Since we last reported on the rise of tumor mutational burden (TMB) in immuno-oncology (I/O) a year ago, there have been numerous material developments of this biomarker in the I/O space. Just in the past year, we have seen advancements, setbacks, and added complexity to this marker.

In our initial TMB blog entry from last year, we posed several questions as a means of thinking about the future trajectory of this technology. Revisiting these same questions a year later sheds light on the growth we have already seen in this space while highlighting future unknowns and unresolved hurdles.

Will TMB testing bring the same challenges as PD-L1?

While PD-L1 testing has become commonplace with I/O treatment, the multitude of available antibodies, cutoffs, and scoring algorithms make this marker anything but uniform and consistent. Developments over this past year in the diagnostics market hint that TMB may face the same variability challenges as PD-L1; the results from one TMB test may not be interchangeable with the next. At SITC 2018, an abstract co-authored by numerous members of the Friends of Cancer initiative1 outlined how in-silico TMB assessments of 11 panel-based diagnostic tests correlated well with a WES ‘gold-standard’. However, when comparing panel-based tests against one another across indications, their interchangeability warranted further research since panel tests may be prone to indication-specific variations. More recently, research from AACR this past month2 reveals that, when compared to whole exome sequencing, TMB panel tests provide highly varied results, justifying either a move to WES to ensure consistency of results or rethinking the scoring algorithm applied.

Continuing on the vein of test and cut-off variability, research published by Memorial Sloan Kettering earlier this year3 presents data on 7,000+ cancer patients across indications who underwent TMB testing with the MSK-IMPACT test. Key findings highlight that while TMB-high cancers correlated with improved outcomes for multiple tumor types, the TMB cutoffs to predict this survival benefit varied significantly across indications, making a single pan-cancer TMB cutoff unlikely.

Main Takeaways

Between factors including the wide variety of TMB tests available today (see Table 1), questionable interchangeability between tests, and variability in scoring across indications, it is conceivable that TMB, like PD-L1, will face the challenge of segmentation across indications and therapies. However, many of the TMB concordance studies to date have been conducted in-silico and the need for further validation is acknowledged.

Despite the PD-L1-esque trend towards segmentation of tests, there is a key area where TMB does not experience the same challenges as PD-L1: quantitative analysis. As pointed out in work conducted by Foundation Medicine4, TMB is a separate biomarker that is orthogonal to PD-L1, and by nature of the fact that TMB is quantifiable, it enables more objective analysis compared to the subjective interpretation of PD-L1 immunoassays.

Will blood-based TMB overrule tissue sampling?

With liquid biopsy tests continuing to attract significant research, investment, and media attention, it is still uncertain if blood-based analyte tests will complement tissue testing or simply replace it. There is mounting evidence to suggest that blood provides separate valuable information on a patient’s disease process, such as a systemic view of the disease’s reach and the body’s immune response. Conversely, the growth of technologies like Definiens’ digital pathology offerings and NanoString’s GeoMx has bolstered the value of profiling the intact tumor microenvironment. A future where both tissue and blood play valuable roles in a patient’s therapeutic journey is certainly conceivable, but when it comes to TMB, ensuring concordance regardless of sample type is a hurdle that must still be overcome.

From research previously conducted at Northwestern University5, concordance between tissue-TMB (tTMB, FoundationOne) and blood-TMB (bTMB, from Guardant360) were low, calling into question if bTMB derived from ctDNA could reliably be used to screen for this marker. It’s crucial to note that the G360 test only is ~70 , potentially limiting its TMB detecting capabilities regardless of sample type. More recent data6 compared TMB-matched outcomes between Foundation’s F1 test and their 394 gene bTMB panel, the results of which showed that tissue and blood have an overall agreement of ~80%. While cutoffs were not defined in this Foundation study, new data from this past month7 highlights an approach where bTMB analysis with the GuardantOMNI and tTMB analysis with F1 had separate cutoffs (16 Mut/Mb and 10 Mut/Mb, respectively). Not only did both the tissue and blood TMB detection methods help predict immunotherapy treatment response, they also correlated to one another. Evidence such as this is key to building a case that concordance between TMB measured in different analytes may be achievable.

