We need the development of new, complex biomarkers to address the increasing complexity in treatment modalities that, in and of themselves, have characteristics of a continuous variable; they require innovation and outcomes data, which perhaps will be partly addressed by some of the emerging real-world evidence databases amassed by pairing sequence information and clinical outcomes. Tumor mutational burden is a great example of this innovation in practice.
https://doi.org/10.37765/ajmc.2020.43625The age of precision medicine has officially entered its adolescent years. A decade ago, we were beginning to unlock the potential of next-generation sequencing (NGS) to inform treatment options for cancer patients that weren’t restricted by the origin of the tumor in the body, but rather informed by the unique genetic mutations responsible for driving tumor growth. This tailored treatment approach made sense: It not only provided the rationale for targeted treatment options, but it also led to new discoveries for drug sensitivity and resistance. Copy number amplifications in MET, for example, were complemented by splicing mutations in exon 141; deletions in EGFR exon 19 added to the plethora of ways to be sensitive to EGFR targeting therapies2; and the number of ways to delete function in BRCA1 or BRCA2 seemed to grow by the day. Further, in a landmark approval from the FDA, all patients with solid tumors that were microsatellite instability—high (MSI-H) could receive the immunotherapy (IO) pembrolizumab regardless of their respective cancer type.3 In another important milestone, the union of molecularly guided targeted therapy with a broadly available commercial assay was finally achieved, with the FDA approval of FoundationOne CDx in 2018.4 Despite these enormous successes in advancing treatment options for patients, there was still much to be learned.
Targeted therapies—that is, treatments developed based on a specific molecular driver of cancer—are no longer the only mechanism to fight cancer in a precise and durable way. Immunotherapies have revolutionized the treatment paradigm and, using perhaps the most complex biological system in our bodies, found a way to provide meaningful benefit in cancers for which no other treatment options previously existed. Additionally, a new mechanism of treating cancer that relied on synthetic lethality involved targeting PARP to allow tumors to effectively destroy themselves by taking advantage of holes in the DNA damage repair pathways.
These additional advancements represented rapid growth in the evolution of cancer treatments, and as such, required new and complicated methods to determine which patients would benefi t most from their application. Yet, despite our initial efforts, we struggled to neatly “fit” these biomarkers into the precision medicine paradigm that worked so cleanly for traditional targeted therapies. A new way of thinking would be required if we were to successfully realize the full potential of personalized health care and, specifi cally, immunotherapy.
Immunotherapy-Guided Biomarkers and Genomic Signatures
Despite its revolutionary eff ect on cancer treatments, immunotherapies weren’t born with companion diagnostics by their side. It took years of trials and failures before we began to focus on PD-L1 immunohistochemistry as a method to identify likely responders. Discrete cutoff s were chosen to segment the cancer population into 2 cohorts: positive (high) and negative (low). However, the application in clinical trials was largely inconsistent, leading to success in some areas and failures in others.5,6 MSI was the fi rst IO biomarker based on the premise that tumors harboring more somatic mutations might be more likely to elicit an immune response through expression of neoantigens, and subsequently quiesce that response by upregulating checkpoint proteins. It also didn’t necessarily require NGS for its application, and in colorectal cancer, the results were staggering.7 Establishing proof of concept for the neoantigenicity hypothesis was crucial; however, MSI still left a gaping sensitivity hole since it was only 1 of many ways for tumors to become hypermutant, and many hypermutant cancers had virtually 0 cases of MSI-H but could still benefit from IO.8 Tumor mutational burden (TMB) eff ectively filled this gap by assessing the overall somatic mutation rate as a proxy for the neoantigenicity of tumors, and preliminary studies using retrospective clinical trial data demonstrated that its clinical utility could be remarkable.9 Disease-specific TMB cutoffs were chosen to evaluate the marker’s performance, and control arms compared with standard-of-care chemotherapy largely demonstrated that TMB was predictive of IO response and not generally prognostic when applied in non-IO treated patients.10
The Evolution of TMB
CheckMate 227 was the first prospective application of TMB in a phase 3, fully randomized clinical trial. One of the coprimary end points in this first-line non—small cell lung cancer (NSCLC) study evaluating the combination of ipilimumab and nivolumab versus platinum-based chemotherapy was based on progression-free survival (PFS), using TMB to defi ne the population. Results presented by Hellmann and colleagues in 2018 at the American Association of Cancer Research meeting, and simultaneously published in the New England Journal of Medicine, showed that the landmark 1-year PFS rate more than tripled for the IO combination compared with chemotherapy, a remarkable and dramatic effect driven by TMB that was not observed in the overall, unselected population.11 Furthermore, results presented that year at the European Society for Medical Oncology (ESMO) by Kim and colleagues from the B-FIRST study evaluating atezolizumab in first-line NSCLC patients identified a rate of enrichment more than 6 times higher in the TMB high vs low subgroups, measured using a using a blood-based TMB calculated from the circulating tumor DNA in plasma.12 Most recently, Mirabelle et al. presented results at ESMO 2019 from the KEYNOTE-158 study evaluating pembrolizumab monotherapy in select solid tumors using TMB calculated from tissue biopsies, finding a 29% response rate in TMB-high versus 6% in nonhigh groups, representing a nearly 5-times enrichment in response.13
A Need for TMB Harmonization
These results are potentially practice-changing; however, the need for harmonization of this complex genomic signature was apparent well before these studies ensued. As a result, a TMB Harmonization Consortium was created and led by Friends of Cancer Research (Friends) to focus on 3 main aspects of TMB: common defi nitions across diagnostic platforms; consistent methods for analytic validation, including the use of reference standards; and proposals for the clinical application of TMB in tumor agnostic studies. The last component followed the guidance of FDA’s Oncology Center for Excellence, which stated that if a biomarker would be used to define a disease category, such as TMB, then it should include participation across both industry and academia.14 These eff orts were recently published by Merino and colleagues in the Journal for Immunotherapy of Cancer15 and guidelines on the tumor-agnostic application of TMB in clinical trials were published on the Friends website,16 with a similar approach being adopted and published by our European counterparts.17 This enormously complex, precompetitive harmonization on TMB required a thoughtful approach that included multiple key stakeholders, and despite the meaningful outcome that should serve to create consistency of TMB across different diagnostic platforms, there is still the unmet need associated with the continuous variable nature of this biomarker.
