We’re showing biomarker data at this AACR meeting 2018 for head and neck cancer patients treated with a PD-1 inhibitor, treated with pembrolizumab. These are data from the KEYNOTE-012 and KEYNOTE-055 studies. So it’s a relatively large cohort, about 250 patients treated with a PD-1 inhibitor. We’re looking at three biomarkers, all of which have potential clinical relevance. We’re looking as especially tumour mutational burden and then an inflammation signature, generally abbreviated as GEP or Gene Expression Profile. I don’t think we have any PD-L1 data, that’s already public, but it’s the largest dataset for head and neck cancer looking at tumour mutational burden and the inflammation signature.
So what does the data from these trials show so far?
KEYNOTE-012 and KEYNOTE-055 are studies that show efficacy of pembrolizumab for head and neck cancer, in fact KEYNOTE-012 led to approval of pembrolizumab for head and neck cancer in the United States. These are now the biomarker data looking at the tissue samples, trying to see if there’s maybe a way to better predict which patients benefit. The problem, obviously, is that only about 16-20% of patients benefit from pembrolizumab and this dataset looks at RNA, looking at inflammation genes, gamma interferon etc., trying to predict which patients benefit on inflamed tumours and then also looks at the number of mutations. Again, the suggestion is that maybe patients who have higher mutational burden are more likely to have benefit from immunotherapy, they might have more antigens and both of these actually work quite well.
What does this mean for the patient when it comes to treatment choices?
Right now I’m not sure that for head and neck cancer it’s quite yet ready for prime time but I think we’re getting close. In fact, at this AACR meeting we’re starting to see data for tumour mutational burden in lung cancer that maybe we can predict which patients have more benefit. Our data is interesting in the sense that all three biomarkers work well and can actually enrich for patients who have benefit. Neither of them are perfect in the sense that you cannot for sure say you have a lot of mutations, you will benefit or not benefit, but they all enrich for benefit and they’re actually fairly good at enrichment.
What is interesting is that they are not necessarily perfectly correlated. So inflammation, GEP and PD-L1 have some level of correlation, they measure in the tumour presence of inflammation, presence of immune cells. By contrast, tumour mutation burden is not well correlated, it measures something very different. Both GEP and PD-L1 correlate with benefit in respect to progression free survival and overall survival; tumour mutational burden only seems to correlate with benefit from progression free survival and not from overall survival and I’m not sure that I fully understand why that’s the case. So they’re different in their flavour. What we don’t quite yet know is how do we put them together, is there a way to actually make a composite biomarker. But all three are potentially ways to select patients for studies and we’re already seeing this. Some studies, like in lung cancer, enrich for PD-L1, enrich for tumour mutational burden and I think that’s what we’ll see for head and neck cancer as well and this is the dataset that provides the basis for that.
Does lead on to the next steps for this research?
I think we will see data forthcoming in the first line setting for head and neck cancer. There are three large studies in the first line setting right now, there is KEYNOTE-048, there is the KESTREL study and there is CHECKMATE-651, so combinations of a PD-1 or PD-L1 agent with CTLA-4 or combinations of a PD-1 agent with chemotherapy. And also single agent first line data. So these studies will likely employ biomarkers such as PD-L1 or maybe tumour mutational burden, we’ll have to see what actually comes out, to select the right patients for those treatments. So I think within a year or maybe two years we’ll actually get those data and these biomarkers will become clinically relevant I would predict.
Anything you’d like to add?
These biomarkers are also important for the future in the sense that not only are PD-1, PD-L1 agents clinically meaningful but we’re also starting to think about combination strategies. I believe these biomarkers will be key in identifying which combinations we should use for which patients. Maybe tumour mutational burden is particularly good for CTLA-4 combinations, at least the lung cancer data that we’ll see here at AACR may suggest it. I haven’t seen the details but it certainly looks like this. I would predict that there is a good chance that tumour mutational burden will find a role for CTLA-4 more commonly, I think there’s a chance. What about other combinations? We just heard that the ECHO-301 study was negative, that’s an IDO combination study. We haven’t looked at these biomarkers – maybe GEP will be helpful there; IDO is certainly one of the genes in the GEP signature. So I believe that these biomarkers will be helpful not just in enriching patients for the first line setting but I think that they will also be helpful in identifying the right patients for which combination and will lead to more personalising immunotherapy – which treatment do you get if you have which immunophenotype.