What Is Tumor Mutational Burden, and Why Does It Predict Immunotherapy Response?

Tumor mutational burden (TMB) — the number of mutations packed into every megabase of a tumor's DNA — has become one of oncology's most closely watched biomarkers because it offers a biologically coherent, testable answer to a hard question: which patients are likely to benefit from checkpoint inhibitor immunotherapy? This article explains what TMB is, why more mutations can mean more targets for the immune system, how the FDA's 2020 tissue-agnostic approval of pembrolizumab for TMB-high tumors came about (and its honest regulatory status today), how TMB is actually measured in a lab, where it falls short as a predictor, and how it fits alongside PD-L1 expression and microsatellite instability testing in real-world decision-making.

What TMB Actually Measures

Tumor mutational burden is a count: the number of somatic (tumor-acquired, non-inherited) mutations found in a defined stretch of tumor DNA, expressed as mutations per megabase (mut/Mb). A foundational 2017 analysis of roughly 100,000 human cancer genomes in Genome Medicine mapped how TMB varies enormously by cancer type — from very low in tumors like pediatric gliomas to very high in cancers linked to strong mutagenic exposures, such as melanoma (driven by UV radiation) and smoking-associated lung cancer (Chalmers et al., Genome Medicine, 2017). TMB is not a measure of tumor size, stage, or aggressiveness on its own; it's a proxy for how genetically "noisy" a tumor's DNA has become, usually because of UV damage, tobacco carcinogens, prior chemotherapy, or a broken DNA-repair system.

The Biological Rationale: More Mutations, More Targets

The logic connecting TMB to immunotherapy response is fairly intuitive once you see the chain of events. Somatic mutations can alter the proteins a cell produces. Some of those altered proteins get chopped up and displayed on the tumor cell's surface as "neoantigens" — flags that look foreign to the immune system because they don't match the body's normal self-proteins. The more mutations a tumor carries, the more neoantigens it is statistically likely to generate, and the more potential targets are available for cytotoxic T cells to recognize. Checkpoint inhibitors like pembrolizumab work by blocking PD-1/PD-L1 signaling, a brake that tumors exploit to keep T cells from attacking. In a high-TMB tumor already displaying abundant neoantigens, releasing that brake gives primed T cells more to actually target — which is the mechanistic reason high-TMB tumors, as a population, tend to respond better to checkpoint blockade than low-TMB tumors (National Cancer Institute, Cancer Currents Blog, 2020).

The 2020 FDA Approval — and Its Status Today

On June 16, 2020, the FDA granted pembrolizumab (Keytruda) accelerated approval for adult and pediatric patients with unresectable or metastatic TMB-high (TMB-H) solid tumors — defined as ≥10 mutations per megabase by an FDA-approved test — that had progressed after prior treatment with no satisfactory alternatives. This was the second tissue-agnostic approval for pembrolizumab (after the 2017 MSI-H/dMMR approval) and meant a patient could qualify based on their tumor's mutation count regardless of where the cancer originated. The approval rested on KEYNOTE-158, a multicohort, open-label basket trial: among 102 patients identified as TMB-H across 10 tumor types, the objective response rate was 29%, with 57% of responses lasting at least 12 months and 50% lasting at least 24 months (FDA Approval Summary, PMC, 2021; FDA.gov, 2020). FoundationOne CDx was simultaneously approved as the companion diagnostic.

It's worth being direct about what "accelerated approval" means and what has — and hasn't — happened since. Accelerated approval is conditional: it's granted on a response-rate surrogate with the expectation that confirmatory trials will verify real clinical benefit (like survival) later. Pembrolizumab's TMB-H indication remains listed as an accelerated approval, with continued approval contingent on verification of clinical benefit in confirmatory trials — language that has persisted for years since 2020. That's a meaningfully different trajectory than pembrolizumab's earlier MSI-H/dMMR tissue-agnostic indication, which the FDA converted to full, traditional approval in March 2023 based on pooled data from over 500 patients across three trials (OncLive, 2023). The TMB-H indication has not yet made that transition, and independent commentary — including from Friends of Cancer Research and coverage in OncLive — has raised real questions about whether a single universal 10 mut/Mb cutoff generalizes well across very different cancer types (OncLive, "Tumor Agnostic Role of TMB Biomarker Faces Challenges").

