Yong Hee LeeGrace DyPaul DePietroJeffrey ConroySarabjot Pabla and Mary Nesline


Background PD-L1 immunohistochemistry (IHC) testing is suboptimal for predicting patient clinical benefit for checkpoint inhibition, while PD-L1 liquid biopsy is not clinically validated and lacks sensitivity, underscoring the need to include PD-L1 testing in more robust, tissue-efficient, comprehensive, scalable next generation sequencing (NGS) tests.

Methods To assess comparability and efficacy of PD-L1 testing by NGS with IHC, we identified NSCLC patients treated by first-line pembrolizumab alone (n=54) or pembrolizumab + chemotherapy (n=49) whose tumors underwent companion diagnostic PD-L1 testing by IHC antibody 22C3 testing (high≥50%; low=1–49%, or negative=0% tissue proportion score), and also by RNA-seq, as part of a comprehensive immune profiling panel. PD-L1 expression by RNA-seq, was measured as a percentile rank, with ≥75 considered ‘high’, and <75 considered ‘not high’, based on comparison to a reference population and normalized to a value of 1–100. All testing was performed in a CLIA certified laboratory prior to treatment initiation (any line) at Roswell Park Comprehensive Cancer Center (June 2017-March 2019, with a minimum of 1 year of follow up). Assay equivalence was assessed by proportion analysis using Fisher exact test comparing IHC versus to RNA-seq, and Bonferroni pairwise post-hoc analysis of IHC (high vs. low, high vs. negative, low vs. negative) with RNA-seq (high vs. not high). A Cox regression model evaluated associations between IHC and RNA-Seq with OS from first dose of pembrolizumab.

Results More than 75% of IHC high cases were classified as high by RNA-Seq for both treatment groups (p<0.001). Post-hoc pairwise comparisons showed PD-L1 IHC and RNA-Seq ‘high’ results were significantly associated with each other, and PD-L1 IHC low/negative results were associated with RNA-seq ‘not high’ results. In the pembrolizumab monotherapy group, RNA-seq high was associated with improved survival for pembrolizumab compared to RNA-seq not high status (HR=3.96; CI=1.22–12.87; p=0.02), while PD-L1 IHC high status was not associated with survival benefit in this group (p=0.63). In the pembrolizumab + chemotherapy group, as expected, neither IHC (high versus low), nor RNA-seq (high versus not high) status was associated with survival benefit (p=0.81 and p=0.76, respectively). These findings are consistent with our previous work demonstrating PD-L1 RNA-seq was predictive of CPI response in multiple tumor types.

Conclusions PD-L1 status by RNA-seq and IHC appear to be comparable. Unlike PD-L1 IHC however, PD-L1 RNA-seq high status versus not high status is associated with greater survival benefit, indicating PD-L1 by NGS may have utility for pembrolizumab selection.