The PD-1/PD-L1 pathway mediates immunosuppression in the tumor microenvironment. Therapies targeting this pathway have been approved along with corresponding diagnostic assays for selected indications. These assays predict patient responses to therapy by measuring the percentage of
tumor cells (pembroluzimab, nivolumab) or immune cells (atezolizumab) that stain positive for PD-
L1. The AACR-ASCO-FDA Blueprint Project seeks to facilitate harmonization of PD-L1 tests in order
to enable their practical use, with the goal of building a method for objective, consistent, proper test interpretation. In response, we have developed computational Tissue Analysis (cTA) approaches that enable objective, consistent assessments of staining. This approach allows the evaluation of multiple cell types (eg, tumor epithelium, immune in ltrate) concurrently and ensures reproducible scoring for the various PD-L1 assays. The present study evaluated Flagship’s cTA approaches to scoring samples stained with the Dako PD-L1 pharmDx (28-8) immunohistochemistry complementary diagnostic assay. The current manual scoring method considers any patient with at least 1% PD-L1–positive tumor cells a candidate for Opdivo treatment. However, a trend toward increased overall survival was observed in patient groups with greater PD-L1 positivity (5% and 10%), suggesting the importance of reliably distinguishing between graded thresholds to predict patient response to nivolumab.
We developed a cTA approach to test the hypothesis that it could provide more consistent scoring around these low thresholds, which are challenging for pathologists. We stained 40 formalin- xed, para n-embedded non-squamous non–small cell lung cancer samples with the aforementioned Dako assay; a subset of these we stained on consecutive days to evaluate assay and scoring precision. The cTA strategy digitally identi ed tumor cells and quanti ed membrane staining intensity in a manner which best mimicked the manufacturer’s scoring guidelines.
The cTA membrane scoring approach for tumor cells met Flagship’s Clinical Laboratory Improvement Amendments (CLIA) validation criteria, with the manually scored assay as the reference standard. This approach was accurate and provided greater repeatability of the assay over the dynamic range of PD-L1–positive cell frequencies. Additionally, cTA enabled better reproducibility for intra- and inter- pathologist scoring of the tissues.
This study demonstrated the utility of using cTA approaches to assist the pathologist to consistently and objectively score PD-L1 expression in tumor cells. It also highlighted the challenge pathologists face in di erentiating between 1%, 5%, and 10% positivity. CTA approaches could be used for consistent scoring of a single assay or to understand di erences between assays in performance or relationship to clinical response.