Phase 3 KEYNOTE-355 Overview

Joyce O’Shaughnessy, MD: Let's turn to some new data. I'm going to turn us over to Hope Rugo, who is going to take us through the KEYNOTE-355 study, and also very importantly, how we find these patients who are PD-L1 positive.

Hope S. Rugo, MD, FASCO: Thanks very much, Joyce. This was a very interesting presentation at ASCO [the American Society of Clinical Oncology annual meeting] this year. We've been waiting for additional data after the striking and practice-changing data from IMpassion130. KEYNOTE-355 differed in a couple of ways. One was that patients were included who had developed metastatic disease, 6 months or greater, so greater than 6 months after they had received their last treatment with curative intent chemotherapy. For example, neoadjuvant or adjuvant taxane or anthracycline. The second difference is that in addition to nab-paclitaxel as a chemotherapy partner, patients were also allowed to receive paclitaxel, and gemcitabine and carboplatin.

Those 2 key differences in design are important for us clinically because we do see quite a number of patients who relapse within that 6- to 12-month period where we believe they have relative taxane resistance. I think we've all been interested to find out whether checkpoint inhibitors can be combined with different agents because essentially all of the data with the exception of a small trial with eribulin has been with nab-paclitaxel, or in some cases paclitaxel. If we think about the single-agent data, I think it's clear to us all that we did need more information about combinations with chemotherapy. The monotherapy data showed response rates in the mid-20%, low-20% range for patients in the first-line setting, but in the second- or greater-line setting, response rates were very low, even when PD-L1 positivity was selected for. Like IMpassion130, the KEYNOTE-355 trial was designed to include all-intent-to treat patients.

Patients were randomized 2:1 to receive pembrolizumab or placebo with a chemotherapy partner of choice, but you basically broke down the study like Impassion, into the intent to treat and then the patients who had PD-L1 positivity. Here's where we get into another major difference, which for the clinician is complex in terms of clinical decision-making. There are a number of different antibodies that are commonly used and different mechanisms of scoring to determine PD-L1 positivity. For pembrolizumab, 22C3 is the antibody, compared to SP142 with IMpassion130, and they used a scoring system called CPS [combined positive score].

A score of 1 or greater has been positive, and in the entire population of KEYNOTE-355, 75% of patients had a score of 1 or greater. When the trial had finished accrual, and before the analysis was done and there was any unblinding, they got the results from KEYNOTE-119 with single-agent pembrolizumab and IMpassion130 and thought, “Well we should really analyze the group of patients who have stronger PD-L1 positivity.” They chose a CPS score of 10 or greater.

That was a little under 40%, almost identical to the number of patients who are PD-L1 positive in the immune cells for IMpassion130. When they looked at the patients who were at a CPS score of 10 or greater, they saw a significant difference in progression-free survival [PFS], with an absolute difference of 4.1 months. It went from 5.6 to 9.7 months, and a significant number of patients met their threshold boundary that they set up.

In CPS of 1 or greater, it was still what we would think significant, 0.0012, but their boundary had been 0.00411. They met that for a CPS 10 or greater, but for a score of 1, the boundary was actually point 0.0011 and the P value was .0014. I think for the clinician it's frustrating because we don't totally understand that, but the absolute difference in PFS was clearly shorter at 2 months.

It's interesting, it didn't matter what the chemotherapy partner was either, which is really helpful. When you look at the forest plot, and I'm interested in people's thoughts about this, although we don't have a good mechanism for choosing patients other than PD-L1 positivity, and now we have 2 potential ways of testing that. The interesting thing is if you looked at patients who had not received any prior chemotherapy and those who presented in the de novo setting, it looked in the forest plot like those patients benefited the most, and that's interesting. Presumably you haven't split out the people who already have chemotherapy resistance.

There was no differential toxicity to consider, and I think that what we're really waiting for, and as Joyce explained, we have this 7-month survival difference in IMpassion130, and we don't yet have survival data for KEYNOTE-355. I think these are fascinating data that add to our understanding of how we can use checkpoint inhibitors in the metastatic setting and potentially the early stage setting.

Transcript Edited for Clarity