Paul K. Paik, MD:One of the questions that we’ve had, because angiogenesis inhibition is a standard-of-care treatment option in our patients with nonsmall cell lung cancer in the form of ramucirumab and bevacizumab, is whether or not there might be some synergistic benefit to giving angiogenesis inhibitors along with immune checkpoint inhibitors. There is some preclinical or scientific rationale in doing this. We do know, based on existing literature, that angiogenesis inhibition, or the VEGFR signaling pathway, does seem to have immunomodulatory effects. We do know that this can alter the immune microenvironment in a couple of different ways; one from the most straightforward way in terms of immune cell penetration by impeding blood vessel formation, for example. The other has to do with some interesting data with regard to particular T-cell subsets; for example, the activation of these pathways and recruitment of regulatory T cells that clamp down on the immune response.
So, there is some preclinical rationale for this, and because of that, there have been a number of trials that have launched taking a look at different angiogenesis inhibitors plus immuno-oncology drugs, either with chemotherapy or without chemotherapy. The one that has read out with positive results, as of the end of last year, was IMpower150. This was actually a 3-arm trial with 3 different regimens given in our newly diagnosed lung adenocarcinoma patients who have metastatic disease.
The first arm was carboplatin/paclitaxel/atezolizumab. That arm has not yet read out. We’ll see what those data look like later on, probably sometime later this year. The control arm for the results presented so far was the backbone chemotherapy arm: carboplatin/paclitaxel/bevacizumab followed by maintenance bevacizumab. And then the experimental arm was carboplatin/paclitaxel/bevacizumab/atezolizumab, or quadruplet therapy, which we never had before up until the IMpower150 trial. The results were positive.
The trial is somewhat complex because there were a number of different things that the investigators were looking at. They were looking at the intention-to-treat, or overall population, for the 2 subgroups, but they were also looking at some exploratory biomarkers. One of them is the standard PD-L1 expression biomarker based on the TCIC score, and the other was this new biomarker that has not been presented before, this T-effector panel. The T-effector panel was a small panel taking a look at 3 different biomarkers combined in 1 score. One of them was PD-L1, which we’re all used to. Another was interferon-gamma, which we know is important in terms of priming the immune system. And the third was a chemokine, CXCL9, and this score was read out as high or low. So, it was a stratifying test that was binary, which is somewhat different than the PD-L1 expression that we know is a continuous variable with cutoffs that we have created at different points.
Again, the intention-to-treat results were positive. In terms of the primary endpoint of median PFS, the results were something like a median PFS of 8.9 months versus 6.8 months. So, the standard control arm had a PFS of 6.8 months. Again, the quadruplet therapy had a median PFS of around 8 1/2 months. Now, it’s important to mention this because while those results are positive, and so definitely we see in the intention-to-treat population activity when you add atezolizumab, that number is still not any better than, for example, pembrolizumab up front in the high PD-L1 expressers where the median PFS is 10 months.
So, it’s important to keep that existing still relatively new data in the back of our mind in terms of contextualizing the results of the IMpower150 study. So, again, they took a look at other subgroups in terms of PD-L1 expressers and high T-effector expressers, and here the biomarker results were positive. In terms of the T-effector score, it did predict, or identify, a population that benefitted even more. Here the median PFS was somewhere around 11.8 months or 11 1/2 months in this subgroup, which was substantially higher than in the intention-to-treat population and much higher than in the control arm, which was still about 6.8 months. PD-L1high patients, this TC3/IC3 subgroup of patients, had a similar benefitactually a little bit more. The median PFS in this group of patients was actually 12.6 months, so even then a little bit higher than the T-effector score, in terms of identifying a subgroup of patients who seem to benefit more.
And, again, I had mentioned the KEYNOTE-024 results as a baseline, a median of 10 months, which is sort of what we think of as the results for single-agent immunotherapy in this arena. And these data are relatively promising in comparison to that. Again, cross-trial comparisons make it very difficult to figure out whether or not any of this stuff is significantly different. And it’s important to note also that pembrolizumab is a single drug, and, of course, the IMpower regimen is quadruplet therapy, so there are differences with regard to that. Although we really did not see too much in the way of concerning additional toxicity signals with the quadruplet regimen.
And then finally, overall survival is something that we need to really take a look at to see whether or not it’s mature. Those data, in terms of the mature data, have not been read out yet but are a concern in terms of consuming lots of different agents in the first-line setting, whether or not giving it all up front is really going to be better than sequencing our regimens. This is an important consideration for our patients, and we’re going to have to really carefully consider and try to tease apart as we go forward.
The final thing I wanted to mention structurally about the trial is that it’s always important to take a look at how the control arm performed. Again, the control arm here was carboplatin/paclitaxel/bevacizumab, a regimen that we’re all very familiar with. And the control arm actually performed substantially better than what we’re used to. The overall response rate was 48%. The median PFS was 6.8 months. These are numbers that are better than, for example, the randomized phase III ECOG trial, that was read out in the middle of the 2000s, where the response rate really was no better than around 40%, somewhere in the 30% range.
Now, I mention this because the control arm is what we use to contextualize the experimental arm. So, for example, the experimental arm in the intention-to-treat population had an overall response rate of 64%. When you see that in isolation, you think, “Wow, that’s phenomenal for a first-line regimen. We really haven’t seen that before.” But in terms of the proportional increase in response, it’s actually not sort of as high as one might hope for. And the best way to see this is that if you took the control arm response rate to 30%, then the proportional increase in response that you would see with the quadruplet therapy would be 40%. And 40% doesn’t sound that much greater than 30%. So, we have to be careful about the control arms and how they perform because the concern would be that for whatever reason, reasons that we can’t control, the groups of patients who were enrolled in these studies might be a different subset of patients who maybe would have performed better, irrespective whatever trial they went into.
Transcript edited for clarity.