Tumor mutational burden (TMB) has emerged as one of the most widely discussed biomarkers in immuno-oncology. Its biological rationale is intuitive: the more somatic mutations a tumor carries, the greater the potential number of neoantigens, increasing the probability of T-cell recognition and response to immune checkpoint inhibitors (ICIs).
When the FDA granted tissue-agnostic approval to pembrolizumab for solid tumors with a TMB of at least 10 mutations/Mb, it effectively enshrined a single number as a pan-cancer passport to immunotherapy.
Yet in the years since, the field has learned that high TMB is a noisy, context-dependent surrogate, and treating it as a binary switch may obscure more than it reveals.
10-Minute Immunotherapy Infusion: Safe, Efficient Delivery of Nivolumab and Pembrolizumab
How Tumor Mutational Burden Is Measured Matters
TMB is not an intrinsic property of a tumor so much as an estimate that is strongly influenced by the assay used to measure it.
Whole-exome sequencing remains the reference standard, but clinical practice largely relies on targeted next-generation sequencing panels. Their size, gene content, sequencing characteristics, and bioinformatics pipelines can generate substantial variability in reported TMB.
Harmonization work from Friends of Cancer Research and QuIP has shown that panels smaller than 667 kb lose reliability, that failure to appropriately filter pathogenic and germline variants can inflate estimates, and that panel-derived values require statistical calibration to align with exome-based TMB.
Preanalytical factors further complicate interpretation. Formalin-fixed tissue older than five years can generate spurious deamination artifacts, while low tumor purity may attenuate TMB estimates. Blood-based measurements using circulating tumor DNA can both under- and overestimate tissue TMB depending on tumor shedding and subclonal architecture.
The practical consequence is important: the same tumor specimen may cross the 10 mutations/Mb threshold—or fail to cross it—depending on the assay and vendor used.
A biomarker whose classification can hinge on assay choice is a fragile foundation for a treatment decision.
Not All Mutations Are Created Equal
Even perfectly measured TMB conflates mutations that matter immunologically with those that do not.
Only a small minority of somatic variants ultimately generate peptides that are successfully processed, presented through HLA molecules, and recognized by T cells. Yet conventional TMB counts mutations indiscriminately.
It does not adequately capture neoantigen quality, clonality, or the integrity of antigen presentation.
A tumor may therefore be TMB-high while remaining functionally invisible to the immune system. Alterations such as B2M loss, HLA loss of heterozygosity, or defects in interferon signaling can interfere with effective immune recognition despite the presence of a large mutational burden.
This is why distinguishing clonal, immunogenic, non-immunoedited mutations from passenger events is important. The raw mutation count should also be interpreted against the tumor’s underlying mutational processes and co-mutation landscape rather than viewed as an isolated number.
In other words, two tumors with the same TMB may have profoundly different immunological potential.
The Cause of Hypermutation May Matter More Than the Count
Perhaps the strongest challenge to a universal TMB threshold comes from the biology underlying hypermutation itself.
In their analysis, Rousseau et al. dissected the tissue-agnostic TMB approval and showed that survival benefit in mismatch-repair–proficient tumors was largely confined to cancers driven by established mutagens, including ultraviolet radiation in melanoma and tobacco exposure in lung and head and neck cancers.
Most other mismatch-repair–proficient TMB-high tumors did not demonstrate the same immunotherapy benefit.
The colorectal cancer experience further illustrates this complexity. An apparent association between TMB and immunotherapy response largely disappeared once mismatch-repair deficiency and POLE alterations were accounted for.
Similarly, TMB-high microsatellite-stable gastrointestinal cancers represent a heterogeneous population and often fail to respond despite meeting a numerical definition of high TMB.
These observations suggest that why a tumor accumulated mutations may be at least as important as how many mutations it contains.
The Tissue-Agnostic Threshold Has Biological Limitations
The evidence supporting tissue-agnostic TMB also requires context.
KEYNOTE-158, which supported the regulatory approval, did not include colorectal cancer, evaluated response rather than survival as the central efficacy signal, and demonstrated considerable variation in benefit across tumor histologies.
A universal threshold therefore risks imposing a single numerical definition on cancers with fundamentally different biology.
A TMB of 10 mutations/Mb in one tumor type may not carry the same immunological meaning as the identical value in another.
The concept of a tissue-agnostic threshold is clinically attractive because of its simplicity. Tumor immunogenicity, however, is not necessarily tissue-agnostic.
Where Does This Leave TMB in Clinical Practice?
None of these limitations mean that TMB is biologically irrelevant.
TMB represents a real and reproducible population-level signal. Across 27 tumor types, mutational burden has been correlated with response rates to immune checkpoint blockade, and TMB may retain clinical value in settings where stronger orthogonal biomarkers are unavailable.
The problem arises when a continuous and biologically complex variable is converted into a context-free binary decision.
Several principles can make its clinical interpretation more meaningful:
- Use the approved companion assay and its validated cutpoint and vendor, rather than treating TMB values generated by different platforms as interchangeable.
- Confirm blood-based high TMB in tissue before acting on it, when feasible.
- Interpret borderline values of approximately 10–13 mutations/Mb cautiously, as response rates are markedly lower than in clearly hypermutated tumors.
- Interpret TMB together with MSI/MMR status, the underlying driver mutational process, and tumor histology, rather than as an isolated biomarker.
The clinically relevant question is therefore not simply “Is the TMB ≥10 mutations/Mb?” but rather “What does this TMB value mean in the biological context of this particular tumor?”
Beyond a Single Number: The Future of Immunotherapy Biomarkers
The next generation of immunotherapy biomarkers will likely move beyond mutation counting toward composite biological models.
Such approaches may integrate neoantigen quality and clonality, antigen-presentation integrity, PD-L1 expression, tumor genomics, and the architecture of the tumor microenvironment.
This would better reflect the multiple biological steps required for successful antitumor immunity: generating immunogenic antigens, presenting them effectively, recruiting and activating immune cells, and maintaining their function within an immunosuppressive tumor microenvironment.
TMB captures only one part of that process.
The OncoDaily IO Perspective
TMB remains an important immuno-oncology biomarker, but its greatest weakness may also be the feature that made it attractive in the first place: its apparent simplicity.
A single numerical threshold cannot fully capture neoantigen quality, clonality, antigen presentation, the biological cause of hypermutation, tumor histology, or the immune context in which those mutations occur.
The future is therefore unlikely to be TMB versus another biomarker. More likely, TMB will become one component of increasingly integrated models that combine tumor genomics with immune biology and the tumor microenvironment.
Until then, TMB is best understood not as an answer, but as one imperfect variable in a still-unfinished equation.
The number is easy to report; its meaning is not.
Written by Ioannis Stouras, MD