Why TMB ≥10 mutations/Mb is not a universal biological line
The FDA’s tumour-agnostic TMB-high indication for pembrolizumab used a particular assay and counting rule. The same numeric threshold cannot simply be copied to every sequencing panel or every tumour type.
Written retrospectively and published on August 3, 2026.
Background reading, not clinical guidance. This describes published research; it is not a recommendation for any individual case, and the cited studies may have been superseded.
In KEYNOTE-158, previously treated advanced solid tumours with tissue tumour mutational burden of at least 10 mutations per megabase had a 29% objective response rate to pembrolizumab, compared with 6% below that cut-off. The single-arm study showed enrichment for response; it did not establish a randomised survival benefit.
The number 10 belonged to FoundationOne CDx, the assay used in that trial and approved as its companion diagnostic. Tumour mutational burden depends on panel size, which variants are counted, allele-frequency cut-offs, filtering and bioinformatic processing. It is a measurement, not one invariant biological substance.
Same number, different test, different classification
Harmonisation studies show that applying the same number across panels can misclassify patients. Other panels may need a lower or higher calibrated threshold to approximate the FoundationOne result, and small panels have more variability at clinically used cut-offs.
The threshold is also biologically imperfect. Within KEYNOTE-158, response was lower in the 10 to under 13 mutations/Mb group than at higher TMB. Some tumour types do not show the same relationship between mutation burden, immune infiltration and response. Regulatory adoption therefore differs across regions.
Comments
Threads are public and indexed — please keep personal medical data out of them. Found a clinical error? Open an issue instead.
Comments are not switched on yet — the backend still needs to be configured.