Karen Titus
September 2022—It may not be the oldest story in the world, but in clinical laboratories it’s an oft-told tale: Tumor meets biomarker; drug meets companion diagnostic; both meet FDA approval; clinicians meet with patients offering new hope—and those in the lab are left trying to figure out how to make it all work.
That story is playing out again in the realm of measuring tumor mutational burden. In mid-2020 the FDA approved pembrolizumab as a new treatment option in adult and pediatric patients with TMB-high (≥10 mutations/megabase) solid tumors, as determined by the FDA-approved FoundationOne CDx assay.
“That doesn’t sound too controversial, right?” says Alain Borczuk, MD, vice chair of anatomic pathology and director of oncologic pathology, Northwell Health Cancer Institute. “It’s not the only way in, but it’s one of the ways in. If you’re arguing for your patient that this is the biomarker that makes them eligible for the drug, then the next questions will be, What was the number? And what was the test?”
And it’s off to the races.
Rollouts of earlier agents and companion diagnostics encountered numerous speed bumps, Dr. Borczuk says; the hope was that TMB would have a smoother journey. PD-L1 in particular has been a quagmire, involving very specific instrumentation and different cutoffs, potentially, for different diseases. “We were hoping TMB would be simpler and a little more straightforward,” he says.
Identifying the same group of eligible patients with a laboratory-developed test might require a different cutoff, requiring validation against the putative gold standard of the companion diagnostic, Dr. Borczuk says. “And that’s an issue.”
The primary TMB study remains Keynote-158 (Marabelle A, et al. Lancet Oncol. 2020;21[10]:1353–1365), but the TMB literature is filled with a variety of cutoffs, some less than 10 mut/Mb, others greater, depending on the setting and platform.
Says Lynette Sholl, MD: “At tumor boards, the existence of this cutpoint definitely triggers conversations.” A clinician may be ready to act on a cutpoint of 11 mut/Mb, at which point Dr. Sholl, associate pathologist and associate director, Center for Advanced Molecular Diagnostics, Brigham and Women’s Hospital, will weigh in. “I remind them to take it with a grain of salt.”
The Foundation assay is a tumor-only test, Dr. Sholl notes, and thus is similar to other available panel assays that use bioinformatics strategies for subtracting out germline variants. “Many academic centers will employ a similar strategy with their tests,” she says, using at least several hundred genes. Anything smaller than one megabase of coverage makes it difficult to derive reliable TMB results.
“So although there’s ample evidence that suggests there’s not great correlation between labs right at that 10 mutations/megabase cutpoint,” she says, “I think people will still have that cutpoint in mind,” regardless of what assay they’re using. “And they’ll use that as an indicator of whether a tumor has a high TMB or not.”
“Whether or not it’s prudent for people to interpret results taken from other assays and apply that cutpoint—I’m not sure. There’s a good chance,” she says, a tumor might measure 8 mut/Mb on one assay and 12 mut/Mb on another, or even 6 mut/Mb on one and 15 mut/Mb on another.
[dropcap]“T[/dropcap]he technical limitations of this biomarker can be a little hard to understand,” Dr. Sholl says, because the published analyses are quite complex.
Nor is it obvious that a single cutpoint is relevant for all tumor types, she adds. A number of studies have shown that 10 mut/Mb can be an iffy predictor. With lung cancer, for example, a higher threshold is probably a more reasonable predictor of response to immune checkpoint therapy. She also notes that some major tumor types (including non-small cell lung cancer) were either not included or not well represented in the Keynote-158 study. “Yet we’re extrapolating results from it to consider these patients for therapy.”
Then again, NSCLC hardly requires a high TMB result to qualify patients for immune checkpoint therapy, she says. “Honestly, you can just walk in the door and get immune checkpoint therapy in the first-line setting.” But, she adds, TMB adds context. A NSCLC patient with low or negative PD-L1, for example, might have contraindications that make physicians nervous about immunotherapy, but not to the point of ruling it out. A high TMB could help them decide to try immunotherapy in these edge cases, she suggests. Similarly, a low TMB, possibly indicating nonresponse, may help them decide the treatment is not worth the risk of immune-related adverse events in a patient with autoimmune disease.
Beyond NSCLC, certain unusual tumor types might also benefit from the context TMB could provide. A metastatic atypical carcinoid tumor, for example, might not be expected to respond robustly to immunotherapy, Dr. Sholl says. “But if they have a really high TMB, you might have an indication” that immunotherapy might be a second- or third-line treatment option.
