It’s interesting, there has been a lot of excitement over the last five or six years about immune checkpoint blockade in multiple cancers including kidney cancer. Nivolumab has been approved in the US for several years for patients who have failed VEGF therapy. So there was a lot of interest about trying to bring immune checkpoint blockade from second line into first line and there have been a number of different strategies to try to get that accomplished by essentially combining first line and second line with VEGF and PD-1. There’s a lot of interesting data at this meeting about those approaches and also combining PD-1 with CTLA4 blockade with ipilimumab and nivolumab, which is obviously now approved in the US and was recently granted a step towards approval in Europe.

But little was known about single agent PD-1 blockade so we said let’s try to figure out how active is PD-1 in untreated patients, how much of this benefit that we’re seeing with combinations is driven by PD-1 and it turns out that it’s a fair bit. So last year at this meeting we presented cohort A of the KEYNOTE-427 study which looked at clear cell patients. About 105 patients or so were enrolled on that study with single agent pembrolizumab in untreated patients and the response rate there was 36% which was actually higher than we expected, higher in intermediate and poor risk patients and in PD-L1 high patients the response rate was actually 50%, suggesting maybe you might even be able to select for those patients. So that was impressive and it helps put some context around what we’re seeing with combinations.

This year we’re taking it a step further, looking at patients with non-clear cell kidney cancer. So these are patients who truly don’t have many therapies. In fact, when it comes to immune therapy they’ve been excluded from most of the trials we’ve seen so far, in part because historically they tended to respond less well to cytokines so people assumed they’d be less likely to respond to immune checkpoints. It turns out that there’s some activity there. So when we looked at single agent pembrolizumab in cohort B of this study, same design, same dosage, just different types of patients, we actually looked for non-clear cell and then we confirmed it with an independent review; most of the patients on the trial were papillary but we had chromophobe patients and unclassified patients. When you take everyone together the response rate was 25%: 25% in papillary, lower in chromophobe, 10%, higher in unclassified, 34%, overall 25%. We actually saw a CR rate of 5% which was interesting – we were seeing complete responses in non-clear cell kidney cancer. We took it a step further looking at, once again, PD-L1 in the microenvironment by immunohistochemistry, and there if you were PD-L1 low, less likely to respond, 10% response rate. But PD-L1 high the response rate was 30%, so three times higher. So in some ways the IHC test actually performed better in non-clear cell than in our clear cell patients suggesting even here that it’s probably more the biology, the reaction to the tumour, that drives the benefit of these checkpoints rather than the actual tumour histology.

So interesting, we saw survival – median survival hasn’t been met in this cohort – reasonable tolerability, as seen in many pembrolizumab trials. The percent of patients who came off drug for toxicity was relatively low, only 6%, however there were two treatment related deaths so we obviously have to monitor our patients very closely. But it suggests that for these patients who have very few options that single agent PD-1, in this case with pembrolizumab, might be an effective option for them and we’re going to try to make it more available to them in the future.

How does the efficacy of pembro compare with other PD-1/L1 pathway inhibitors?

It’s a very interesting question. There’s not a lot of head-to-head data so most of what you say is sort of speculative. But essentially the data in the first line single agent experience with pembrolizumab was better than we expected. We need to get more single agent experience with these other agents and obviously head-to-head trials would be nice but we’re stuck making cross-trial comparisons which are fraught with all sorts of potential biases. But, in general, the data with single agent PD-1s is hard to beat when you do a cross-trial comparison. It doesn’t seem like, at least in kidney cancer, that PD-L1 strategies produce much efficacy as PD-1 although that’s a speculative stance going forward. That’s going to be a controversial issue going forward. It’s like which are the preferred combos – combos that are based on PD-1 or PD-L1? There will be a lot of debate around that question going forward.