Paul K. Paik, MD:The role of PD-L1 as a biomarker for treatment selection shifted really at the end of 2016. The biomarker itself had been tested in clinical trials and was available, but it really didn’t matter because its use in the second-line setting was, in some ways, immaterial. Patients seem to benefit, irrespective of PD-L1 status, from treatments like nivolumab in the second-line setting and beyond. At the end of 2016, there were a couple of trials that were released. One of them was KEYNOTE-024, which was a randomized phase III trial of pembrolizumab versus dealer’s choice chemotherapy in the first-line setting across histologies. These were lung adenocarcinoma patients and squamous cell lung cancer patients.
And it was at that point that we realized that while imperfect, PD-L1 at least seemed to identify a subgroup of patients who would benefit from pembrolizumab up front more than chemotherapy. And this was in the ≥50% expressing patients, and the data were not subtle. The median PFS was 10 months versus 6 months for chemotherapy. Hazard ratio was something like 0.55 in that situation. There was also further update in overall survival improvement. The response rate was better. And so, it really became quickly, after its presentation in 2016, the standard treatment option for patients with high-expressing PD-L1 tumors, irrespective of subtype.
It’s important to note, as I said before, that PD-L1 immunohistochemistry is an imperfect biomarker. It has found its usefulness in one setting, but it’s problematic for a few different reasons. One reason is that there are a number of different PD-L1 immunohistochemistry tests that are available. Different laboratories across the United States and across the world use different assays. And while we have some assurance by work that has been done by different groups, that there’s a good correlation across most of these, there are issues of intralaboratory variability. There are also issues of intrapatient differences across time and within tumors, where we do know that PD-L1 expression seems to be dynamic insofar as in one section of the tumor may be low and another section may be high.
And in addition to that, again, it works for pembrolizumab in the first-line setting. But for other studiesand we’ll talk about the BIRCH study later on, as well as for nivolumab in the CheckMate-026 study in the first-line setting—it was not actually predictive of good benefit versus chemotherapy. And so, we have been looking for better biomarkers, and there are a couple on the horizon. One of them we know is already FDA approved for pembrolizumab, and that is mismatch repair-deficient tumors. And that, in and of itself, is sort of a stand-in for the second biomarker, which is high tumor mutation burden. The idea behind tumor mutation burden is that some tumors, in the process of becoming cancerous, have a lot of different genetic alterations that end up accumulating.
And we think, in theory, that this opens the door for a lot of different antigens that the immune system can be exposed to and recognize and, in that way, have increased surveillance against cancers. This is born out in the MMR-deficient tumors, this is born out in a subset of colorectal cancers that have mismatch repair deficiencies; Lynch syndrome, for example, where there are lots and lots of different mutations that are present. And it looks like this is beginning to bear out also in lung cancer.
I had mentioned that the CheckMate-026 study, which was the randomized phase III trial of nivolumab versus chemotherapy, was negative by PD-L1 expression. They took a look at ˃5% as a cutoff. They took a look even at ˃50%, and it was still negative. They went back and took a look at tumor mutation burden based on a subset of patients who had tumors that were sequenced, and here a cutoff of ˃10 mutations per megabase seemed to predict for the kind of response difference that we saw actually with pembrolizumab in the first-line setting. So a median of something like 9.8 months versus 6 months, for example. And so, it looks like that has some amount of promise going forward. The issue, of course, is that trials, by and large, have not been constructed around this in some of the trials that we’ll talk about later on, for example. And so, we’re going to have to go back and take a look at whether or not this bears out across other kinds of inhibitors in other kinds of settings.
It’s important to note that while there are other biomarkers that are being investigated, specifically tumor mutation burden, and there are also RNA scores that people are taking a look at that are built in to some of these clinical trials, none of these things are standard of care as of yet. None of these things has, as of yet, been validated prospectively. And so, PD-L1 remains the single biomarker that we have to be able to make a decision upon. And I think it is, at this point, a little bit premature to make clinical decisions for our patients based on some of these exploratory markers. I am hopeful that we will, in a very short period of time, see some prospective data confirming its use. But even then, we will have to be careful about its use in a single trial and as a single drug, and then abstracting that to the field as a whole.
Transcript edited for clarity.