Richard S. Finn, MD: Regarding the new second-line setting, in the post–atezolizumab-bevacizumab era, Amit. We have the REFLECT trial, which proved that lenvatinib was active in the frontline setting. Then we had the CELESTIAL trial, which proved that cabozantinib is active in second line vs placebo in patients previously treated with sorafenib. We have the RESORCE study, which proved that regorafenib is active in patients previously treated with sorafenib who tolerated it for a minimal period of time and showed a nice overall survival for that sequence of up to 28 days. Then we have ramucirumab for patients who have an elevated AFP [alpha fetoprotein] after being treated with sorafenib. It’s become a real-world issue. Doctors start their patients on atezolizumab-bevacizumab, and at some point patients will progress. What do you do next?

Amit G. Singal, MD, MS: Rich, it’s a good question, and I think Dan and Josep [Llovet] started to touch on this in terms of the 2 approaches. All of us call it a different term. I think of it like a T-minus-1 approach, where you start with atezolizumab-bevacizumab and then treat with sorafenib-lenvatinib. The prior second-line therapy then moves to the third-line setting. In the alternative approach, which I call a grab-bag approach, you just put the rest of the drugs in 1 big bag and choose from them however you wish.

When I think of the grab-bag approach, the first thing I think about is trying to have some mechanism of action [MOA]. Of all the drugs that are approved in the United States, we’ve acted on the PD-1, PD-L1 access, and the VEGF access, so I am less enthused about using—for example—single-agent PD-1. Therefore, I take out nivolumab-pembrolizumab. We’ve also acted on the VEGF access, so I am less enthused about using ramucirumab after atezolizumab-bevacizumab, which really leaves you with 4 agents. It leaves you with sorafenib and lenvatinib, and then it leaves you with regorafenib and cabozantinib, in terms of the TKI [tyrosine kinase inhibitor] therapies. Of course, the other treatment that we have available in the United States is ipilimumab-nivolumab, so that can be considered.

When I think of regorafenib, I think of careful patient selection. In my mind, patient selection was 1 of the reasons regorafenib was well tolerated in the RESORCE trial. Patients were required to be on sorafenib for at least 20 of 28 days coming into the trial, so I continued to use that sorafenib tolerability as something that I consider using regorafenib. You whittle your box down even further.

Those are the agents I’m left with. In terms how we choose among those different agents, it’s similar to how we’ve been choosing for the last several years. When we look at the efficacy signals from an overall survival standpoint relatively similar among the drugs, there are small differences in terms of AE [adverse event] profiles. We tend to consider higher hand-foot skin reaction with sorafenib, for example, and associate higher rates of anorexia, proteinuria, and hypertension with lenvatinib. We’ve talked about the differences in terms of the secondary outcomes, objective responses, and progression-free survival.

These minor differences are how we end up choosing among these different therapies in the second-line setting after atezolizumab-bevacizumab. I don’t think that necessarily changes dramatically.

The last thing that I’d say is the CELESTIAL you mentioned was after sorafenib and led to the approval of cabozantinib in the second-line setting. One of the things about that trial was that 27% of the patients had received 2 prior lines of therapy, so it’s 1 of the only agents that we have evidence for in terms of third-line therapies. We’ve already mentioned trying to get people onto systemic therapy earlier with the hopes that we can use multiple lines of therapy. Cabozantinib is a great drug, and I’m not saying we should only reserve it for the third-line setting. But as much as you can predict these things, if you really feel that somebody can get to 3 lines of therapy, it’s 1 of the only drugs that we have evidence for in that third-line setting.

Richard S. Finn, MD: Daniel, do you see a role for I/O [immuno-oncology] beyond progression?

Daniel H. Ahn, DO: Yes. First, to Amit’s point: In addition to looking at the data for the CELESTIAL trial in the refractory setting, they actually had some data that were presented in 2019, post–I/O. As a lot of us had mentioned that there are not much data for how to treat patients post–atezolizumab-bevacizumab. They did show some signals in the CELESTIAL trial that giving cabozantinib is not only safe post–I/O but also effective. It is a small subset of patients. As Dr Li mentioned, as more real-world data come out, we’ll understand the safety and efficacy. To your points, I agree MOAs need to be taken into consideration in terms of treatment paradigm, especially when you start with an anti-VEGF I/O. From there, we need to determine how to best sequence treatment.

Rich, I didn’t mean to shift from the question you asked.

Richard S. Finn, MD: Not at all. It’s good discussion.

Daniel H. Ahn, DO: There is a lot of interest in salvage I/O for patients who had received prior I/O therapy. A lot of the data that are out there has been led by our GU [genitourinary] colleagues in renal cell carcinoma. When they looked at giving combination dual I/O checkpoint inhibition post-nivolumab, which is not necessarily the standard of care for renal cell carcinoma, the response rate can be 20% to 29%.

There was an interesting abstract presented this year at ASCO GI [American Society for Clinical Oncology Gastrointestinal Cancers Symposium] 2021 that looked at giving the combination of nivolumab or pembrolizumab in combination with ipilimumab post-nivolumab. What they showed was the response rate was about 16%, so that’s pretty consistent and has a similar efficacy from what we see from renal cell carcinoma. They didn’t explain in too much detail whether patients had prior TKI before they went to salvage I/O therapy, but it does look as if it’s a potential effective strategy, especially when treatment options are limited.

We did have the updated results from CheckMate040 from Anthony El-Khoueiry, MD, and colleagues, where they showed the response rates were consistently 30% to 31%. In terms of treatment arm A, which looked at nivolumab 1 mg/kg plus ipilimumab 3 mg/kg dosing, it did show a present median overall survival of about 20 months. As more data come out, we’ll understand the role of salvage I/O post–I/O failure. But with the data being uncertain, we should look at TKI sequencing after prior I/O and consider the MOA of the prior drugs that the patient received. Then again, you have the other camp that says it’s OK to use sorafenib or lenvatinib. But from a rational approach, it makes a lot of sense.

Richard S. Finn, MD: Some of us wrestle with the high expectations for I/O. With atezolizumab-bevacizumab, it does have a high disease control rate, especially for patients with big tumors if they’re not 1 of the responders per se. They achieve stable disease with atezolizumab-bevacizumab, and that’s of value because we want to improve survival, but sometimes I’ve gotten a little concerned with how much will I achieve with a TKI if I didn’t see that big response up front. We know generally in oncology our biggest benefits come in the frontline setting, and progressive benefits from sequential drugs is less than we see in the frontline setting. Some patients progress very fast on the front line, and their benefit from subsequent treatments is not as strong. Sometimes, if they progress quickly on the front line on atezolizumab-bevacizumab, maybe ipilimumab-nivolumab is what they need, hoping for that big response. The response rates with that combination are in the 30% range as well in patients previously treated with sorafenib post–atezolizumab-bevacizumab, but that’s an area of unmet need.

Transcript Edited for Clarity