955PD - A comparative study of PD-L1 diagnostic assays in squamous cell carcinoma of the head and neck (SCCHN)

Date 08 October 2016
Event ESMO 2016 Congress
Session Head and neck cancers
Topics Head and Neck Cancers
Presenter Marianne Ratcliffe
Citation Annals of Oncology (2016) 27 (6): 328-350. 10.1093/annonc/mdw376
Authors M.J. Ratcliffe1, A. Sharpe2, M. Rebelatto3, M. Scott1, C. Barker2, P. Scorer2, J. Walker2
  • 1Personalised Healthcare And Biomarkers, AstraZeneca, CB4 0WGT - Alderley Park, Macclesfield/GB
  • 2Personalised Healthcare And Biomarkers, AstraZeneca, Cambridge/GB
  • 3Translational Medicine, MedImmune, Gaithersburg/US



PD-1/PD-L1 directed antibodies are emerging as effective therapeutics in multiple oncology settings. In the SCCHN Checkmate 141 study, improved efficacy with nivolumab, a PD-1 targeted therapy, was observed in pts with tumour PD-L1 expression ≥1% vs pts with PD-L1 expression below this cut off. Multiple diagnostic PD-L1 tests are available using different antibody clones, different staining protocols and different cut offs. A better understanding of the technical performance of these assays will allow appropriate interpretation of clinical outcomes with different drugs.


108 tumour biopsy samples from stage I–IV SCCHN pts, obtained from a commercial source and including HPV positive and HPV negative, were assessed using 3 PD-L1 diagnostic assays: the Ventana SP263 assay currently being used in durvalumab (anti-PD-L1) clinical trials, the Dako 28-8 and Dako 22C3 assays, commonly used in nivolumab (Opdivo®) and pembrolizumab (Keytruda®) trials, respectively. Assays were performed in an accredited laboratory, following the device protocol. Concordance between tumour membrane staining was assessed across a range of clinically relevant cut offs, including ≥1%, ≥10% and ≥25%. Lower 95% CI were calculated using the Clopper-Pearson method.


Data indicated strong association, with a Spearman correlation coefficient of ≥0.9 for each pairwise comparison. Overall percent agreement (OPA) of >90% was seen between the three assays across multiple clinically relevant cut points. Assessment of a further 392 samples is ongoing to complement the current data set with a larger dynamic range of PD-L1 expression.

Ventana SP263 vs Dako 28-8 Dako 22C3 vs Dako 28-8
Assay cut off OPA (%) Lower 95% CI OPA (%) Lower 95% CI
≥1% 91.7 85.9 96.3 91.7
≥10% 92.6 87.0 95.4 90.5
≥25% 94.4 89.3 97.2 93.0


This study indicates that the SCCHN patient population defined by Ventana SP263, Dako 28-8 and Dako 22C3 assays is similar where an identical cut point is used. The findings align with those of a similar study in NSCLC, and build optimism that it may be possible to compare studies using different PD-L1 tests and deliver harmonization of PD-L1 diagnostic testing.

Clinical trial identification


Legal entity responsible for the study

AstraZeneca PLC




M.J. Ratcliffe, A. Sharpe, M. Scott, C. Barker, P. Scorer, J. Walker: AstraZeneca employee and holds stocks/shares. M. Rebelatto: Employee of MedImmune LLC and hold stocks or shares in AstraZeneca.