Reevaluating Tissue-Agnostic Therapies: Insights From a Real-World Genomic Analysis

In this interview, George W. Sledge, MD, explores the promise and limitations of tissue-agnostic cancer therapies, highlighting real-world disparities, biomarker variability, and the urgent need for better clinical decision support to guide precision oncology.

Please introduce yourself by stating your name, title, and any relevant clinical experience you’d like to share.

George W. Sledge, MD: I'm George Sledge, MD. I'm a breast medical oncologist. I am the former chief of oncology at Stanford University and the former president of the American Society of Clinical Oncology. I am currently chief medical officer and executive vice president at Caris Life Sciences.

What were the primary objectives behind conducting this large-scale analysis, and what key questions were you hoping toanswer about the role of genomic profiling in cancer care?

Dr Sledge: By way of background, tissue-agnostic therapies are a relatively new thing. The very first tissue-agnostic therapy was approved by the US Food and Drug Administration (FDA) late in the last decade. They've become increasingly more common over the years, but no one had actually looked at them systematically. They had been approved, but not really observed very well.

If you were to look at the approvals of the tissue-agnostic therapies, almost without exception, they were approved based on studies with relatively small numbers, in some cases, fewer than 100 patients. Because they were tissue-agnostic, in theory applicable to all cancers, and because they were so small, in most cancers, we had no strong sense of how well they worked.

In every cancer we had a few mutations that we had a lot of data for, but we had a lot of less common ones where we had very little data. We hoped, in doing this study, to provide a little bit more certainty to the field about how tissue-agnostic therapies actually work. That is to say, did they work the same across all cancers? Were there any new lessons that you could learn given that we've been able to accumulate a fairly large amount of data about tissue-agnostic therapies?

Caris Life Sciences does a lot of whole exome and whole transcriptome sequencing. We've done over a half a million at this point in terms of whole exomes and whole transcriptomes. Attached to that, we have a lot of claims data and a lot of immunohistochemistry data. Having multiple platforms allows us to answer interesting questions about this rapidly emerging and rapidly growing area of tissue-agnostics.

How do you see the findings from this study accelerating the adoption of precision medicine in oncology practice, particularly outside of major academic centers?

Dr Sledge: Amongst all of the different cancer types for tissue-agnostic therapies—particularly where we have really good data in, I would say, microsatellite instability (MSI)-high cancers and tumor mutational burden (TMB)-high cancers, probably to the greatest amount, and to lesser amounts for some of the other tissue-agnostics—what we discovered is that tissue-agnostic therapies weren't particularly tissue-agnostic in terms of outcomes. If you look cancer by cancer and ask, how long do patients respond? How long do patients live after receiving these therapies? The answer is that there's huge variation.

Why is this important? If you have an FDA indication that basically says you can use a drug to treat any cancer, then the temptation is to use that drug to treat any cancer. However, if a drug doesn't work particularly well for a particular cancer, maybe you should be looking at something else.

If you compare, as an example, non-small cell lung cancer to small cell lung cancer, and you look at how well a checkpoint inhibitor such as pembrolizumab works, the answer is there are huge and distinct differences between the cancers, even though they're in the same organ. There is nothing particularly tissue-agnostic about their response to therapies.

We dove a little bit deeper. If you look at something like TMB-high cancers, where the FDA accepted—based on company studies—a cutoff of 10 mutations per megabase, and bearing in mind that the approvals were done based on fairly small numbers across most cancers, do you actually see the same cutoffs for tumor mutational burden from cancer to cancer?

I will add that this was a separate paper that we did at Caris, along with colleagues in our Precision Oncology Alliance, but the answer is that not only are they tissue-agnostic in terms of responses, these drugs are tissue-agnostic in terms of where the cutoffs ought to be. Some have adequate cutoffs at a much lower level, some have adequate cutoffs at a much higher level, but there's certainly nothing consistent.

In essence, tissue-agnostic isn't really tissue-agnostic. That's a basic concern. How does this help physicians going forward? If you're a physician looking at this data, the first thing that you have to say is, "I probably need to do more homework before assuming that any particular drug is actually tissue-agnostic. I have to be thinking about this in the context of the other therapies that are available to the patient. Is this actually the best choice or not?"