How TMB Is Actually Measured

In practice, TMB is not measured by sequencing an entire genome for routine clinical care — that's expensive and slow. Instead, most clinical labs use comprehensive genomic profiling (CGP): targeted next-generation sequencing (NGS) panels that read several hundred cancer-relevant genes (FoundationOne CDx panels roughly 300+ genes) and extrapolate a mutations-per-megabase estimate from that sampled portion of the genome. Whole-exome sequencing, which reads all protein-coding DNA, is considered a research-grade reference standard but is impractical for routine same-week clinical turnaround. Because different panels cover different gene sets, use different bioinformatic filters (for example, how they handle germline variants or synonymous mutations), and calibrate their mut/Mb math differently, two labs testing the same tumor sample can report meaningfully different TMB values — a problem serious enough that Friends of Cancer Research launched a multi-phase TMB Harmonization Project specifically to align results across platforms (Friends of Cancer Research, TMB Harmonization Project; Annals of Oncology, Phase II results, 2021).

The Real Limitations

TMB is a genuinely useful signal, not a guarantee. Even within the KEYNOTE-158 population that defined the FDA approval, the majority of TMB-H patients — roughly 71% — did not achieve a response; a 29% response rate is clinically meaningful in a heavily pretreated population with no good options, but it is far from a coin flip in the patient's favor. The NCI's own summary of the approval is blunt about this: most patients with TMB-H tumors did not benefit from pembrolizumab, and conversely some patients with tumors below the 10 mut/Mb threshold still responded. Response rates within KEYNOTE-158 also varied dramatically by tumor type, undercutting the idea of a single cutoff behaving identically everywhere. There are also documented edge cases where high TMB does not track with better outcomes at all — glioma is a notable example, where elevated TMB is sometimes a byproduct of prior chemotherapy-induced hypermutation rather than a sign of an immunologically "hot," attackable tumor, and has been associated with worse, not better, outcomes. Measurement inconsistency compounds all of this: without standardized panels and cutoffs, a patient's TMB-H or TMB-low classification can, in principle, depend partly on which lab and which assay was used.

How TMB Relates to PD-L1 and MSI-H/dMMR

TMB is one of several biomarkers oncologists use to estimate the odds a tumor will respond to checkpoint inhibition, and none of them work perfectly in isolation. PD-L1 expression, measured by immunohistochemistry on tumor or immune cells, reflects whether a tumor is actively exploiting the PD-1/PD-L1 brake — high expression is associated with better odds of response, but PD-L1-negative tumors can still respond, and expression can vary by biopsy site and timing. Microsatellite instability-high (MSI-H) and mismatch repair-deficient (dMMR) status identify tumors with a broken DNA proofreading system, which causes errors to accumulate rapidly — these tumors are, almost by definition, usually also TMB-high, and dMMR/MSI-H is itself an FDA-recognized, now fully approved, tissue-agnostic biomarker for pembrolizumab. But TMB-high and MSI-H are not synonymous: many TMB-high tumors are microsatellite-stable (their mutation burden comes from other sources like UV or tobacco exposure, not repair failure), and the two tests can classify the same patient differently. In practice, oncologists increasingly view PD-L1, TMB, and MSI-H/dMMR as complementary, partially overlapping pieces of evidence rather than a single decisive test — each captures a different facet of what makes a tumor visible and vulnerable to the immune system (Cancer Research Institute, Biomarkers in Cancer Immunotherapy).

Bottom Line

TMB testing is a legitimate, FDA-recognized tool for identifying patients who may benefit from checkpoint inhibitor immunotherapy when other options have run out — but it should be discussed with real numbers, not treated as a yes/no verdict. Patients and caregivers can reasonably ask their oncologist: What specific TMB test or panel was used, and what mut/Mb value did my tumor score? Was PD-L1 expression and MSI-H/dMMR status also tested, and do the results agree or conflict? Given that roughly 7 in 10 TMB-H patients in the pivotal trial did not respond, what is the actual expected likelihood of benefit for my specific cancer type, not just the overall trial average? Since this pembrolizumab indication remains under FDA accelerated approval pending confirmatory trial data, is there a clinical trial or an alternative, better-established biomarker-matched option that might be more appropriate? And practically: will insurance cover TMB testing, and how will the result change the actual treatment plan? A precise, tumor-type-specific conversation grounded in these questions will do far more than the TMB number alone.