Adding to the complications, new trials have added liquid biopsies to the mix. “So all of a sudden,” Dr. Borczuk says, “when looking at responders and nonresponders, there’s not only heterogeneity in the platform and the cutoff, but even in the type of specimen used.”
“So many different things happen in the research literature simultaneously,” he continues, making it difficult to draw conclusions about how to use TMB. Initial studies may show a benefit across different tumor types, “and therefore it’s often stated that it’s agnostic to tumor type or site of origin.” That’s often refined over time, he adds, “but the problem is as those studies come out, readers have to be very careful” and check to see whether the same cutoffs and assays have been used.
Is that picture coming into focus?
“Yes and no,” says Dr. Borczuk. Pathologists are gaining some indication of tumor types that are more likely to be high TMB. When it comes to determining eligibility, however, “In theory, many patients may be eligible.”
In his experience, solid tumor stage, rather than histologic type, is driving the decision for TMB testing. “For patients who have high-stage solid tumor—and certainly non-small cell lung cancer is an example—we’re being asked to do FoundationOne’s CDx” along with PD-L1.
Larissa Furtado, MD, molecular pathologist, St. Jude Children’s Research Hospital, also looks beyond tumor type when she considers the value of TMB. While having more tumor-specific studies is important, she says, other factors may be equally valuable. Knowing not only the number of mutations but also the patterns of mutations might be useful, she suggests. “It’s going to be very interesting to look at different tumor types and see if we can tease out some sort of pattern.” Research efforts have begun to look at mutation signatures, she says. “If we can put together the signatures with mutation load, and with more information about microenvironment markers, we’ll get a bigger picture of how tumors behave when they’re treated with immune checkpoint inhibitors.”
[dropcap]L[/dropcap]aboratories are also being asked to consider the role of other biomarkers alongside TMB. With biomarkers, there are no Wally Pipps, losing their starting jobs when more enduring players come along. It makes for a crowded lineup.
In addition to PD-L1 and TMB, oncologists typically want to know about microsatellite instability and mismatch repair deficiency. There is some overlap with MSI, MMR, and TMB, says Dr. Borczuk, “but it’s not pure overlap.”
Published in Archives of Pathology & Laboratory Medicine last month is a new CAP guideline on MMR and MSI testing for immune checkpoint inhibitor therapy, developed in collaboration with the Association for Molecular Pathology (Bartley AN, et al. Arch Pathol Lab Med. Published online Aug. 3, 2022. doi:10.5858/arpa.2021-0632-CP). Its recommendations also address the role of TMB. “It’s an important document,” says Dr. Borczuk, the journal’s editor-in-chief, though he notes that TMB is a biomarker with slightly different goals.
Another CAP guideline (not yet published), on PD-L1 in NSCLC, includes information on TMB, says Dr. Furtado, who co-chairs the guideline with Dr. Sholl. Dr. Furtado is also involved in writing and chair of an AMP guideline on best-practice recommendations for TMB testing, validation, and reporting. “The focus is on the technical, or analytical, aspects of TMB testing,” she says. “We’re not addressing clinical validation or clinical utility of TMB.”
The technical guidance is needed, Dr. Furtado says, as more labs start offering TMB testing. In looking at the literature systematically, as well as relying on the expertise of those in the guideline group, “We noticed that validation can be particularly difficult, because there is no established standard for calculating TMB,” she says, nor for reporting and interpreting it. “There is so much variability in how TMB is measured across different laboratories.”
Much of this relates to assay-specific features, she says, including the genomic territory. Is the lab using tumor-only or paired-normal tissue? What type of informatics pipeline is being used for variant calling? What types of variants are being included in the calculation? Labs also need to consider preanalytic factors, such as DNA quantity and quality and tumor purity.
Labs may look to panel size to draw conclusions about validation, she says—for example, how a panel-based test compares to whole exome sequencing. “But sometimes due to the incomplete or absent methodologic descriptions in the literature,” Dr. Furtado says, “it’s been harder than you think it would be to draw those conclusions.”
Dr. Furtado says she’s particularly interested in assessing information from an international survey of labs that perform TMB testing. Coupled with the standard literature review, this should give the guideline writers added insight. “Having an idea of how things are being done clinically should be very interesting from a practical standpoint,” she predicts.
[dropcap]G[/dropcap]iven the availability of different biomarkers, it might make sense to test tumors sequentially, moving to TMB in patients who are negative for, say, PD-L1. “But from a practical point of view, most of our treating physicians want most of the information up front,” Dr. Borczuk says. “If there are two routes to eligibility, no one’s going to complain about that.”
