We have known for some time that molecular profiles can be useful for predicting response to therapy and for tailoring therapy in the first line. In the absence of KIT/PDGFRA mutations, the role of imatinib is not well defined. Additionally, primary mutations in KIT exon 11 respond well to imatinib 400 mg, but there is little benefit to escalating the dose to 800 mg. In contrast, primary mutations in KIT exon 9 respond better to imatinib 800 mg compared with the 400-mg dose. 

In the post-imatinib setting, we are still learning about molecular profiling. The challenge is that, after imatinib treatment, tumors can become quite complex and heterogenous. There may be more than 1 mutation, and there may be mutations in different parts of the KIT gene. Mutations tend to cluster in both the activation loop and the ATP-binding pocket regions of the KIT gene. Preclinical data suggest that some of the available TKIs may be preferentially beneficial for treating GIST with resistance mutations in particular regions of the gene (eg, activation loop vs binding pocket). The question becomes whether it may be possible to fine-tune subsequent lines of therapy, post imatinib, based on the resistance mutational profile of the tumor, as determined by circulating tumor DNA (ctDNA). And I think that the answer is becoming clearer; however, the data are still emerging. 

For example, at the 2022 American Society of Clinical Oncology (ASCO) Annual Meeting, Bialick et al analyzed ctDNA retrospectively to see links between select resistance mutations and the activity of particular TKIs in the second-line setting and beyond. They reported that patients with ATP-binding pocket mutations appeared to do well with sunitinib, which is consistent with our understanding of how sunitinib works.

Our team also presented data at the recent ASCO conference from the larger phase 3 VOYAGER study, from which we explored the mutational landscape of KIT-mutant GIST. That study randomized patients to avapritinib or regorafenib in the third line. We found that patients with ctDNA positive for ATP-binding pocket mutations, and no other resistance mutations, did very poorly on avapritinib, an effect that could be predicted based on the understanding that this agent is a potent KIT inhibitor, but in the activation loop region (ie, exon 17). In contrast, responses in patients treated with regorafenib were independent of the secondary resistance mutational landscape. This was somewhat unexpected because some preclinical data had suggested that regorafenib was not very active in the ATP-binding pocket. So, I think that this tells us that ctDNA profiling may work in some settings but not in others, and that perhaps GIST mutational heterogeneity, as well as the narrowness vs breadth of the TKI activity, come into play.

Analyses of ctDNA from trials with ripretinib have also been performed. Based on samples from the INVICTUS trial prior to enrollment, ripretinib did show benefit compared with placebo in all of the molecular groups. Perhaps more interesting were the ctDNA analyses from the phase 3 INTRIGUE trial, which randomized patients to ripretinib or sunitinib in the second line. That study was powered for superiority, which was not shown, but in terms of efficacy, when looking at the overall population, the progression-free survival curves for ripretinib and sunitinib were essentially overlapping; some subgroups showed numerical—but not statistically significant—differences. 

Taken together, these findings suggest that ctDNA may be predictive in some scenarios, but not in others. There is still a lot more to be done in this space, and we are starting to build a library of molecular findings to help inform therapeutic decision making.