Abstract 4037: Analytical and inter-observer comparability of 4 clinically developed programmed death-ligand 1 (PD-L1) immunohistochemistry assays in advanced clear cell renal cell carcinoma (CCRCC)

Background: PD-L1/programmed cell death-1 (PD-1)-targeting checkpoint inhibitors have shown clinical activity in advanced and metastatic RCC, which correlates with PD-L1 expression on tumor cells (TC) and/or tumor-infiltrating immune cells (IC). We examined the analytical performance and inter-observer agreement of 4 clinically developed assays for PD-L1 expression on IC and TC in locally advanced CCRCC. Methods: Archived formalin-fixed paraffin-embedded nephrectomy specimens from 30 patients with locally advanced (T2a+) CCRCC were selected from 201 cases based on PD-L1 status per VENTANA SP142 (IC 5%; 10 cases each). All selected cases were stained for PD-L1 using VENTANA SP142 and SP263, and DAKO 22C3 and 28-8, each at one site, according to manufacturer protocols. Stainings were blinded with respect to both assay and sample information and scored at 5 sites for PD-L1 expression on IC (% per tumor area) and TC (% of tumor cells). All readers had been trained in scoring IC per tumor area. Results: Adjusted means of IC staining for PD-L1 varied from 4.00 to 4.90%, and TC from 1.32 to 10.68%, depending on the assay used (Table). Pairwise comparisons of adjusted means between assays revealed small, non-significant differences for IC (-0.9 to 0.3%), while for TC there were significant differences for SP142 v other assays (-9.4 to -5.7%) and for SP263 v 28-8 (-3.7%), and small, non-significant differences for other comparisons (-2.9 to -0.8%). The pre-specified allocation to binary cutoffs (1% or 5%) for IC or TC alone predominantly showed moderate-to-substantial Kappa agreement scores (0.4 to 0.8) for IC and TC between assays for each reader. Conclusions: This is the first multicenter analytical comparison study of PD-L1 assays in CCRCC. Good-to-high concordance rates across all assays were achieved between trained readers for scoring PD-L1 on IC and for some assays for PD-L1 on TC. AssayPD-L1 on IC, % (95% CI)*Reader ICC†PD-L1 on TC, % (95% CI)*Reader ICC†VENTANA SP1424.10 (3.33 to 4.86)0.7251.32 (-2.14 to 4.79)0.142VENTANA SP2634.26 (3.49 to 5.02)0.4946.98 (3.52 to 10.45)0.823DAKO 22C34.00 (3.23 to 4.76)0.7019.87 (6.41 to 13.34)0.758DAKO 28-84.90 (4.14 to 5.60)0.42210.68 (7.21 to 14.14)0.778* Sample adjusted means for each assay; † Intra-class correlation per assay between 5 readers. Citation Format: Ulrich Sommer, Markus Eckstein, Johannes Ammann, Till Braunschweig, Stephan Macher-Goeppinger, Kristina Schwamborn, Stefanie Hieke-Schulz, Bernd