The VENTANA PD-L1 (SP263) assay has been developed as a companion diagnostic for anti-PD-L1 immune checkpoint inhibitors. Here we investigate assay inter-rater reliability, applied to PD-L1 scoring of tumour cells (TCs) and immune cell (IC) infiltrates in NSCLC.
Six expert European pulmonary pathologists independently scored 200 NSCLC samples stained using the VENTANA PD-L1 (SP263) assay. Archival, commercially-sourced formalin-fixed paraffin-embedded resections were selected to represent the dynamic range of PD-L1 expression. Each pathologist scored the proportion of TCs expressing PD-L1 (TM score), tumour-associated IC population as a percentage of total tumour area (PIC value), and percentage of ICs expressing PD-L1 (IC score). Scores were analysed using intra-class correlation coefficient (ICC) and patient classification using Fleiss’ Kappa.
Interim results were available for 3 pathologists and 180 cases. TM scoring between pathologists showed strong pair-wise correlations between individuals (R2>0.90) with an ICC >0.95. Pair-wise and overall agreement was ≥85% for TC ≥ 1% and >93% for TC ≥ 20%, TC ≥ 25%, and TC ≥ 50%. Fleiss’ Kappa showed substantial agreement for TC ≥ 1% and excellent agreement for TC ≥ 20%, TC ≥ 25%, and TC ≥ 50%. There were systematic and substantial differences in PIC values and IC scores between pathologists with poor pair-wise correlations. ICC indicated poor reliability for both IC score (0.36) and PIC values (0.044). Fleiss’ Kappa showed poor agreement for IC ≥ 25% (0.183).
Assessment of TM score in NSCLC was highly reproducible using VENTANA PD-L1 (SP263) assay, building confidence in the accuracy of this assay in patient selection for anti-PD-L1 therapy. However, expert pathologists were unable to reproducibly assess IC score in NSCLC suggesting assessment methodology is unreliable for this tumour type and assay. This contrasts with urothelial cancer (UC) in which pathologist agreement for PIC values and IC scores was generated as part of UC VENTANA PD-L1 (SP263) IHC assay CE marking and FDA approval. This difference in pathology of the different tumour types requires further investigation.
Clinical trial identification
Legal entity responsible for the study
Editorial assistance, which was in accordance with Good Publication Practice (GPP3) guidelines, was provided by Kaveri Sidhu, PhD, of Cactus Communications (Mumbai, India).
A.G. Nicholson: Consultant: Abbvie and Oncologica; Educational grant: Pfizer. E. Thunnissen: Corporate sponsored research: Oncologica; Honorarium: Roche Ventana, Histogenex; Advisory boards: Roche. P. Cane: Speaker fees: Pfizer, Roche; Fees for interview: AstraZeneca. K.M. Kerr: Advisory boards: Roche, AstraZeneca, BMS, MSD, Pfizer, Merck Serono, Roche Diagnostics. M. Scott, P.W. Scorer, C. Barker: Employment: AstraZeneca; Stockholder: AstraZeneca. All other authors have declared no conflicts of interest.