Antibodies targeting programmed cell death-1/PD-L1 (PD-1/PD-L1) have shown clinical activity in advanced UC. The ability of PD-L1 to predict response has been investigated using different antibody clones and scoring algorithms. It is important to understand if these assays/algorithms identify the same patients and how they compare in predicting response.
Archival UC tumour samples from 335 patients from a commercial source were stained with VENTANA SP263, VENTANA SP142, PD-L1 IHC pharmDx 28-8 and PD-L1 IHC pharmDx 22C3 Assays; classified according to their respective algorithms: tumour cell (TC) or immune cell (IC) staining ≥25% (TC/IC≥25%), PD-L1 staining IC area ≥5% of tumour (IC ≥ 5% ), TC staining ≥1% or combined positive score (CPS ≥10). Overlap between populations was assessed by overall percent agreement (OPA), negative percent agreement (NPA) and positive percent agreement (PPA). UC samples from study 1108 (NCT01693562) were stained using VENTANA SP263 and TC/IC≥25%, IC ≥ 5% and CPS≥10 algorithms were applied. Objective response rates (ORR; data cutoff Oct 2017) in patients classified as PD-L1 high or low by these algorithms were investigated.
There was moderate overlap between populations identified by VENTANA SP263 (TC/IC≥25%) and PD-L1 IHC pharmDx 28-8 (TC ≥ 1%) or PD-L1 IHC pharmDx 22C3 (CPS≥10) and minimal overlap between VENTANA SP263 (TC/IC≥25%) and VENTANA SP142 (IC ≥ 5%) (Table). Applying different algorithms to data from study 1108 also gave differences in patient classification. ORR in patients determined as PD-L1 high vs low/negative were as follows: TC/IC≥25%: 28% vs 6%, IC ≥ 5%: 48% vs 14%, CPS≥10 25% vs 13%.
The TC/IC≥25% algorithm identifies a different population to IC ≥ 5% or CPS10. In CD-ON MEDI4736-1108, highest response rates were seen in PD-L1 high patients determined by IC ≥ 5%, whereas TC/IC≥25% was optimal in predicting non-responders to durvalumab.Table: 904P
Overall (OPA), negative (NPA) and positive percent agreement (PPA) between PD-L1 assays
|Clinical Algorithm (assay)||VENTANA SP263 (TC/IC 25%) Assay used as reference, % agreement (95% CIa)|
|CPS ≥1 (PD-L1 IHC pharmDx22C3||77.0% (72.9%)||90.7% (85.0%)||69.6% (64.0%)|
|CPS ≥10 (PD-L1 IHC pharmDx 22C3)||81.5% (77.6%)||62.7% (54.8%)||91.7% (87.9%)|
|TC ≥ 1% (PD-L1 IHC pharmDx 28-8)||75.5% (71.3%)||66.9% (59.1%)||80.2% (75.2%)|
|ICTumorArea ≥5% (VENTANA SP142)||69.9% (65.5%)||15.3% (10.1%)||99.5% (97.8%)|
For each metric, lower boundary of 95% confidence interval (CI) was calculated with no upper bound using the Clopper-Pearson method
Clinical trial identification
CD-ON MEDI4736-1108 NCT01693562.
Legal entity responsible for the study
Editorial support, which was in accordance with Good Publication Practice (GPP3) guidelines, was provided by Anne-Marie Manwaring of Parexel, and was funded by AstraZeneca.
J. Walker, M. Zajac, J. Ye, M. Scott, P. Scorer, C. Barker, A. Gupta, P. Mukhopadhay, S. Ferro, J.A. Williams: Employee and shareholder of AstraZeneca. M. Ratcliffe: Consultant for AstraZeneca. H. Al-Masri: Employee of Hematogenix. M.C. Rebelatto: Employee of Medimmune. T. Powles: Honoraria for advisory boards: Novartis, Roche and Pfizer, Bristol-Myers Squibb; Research grant: AstraZeneca.