Abstract
Simple Summary Diagnostic PD-L1 assessment of urothelial cancer to predict a patient's immune therapy response remains a matter of controversy. Several contributing factors have been discussed; however, systematic studies are lacking. The present study demonstrates that clinically applied PD-L1 scoring algorithms are influenced by inter-algorithm variability and result in the selection of ...
Abstract
Simple Summary Diagnostic PD-L1 assessment of urothelial cancer to predict a patient's immune therapy response remains a matter of controversy. Several contributing factors have been discussed; however, systematic studies are lacking. The present study demonstrates that clinically applied PD-L1 scoring algorithms are influenced by inter-algorithm variability and result in the selection of different "PD-L1" positive populations within the tumor immune microenvironment (TIME). The results further demonstrate that specific immune phenotypes of muscle-invasive urothelial cancer are associated with very different clinical outcomes, which cannot be resolved by PD-L1 testing. Thus, PD-L1 alone not only fails to reflect the TIME, but also has implications for patients. We conclude that a comprehensive integration of PD-L1 expression and immune phenotypes is superior to PD-L1 testing. This might be a novel strategy to predict a patient's response to immune therapy. Background: Immune therapy has gained significant importance in managing urothelial cancer. The value of PD-L1 remains a matter of controversy, thus requiring an in-depth analysis of its biological and clinical relevance. Methods: A total of 193 tumors of muscle-invasive bladder cancer patients (MIBC) were assessed with four PD-L1 assays. PD-L1 scoring results were correlated with data from a comprehensive digital-spatial immune-profiling panel using descriptive statistics, hierarchical clustering and uni-/multivariable survival analyses. Results: PD-L1 scoring algorithms are heterogeneous (agreements from 63.1% to 87.7%), and stems from different constellations of immune and tumor cells (IC/TC). While Ventana IC5% algorithm identifies tumors with high inflammation and favorable baseline prognosis, CPS10 and the TCarea25%/ICarea25% algorithm identify tumors with TC and IC expression. Spatially organized immune phenotypes, which correlate either with high PD-L1 IC expression and favorable prognosis or constitutive PD-L1 TC expression and poor baseline prognosis, cannot be resolved properly by PD-L1 algorithms. PD-L1 negative tumors with relevant immune infiltration can be detected by sTILs scoring on HE slides and digital CD8(+) scoring. Conclusions: Contemporary PD-L1 scoring algorithms are not sufficient to resolve spatially distributed MIBC immune phenotypes and their clinical implications. A more comprehensive view of immune phenotypes along with the integration of spatial PD-L1 expression on IC and TC is necessary in order to stratify patients for ICI.
Altmetric