Intraobserver and Interobserver Agreement in the Scoring of PD-L1 (SP142) and Tumor-Infiltrating Lymphocytes in Triple Negative Breast Cancers

Abstract

Objectives. Known for their poor outcomes, triple negative breast cancers (TNBCs) have been investigated for immune checkpoint inhibitors that target Programmed death ligand 1 (PD-L1). In the recent decade, tumor-infiltrating lymphocytes (TILs) have also become potential biomarkers. The aim of the study is to determine the reproducibility of PD-L1 scoring system for TNBC (SP142 clone) and TILs interpretation in the local setting through intra- and interobserver agreement. Methodology. Forty-three primary resection specimens TNBC were evaluated on two occasions with PD-L1 (Roche VENTANA SP142 assay) and TILs by two breast pathologists and one general pathologist on physical glass slides. PD-L1 expression was determined by at least 1% positivity among immune cells within the tumoral area and contiguous peritumoral stroma while TILs was assessed based on International Immuno-Oncology Biomarker Working Group on Breast Cancer. Kappa statistic for PD-L1 and TILs categories while intraclass correlation coefficient (ICC) were assessed, with cutoffs of 0.80 and 0.70, respectively. Results: The overall interrater kappa statistic for PD-L1 on the first and second rounds were weak at 0.506 (95% CI: 0.334-0.679) and minimal at 0.314 (95% CI: 0.142-0.487), respectively. Intraobserver kappa statistic for PD-L1 were varied across the three readers while interobserver kappa values for PD-L1 showed none (0.181) to moderate (0.789) agreement. The TILs intraobserver reliability showed poor to good agreement, with the highest ICC of 0.889 (95% CI: 0.805-0.938). Conclusion. This study demonstrated variable intra and interobserver agreement for both TILs and PD-L1 expression. Although it is desirable to have strong to almost perfect agreement, the kappa and ICC values suggest additional room for improvement. In light of the repercussions in management of patients who will undergo immune checkpoint inhibitor therapy, regular training sessions, concurrences of equivocal results, and possible use of digital pathology as a medium in interpreting TILs and PD-L1 stains to achieve consistent results.