Predictive value of WHO classification for PD-L1 and Her2/Neu expression and distinct associations with protein expression based classification in gastric carcinoma.

The Cancer Genome Atlas data on gastric carcinoma has identified four biologic pathways as potential drivers of gastric carcinogenesis and suggested targeted therapies based on the genomic alterations underscoring each subset. The correlation between morphology, biologic groups and their corresponding biomarkers has been eluded in several previous studies; however, a comprehensive analysis in consideration of the recent advancements has not been performed. In this study we explored the predictive value of morphology for biomarker expression and its association with protein expression based classification of gastric carcinoma. Four hundred eighty six gastric carcinomas which had been classified into protein expression-based groups formed the case cohort. Upon analysis, we found a low positive predictive value of an individual morphologic pattern for biomarker-expression, indicating that an individual morphologic pattern alone cannot predict PD-L1, Her2/neu expression and EBV- or MSI-gastric cancer. A combination approach targeting the test in certain WHO patterns can be employed for maximizing the positive predictive values. These include, PD-L1 testing in tubular, carcinoma with lymphoid stroma, undifferentiated and poorly cohesive patterns and Her2/neu testing in tubular, mixed, papillary, mucinous and solid patterns. The predictive values and morphologic associations presented here have the potential to select patients for personalized therapy.