57P - Inter-observer variability in programmed death-ligand 1 (PD-L1) immunohistochemistry scoring in non small cell lung cancer (NSCLC)

Date 17 December 2016
Event ESMO Asia 2016 Congress
Session Poster lunch
Topics Biomarkers
Thoracic Malignancies
Presenter Alesha Thai
Citation Annals of Oncology (2016) 27 (suppl_9): ix9-ix18. 10.1093/annonc/mdw574
Authors A. Thai1, G. Rivalland1, K. Asadi2, C. Murone3, T. John4, P. Mitchell1
  • 1Medical Oncology, Austin Hospital, 3084 - Heidelberg/AU
  • 2Anatomical Pathology, Austin Hospital, 3084 - Heidelberg/AU
  • 3Victorian Cancer Biobank, Austin Hospital, 3084 - Heidelberg/AU
  • 4Medical Oncology, Austin Helath/ Olivia Newton-John Cancer Research Institute, 3084 - Melbourne/AU



Inhibitors of the PD-1 pathway are effective in treatment of NSCLC. Expression of PD-L1 on tumour tissue can help predict response to therapeutic antibodies. To date, most trials involving targeted antibodies to PD-L1 in NSCLC have variable definitions of PD-L1 “positivity” on immunohistochemistry (IHC) with cut-offs ranging from ≥1% to ≥ 50% of cases. If patients are selected for, or are denied access to, a potentially effective treatment, it is important to have a biomarker that is reproducible between assessors. We evaluated the inter- observer agreement for PDL-1 IHC staining in a large cohort of resected NSCLC.


A pathologist selected the site for triplicate cores of the primary tumour block from sequential patients with stage I-III NSCLC undergoing pneumonectomy or lobectomy at Austin Health. Tissue arrays were stained using the Dako 28-8 antibody inhouse by Bristol Myers Squibb (BMS). Arrays were scored by two independent, blinded assessors including one pathologist trained in scoring by BMS. The proportion of tumour cells with membranous staining of any intensity was scored. The Cohen’s kappa coefficient was calculated for each category to determine inter-observer agreement.


Tumour cores were from 517 patients, with 1503 evaluable samples from 1912 cores collected (78.6%). Preliminary analysis demonstrated agreement of the two scorers for 99.9%, 99.8% and 98.7% for tumours staining 1%, 5% and 50% of tumour cells respectively. Tumours staining for >1%, >5% and >50% had a kappa coefficient of 0.996 (95% CI 0.991 - 1.000), 0.994 (95% CI 0.987 - 1.000) and 0.920 (95% CI 0.885 – 0.956) respectively.


Inter-observer agreement for PD-L1 IHC staining using the Dako 28-8 antibody is strong across all categories. The findings of this study suggest reporting of PD-L1 staining is reproducible between assessors.

Clinical trial indentification

Legal entity responsible for the study

Austin Hospital Ethics Committee


Austin Hospital


G. Rivalland: Honoraria (for speaking engagements) AstraZeneca Roche. P. Mitchell: Honoraria- Roche, MSD Advisory board- BMS, Roche, AstraZeneca, MSD, Boehriner Ingelheim Other- Roche, BMS, AstraZeneca- travel and accommodation. All other authors have declared no conflicts of interest.