But even if both roads will get you to Scotland, the trip might require further planning. Does it matter if patients are eligible for immunotherapy by one test or by several? Dr. Borczuk asks. If a patient’s tumor is PD-L1 negative and TMB high, will the response rate be different than a case that is PD-L1 positive and TMB high? While there have been papers to address that, the complexity increases depending on the patient population studied. “If you’re focusing on, say, tumors that were not eligible for immunotherapy based on PD-L1 status, you may end up with a selected cohort where the cutoff for TMB may not be the same in terms of responders and nonresponders,” he says.
That moves the question back to square one: how TMB is determined and whether that one study was deterministic of a cutoff, Dr. Borczuk says. “The answer is: probably not. You probably need multiple studies, and they need to be designed properly.”
There’s a perfectly good biologic rationale for a high TMB indicating a tumor is likely to respond to immunotherapy. But data from study cohorts are hard to parse, Dr. Sholl says. For example, did a responder have a TMB of 200 mut/Mb, or 11 mut/Mb? Complicating matters further, she notes, is that the need to study rare tumor types leads to basket trials that incorporate all sorts of tumors. “You look for these coarser biomarkers, but it’s hard to then use that to predict, on an individual basis, who’s actually going to respond.”
Dr. Sholl admires the work being done by the U.S. group Friends of Cancer Research TMB Harmonization Project and its European counterpart, Quality in Pathology, which published results of their cross-laboratory validations/technical comparisons in December 2021 (Vega DM, et al. Ann Oncol. 2021;32[12]:1626–1636).
Dr. Sholl calls their efforts useful, particularly in highlighting the degree and sources of variations between labs, including specimen degradation. “They also point out the variations in TMB that are dictated by the contribution of germline variants and how patient ethnicity can strongly influence that number, because our approaches to informatics filtering of germline variants are not as robust” for non-Caucasian populations.
[dropcap]I[/dropcap]t’s possible that few of these details will register with physicians outside the laboratory the way they do within the laboratory. Oncologists may simply opt for the companion diagnostic, Dr. Borczuk suggests, trying to match what was done in the clinical trials.
A laboratory may opt not to develop a TMB assay, given the challenges of doing so. His oncologist colleagues might prefer to use the companion diagnostic, with its clear cutoff of ≥10 mut/Mb. “And then they have no questions,” he says.
“Now, whether that’s good or bad is a completely different conversation,” he adds with a laugh. The number is clear and easy to act on, and it may pass muster more easily with insurers when it comes to seeking approval for the drug.
But that still doesn’t ease the burden on pathologists. “We want to offer not only diagnostic testing,” Dr. Borczuk says, “but also advice as to some of these predictive tests.” With tests such as FoundationOne’s, “It does get taken out of our hands,” especially if the TMB results go directly to the oncologists.
“As we develop molecular platforms in our own in-house laboratories,” Dr. Borczuk adds, “all of us want to try to create some sort of one-stop shopping approach. The more laboratories that are involved, the more complicated it is for clinicians to read all the different reports, and the more complicated it is for pathologists to coordinate utilization of tissues, some of which are not plentiful.”
“It’s also a little uncomfortable,” he says, since it tends to direct testing toward a particular laboratory. “If you don’t have much choice about one of the tests on the panel, then you’re almost obligated to use the rest of the panel. Sometimes we like that and sometimes we don’t,” he says. “But to get the results from the FDA-approved diagnostic, it’s the route we have to take.”
Moreover, when testing analytes and methodologies can’t be reproduced because of a proprietary algorithm, “we end up in a position where there really isn’t an easy alternative,” Dr. Borczuk says.
Could this happen with TMB? It seems like a technology and an approach that shouldn’t need to be proprietary, he says. If sequencing is open to everyone, and labs sequence a sufficient number of megabases, “you should be able to get to a number that’s reproducible.
“However, it turns out that simplistic point of view is not correct,” he continues. Depending on the variables used in the algorithm, “You need your own training and testing sets to prove that your number, for your assay, has the true biological relevance to the endpoint—in this case, therapeutic response.”
Labs may instead compare their cutoff to that of the FDA-approved companion diagnostic. But again, the question remains: Does the LDT capture the same thing as the companion diagnostic that was linked to the clinical trial? “You’re always going to be left with some level of doubt,” Dr. Borczuk says, “because you’re one step removed from the actual endpoint.”
