Is Tumor Mutational Burden Really 'Ready for Primetime' as a New Immunotherapy Biomarker?

H. Jack West, MD; Hossein Borghaei, DO


May 11, 2018

H. Jack West, MD: Hello. I'm Jack West, medical director of the Thoracic Oncology Program at Swedish Cancer Institute in Seattle. Welcome to Medscape Oncology Insights. Joining me today is Hossein Borghaei, chief of thoracic medical oncology at Fox Chase Cancer Center in Philadelphia. Welcome, Hoss.

Hossein Borghaei, DO : Thank you.

West: One of the most interesting trials presented at the American Association for Cancer Research (AACR) meeting just a few weeks ago was by Matt Hellmann and colleagues on the CheckMate 227 trial,[1] which was also published concurrently in the New England Journal of Medicine.[2] It gave us some interesting insights about combination immunotherapy, specifically with nivolumab and ipilimumab (nivo/ipi).

CheckMate 227 Trial

This study was quite complex and it evolved over the course of its development. Initially it had over 1700 patients with either squamous or nonsquamous advanced lung cancer, with the division between detectable or undetectable programmed death ligand 1 (PD-L1). Chemotherapy doublet was compared with nivolumab alone in the PD-L1-positive group and nivolumab with chemo in the PD-L1-negative group. The third arm was nivo/ipi. As the study evolved, the sponsor company, Bristol-Myers Squibb, learned more about tumor mutational burden (TMB) as a potentially helpful biomarker for predicting the beneficiaries of at least nivo/ipi. We saw some of this in non–small cell lung cancer[3] as well as small cell lung cancer.[4]

In a subset of 299 patients regardless of PD-L1 level, nivo/ipi was compared with chemo doublet, and they found a pretty impressive improvement in progression-free survival (PFS). At 1 year, PFS was 13% for the chemo-alone arm versus 43% for nivo/ipi.

These results are certainly provocative, but we have not seen results on overall survival (OS). There was a hint of a plateau on that curve, so we will need to see whether it translates in OS. Notably in the CheckMate 026 trial,[5,6] PFS looked that way for high TMB but it did not translate to OS.

Nivo/ipi is not approved by the US Food and Drug Administration (FDA) in lung cancer right now and certainly not in the first-line setting.

'Practical Issues' With TMB

There are practical issues with TMB. It takes weeks to get back. Cost may be an issue unless it is done in the context of FoundationOne® testing, which you might already be doing; then it's not as much of an issue. This was just the subset with high tumor mutation burden, and the cut-point, or definition of "high" TMB in this subset of 10 mutations per megabase, has not been used before. There has been some criticism about the validity of that cut-point. Drs Hellmann and Ramalingam, who presented on the CheckMate 568 single-arm trial,[3] showed some of the justification for the threshold of 10 mutations per megabase as "high TMB." Yet, in the same presentation, Hellmann showed that a 13-mutations-per-megabase cut-point for monotherapy with nivolumab was not predictive.

There are many potential barriers to using TMB. What do you think is the role of TMB now, and where does this leave nivo/ipi as an option?

Role of TMB

Borghaei: You have raised several great issues that we need to discuss. Just so that we're clear, I'm completely biased towards ipi/nivo. I've been doing the studies with CheckMate 012[7] and I've been involved with CheckMate 227 and CheckMate 568, so I sort of lean towards that.

It's hard to argue with the success of chemo plus immunotherapy (IO) in the context of KEYNOTE-189.[8] We're not arguing that. To some extent these might have been completely different patient populations, and they are. Let's start with the TMB issue.

I would like to remind everybody that back when we started doing PD-L1 testing, everybody who had a drug in this category had their own PD-L1 [threshold]. We talked exactly about the same issues. The cut-offs were different from one company to another. The cut-points were selected based on something that we did not quite understand. Issues with new biomarkers always come up and they need to be discussed. If TMB becomes a biomarker, we need to standardize it so that we do not end up in a place like we were with PD-L1 until the Blueprint Project[9] came and tidied everything up. Perhaps a Blueprint 2.0 would be needed for TMB issues.

The finding of a new biomarker, particularly one as complex as TMB, is appropriate in light of CheckMate 227, and I have a feeling that we're going to see more data from other studies concentrating on TMB. It might be a better biomarker, but we do not know for sure right now.

CheckMate 227 gives us a chemo-sparing regimen for a very particular patient population regardless of the level of PD-L1 expression—that is TMB high. I'm not going to go into whether 10 mutations per megabase is the appropriate cut-off. Based on CheckMate 568, that definitely was the cut-off for this study. Because the study was done in two separate cohorts of patients, one with CheckMate 568 and now with CheckMate 227, I think that is where the data are leading us.

West: We should clarify that. In this study, 44% [had high TMB] and that was completely independent of PD-L1 [expression]. They showed this in a few different cohorts but essentially they tracked totally independently.

