IPM Take
This is precision oncology correcting itself in public, which is exactly what a mature evidence system should do. A threshold that becomes familiar can quickly start looking biologically absolute. FDA’s revision shows why accelerated approvals need confirmatory evidence and why molecular reports should show the underlying value, not only the label “TMB-high.”
Executive Summary
FDA converted pembrolizumab’s tissue-agnostic TMB-high indication to traditional approval while redefining TMB-high from ≥10 to ≥13 mutations per megabase. FDA stated explicitly that efficacy sufficient to support traditional approval had not been confirmed in tumours with TMB between 10 and <13 mut/Mb. Among 220 patients with tumours at or above the revised threshold in KEYNOTE-158, objective response rate was 36%, with 58% of responders maintaining response for at least 12 months.
Why it matters
- Patients / advocates: A few mutations per megabase can now change access to a tissue-agnostic therapy.
- Diagnostics / pathology: Assay methodology and the actual numerical TMB value matter.
- Clinicians: TMB should not be interpreted as a biologically universal binary marker.
- Regulators: Confirmatory evidence can legitimately narrow an accelerated indication.
The FDA’s updated pembrolizumab TMB-high approval is a small regulatory change with a much larger precision-oncology lesson. When pembrolizumab first received accelerated approval for this tissue-agnostic indication, the threshold was set at 10 mutations per megabase. Confirmatory evidence has now failed to establish sufficient benefit in the 10 to <13 mut/Mb range, so traditional approval begins at 13. The implication is not that TMB has suddenly changed biologically; it is that the evidence supporting the clinical decision has become more precise.
That distinction matters because molecular thresholds rapidly become embedded into clinical software, diagnostic reports, guidelines, reimbursement criteria and tumour-board language. Once that happens, a number can begin to look like a natural boundary rather than a cut-off derived from imperfect evidence. TMB has always been especially vulnerable to this problem because different assays, tumour types and mutational processes can produce similar numerical values with different biological implications. A patient at 12 mut/Mb is not biologically transformed by becoming 13 mut/Mb, even though the regulatory pathway necessarily has to draw a line somewhere.
The importance of this decision is therefore methodological as much as therapeutic. FDA’s verified accelerated-approval record for pembrolizumab shows the pathway functioning as intended: grant earlier access when evidence is promising, then use additional data to verify, amend or narrow the original claim. Precision medicine becomes more credible when it is willing to revise its definitions rather than defend them indefinitely.

