PD-L1 Testing and Squamous Cell Carcinoma of the Head and Neck: A Multicenter Study on the Diagnostic Reproducibility of Different Protocols

Abstract

Immune checkpoint inhibitors for blocking the programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) axis are now available for squamous cell carcinoma of the head and neck (HNSCC) in relapsing and/or metastatic settings. In this work, we compared the resulting combined positive score (CPS) of PD-L1 using alternative methods adopted in routine clinical practice and determined the level of diagnostic agreement and inter-observer reliability in this setting. The study applied 5 different protocols on 40 tissue microarrays from HNSCC. The error rate of the individual protocols ranged from a minimum of 7% to a maximum of 21%, the sensitivity from 79% to 96%, and the specificity from 50% to 100%. In the intermediate group (1 ≤ CPS < 20), the majority of errors consisted of an underestimation of PD-L1 expression. In strong expressors, 5 out of 14 samples (36%) were correctly evaluated by all the protocols, but no protocol was able to correctly identify all the “strong expressors”. The overall inter-observer agreement in PD-L1 CPS reached 87%. The inter-observer reliability was moderate, with an ICC of 0.774 (95% CI (0.651; 0.871)). In conclusion, our study showed moderate interobserver reliability among different protocols. In order to improve the performances, adequate specific training to evaluate PD-L1 by CPS in the HNSCC setting should be coordinated.

Automated PD-L1 Scoring Using Artificial Intelligence in Head and Neck Squamous Cell Carcinoma

“…This was also reflected in the concordance for the CPS cutoffs, which was notably better for human-machine (κ = 0.58) than that for human-human (κ = 0.43). Our kappa values were in the range to a multicenter inter-rater study for HNSCC, which reported a Fleiss' kappa value between 0.303 and 0.557 for the CPS [51]. Overall, this supports the potential clinical value of machine-based decision finding for the treatment of HNSCC based on a CPS cutoff of ≥1.…”

Concordance between Three PD-L1 Immunohistochemical Assays in Head and Neck Squamous Cell Carcinoma (HNSCC) in a Multicenter Study

“…Although our study was conducted on tissue-microarray (TMA) and not on whole tissue sections, our results are in line with the previous observations. One previous study [ 22 ], reporting low concordance between the two Dako platforms was performed using 22C3 antibody as LTD on Dako Omnis before the PharmDX kit had been optimized for use on this platform. We providednewevidence that validates the PharmDX assay optimized for use on Dako Omnis platform in comparison to Dako Autostainer link48, in the light of the ongoing shift between the two platforms.…”

Challenges facing pathologists evaluating PD‐L1 in head & neck squamous cell carcinoma

“…compared the reference 22C3 assay with a 22C3 laboratory developed test (LDT) and the SP263 assay in 143 cases. A smaller series published by Crosta et al 21 tested the performance of five different PD‐L1 protocols with clones SP142, SP263, and 22C3 on various platforms against the 22C3 pharmDx on 15 cases/30 cores. The aforementioned studies utilized a tissue microarray (TMA) and were comprised of cases from several sites in the head and neck region, with PD‐L1 assessed centrally by trained pathologists.…”

“…Not surprisingly, training in this regard has been shown to be important 35 . However, when pathologists are trained, the reproducibility among them appears to be high in assessing CPS, with ICC≥0.70 21,29 or excellent ≥0.90 20 with all of the assays including 22C3 pharmDx, SP263, and SP142. Some studies suggest better concordance for pathologists with the SP263 assay when assessing both CPS or separately counting tumor and immune cells, with an almost perfect kappa of 0.836 and in the range of moderate to substantial for the clones SP142, 22C3, and E1L3N 24 .…”