One final factor to consider is explored in a 2019 study led by Merck where bTMB analysis from both FMI and Guardant health are compared to tTMB as measured in WES8. In this study, bTMB and tTMB were highly correlated in TMB high patients, but poorly correlated when patients were TMB low. The study concludes that there may be distinct biological factors leading to discrepancies between bTMB and tTMB and each should be evaluated independently. Furthermore, research from the recent AACR conference9 suggests that cell-free DNA (cfDNA) may contain a more complete view of a tumors heterogeneity compared to a tissue biopsy, and it may be appropriate to profile both to effectively characterize acquired resistance mechanisms.

Main takeaways

With the rise of large panel ctDNA tests, historical limitations associated with small genomic coverage are less likely to be a barrier in attaining bTMB results. The body of evidence in the diagnostics market comparing these techniques continues to grow and there is the potential that, like other liquid biopsy markers, bTMB will offer unique information that complements the results of tTMB. In short, rather than blood overruling tissue testing in this marker (or others), we see mounting evidence that patients should receive both tumor and blood NGS testing to provide the most complete picture of their overall disease.

Do driver mutation panel tests confer TMB detection bias?

Over the past year, the list of companies offering TMB-enabled panel tests continued to rise, with at least 14 known tests currently available or under development. (See the table below for a list of clinical TMB assays. If we’ve missed your clinical TMB test, let us know and we'll add it here.) With the exception of a single test using WES (Tempus xE), the remaining commercial tests are panels of 300+ genes, all of which are designed around profiling driver mutations and genes associated with oncogenesis.

ManufacturerTest NameTotal Genes
Tissue
TempusTempus xEWES
TempusTempus xT596
CarisMI Tumor Seek592
IlluminaTrusight 500523
PGDxPGDx Elio500+
QiagenQiAseg TMB Panel486
Memorial Sloan KetteringMSK-IMPACT468
KEWCancerplex435
ThermoFisherOncomine TMB409
NeogenomicsNeoTYPE Discovery Profile326
Foundation MedicineF1CDx324
Blood
Guardant HealthGuardant OMNI500
Foundation MedicinebTMB Assay (trade name TBD)394
PredicinePredicineATLAS600

At AACR this past month, an abstract published by Roche2 questions the consistency of panel-based TMB analysis given the enrichment of driver and hotspot mutations as well seemingly arbitrary cutoffs for TMB-high calls. The authors in this study propose a new scoring system that they call ‘ecTMB’. This system takes the broader heterogeneous mutation context into account by normalizing TMB scores from panels to align with WES results.

As an extreme example of the discrepancy that can arise in TMB scoring between targeted vs WES analysis, NantOmics presented data at ASCO 201810 in which a 248-gene panel overestimated TMB by three-fold compared to WES. NantOmics cautioned against relying on TMB estimates from panels with less than 500 genes due to this overestimation. This sentiment was shared in a recent ESMO publication11 wherein the authors claim that panels of 0.8 Mb or more are needed to reduce unwanted bias in genomic coverage, although other research puts this cutoff at 1.1 Mb, or even upwards of 1.5 Mb to ensure accurate results.

Main Takeaways

Despite the price of sequencing decreasing significantly in recent years, WES is still largely seen as impractical for routine clinical testing. In the near-term, panel tests will bring TMB testing to the diagnostics market, the most reliable of which will likely attempt to reduce bias by including large gene sets and/or adjusting their TMB algorithms to account for tumor suppressor genes and germline mutations4. Overall, our findings support that panel tests containing oncogenes can still produce accurate TMB results as long as there is sufficiently large genomic coverage that includes non-cancer driver regions.

