Our work that we are presenting in ASCO is about GIST, GIST is gastrointestinal stromal tumour. It’s a type of sarcoma that is driven by mutations in KIT or PDGFRA. We used targeted agents that actually do a fantastic job for patients against these two receptor tyrosine kinases but eventually most of the patients will progress to first line treatment that is imatinib. 

Something that we don’t know very well is how these tumours evolve over the time of the disease but we know that they have secondary mutations in KIT or PDGFRA. Actually the agents that we use in these patients target these secondary mutations but only against specific subsets and that explains probably the modest activity that they have. 

So the idea here for this study was, first, trying to understand a little bit more the mutational landscape in these patients through circulating tumour DNA, liquid biopsy, and, second, see if we can use the ctDNA assessment to predict outcomes and the function of these TKIs against GIST.

What did you look at?

What we did was we used the data collected from the phase III VOYAGER trial. This was an international randomised clinical trial that checked for activity between avapritinib and regorafenib. The endpoint wasn’t met, meaning that both agents had similar activity against this population but in this subset of patients that were 476 there were plasma samples collected at baseline and other time points. So what we did was analysing the plasma samples through a 74 gene panel from Guardant, trying to understand this in a very relevant manner because since GIST is a rare disease we only have a small series. This is the first time that these questions are addressed in a prospective randomised clinical trial. 

What did you find?

First of all we just looked for mutations in KIT and PDGFRA, they are the driver mutations. We were able to detect them in 80% of all the patients which is actually numbers that we like because, overall, GIST are tumours that are thought to not shed that much DNA into the blood stream compared to other tumours. 

Then we had a mutational landscape. Actually, the proportions of primary mutations and secondary resistant mutations in KIT were much what we know from tumour tissue. But we were able to find, for instance, 25% of mutations that we found were completely new and not reported in COSMIC or elsewhere. We checked for the pathogenicity of these mutations and they were pathogenic. So this somewhat highlights that ctDNA assessment in GIST patients can be clinically meaningful and also from a biological standpoint.

That was the first part; the second part was much more focussed on outcomes. Here we wanted to check how avapritinib and regorafenib worked in the context of ctDNA. So avapritinib is a highly selective tyrosine kinase inhibitor, type 1, against the activation loop mutants and regorafenib is a multi-kinase inhibitor that also by preclinical studies is supposed to target only specific subsets of secondary mutations.

What we found first was that secondary mutations in one region of the KIT kinase, the ATP binding pocket, predicts a negative response to avapritinib compared to regorafenib. Second, we also observed that regorafenib is an agent that has activity against all secondary mutations irrespective of their type, if they are ATP binding pocket or activation loop, which for one side is slightly different to the activity of avapritinib – multi-kinase versus a highly selective inhibitor. Second, this also argues against the preclinical data that we have saying that regorafenib is an agent effective against a subset of mutations. 

Third, we also checked for ctDNA negative versus positive in the blood. This is important because in an important subset of cancer when there is a ctDNA negative in the plasma it usually means that it’s a good prognosis for the patients. Here again we found differences. That was true for avapritinib – ctDNA negative patients on avapritinib performed better – but ctDNA negative patients with regorafenib were the same as ctDNA positive. So, again, we found a different behaviour between a multi-kinase inhibitor and a highly selected tyrosine kinase inhibitor.

So overall that’s what we found in our study.

What could be the impact of these findings?

The impact that’s relevant, first because there were many questions and doubts about the true clinical utility of ctDNA in GIST, if we can use it for monitoring the disease or having therapeutic decisions. So we can say that, first, yes, we can use it. This is a subset of patients that are resistant to imatinib and sunitinib so when patients are metastatic and have progressed to at least one line of treatment we have shown that it’s highly possible to find mutations, always the patients had to be progressing on therapy. 

Second, this ctDNA assessment would be able to detect mutations or mutants that can be useful for directing some types of TKIs against GIST. Third, although avapritinib is not approved for KIT mutant patients, it’s approved for PDGFRA mutant patients, still this is the first study that shows that we can use ctDNA to predict outcomes. So it’s highly likely that we can use it as well for other TKIs that are already used in the GIST field for treating these patients. 

Will people be looking at that in the future?

All of us daydream with a clinical trial in which we can just randomise patients to treat them sequentially as drugs are approved by regulatory agencies. The other arm would be trying to use ctDNA to match the best drug, depending on the mutations of the patient, trying to not deliver drugs and toxicity in a subset of resistant population that won’t be effective.

Probably the most important thing in rare tumours but in GIST in particular, GIST is a phenomenal model for using ctDNA. It has primary mutations, a predicted pattern of secondary mutations and drugs that apparently are active against some subsets. So it’s such a cool model, actually, to use ctDNA to select the best drug candidates. But we need these efforts in which we do these academic collaborations with that much data to really reach meaningful conclusions.