What Tumor Mutational Burden Does and Does Not Tell You

Tumor mutational burden, usually shortened to TMB, is a count normalized by the amount of tumor DNA evaluated. A report may express it as 4, 10, or 25 mutations per megabase. The concept is that a tumor with more somatic mutations may produce more altered proteins, some of which can become neoantigens that the immune system recognizes.

That biological chain is plausible but incomplete. Not every mutation is expressed. Not every altered peptide is presented by human leukocyte antigen molecules. Not every presented peptide recruits an effective T-cell response. Tumors also suppress immunity in many ways.

TMB is therefore an indirect genomic biomarker. It can add useful information, especially in defined immunotherapy settings, but the number is inseparable from the test, specimen, cancer type, and treatment indication.

What is actually counted#

A tumor contains inherited germline variants and acquired somatic variants. TMB aims to count the somatic component. A laboratory sequences tumor tissue, and sometimes a matched normal sample, identifies eligible variants, filters likely germline changes and artifacts, then divides the mutation count by the number of megabases that met the assay's analytic requirements.

Whole-exome sequencing surveys most protein-coding regions. Clinical laboratories more often use targeted next-generation sequencing panels because they require less tissue, can achieve deeper coverage, and also identify actionable driver alterations. A panel-derived TMB is an estimate of a broader genomic burden.

The numerator differs among methods. Some include synonymous and nonsynonymous substitutions and short insertions or deletions; others use a narrower set. Known driver mutations may be treated differently. The denominator may be total panel territory or only adequately sequenced territory. The report should spell out the assay and the method.

Why mutation burden might predict immune response#

Checkpoint inhibitors remove inhibitory signals that restrain T cells. If a tumor presents recognizable mutation-derived neoantigens, releasing those brakes can permit an antitumor immune response. Cancers caused by ultraviolet light, tobacco carcinogens, mismatch-repair deficiency, or defective DNA polymerase proofreading can carry high mutation burdens.

Mutation quantity is only one link. The mutation must generate a peptide, the cell must express and process it, a compatible HLA molecule must present it, and functional immune cells must reach and attack the tumor. Antigen-presentation loss, T-cell exclusion, immunosuppressive cells, and other escape mechanisms can interrupt the pathway. That is why some TMB-low tumors respond and many TMB-high tumors do not. TMB is correlated with one aspect of immunogenic potential rather than a direct immune-function test.

The FDA tissue-agnostic indication#

In 2020, FDA granted accelerated approval to pembrolizumab for adult and pediatric patients with unresectable or metastatic TMB-high solid tumors, defined as at least 10 mutations per megabase by an FDA-approved test, whose disease progressed after prior treatment and who have no satisfactory alternative options.

The FDA approval summary describes the evidence. In 102 patients with TMB-high tumors across nine tumor types in KEYNOTE-158, the objective response rate was 29 percent, with a 95 percent confidence interval from 21 to 39 percent. Fifty-seven percent of responses lasted at least 12 months.

Those figures show both signal and limitation. A durable response occurred in a subset, while most patients did not have an objective response. The analysis was a prospectively planned retrospective biomarker analysis within a nonrandomized trial, not a randomized comparison showing survival benefit in every tumor type.

Accelerated approval is a precise regulatory status#

The current Keytruda label states that the TMB-high indication is approved under accelerated approval based on tumor response rate and response durability. Continued approval may depend on confirmation and description of clinical benefit.

FDA's ongoing cancer accelerated approvals page still lists the indication as ongoing on the source-check date for this draft, July 16, 2026. That is not the same as withdrawal or conversion to traditional approval; regulatory status can change, so verify the current label and the FDA listing before you act on it. The label conditions also matter. “TMB-high” alone does not establish eligibility. Disease must be unresectable or metastatic, have progressed after prior treatment, lack a satisfactory alternative, and be identified with an FDA-approved test for the indication.

Why 10 is not a universal biological switch#

The cutoff of 10 mutations per megabase serves a regulatory decision tied to evidence and an approved companion diagnostic. Nature does not change abruptly between 9.9 and 10.0. Analytic variability can move a borderline specimen across the threshold, and predictive value can differ across histologies.

Some cancer types normally have very low mutation burdens; 10 may be unusually high within that distribution. Others commonly have higher burdens, making the same value less distinctive. Tumor microenvironment, prior therapies, and the genomic process that generated the mutations can modify response.

A 2025 analysis of more than 8,000 patients across 24 cancer types examined real-world survival after single-agent checkpoint inhibition. It reinforces that platform consistency and cancer-specific context matter. A pan-cancer threshold should not erase tumor-specific evidence.

The companion diagnostic anchors the claim#

FDA approved a FoundationOne CDx supplement to identify TMB at 10 or more mutations per megabase for patients who may benefit from pembrolizumab. The PMA record and technical documentation describe a tissue assay using formalin-fixed, paraffin-embedded specimens and targeted sequencing.

