The desire to understand which patients with NSCLC will receive the most benefit from immune checkpoint inhibitors (ICIs) is what drives much of our thinking about diagnostic testing. We know that high PD-L1 scores, using tumor proportion scores of 50% or greater, is an excellent predictor for patients who will derive substantial benefit from single-agent ICIs. However, what we have also observed is that there are patients who do not have very high levels of PD-L1 but still derive significant benefit from ICIs. For many years, there has been a notion that high TMB is a predictor of response. The thinking is that, if there are more mutated proteins in a tumor, there is a higher likelihood that harnessing the immune response to treat the cancer will be an effective approach. 

One of the challenges of TMB is defining how it should be calculated and what numbers are representative of a high TMB. It is generally calculated by sequencing as much of the genome as you can and then dividing by the number of mutations that you identify. This is a good approach if you are able to do whole exome sequencing or large panel testing that examines DNA. With small panels of genetic testing, however, you are not going to sequence enough genes to be able to accurately determine the TMB; thus, you would have to perform another test. 

In addition, the clinical data to date have not really shown that using TMB is superior to PD-L1 testing in evaluating patients with NSCLC. Further, the field has moved toward combination therapy with results such as those from KEYNOTE-189, where pembrolizumab plus chemotherapy was associated with better outcomes compared with chemotherapy alone, regardless of PD-L1 status and potentially regardless of TMB status. So, at this point, my decisions regarding whether to give chemotherapy along with an ICI for a patient with no actionable genomic alterations are guided by PD-L1. In that regard, I do not think of TMB as an important additional piece of information at this time.