Borghaei: They did. The important thing is that PFS is better, but is survival also going to track with that? We have KEYNOTE-189, which showed a significant OS benefit. If a patient has the potential of getting [combination immunotherapy] over chemo/immunotherapy and one study shows significantly better OS and the other one does not even have data on OS yet, it's kind of a difficult argument for a patient.

West: Like I said before, you can't take a knife to a gun fight. So, you can't take PFS to an OS fight.

Borghaei: Although I think it is a little premature to say that this is a gunfight with PFS versus OS, I loved your comment on Twitter. We have to give CheckMate 227 more time to see where it goes. Early on, chemo with IO seems to be the winner—there's no question. But what is going to happen 3, 4, 5 years down the road? What will that survival curve look like when we have longer follow-ups of both groups of patients that we have been discussing?

If a patient comes in and, for whatever reason, does not want chemotherapy or cannot have chemotherapy, and you have TMB data, having an IO/IO option would be a viable option if there is FDA approval and you can get the drug.

Overcoming Barriers to TMB

Borghaei: Expense is a concern, absolutely. It is expensive because you have to do extra sequencing to get that level of data. Maybe with time and improvement, there will be lowering of the cost.

West: There is the promise of many things. If the turnaround time gets reduced to under 2 weeks, that will be much more viable than waiting 3 or more weeks for a first-line therapy option. Today, turnaround time is a real limitation. The test requires significant tissue. Particularly in rank-and-file community centers where tissue may be hard to come by, you potentially would be requiring another biopsy. And it takes time to get from the pathology lab to Foundation. These are real challenges to overcome, but it should only improve.

We do not have data on OS yet and we need to see if there really is a plateau. The conversation would change significantly if a subset of patients does astonishingly better and they are different from patients who are well served by pembrolizumab monotherapy. It is still a viable issue.

Is TMB Ready for Primetime?

West: The real question now is whether TMB is a tool that should be used right now. You were there [at AACR] for David Rimm's discussion on the topic. He certainly has an expertise and came to a damning conclusion when he ended by saying that this is not ready for clinical use right now. Do you agree with that? What would you say? If not, what is the use of TMB today?

Borghaei: I don't completely agree with Rimm's assessment that TMB is not an appropriate biomarker. How do you define whether it is ready for primetime? You need multiple studies—kind of like what we did with PD-L1. After the first few publications on PD-L1, everybody said it was not ready for selecting patients until we saw more data. Clearly, we need to see more data.

Gandara[10] previously presented a blood-based TMB assay, which could be a viable thing down the road if it develops and everybody can access it. Then you eliminate the need for tissue, and the cost and wait time could be a little bit different.

I think it is premature to dismiss TMB as a potential biomarker. As you pointed out, if you look at the subgroup analysis of CheckMate 026,[6] TMB can be predictive of clinical benefit when it comes to the use of IO. I agree that we did not see OS advantage in that retrospective look at CheckMate 026. Remember that with CheckMate 227, we're talking about an IO/IO combination, and from Hellmann's presentation there is at least a hint suggesting that the OS curve was separating.

To be honest with you, if I have a patient where I am hesitating whether to use IO, for whatever reason, and I see data that says TMB is high, I feel a little bit more comfortable using IO in that patient; because even if I do not necessarily get the OS advantage, I'm getting a significant PFS advantage. At least that's the way I'm interpreting the data.


West: In closing, let's turn to toxicity. Although this is "chemo free," it's not toxicity free. It is really a lateral move in terms of general numbers but with different and perhaps less predictable toxicities, which I think is unsettling for many folks. I was pleasantly surprised that it was not more toxic when brought out into a broader population. What would you say about the toxicity challenges with this regimen in broader use?

Borghaei: Remember that what we saw in terms of the ipi/nivo combination for CheckMate 227 came after a dose optimization phase 1 study where multiple doses and schedule of delivery of both drugs were studied. We came up with this particular regimen with all of the other investigators and the sponsor because this one was the one that was best tolerated and gave us the better clinical data that we wanted to see.

I think the toxicities are real. I cannot discount the fact that you have up to a 30% chance of having toxicities with the combination of ipi/nivo, kind of like what you saw with the chemotherapy. We've learned to manage these kinds of toxicities. There were no new safety signals. There was not significantly more death as a result of the combination in the [nivo/ipi] arm. There are the usual skin thyroid, pulmonary, and [gastrointestinal toxicities] that we've all learned to manage, for the most part, with steroids in this patient population.

I think that the time where we were worried about the toxicities of IO has passed. Patients should be aware of the increased risk for toxicities with a combination, but if the clinical benefit is there then I think it justifies it.

West: Thank you very much, Hoss, for a great discussion. Thank you, the audience, for joining us. This is Jack West for Medscape Oncology, and I look forward to our next discussion.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.