The evaluation of biomarkers for assessing immunotherapy is somewhat different than the evaluation of biomarkers for assessing targeted therapy. With targeted agents, the genetic alterations are dichotomous variables, with the alteration being either present or absent. In contrast, with biomarkers for immunotherapy, there is more of an incremental effect. Increasing levels of response are seen as levels of the biomarkers increase. With PD-L1, approximately one-third of patients will have no detectable expression, and it is relatively rare for these individuals to respond to an ICI. As expression increases, response rates also increase so that when you reach 90% PD-L1 expression or higher, which occurs in approximately 10% to 15% of patients, you achieve response rates of 60%. 

It is a similar situation with TMB, which has emerged as a proxy for neoantigen burden. TMB is estimated using algorithms from next‐generation sequencing assays, and they extrapolate a number. Generally, the higher the TMB, the more likely the patient is to respond to ICI therapy. The cut-point for a high TMB is typically 10 somatic mutations per DNA megabase.

Evidence suggests that TMB burden in NSCLC changes as a function of smoking. Together with her team, my colleague Xinan Wang recently found that there was a dose-response association of smoking history with TMB in patients with NSCLC, with the number of mutations related to not only how much you smoked and how long you smoked but also to how long ago you quit. If you quit smoking years ago, you do not accumulate those additional mutations. Thus, identifying patients as being either current, former, or never smokers will give you incomplete information. Obtaining a better smoking history may improve the ability to predict response to checkpoint inhibition.

One of the challenges in predictive testing for available ICIs is that the use of biomarkers is different between agents. Studies of nivolumab and ipilimumab commonly use TMB as a potential predictive marker of response to ICI therapy, while studies of atezolizumab and pembrolizumab used PD-L1 values. The anti–programmed cell death protein 1/PD-L1 agents also have different clinical applications in the US Food and Drug Administration–approved indications for advanced NSCLC either alone, with chemotherapy, or after chemotherapy plus chest radiotherapy for stage III NSCLC. Whether these different ICI agents will produce similar outcomes if they are tested head to head in the same setting is difficult to predict. Thus, in the absence of such data, we generally use the different agents based on the experience from clinical trials (ie, in the clinical scenarios that led to their approvals).