Measuring TMB as a Continuous Variable
Alignment on measuring TMB is even more complex because of its nature as a continuous variable genomic signature. An inherent property of continuous variables is their ability to provide more information beyond restricting populations into 2 discrete segments based on a singular cutoff . TMB produces a somatic mutations-per-megabase score that falls along a continuum that can be associated with a probability of achieving clinical outcome. It’s important to understand that continuous variables don’t begin and end with TMB. For instance, another common continuous variable is age, one that can be generally associated with overall risk of death.
Contrasting this against a discrete biomarker, like a pregnancy test result or EGFR L858R, and it becomes more obvious how, in these latter scenarios, the practical application lends itself to defining the population according to 2 groups: positive or negative. However, forcing continuous variables into an oncology infrastructure that was designed for discrete biomarkers leaves behind valuable information that can’t be gleaned from restricting populations into 2 broad segments. For example, a singular TMB cutoff to define high versus low doesn’t account for the patient who falls just below or just above the cutoff and would likely have statistically equivalent chances of response. In this way, we are ignoring the probabilistic relationship associated with the continuous variables that can provide significantly more information at the point of care for clinicians, who themselves are facing ever-increasing challenges in determining a treatment plan amid the multitude of options available for their patients.
Closing the Gaps in Precision Medicine
In order to fully realize the future of precision medicine, we must acknowledge the that the current infrastructure falls short in several respects. First, we need the development of new, complex biomarkers to address the increasing complexity in treatment modalities that, in and of themselves, have characteristics of a continuous variable; they require innovation and outcomes data, which perhaps will be partly addressed by some of the emerging real-world evidence databases amassed by pairing sequence information and clinical outcomes.18 TMB is a great example of this innovation in practice; however, multimodal signatures that account for genomic predictors of response, as well as potential resistance, will help to more fully address this gap. Second, clinical trials that rely upon these signatures should begin to adopt methods that account for the probabilistic nature of the biomarker, rather than force-fi tting them into discrete biomarker subtypes based upon a single cutoff . Predefi ned end points—such as hazard ratios that are based on the classification of cohorts according to 2 definitions, high versus low or positive versus negative—will fail to capture the full potential of the genomic signature—associated outcomes. New trial designs, like adaptive clinical trials,19 will be required to more fully account for these complex signatures. Third, health authorities need new models and statistical tools with which to make regulatory decisions, ones that aren’t confined to predefi ned end points based on the comparison of 2 cohorts. This will require new, innovative approaches for the FDA to consider when evaluating the outcomes of a trial. Complementary diagnostics are the closest defi nition to this type of strategy, but when approved as such, they aren’t required to inform treatment decisions. Finally, device companies need to integrate the probabilistic nature of these complex signatures into medical reports in a way that captures the totality of the information clearly and concisely, ensuring that the best point-of-care decisions are made, despite the complexity of treatment options. Addressing these gaps will require a harmonized approach that includes stakeholders across all segments of the industry, including biopharma, diagnostic companies, academia, health authorities, and patient advocacy organizations. Only then can we fully realize the potential of precision medicine and help to shepherd it through the some of the inevitable awkwardness that comes with the transition into biomarker adulthood.Author Information
David A. Fabrizio, BS, is the vice president of translational strategy at Foundation Medicine.References
1. Frampton GM, Ali SM, Rosenzweig M, et al. Activation of MET via diverse exon 14 splicing alterations occurs in multiple tumor types and confers clinical sensitivity to MET inhibitors. Cancer Discov. 2015;5(8):850-859. doi:10.1158/2159-8290.CD-15-0285
2. Shrock AB, Frampton GM, Herndon D, et al. Comprehensive genomic profiling identifies frequent drug-sensitive EGFR exon 19 deletions in NSCLC not identified by prior molecular testing. Clin Cancer Res. 2016;22(13):3281-3285. doi:10.1158/1078-0432.CCR-15-1668
3. FDA grants accelerated approval to pembrolizumab for first tissue/site agnostic indication. News release. FDA; May 23, 2017. Updated May 30, 2017. Accessed May 15, 2020. fda.gov/drugs/resources-information-approved-drugs/fda-grants-accelerated-approval-pembrolizumab-first-tissuesite-agnostic-indication
4. FoundationOne CDx technical information. Roche Foundation Medicine; Cambridge, MA. rochefoundationmedicine.com/f1cdxtech.
