Recently, two diagnostic tissue-based assays were developed and clinically validated to identify patients likely to respond to anti-PD-1 antibody therapies. Both assays evaluate PD-L1 expressing cell frequencies in neoplastic cells via a manual estimate, and patients with above threshold levels of %PD-L1+ cells are recommended/eligible for receiving therapy. PD-L1 expression in tissues, such as non-small cell lung cancer or melanoma, can be heterogeneous in staining intensity and have variable distribution throughout an individual tissue. The heterogeneity in staining intensity and the necessity to estimate a total percent positive endpoint value for PD-L1+ cells over tissues with non-uniform distribution of positive cells can pose challenges for manual assessment by a pathologist due to known biases and limitations.
Tissue image analysis (tIA) is a potential tool to complement the reviewing pathologist’s skill, and can enable objective and quantitative assessment of cells within a tissue. In the present study, Flagship performed an analytical validation study of a tIA analysis solution and workflow to evaluate PD-L1 expression in tissues stained with the Dako pharmDx® PD-L1 (28-8) IVD assay. tIA solutions were developed under the oversight of the study pathologist, and analysis mark-ups were reviewed on each slide by a skilled pathologist to confirm appropriate identification of neoplastic cells within the tissue. In parallel, each tissue was scored according to the manual paradigm outlined by the manufacturer (note: the manufacturer had not provided specific training on the scoring paradigm at the time of this study). Assessment of the tissues by tIA demonstrated concordance with the manual scoring paradigm and the tIA scoring approach had improved reproducibility compared to the manual scoring paradigm. The results of this study suggested that tIA scoring can be used to complement the reviewing pathologist’s skill by providing a consistent and objective assessment of PD-L1 expression on a cell-by-cell basis.