If I'm a drug developer I have to be thinking, "I'd love to follow a tissue-agnostic pathway because I'd love the FDA to approve every single cancer for me." At the same time, if you're an honest and decent drug developer, and I think most of them are, you also probably want to look at this and say, "What going to be the biggest bang for the buck? What's going to help the most patients with a particular disease or across diseases, if I'm using a particular drug. I really want the patients who are being exposed to my drug to be getting the right therapy." In the long run, that's better for patients and better for the reputation of the company.

Despite eligibility, many patients—especially those with rare cancers—are not receiving these therapies. What are the keybarriers contributing to this gap?

Dr Sledge: That's a really great question, and we looked at it in the context of a rarer mutation—an NTRK mutation. You can have NTRK 1, 2, or 3 mutations. If you have an NTRK mutation, you are FDA-approved to receive one of the NTRK drugs, such as larotrectinib or entrectinib.

This is a classic case of what's not to like. These are oral medications or pills. They're pretty nontoxic compared to a lot of the other therapies that doctors prescribe. Overall, in the initial studies, they have a very high response rate—around three-quarters of patients will have an objective response—and then, frequently, you can cross over to another drug in the same class and also have an objective response. What's not to like?

However, when we looked across multiple years, only a minority of patients who had an NTRK fusion gene ever ended up getting an NTRK drug. That surprised us because the alternatives were frequently much more toxic agents that were reasonably likely to be less active or less effective.

What was going on there? When you're looking at claims data, which is our starting point, it's really hard to read doctors' minds. However, there were some interesting clues. Probably the most interesting one was that if you looked at the patients who are being treated with these NTRK drugs or not being treated with the NRK drugs, what you found is that a fair number of patients with an NTRK fusion gene are either TMB-high or MSI-high.

If you were TMB-high or MSI-high, you almost always would get a checkpoint inhibitor. But if you were TMB-high or MSI-high, you were significantly less likely to get an NTRK drug. Why is this the case? My suspicion—I don't have any proof of this, but it's a suspicion—is that this might be a case where rarity is a form of disparity. Checkpoint inhibitors are widely used in the oncology space, maybe that's their go-to drug. We're talking about 2 patients in 1000 having one of these mutations. Maybe for a 2 patient in 1000 drug, they've never actually used the drug before. They don't have comfort with it, and there's a learning curve for every drug.

Maybe it's just mentally easier for the medical oncologist to stay with their tried and true and not use this new drug that they've never used before and might not use it again for another 5 or 10 years. That's guessing on my part. But the data would support that if you’re TMB-high or MSI-high, your physician just doesn't use an NTRK drug despite having a mutation.

Do you believe payer policies are aligned with the promise of pan-cancer therapies? What needs to change to improve coverage and access?

Dr Sledge: I don't think the lesion is, by and large, at the insurance level. If the FDA approves something, most payers are going to pay for it. I'm sure there are exceptions, but it's really hard to argue with an FDA approval.

I suspect, again, it may be at other levels. I think some of this may be on doctors who are having to deal with an overwhelming number of new drug indications and have a hard time keeping up. If you, for instance, look at last year's FDA list for new oncology and hematology drugs, there were 58 new indications last year. Try to imagine being a general medical oncologist and having to learn more than 1 new indication every week of your life. I think part of this is at the level of education or familiarity. It's hard to learn new drugs.

A related issue is that we don't yet have the decision support that we need built into the electronic health records. This is on us as a community to look to try and provide better education and clinical decision support that will support objective, evidence-based decisions that will help patients.

Is there anything else you hope audiences will take away from this study?

Dr Sledge: The FDA has indicated in the past that you might be able to use real-world data to support new indications. I think that's quite reasonable when we've looked at some of the data that we have. For instance, there's a tissue-agnostic indication for the most common checkpoint inhibitor, but not for the second most common checkpoint inhibitor.

However, when we look at our databases, the answer is that they both appear to have similar efficacy on a pan-tumor basis. My suspicion is that we're leaving some cards on the table here, where we could be using them to support new approvals for patients with more than one drug.

That will be very interesting to see, going forward, whether or not we start looking at new indications and new approvals on the basis of real-world data.