An approved test is more than a gene list. Its specimen requirements, wet-laboratory procedure, sequencing performance, variant pipeline, TMB calculation, quality controls, and clinical cutoff form one validated system. A numerically similar result from another platform is not automatically analytically equivalent for the FDA-labeled claim. Laboratory-developed tests may provide clinically useful TMB estimates in other settings, but their reports require attention to validation, calibration, and the treatment framework. You need to know whether the result came from the companion diagnostic the label specifies.

Why two assays can disagree#

Panel size matters because small panels sample fewer mutations and produce less stable estimates, particularly near a cutoff, and gene selection matters because cancer panels enrich regions that are not a random genomic sample. Inclusion or exclusion of synonymous changes and driver hotspots affects the numerator.

Germline filtering matters. Matched normal DNA can directly identify inherited variants, while tumor-only pipelines infer them from population databases and variant features. Under-filtering inflates TMB; over-filtering can remove true somatic variants. Ancestry underrepresentation in germline databases can produce unequal error. Tumor purity, DNA damage from fixation, sequencing depth, minimum allele fraction, artifact suppression, and bioinformatic version also matter. The 2024 AMP, CAP, and SITC consensus recommendations call for transparent validation and reporting across preanalytic, analytic, and postanalytic stages.

Tissue quality and tumor heterogeneity#

A small biopsy may contain few malignant cells, necrosis, inflammatory tissue, or damaged DNA, and low purity reduces the fraction of tumor-derived variants and can make mutations harder to detect. Laboratories set minimum quality criteria and may report failure, reduced confidence, or limitations.

One specimen also samples one place and time. Metastases can differ from the primary tumor, and treatment can select new clones. A years-old specimen may not fully represent current disease. Rebiopsy can provide newer tissue but adds procedural risk and may still miss spatial variation.

Blood-based TMB attempts to use circulating tumor DNA. It has different analytic behavior and depends on how much tumor DNA is shed into plasma. Tissue TMB and blood TMB are related but not interchangeable unless the specific assay and clinical evidence support that use.

Relationship to MSI and mismatch-repair deficiency#

Mismatch-repair deficiency can cause microsatellite instability and a high mutation burden. FDA has a separate tissue-agnostic pembrolizumab indication for MSI-high or mismatch-repair-deficient tumors. The biomarkers overlap but are not synonyms.

A tumor can be TMB-high without MSI-high because ultraviolet damage, tobacco-associated mutagenesis, POLE or POLD1 proofreading defects, or other processes generated mutations, and a tumor can also meet an MSI or mismatch-repair biomarker indication even if a particular panel's TMB estimate does not cross 10. Reports should display both when tested, and the treatment team should use the biomarker linked to the strongest evidence and current label for that cancer and setting, not assume that one result substitutes for another.

Relationship to PD-L1 and driver alterations#

PD-L1 immunohistochemistry measures protein staining in tumor and immune cells using indication-specific scoring systems. It captures a different part of tumor-immune biology from TMB. The markers can agree or conflict, and neither one universally overrides the other.

Oncogenic drivers can be more actionable than a burden score, and a lung cancer with an EGFR or ALK alteration, for example, may have a targeted therapy sequence and a different expected checkpoint-inhibitor response. Histology-specific guidelines integrate these facts. Treatment choice also includes disease tempo, autoimmune conditions, organ transplant, prior immunotherapy, performance status, toxicity risk, and patient preferences. Biomarkers narrow uncertainty; they do not make the decision alone.

Reading a TMB report#

Start with the specimen date, site, tumor content, assay name, and whether tissue or blood was tested. Identify the numerical score, categorical call, cutoff, and quality limitations. Ask what mutation classes and genomic territory were used and how germline variants were filtered.

Then connect the report to a specific treatment label or guideline. Confirm cancer stage, prior therapy, alternatives, and whether the test is approved for the intended claim. Review MSI or mismatch-repair status, PD-L1 when relevant, actionable genomic alterations, and contraindications.

For a borderline result, avoid false precision. Analytic variation, tumor sampling, and platform differences may be clinically relevant. Molecular tumor board review, or a conversation with the laboratory, can clarify for you what the number supports.

What TMB cannot promise#

TMB cannot promise you response, duration, survival improvement, or absence of immune toxicity. It cannot tell you whether every mutation is clonal, expressed, or immunogenic, and it does not measure T-cell function or tumor immune evasion directly.

It also should not be used to describe a tumor as more dangerous merely because the number is high. Prognostic meaning, the likely untreated course, and predictive meaning, differential benefit from a therapy, are different questions.

The responsible summary is conditional: this assay found a burden estimate that may support a defined immunotherapy option, subject to regulatory status, tumor-specific evidence, other biomarkers, prior treatment, and the patient's circumstances.

References#

1. FDA approval summary for pembrolizumab in TMB-high solid tumors
2. Current FDA Keytruda prescribing information
3. FoundationOne CDx TMB companion diagnostic PMA
4. AMP/CAP/SITC TMB assay recommendations
5. TMB and survival across 24 cancer types
6. FDA ongoing cancer accelerated approvals