High AFP as Deciding Factor for Treatment

Richard S. Finn, MD: When we talked before, Pierre mentioned that you don’t need a biopsy to make the diagnosis. Generally, most patients don’t get biopsies. I very rarely find myself sending Foundation Medicine, or some other genomic profiling, for these patients, or some peripheral ctDNA [circulating tumor DNA] because it hasn’t had an impact on treatment decisions. We talk about NTRK, and those are certainly very uncommon in liver cancer. To your point, PD-L1 expression is not included as an entry criterion in any of our phase 3 studies because in the phase 2 studies, it hasn’t correlated.

This ESMO [European Society for Medical Oncology Virtual Congress] data were not was based on patient-level data as much as population data. But it was certainly hypothesis-generating, as well as this idea that etiology might play a role in response to PD-1 inhibition. There have been trends about hepatitis B and hepatitis C maybe doing better than those that are nonviral, but we’re not denying patients a certain treatment based on those criteria. The 1 drug we have that has had a biomarker has been ramucirumab, and that’s elevated alpha-fetoprotein [AFP]. Peter, you’ve been extensively involved in that program with me for some time. Is there still a role for ramucirumab in high AFP patients in the modern era? What’s your take?

Peter Galle, MD: I believe there is a role. To quote Tony, “Ramucirumab is dead;” I think it’s still alive, and there are 3 reasons for it. First of all, it has been demonstrated in that trial to be effective with a hazard ratio of 0.7. Secondly, it is not the same as bevacizumab: it is tackling the ligand with bevacizumab and the receptor with ramucirumab. That might be quite a distinct mode of action. If you think about it, the single-agent activity for bevacizumab is quite disappointing, so it’s the combination that matters.

In that setting, I would say that is still alive and that we need to see what is coming up. There is an urgent need, as you rightly point out Rich, that biomarkers need to be developed. It’s quite disappointing that, in a field where we now have signatures that are clearly prognostic that have been around for about 20 years or so, that they have never made it into the clinic and turned out to be predictive, which is what we need for all patients. They don’t want to know how long they’ll live; they want to be helped, and this is by prediction of outcome. Here is neglect, and it’s explained by neglect of biopsy, tissue, assessment, and upfront assessment in clinical trials. That’s something that we need to get much better at in the future. Some other aspects to mention where there is a lot of uncertainty and for which we should be open would be the TKI [tyrosine kinase inhibitor] issue in addition to whatever you are using, and that includes the IO [immune-oncology] treatment.

It is immune modulation when we get away from single-agent activity. With single-agent activity, here I believe Tony is right: both are created equal, but if you combine them, then you do something to the immune system, to the microenvironment, where we have a lot of uncertainties and unknowns. That is probably very distinct. Here, there are some theoretical arguments for cabozantinib as a combination or partner. What happens with ramucirumab as a combination or partner is completely unknown.

Richard S. Finn, MD: Those are some important ideas to build upon in regard to combinations. To wrap up the AFP issue, when we looked at the regorafenib data set, AFP didn’t seem to make a difference in the second-line population. Katie, just briefly, you also did some interesting AFP analyses with the cabozantinib data set.

Katie Kelley, MD: Yes, we looked at the question of baseline AFP and saw that cabozantinib had benefit across baseline whether elevated greater than 400 ng/mL, or 200 to 400 ng/mL, or even less than 20 ng/mL. There were benefits in terms of the primary end point and secondary end points for the use of cabozantinib. What we also looked at was using AFP as a response marker, not just for the question of whether it change prognosis. I should backtrack and say that cabozantinib showed symmetric benefit over placebo in all of those AFP cut points that we looked at, but patients with high AFP greater than 400 ng/mL certainly have a worse prognosis, so we know that already. That was borne out in the CELESTIAL trial as well, though cabozantinib benefitted those patients equally. I’m also, like Pierre, an AFP aficionado: AFP has other value, especially in tumors that are difficult to image in many cases. What we looked at is patients with evaluable AFP with a baseline level greater than 20 ng/mL and with serial time points measured, and these were blinded results at symmetric time points on a clinical trial.

Patients who had an AFP that did not go up greater than 0 ng/mL, which was if patients had AFP that was static such that it did not increase from baseline, or better yet decreased at whatever cut point we looked at such as a 20% or 50% decrease, those patients had a substantially better PFS [progression-free survival] and OS [overall survival] than patients who had a progression of AFP of greater than 20% or 50%. The point is that AFP can also be used as a surrogate response biomarker in the carefully selected patients who have an evaluable AFP at baseline and have serial time points to look at.

Richard S. Finn, MD: Yes. AFP is certainly prognostic. I don’t think we’re at the level of PSA [prostate-specific antigen] and using AFP as end points, but it is something reassuring to see, an AFP that isn’t rapidly rising, or staying the same, or even going down. With some of these novel combinations, we see things drop fairly quickly. I’ve now had a collection of patients with AFP in the thousands who normalize. To think that we’re seeing that when we never had drugs that did that for so long.

Transcript edited for clarity.