Q: Under what circumstances and to what extent are you willing to take clinical actions on a cancer patient based primarily on preclinical data?
A: There are two scenarios that come to mind when thinking about reliance on preclinical data for treatment decision-making: (1) use of an agent that is an established treatment for a noncancer indication and (2) use of an established cancer therapy outside of the context(s) for which there is known clinical efficacy.
In the first scenario, I am essentially NEVER willing to accept a preclinical finding that suggests efficacy as a basis for prescribing use of such an agent. This perspective comes from a career focused on performing early phase clinical trials with agents that demonstrated promising in vitro and in vivo effects that failed to exert any discernible effect in cancer patients.
The explanations for these failures are numerous, but the most common explanations are inability to achieve the drug concentrations/exposures needed to match those used in preclinical experiments and lack of fidelity of the preclinical model systems for predicting outcomes in patients.
This second category remains a major challenge for many cancer therapeutics in that we simply do not have a repertoire of ex vivo or in vivo model systems that fully recapitulate the complexity of human cancer with regard to the molecular features of the cancer cells themselves, the tumor microenvironment, and an intact immune system. This is a greater or lesser liability for certain classes of cancer therapeutics, but we generally cannot be dogmatic about which mechanisms of action will translate across preclinical and clinical settings.
Regarding the scenario of using a cancer therapy with known efficacy in some context for an off label indication based on preclinical data, I am generally more willing to consider this possibility, though it is infrequent in clinical practice that such an approach trumps direct clinical evidence for a repertoire of therapies in a given cancer indication.
When thinking about specific types of systemic cancer therapy (conventional cytotoxic chemotherapy, oncogene targeted therapy, and immunotherapy) there are very real differences in cancer biology that relate to variable efficacy for each across Cancer types. Several years ago, we had hoped that this would not be true for the activated oncogenes for which molecular targeted therapies have been established as effective and at least one contact. BRAF mutations are, perhaps, one of the best examples here. We know that BRAF mutations are found across the majority of cancer types, though at very low rates in many of them. Going into the first clinical trials with selective BRAF inhibitors, we were equally optimistic regarding potential efficacy in melanoma and colorectal cancer, but came away with enormous enthusiasm in melanoma and essentially no efficacy in colorectal cancer.
Extensive laboratory research subsequently help to explain the relative resistance of BRAF mutant colorectal cancer, but it was not a phenomenon predicted by preclinical models prospectively. In fact, it is this type of example that motivated the NCI MATCH trial which is broadly exploring molecular targeted therapies in cancer patients where the molecular features are the basis for inclusion in a given therapeutic arm, not cancer type.
In other words, we take it as conventional wisdom currently that we do not have a basis for predicting efficacy when exploring these types of approaches across the spectrum of cancer types.