Comparison of patient populations identified by different PD-L1 assays in in triple-negative breast cancer (TNBC)

Abstract

Background: TNBC represents an unmet clinical need; the IMPassion 130 and KEYNOTE-086 trials have demonstrated efficacy of PD-1/PD-L1 targeted therapeutics in the metastatic/unresectable first line setting, with increased response in PD-L1 high populations, defined using different diagnostic assays and algorithms. This work addresses the need to establish PD-L1 assay concordance in TNBC. Methods: 196 TNBC cases were stained by IHC using the VENTANA PD-L1 (SP263), VENTANA PD-L1 (SP142), Dako PD-L1 IHC 28-8 pharmDx and Dako PD-L1 IHC 22C3 pharmDx assays. A single pathologist scored the proportions of membrane staining tumour cells (TC), staining immune cells (IC) and tumour occupied by immune cells. The proportion of tumour occupied by staining immune cells (IC TA) and Combined Positive Score (CPS) were derived. Concordance between assays was evaluated using the Spearman coefficient q. Concordance in patient status was assessed using overall, positive and negative percent agreement (OPA/PPA/NPA) at matched algorithms and cutoffs. Results: SP263, 22C3 and 28-8 assays showed good analytical correlation for TC staining (q ¼ 0.84-0.89), but lower for SP142 (0.44-0.46), whereas all assays showed good correlation for IC TA (0.77-0.96) and CPS (0.78-0.95). OPA ranged from 58%-97% at matched algorithms (Table). Prevalence for SP142 IC TA !1% and 22C3 CPS!1 was in-line with IMPassion 130 and KEYNOTE-086, respectively.