I talked about tools to try and work out how we might make decisions about immunotherapy. The problem we have at the moment is that we treat lots of patients and only a fraction of them benefit. We make more patients unwell than we need to who don’t benefit and so the real puzzle for us is whether you can work out who will have side effects on the one hand and who will have benefit on the other hand. If it was possible to work that out before we start then we would not only save toxicity and, of course, drug cost but also spare patients unnecessary hospital attendance. Furthermore, we could make smarter decisions about what to do alternatively. So the second part of what I talked about was the attempt to understand the tumour microenvironment using high resolution sequencing tools, RNA sequencing specifically, and by understanding what different cell populations contribute to the overall tumour microenvironment. What we’re doing is to characterise what is there before intervention in order that we can work out what changes after intervention. The intention and the hope and the early data suggest that you can read out what has changed and make an assertion about what will happen once you know this paired before and after in enough patients.

So the intention of this is to, on the one hand, save treatment for patients who will benefit but, on the other hand, make better decisions for those patients who may not benefit from just, for example, one drug such as an anti-PD-1 antibody. Because when we look at global treatment on single agent treatment only about a third of the patients benefit at best; on combinations that are also becoming available up to half of the patients benefit. So that still leaves the majority of patients with no benefit. So if we could remove the group that will be fine on the drugs that we have then we can focus our efforts on the patients that won’t be fine.

Can you give some examples of these prognostic tools?

We have looked at the number and the phenotype of immune cells within the cancer tissue. We have identified that simply counting the immune cells is a really good start, lots of immune cells good, no immune cells is bad. Then working out which type of immune cells and specifically CD8 T-cells seem to be important for protection. Then there are other cells that switch immune responses off such as regulatory cells, stromal cells, so they contribute to the puzzle. Then within the individual cell populations, particularly in effector CD8 T-cells, we have just demonstrated that the ability of these cells to live in the tissue microenvironment, they reside in the tissue so their short name is tissue resident memory cells, that these appear to be the key cells that are protective for the patient. So if a patient has a lot of tissue resident memory T-cells then their outlook is likely going to be very good. If they don’t have those then the outlook is likely poor and that is more accurate than just counting T-cells or even counting CD8 T-cells.

The puzzle now is whether what we do to the patient can enhance the number of these tissue resident memory cells. So what it is doing for us is that it is allowing us to go back to the start of the puzzle with a new set of hypotheses. So our previous start was all the immune protection is in the tumour, understanding the events in the tumour will lead us to understand immune protection. Now we’ve narrowed that down and are able to say that a particular set of cells confers protective features. So we’re now going back to the same puzzle using this iterative approach and that’s what we are also going to examine in our early phase trials.

A concluding message for a doctor?

Immunotherapy is here to stay, we need to do better with less toxicity and make better choices for our patients.