5. Carbone DP, Reck M, Paz-Ares L, et al; CheckMate 026 Investigators. First-line nivolumab in stage IV or recurrent non-small-cell lung cancer. N Engl J Med. 2017;376(25):2415-2426. doi:10.1056/NEJMoa1613493
6. Hellmann M, Paz-Ares L, Bernabe Caro R, et al. Nivolumab plus ipilimumab in advanced non-small cell lung cancer. N Engl J Med.
7. Le DT, Uram JN, Wang H, et al. PD-1 blockade in tumors with mismatch-repair deficiency. N Engl J Med. 2015;372(26):2509-2520. doi:10.1056/NEJMoa1500596
8. Fabrizi DA, George TJ Jr, Dunne RF, et al. Beyond microsatellite testing: assessment of tumor mutational burden identifies subsets of colorectal cancer who may respond to immune checkpoint inhibition. J Gastrointest Oncol. 2018;9(4):610-617. doi:10.21037/jgo.2018.05.06
9. Johnson DB, Frampton GM, Rioth MJ, et al. Targeted next generation sequencing identifies markers of response to PD-1 blockade. Cancer Immunol Res. 2016;4(11):959-967. doi:10.1158/2326-6066.CIR-16-0143
10. Kowanetz M, Zou W, Shames D, et al. OA20.01 Tumor mutation burden (TMB) is associated with improved efficacy of atezolizumab in 1L and 2L+ NSCLC patients. J Thorac Onc. 2017;12(1):S321-S322. doi:10.1016/j.jtho.2016.11.343
11. Hellmann M, Ciuleanu TE, Pluzanski A, et al. Nivolumab plus ipilimumab in lung cancer with a high tumor mutational burden. N Engl J Med. 2018;378(22):2093-2104. doi:10.1056/NEJMoa1801946
12. Socinski M, Velcheti V, Mekhail T, et al. Final efficacy results from B-F1RST, a prospective phase II trial evaluating blood-based tumour mutational burden (bTMB) as a predictive biomarker for atezolizumab (atezo) in 1L non-small cell lung cancer (NSCLC). Ann Oncol. 2019;30(suppl 5):v919-v920. doi:10.1093/annonc/mdz394.081
13. Marabelle A, Fakih MG, Lopez J, et al. Association of tumor mutational burden with outcomes in patients with select advanced solid tumors treated with pembrolizumab in KEYNOTE-158. Ann Oncol. 2019;30(suppl 5):v475-v532. doi:10.1093/annonc/mdz253
14. Lemery S, Keegan P, Pazdur R. First FDA approval agnostic of cancer site—when a biomarker defines the indication. N Engl J Med. 2017;377(15):1409-1412. doi:10.1056/NEJMp1709968
15. Merino DM, McShane LM, Fabrizio D, et al; TMB Harmonization Consortium. Establishing guidelines to harmonize tumor mutational burden (TMB): in silico assessment of variation in TMB quantification across diagnostic platforms: phase 1 of the Friends of Cancer Research TMB Harmonization Project. J Immunother Cancer. 2020;8(1). pii:e000147. doi:10.1136/jitc-2019-000147
16. Tissue agnostic TMB clinical cut-off harmonization initiative. Friends of Cancer Research. Accessed May 15, 2020. focr.org/sites/default/files/Tissue-Agnostic-TMB_Summary.pdf
17. Stenzinger A, Allen JD, Maas J, et al. Tumor mutational burden standardization initiatives: recommendations for consistent tumor mutational burden assessment in clinical samples to guide immunotherapy treatment decisions. Genes Chromosomes Cancer. 2019;58(8):578-588. doi:10.1002/gcc.22733
18. Singal G, Miller PG, Agarwala V, et al. Association of patient characteristics and tumor genomics with clinical outcomes among patients with non—small cell lung cancer using a clinicogenomic database. JAMA. 2019;321(14):1391-1399. doi:10.1001/jama.2019.3241
19. Adaptive Platform Trials Coalition. Adaptive platform trials: definition, design, conduct and reporting considerations. Nat Rev Drug Disc. 2019;18(10):797-807. doi:10.1038/s41573-019-0034-3. Published correction appears in Nat Rev Drug Disc. 2019;18(10):808. doi:10.1038/s41573-019-0045-30