Joyce A. O'Shaughnessy, MD: I want to come back and get everybody’s take on this. But let’s bring in KEYNOTE-355 before we have a larger conversation. Sarah Hurvitz, could you mention KEYNOTE-355? We saw those data at ASCO [the American Society of Clinical Oncology annual meeting], and could you talk a little bit about testing, like is any assay okay for PD-L1?

Sara A. Hurvitz, MD: The 355 clinical trial was also in the first-line setting, as Adam indicated. It had a little different disease free interval that was allowed. Patients could go on if their disease free interval was at least six months on this trial from treatment in the adjuvant setting, as opposed to IMpassion130 where it was 12 months. And they allowed treatment of physician’s choice based chemotherapy, so either nab-paclitaxel, plain paclitaxel, or gemcitabine-carboplatin. And interestingly, but not surprisingly, about 55 % of patients chose to use gemcitabine-carboplatin, which omits the need to lose one’s hair. They didn’t report how many got straight paclitaxel versus nab-paclitaxel, when I looked through the slides.

Maybe somebody else knows what the percentage was, but at any rate, patients received either the placebo or pembrolizumab, and they determined essentially the triple negative status as well as PD-L1 status. They used this thing called the CPS [combined positive score], which is a combined score looking at the number of PD-L1 positive tumor cells, lymphocytes, and macrophages divided by the number of tumor cells. Then they based their outcomes on this. When they looked at the progression-free survival [PFS], it was a hierarchical statistical design. First they looked at the PFS for CPS of 10 or greater, and the hazard ratio was 0.65, and it met their statistical significance. The median PFS was 9.7 months for pembrolizumab vs 5.6 months.

This was a nice 4-month, almost, improved median PFS. Then they go to the second step, which is looking at the median PFS in patients with a CPS of at least 1. Now, although there’s about a 2-month improvement in PFS, the hazard ratio isn’t as good, it’s 0.74. And while the P value was .0014, it didn’t meet the predefined level of significance. They weren’t allowed to statistically look at the next question, which is the overall intent-to-treat population. They looked at it numerically. There was about a 2-month improvement in median PFS, but they wouldn’t give a significance value because they hadn’t met that step before. We see that there is this benefit with pembrolizumab. We don’t know if it’s dependent on which chemotherapy you choose. That wasn’t didn’t appear to be a factor. Some patients did get paclitaxel, and the benefit seems to be in the patients who are CPS 10 or greater.

Then the question is how do we test our patients and decide what therapy to give them? We’ve got the SP142 assay for PD-L1 in the tumor immune cells if we’re going to choose atezolizumab. Do we use the CPS score with the 22C3? This is just mind blowing. If you’re a general medical oncologist, how does one keep up with this? Well, Hope Rugo, MD, actually did a nice analysis and presented at ESMO [the European Society for Medical Oncology Congress] in 2019, looking at the different assays. And actually the SP142 antibody for PD-L1 expression, if you’re positive for SP142, then the other assays are likely to be positive, but the opposite can’t be said true.

If you’re looking just at the IMpassion130 data, and the analysis she did was somewhere around 600 patients where tumors were looked at by these 3 different antibodies: SP142, 22C3, and the SP263. She showed that the PFS was most strongly associated with the SP142, so it was most associated with benefit from atezolizumab. We don’t know that for other immune checkpoint inhibitors. This field is rapidly evolving and rapidly becoming quite confusing, in my opinion. We are going to continue to need to do tissue analyses to better understand how we need to test to identify patients who will benefit.

Joyce A. O'Shaughnessy, MD: Thank you. That’s fabulous, great state-of-the-art summary, honestly. I want to ask Kevin and Sara Tolaney, given all the data we’ve just heard beautifully summarized, where are we right now in practice? We’ve got survival with IMpassion130, final survival data. Albeit, the intent-to-treat population was not positive for survival in IMpassion130, but the PD-L1 subset was. Negative, atezolizumab and paclitaxel, positive progression-free survival, but not yet survival. We don’t have the data yet for KEYNOTE-355, so we have a gemcitabine-carboplatin opportunity there for patients. How do you use this in practice? I’m also curious how you guys test in your institutions. Do you have multiple antibodies, do you tend to use 1 antibody? Let me start with Kevin.

Kevin M. Kalinsky, MD: A lot of really great questions in there. As Sara Hurvitz had mentioned, the antibodies are not all the same. There are different pre-analytic and analytic considerations between the antibodies. To me, the specific biomarker associates with that specific agent. For SP142, that’s the predictor that I utilize for determination of whether to use atezolizumab. As Adam said, I also think we’re all a bit surprised, disappointed, and baffled by the results of IMpassion131. At this very moment, the approval really just is for atezolizumab in patients with PD-L1 positive tumors, but later in the year we’ll see whether this expands to pembrolizumab based upon KEYNOTE-355.

One of the distinguishing features, as was already mentioned, was the 12 vs 6 months and gives the flexibility of a different chemotherapy regimen. At the end of the day, to go back to one of the things that Adam had said, I wish I had a great explanation of why there is this difference between the 2 IMpassion trials. There has been some hand-waving, could it be the difference of the 2:1 randomization, and a smaller study, and then steroids for Taxol? But ultimately, as Sara Hurvitz previously mentioned, in the neoadjuvant setting patients got Taxol and they received steroids. We saw that that was a positive study, so I think that there is still much to learn.

I wonder, in terms of the differences between those 2 studies, whether there is just, by chance, a difference in terms of the biology of the tumors in those patients who were enrolled. There seems to be this subset of patients who are really driving a lot of the benefit. Is it possible that there were more immunomodulatory tumors in IMpassion130 compared to IMpassion131? Is there a difference in terms of the ratio of the subset of patients who are really long-term benefiting with those patients with durable responses? I don’t think we quite know a great explanation for the data that we have right now.

Joyce A. O'Shaughnessy, MD: It sounds like in your practice, the atezolizumab, nab-paclitaxel is the standard of care for the SP142. Is that what you guys use, Kevin?

Kevin M. Kalinsky, MD: Yes, that’s exactly right. For those patients who don’t have a recurrence within 12 months of their operable therapy, it’s to use SP142 and to use nab-paclitaxel along with atezolizumab, with the consideration of neuropathy and things that would concern me about using a taxane.

Joyce A. O'Shaughnessy, MD: Would do ever consider the gemcitabine-carboplatin and pembrolizumab for people who recur more quickly or something, or who decline alopecia-induced chemotherapy?

Kevin M. Kalinsky, MD: If this becomes an approved regimen, for sure. That’s one of the benefits of that study, the lower period from time to reoccurrence, as well as giving drugs that have different mechanisms of action, and patients may not have developed chemotherapy resistance.

Joyce A. O'Shaughnessy, MD: Thank you.

Transcript Edited for Clarity