes more sense that if you have stage III, your likelihood of recurrence is higher. You're more likely to have still residual disease and therefore more opportunity to benefit.


You don't recall the cell type distribution non-squamous versus squamous in 091, do you?


I think it was--


Most of the benefit is in non-squamous, I think.


Right. And I think the studies were, again, similar. And both were about a third squam and about two thirds non-squam. Histology didn't matter as much. But one other thing that did matter were the driver mutations, right? And so there were patients who had EGFR mutations in their tumors in both the big adjuvant trials, some ALK as well. Though it mostly focused on EGFR.


And in both of those trials when you look at patients who had tumors with EGFR mutations, they actually seem to benefit from the adjuvant-- strikingly so in the KEYNOTE study, questionably so in the IMpower study. So what do you think of that and especially in regarding when we've got ADAURA to deal with as well?


Well, so first of all, again, it goes against everything we've learned about how these patients are treated with immunotherapy. So far, I can tell you in our trial, none of the patients-- well, one out of 12 patients responded in our trial. I know that CheckMate 816 excluded those patients from enrollment in the trial. And obviously in the metastatic setting, they're not eligible for immunotherapy.


So it is kind of confusing, what you do with that. I don't think that-- in the early-stage setting, none of us would actually forego adjuvant TKIs for these patients in favor of immunotherapy. But it is food for thought in terms of how we try and develop IO in that cohort of patients. So that's something that will probably have to be worked out in the lab first so that we can better understand what is it about those patients that sort of drove the response more so than others.


But yeah, I know. I heard you say that about IMpower010 and I thought, hmm, I wonder, is she OK?


Yeah, I think--


And then I saw it in 091 and I said, wow, she's so smart.


Well, just looking at the data trying to make sense of it. I mean, when I have a patient who comes to see me who has had resected early-stage lung cancer and there is an EGFR driver mutation, I'm still going to choose osimertinib at this point, based on the ADAURA data where the disease-free survival hazard ratios there are less than 0.2. So that's still much better.


I think the question though is with the TKIs, they work fabulously while you're giving them. But if you stop it, do you get long-term benefit? Is it really going to change overall survival? We of course don't know that yet either with the checkpoint inhibitors.


However, we have the-- for stage III where they've had chemoradiation-- the PACIFIC trial, there we have disease-free and overall survival benefit. So that benefit persists. And we're seeing those hints on the overall survival curves from the trial. So I'm hopeful that we really are having long term benefit with the checkpoint inhibitors.


And I think in CheckMate 816 they presented some data that would suggest that the overall survival is also better.


Right, it makes sense.


Yes, totally, totally. The issue is as the new ADAURA matures and gets reported, let's hypothetically-- let's see if it doesn't work, if the new ADAURA doesn't work. Then what do we do with those patients if we want to give them new adjuvant therapy? Do they gravitate to the old, just give them chemo? And then?


So lots of questions to be thinking about there. That's for sure. So I think that also gets me thinking through the fact that we have so many trials that are still ongoing or are waiting for the results, right? So we've got a couple more adjuvant trials that were just with single agent checkpoint inhibitors. We've got the studies that are chemo plus checkpoint inhibitors in the adjuvant setting, where we're just getting started with those.


And then we've got all the other neoadjuvant trials, which, different than CheckMate 816, all the other neoadjuvant trials also have adjuvant therapy. So it's a lot of exciting things coming forward. And I think I just want to, before we wrap up, just ask you, do you have any particular thoughts about those upcoming trials or just more of a, let's wait and see because it doesn't fully make sense yet?


Well, I thought everything made sense in the KEYNOTE-091. Before that, I thought, OK, IMpower-- all the adjuvant trials will show the same kind of benefit that IMpower010 showed, maybe a little bigger and maybe a little bit less, but whatever. And all the neoadjuvant trials will fall in line with CheckMate 816. Now the excitement is in the air because I really don't know anymore.


Well, what I wanted-- I wanted to ask you two questions. The first one, your thoughts on dual immunotherapy. Where are we with that in the early stage, or in fact, in any stage?


So great question. I mean, with CheckMate 816, of course they did have a nivolumab-ipilimumab part to the trial, which was not-- the early data from that did not look any better than nivolumab alone and with higher toxicity. So that was not pursued. I don't believe any of the other perioperative studies right now include the dual combination.


I do think CTLA-4, PD-1 inhibition can play a role in some patients in the metastatic setting. It's not something I use very often though. And I think we're going to need to better understand where that extra potential toxicity is really justified. And so far, I wouldn't say that it's there in the perioperative setting.


I have the same view on that. And then the next question is we had a big tumor board the other day-- went on for an hour and a half trying to decide how do we now treat patients given IMpower010 and the CheckMate 816. Who do we choose for what? So I told them that you and I will decide that today.


OK, I don't think we have an answer for it yet, though. I think tumor board is key. So I think we really need to be continuing to meet and talk about any stage III patients, especially to make those decisions about neoadjuvant. I'm not really doing neoadjuvant yet or proposing it much for stage II. But for the stage III patients, I think chemo plus nivolumab is absolutely a standard approach to be considered.


And then for patients who have already had surgery and we're trying to figure out what to do afterwards, I think it's really important to look at PD-L1 levels now and to also look at the driver mutations. It is important to note that certain drivers like ALK, ROS, can have high PD-L1, and that's meaningless, should not give checkpoint inhibitors in that setting. We saw that pretty clearly in IMpower010.


But if there's EGFR, we've got osimertinib. If there's not, we've got adjuvant atezolizumab, especially when there's high PD-L1. And I think that the questions about, should we be using adjuvant pembrolizumab in all-comers just with certain PD-L1 levels and what are all those other trials going to show? That to me is sort of where we are at the moment.


No, I agree. I think as I watch the final installment of CheckMate 816, my view is that for patients with stage IIIA for whom we ordinarily give neoadjuvant therapy, the answer is clear. We give a neoadjuvant plus nivo depending what the other-- but I'm not so sure anymore about the big IBs and the IIs. And I'm trying to modulate how I treat them based on the pretreatment PD-L1 expression in the biopsy.


So if I thought, for example, you had a stage II with a negative PD-L1, I might be more inclined if I was sure that it's stage II to give them neoadjuvant therapy. But if it's PD-L1 positive, I might just sort of think that the patient can be resected and treated with adjuvant therapy all the same.


Great.


So who knows what? I mean, ideally we would have a trial to answer the question.


Right. Well, I think your algorithm for now makes sense. I think we'll be learning a lot too with ctDNA and other markers about who--


Oh, yes, definitely.


--needs that therapy in the future. And it's been great talking with you about all this.


Same here. And hopefully our audience learned something from both of us. So great talking to you.