EADO Congress 2015

What is the relationship between PD-L1 expression and antitumor activity?

Dr Paolo Ascierto - National Tumor Institute, Naples, Italy

We’ve been hearing, here at this melanoma meeting, a lot about biomarkers and there is a big need, if you’re treating patients with advanced melanoma, to know which of the drugs will work because these amazing drugs do work but not always. PD-L1 is such a biomarker, can you tell me what you’ve been doing in this particular poster that we’ve been quickly looking through? What did you do in this study?

As you say, the biomarker is an important need, especially for immunotherapy because it’s a high cost treatment and we need to select the patients that can have a benefit. In particular, in the KEYNOTE-002 trial, this is a trial with pembrolizumab in patients who were resistant to ipilimumab treatment, it was retrospectively analysed the impact of the PD-L1 status in the outcome of patients. You know that there is a great discussion about the PD-L1 as a biomarker.

And it has been found to be quite useful, hasn’t it, in other studies?

Yes, there is a great discussion about this. But my feeling, in this study it was found that also the PD-L1 negative patients can have a benefit and this is consistent with all the other findings.

So, tell me what are the data that came out of this because you’ve got a counterintuitive finding in this poster.

Yes, so the data clearly indicated that if a patient is PD-L1 positive the outcome is better but also in the group of patients PD-L1 negative we have some benefit for these patients. For this reason we cannot avoid the treatment to this group of patients.

Now pembrolizumab is, in fact, targeting PD1 so the PD-L1 should be something that tells you whether it’s going to work or not. How do you explain the fact that both groups of patients did well?

First of all it should be clear that the PD-L1 marker is not a molecular marker but an immunological marker. This means that it’s a dynamic marker so can be upregulated or downregulated at either factor. This is the first point, the second point is that, yes, it’s true that the PD1 PD-L1 pathway works more in the tumour microenvironment and this is responsible for the high response rates for this kind of compound in respect to the CTLA-4 pathway. But the PD1 PD-L1 pathway can have also an impact during the priming phase so like CTLA-4. So this means that in the tumour microenvironment the impact in the ligand from the immune resistant cell in the tumour cell is more important; in the priming phase there is another effect, different, that can have an impact just in terms of overall survival more than response rate and this may be the reason because we have benefit also in the negative patients.

In practical medicine, then, if you’re thinking of using an agent like pembrolizumab or nivolumab targeting PD1, should you take note of PD-L1 or not?

Not, from my point of view not. But some interesting data also from the nivolumab and in terms of overall survival the CheckMate-066 clearly indicated that the PD-L1 negative patients had a one year overall survival rate of 68%. Remember historically it’s 25.5% over the era pre-new compound and with ipilimumab we had 44-46%, so 68% is close to the 73-75% also targeted agent so we cannot avoid to treat these patients with PD1.

So the targeted agents are working, what are the lessons coming out of this, finally, in summary, would you say, for doctors?

I think that the PD-L1 is not useful for predicting the patients who can respond with PD1. We need other kinds of biomarkers probably the immuno-filtrate, the immuno-profiling. You know that from other kinds of cancer like colorectal cancer and also the data from the KEYNOTE-001 from pembrolizumab showed that some signatures, and in particular in KEYNOTE-001 the interferon gamma signature, can have a role probably more important than PD-L1. I think that we should forget about PD-L1 as a biomarker for predicting outcome. We should go to the way of the immuno-profiling and then the gene signature.

And is the hunt for biomarkers promising at the moment or are we still a bit worried about this?

If we forget about PD-L1 my answer is yes because all the data that was showed this year at ASCO and ESMO clearly indicated that the signature, as I said, in the profiling is the best way for selecting patients.