Conclusion

From the findings discussed above, it is clear TMB has undergone significant developments over this past year alone, emerging as a new ubiquitous biomarker in the I/O space. There is somewhat of a race underway as numerous diagnostics companies position their tests to address multiple indications and sample types, but it is possible that in the near-term many offerings will coexist like we have seen with PD-L1.

It is likely that near-term, patients will either receive either a single tissue or blood TMB panel-test, but we are eager to see if future reimbursement guidelines will allow patients to receive multiple NGS tests given the mounting evidence of valuable clinical insights attainable from local and systemic samples. And while WES may be a key evolutionary step in profiling this biomarker and improving concordance, TMB is fundamentally a proxy marker for tumor immunogenicity, and may well give way to more direct measures in the future (i.e. T-cell repertoire, neoantigens).

Author | Colin Enderlein

Colin is an Associate at DeciBio with a background in industry CDx commercialization, tech transfer, and preclinical translational research. At DeciBio, many of the projects Colin is involved with relate to I/O therapies and their associated biomarkers, spanning both the pharmaceutical and genomic tools markets. Connect with him on LinkedIn.

Citations
  1. Abstract O48 – In Silico Assessment of variation in TMB quantification across diagnostic platforms: Phase 1 of the Friends of Cancer Research Harmonization Project; https://higherlogicdownload.s3.amazonaws.com/SITCANCER/7aaf41a8-2b65-4783-b86e-d48d26ce14f8/UploadedImages/Annual_Meeting_2018/Annual_Meeting/Abstracts/Abstract_Book_Edited_11_20.pdf
  2. LB-213 / 5 - ecTMB: A robust method to estimate and classify tumor mutational burden; https://www.abstractsonline.com/pp8/#!/6812/presentation/9085
  3. Tumor mutational load predicts survival after immunotherapy across multiple cancer types; https://www.nature.com/articles/s41588-018-0312-8
  4. Approach to evaluating tumor mutational burden in routine clinical practice; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6249621/
  5. Comparison of tumor mutational burden (TMB) across tumor tissue and circulating tumor DNA (ctDNA); https://ascopubs.org/doi/abs/10.1200/JCO.2017.35.15_suppl.e23028
  6. Abstract 5706: A blood-based next-generation sequencing assay to determine tumor mutational burden (bTMB) is associated with benefit to an anti-PD-L1 inhibitor, atezolizumab; http://cancerres.aacrjournals.org/content/78/13_Supplement/5706
  7. CT074 - Tumor mutational burden (TMB) as a biomarker of survival in metastatic non-small cell lung cancer (mNSCLC): Blood and tissue TMB analysis from MYSTIC, a Phase III study of first-line durvalumab ± tremelimumab vs chemotherapy; https://www.abstractsonline.com/pp8/#!/6812/presentation/9830
  8. Measuring Tumor Mutational Burden (TMB) in Plasma from mCRPC Patients Using Two Commercial NGS Assays; https://www.nature.com/articles/s41598-018-37128-y#ref-CR4
  9. Tumor Biopsy Found Less Effective Than Liquid Biopsy for Identifying Resistance Mechanisms in Some GI Cancers; https://www.cancertherapyadvisor.com/home/news/conference-coverage/american-association-for-cancer-research-aacr/aacr-2019/gastrointestinal-cancer-tumor-biopsy-less-effective-liquid-identifying-resistance/
  10. LB-213 / 5 - ecTMB: A robust method to estimate and classify tumor mutational burden; https://www.abstractsonline.com/pp8/#!/6812/presentation/9085
  11. Three-fold overestimation of tumor mutation burden using 248 gene panel versus whole exome; https://ascopubs.org/doi/abs/10.1200/JCO.2018.36.15_suppl.12117
  12. Implementing TMB measurements in clinical practice: considerations on assay requirements; https://esmoopen.bmj.com/content/esmoopen/4/1/e000442.full.pdf