[dropcap]A[/dropcap]t Brigham and Women’s, Dr. Sholl and her colleagues have been reporting TMB since 2017. They provide a calculated TMB—8.6 mut/Mb—which, she says, is likely to vary from a TMB from another lab. “It could be slightly different; it could be very different,” she says.
The lab also indicates where a particular tumor lands in terms of its relative amount of mutational burden, she continues. For example, “Is this tumor in the 20th percentile of TMB for adrenal cortical carcinoma? Or is it in the 99th percentile?” Such disease-specific context should be useful for less common tumors, given that so little information exists about average TMB measurements.
Dr. Furtado reports seeing increased interest from oncologists wanting TMB assessment for immunotherapy clinical trials and in some clinical situations. She and her laboratory colleagues intend to launch testing soon. The plan is to provide for each tumor type a range of TMB in their patient cohort. “We’re understanding more and more that if we just have a number indicating TMB high for all tumors, or if the validation didn’t include different tumor types, it’s hard to come up with a meaningful cutoff for interpretation,” she says. “So we are trying as much as we can to be tumor specific and to keep updating our series.”
[dropcap]D[/dropcap]own the road, labs may also be looking at nontissue-based tests. As Dr. Borczuk notes, blood is an appealing specimen and more convenient from a clinical perspective.
That might also make it easier to fill in research gaps. Says Dr. Sholl, “You look at studies that try to incorporate analyses of tissue TMB, and the drop-out rate before we actually even get TMB from the specimens is incredible—you lose about 60 percent of the cases that you want to test.” That’s mostly because these are retrospective analyses—measuring TMB on samples that may have been biopsied a decade earlier and have already been subjected to testing for numerous other biomarkers is onerous. “Trying to tack these additional analyses on existing data sets is really hard,” she says. “It’s not like a PD-L1 stain that requires one slide.”
Cell-free DNA testing might be useful for that work, as well as for patients whose tumors don’t easily yield tumor tissue for genetic profiling, including lung cancer patients. Dr. Sholl thinks it’s possible, but with caveats. Confounders include clonal hematopoiesis-derived mutations, “including weird things that show up in the blood” and may disappear months later. “Is it related to the tumor or is it some other clonal process that came and went?” she asks. “It’s hard to know. You can’t always attribute all the variants that you find in the patient’s blood to the tumor, which is one of the things that confounds interpretation of blood TMB.”
One recent ESMO abstract tried to shed light on this but didn’t get far, she says (Dziadziuszko R, et al. Ann Oncol. 2021;32[suppl 5]:S950–S951). It serves as a good reminder of why TMB remains a promising but thorny biomarker.
The study, like many in the TMB realm, was complicated. Patients were randomized according to their blood TMB, but the researchers used multiple different cutpoints—not surprisingly, Dr. Sholl says with a laugh. The goal was to see if high blood TMB would predict better progression-free survival. “The bottom line is, it didn’t,” she says. And while there was some signal that there was maybe slightly better overall survival with the atezo versus the chemo in the high-TMB group, she says, it wasn’t significant.
This remains in line with the mixed results in the blood TMB literature, she notes. “It’s a complicated biomarker. Again, there are challenges to what we define as the cutpoint for high. There are challenges around identifying those patients who have an artificially inflated high TMB because of clonal hematopoiesis.”
The biology of TMB is being teased out bit by bit. A few studies have suggested that TMB might be more stable than PD-L1, Dr. Sholl says, but other studies have indicated it’s also dynamic.
“You also see intratumoral heterogeneity in terms of TMB levels. If you do kind of a multitumor block analysis, you’ll find variations in TMB from one region to the other, even just within a primary tumor,” she says.
In her own practice, “We do see patients where you can see TMB really take off in the post-treatment setting. The classic context is you have a patient who, say, receives a targeted therapy and potentially evolves a mutational signature process, such as an APOBEC, and this is a known mechanism of driving resistance to some targeted therapies. So you have a mutagenic process that kind of gets triggered by some pressure you’re applying.”
A rebiopsy will often then show a TMB that’s much higher than the pretreatment measurement. “What do you do now?” Dr. Sholl asks. Will a newly acquired high TMB respond to immunotherapy?
TMB can be very dynamic, in other words. And sometimes the mechanisms behind that are clear. “I guess that’s good and bad, right?” she says. “Maybe you identify points in the tumor’s life where immunotherapy is more likely to be effective. But right now, there’s not a lot to help us know what will happen in that scenario.”
Karen Titus is CAP TODAY contributing editor and co-managing editor.