Poster viewing and lunch

22P - PD-L1 testing in mTNBC: Interplatform and interobserver reproducibility of CE-IVD assays for CPS and IC scores (ID 242)

Lecture Time
12:15 - 12:15
Session Name
Poster viewing and lunch
Exhibition area
Fri, 12.05.2023
12:15 - 13:00
  • Nicola Fusco (Milan, Italy)
  • Nicola Fusco (Milan, Italy)
  • Mariia Ivanova (Milan, Italy)
  • Chiara Frascarelli (Milan, Italy)
  • Bruna Cerbelli (Rome, Italy)
  • Gemma Pignataro (Rome, Italy)
  • Carmen Criscitiello (Milan, Italy)
  • Konstantinos Venetis (Milan, Italy)
  • Elham Sajjadi (Milan, Italy)
  • Elena Guerini-Rocco (Milan, Italy)
  • Paolo Graziano (San Giovanni Rotondo, Italy)
  • Maurizio Martini (Messina, Italy)
  • Giulia D'Amati (Rome, Italy)



PD-L1 test is recommended to select mTNBC patients eligible for immune checkpoint inhibitors (ICIs). Several factors may challenge this test, including: 1) different assays and platforms available, 2) different scoring systems (CPS and IC), 3) different cut-off values (CPS≥10 and IC≥1%), and 4) different indications (CPS→pembrolizumab and IC→atezolizumab). We aimed to characterize the interobserver and interassay reproducibility of PD-L1 testing in mTNBC using the currently validated CE-IVD assays for CPS and IC.


60 mTNBC samples were subjected to PD-L1 immunohistochemistry (IHC) using 22C3 pharmDx on a Dako Autostainer Link 48, and both SP263 and SP142 on a Ventana BenchMark Ultra. CPS was evaluated with both 22C3 and SP263 by 5 pathologists (certified PD-L1 trainers). In addition, IC with SP142 was assessed by 3 of the 5 pathologists, with specific expertise in breast pathology. After intraclass correlation coefficient (ICC) calculation for each assay and corresponding cut-offs (CPS≥10 and IC≥1%), the inter-rater and inter-platform agreements were measured by Fleiss’s κ (95% confidence interval, CI). Statistical analyses were performed with R (v 4.2.2) and irr (interrater reliability) package (CRAN).


Significant (p<0.001) ICC was observed with both CPS assays (22C3=0.939 (CI:0.913-0.96); SP263=0.972 (CI:0.96-0.982); combined=0.909 (CI:0.874-0.938)). Fleiss’s κ confirmed an almost perfect agreement both among pathologists and assays (22C3=0.938 (CI:0.857-1.018); κ=0.972 (CI:0.890-1.052); combined=0.907 (CI:0.869-0.945)). Perfect inter-rater agreement was reached for IC (100%). The correlation between CPS and IC scores, albeit significant, was lower (ICC=0.634 (CI:0.455-0.816); κ=0.619 (CI:0.537-0.700)).


In mTNBC, CPS can be reliably performed either by 22C3 on a Dako platform (as in the KEYNOTE studies) or SP263 on a Ventana platform. CPS and IC are not interchangeable tests in mTNBC but each test is accurate when assessed by trained pathologists using CE-IVD assays, precisely 22C3 and SP263 for CPS and SP142 for IC score.

Legal entity responsible for the study

The authors.


Has not received any funding


N. Fusco: Financial Interests, Personal, Invited Speaker: Merck Sharp & Dohme (MSD), Boehringer Ingelheim, AstraZeneca, Daiichi Sankyo, GlaxoSmithKline (GSK), Gilead; Financial Interests, Personal and Institutional, Research Grant: Novartis. C. Criscitiello: Financial Interests, Personal, Invited Speaker: Pfizer, Novartis, Eli Lilly, Roche, Gilead; Financial Interests, Personal, Advisory Board: MSD, Seagen, AstraZeneca, Daiichi Sankyo. E. Guerini-Rocco: Financial Interests, Invited Speaker: Thermo Fisher, Novartis, AstraZeneca, Roche, Biocartis, Exact Science, GSK, Illumina. P. Graziano, M. Martini: Financial Interests, Invited Speaker: MSD. G. D'Amati: Financial Interests, Invited Speaker: MSD, Roche. All other authors have declared no conflicts of interest.