Prevalence of PD L TPS
Prevalence of PD-L1 TPS ≥ 50% and PD-L1 TPS ≥ 1% was similar across geographic regions, between surgical specimens and biopsies, and irrespective of whether tissue was from the primary tumor or from metastases. Importantly, PD-L1 4μ8C results were available for >90% of patients eligible for the study. The consistency in results across geographic regions and high assay success rate are indicative of the high overall reliability and reproducibility of the PD-L1 IHC 22C3 pharmDx assay when performed by experienced trained pathologists. All contributing pathologists in EXPRESS were required to complete a training program, to standardize the technical procedures of staining, scoring, and interpreting PD-L1 expression in NSCLC at the TPS ≥ 1% and ≥50% cut points, as variability between pathologists’ assessment of PD-L1 expression can result in lower concordance in PD-L1 TPS classification . Notably, the pathologist interpretation training included ≥25% of cases around the PD-L1 TPS ≥ 1% and ≥50% cut points and covered the range of PD-L1 expression in NSCLC. Additionally, the majority of pathologists in the study had significant experience with PD-L1 testing because pembrolizumab had previously been approved in NSCLC, with obligatory PD-L1 testing, when the study was conducted. The prevalence of PD-L1 TPS ≥ 1% and PD-L1 TPS ≥ 50% in EXPRESS was broadly consistent with, albeit somewhat lower than, that reported with central testing of clinical trial screening populations in the KEYNOTE-001, KEYNOTE-010, and KEYNOTE-024 studies.  Because no assay variability was apparent, the pathologist training program was rigorous, and PD-L1 expression prevalence was consistent across regions, other potential factors may have contributed to this slightly lower prevalence, such as the inclusion of patients with sensitizing EGFR mutations and ALK translocations, who were either excluded from pembrolizumab clinical trials (KEYNOTE-024) and/or infrequently enrolled in these trials (KEYNOTE-001, KEYNOTE-010). Notably, we observed higher PD-L1 expression prevalence among patients without EGFR mutations or ALK translocations. Thus, given that PD-L1 expression is lower among patients with these genetic aberrations, inclusion of this patient group in our study may partially explain the difference in PD-L1 prevalence between EXPRESS and clinical trial populations. Notably, although the presence of driver mutations has been hypothesized to result in differences in tumor PD-L1 expression, a number of previous studies have evaluated associations between PD-L1 expression and EGFR mutations with equivocal results . The potential for referral bias, whereby patients with histologically/cytologically confirmed PD-L1–expressing tumors are more likely to be enrolled in clinical trials, might also partially explain the moderately higher PD-L1 prevalence reported in clinical trials. Notably, immunohistochemistry assays that use antibodies other than the 22C3 clone have been developed to evaluate PD-L1 expression and are commercially available (ie, 28-8, SP263, and SP142). An evaluation of these assays showed results that were typically (though not always) concordant with those from PD-L1 IHC 22C3 pharmDx kit . Laboratory-developed tests (LDTs) that employ one of these existing antibodies have also been shown to provide reliable assessments of PD-L1 expression when properly implemented . As demonstrated by Ilie and colleagues, high concordance can be achieved between LDTs and the PD-L1 IHC 22C3 pharmDx assay, at both the PD-L1 TPS ≥ 50% and TPS ≥ 1% cutpoints [17,18]. Together with the low testing failure rate in EXPRESS, these data demonstrate the feasibility of evaluating PD-L1 expression with 22C3 antibody-based assays in the real world to support treatment decisions.