A new paper by Samstein et al. (Memorial Sloan Kettering) reports findings in 1600 patients over a range of cancer types.
Key findings include high overall value of TMB as a predictive biomarker of response, but also, substantial differences among cancer types. Expressing TMB as mutations per megabase, the threshold for decisions may vary considerably among cancers. They recommend setting different cutoffs based on quintiles for different cancer types. In addition, Samstein et al. found that above a certain level, TMB's predictive impact asymptotes.
- Samstein et al. here (Nature Genetics).
- Coverage at STAT PLUS (subscription), here. Also Genomeweb (subscription), here.
- It got coverage at Forbes also, open access, here.
- Prior landmark paper by Chalmers et al. with 100,000 patients, here.
Harmonization. A TMB harmonization project is at Friends of Cancer Research, FOCR, here. Somewhat comically, Genomeweb notes there are multiple different harmonization committees at work in different places, here. I think my first blog on TMB was about two years ago, here.
Tidbit. Regarding FDA's position in November 2017, in FMI F1 CDx test paperwork, P170019C, statement that "the clinical validity of TMB defined by this panel has not been established." Here.
|Top, Kaplan Meier, TMB levels set "within histology" then recombined.|
Bottom, TMB impact flattens out (as hazard ratio) above 15-20.
|Forest Plots for whole cohort (top line) and subgroups|