Paul Scorer

Agreement between Programmed Cell Death Ligand-1 Diagnostic Assays across Multiple Protein Expression Cutoffs in Non–Small Cell Lung Cancer

Purpose: Immunotherapies targeting programmed cell death-1 (PD-1) and programmed cell death ligand-1 (PD-L1) demonstrate encouraging antitumor activity and manageable tolerability in non–small cell lung cancer (NSCLC), especially in patients with high tumor PD-L1 expression, as detected by companion or complementary diagnostic assays developed for individual agents. A laboratory is unlikely to use multiple assay platforms. Furthermore, commercially available diagnostic assays are not standardized, and different assay methods could lead to inappropriate treatment selection. This study establishes the extent of concordance between three validated, commercially available PD-L1 IHC diagnostic assays for NSCLC patients [Ventana SP263 (durvalumab), Dako 22C3 (pembrolizumab), and Dako 28-8 (nivolumab)]. Experimental Design: Five hundred formalin-fixed, paraffin-embedded archival NSCLC samples were obtained from commercial sources. Stained slides were read in batches on an assay-by-assay basis by a single pathologist trained in all methods, in a Clinical Laboratory Improvements Amendments program–certified laboratory. An additional pathologist performed an independent review of 200 stained samples for each assay. Results: PD-L1 expression was evaluable with all assays in 493 samples. The three assays showed similar patterns of tumor membrane staining, with high correlation between percent PD-L1 staining. An overall percentage agreement of >90% was achieved between assays at multiple expression cutoffs, including 1%, 10%, 25%, and 50% tumor membrane staining. Conclusions: This study builds optimism that harmonization between assays may be possible, and that the three assays studied could potentially be used interchangeably to identify patients most likely to respond to anti-PD-1/PD-L1 immunotherapies, provided the appropriate clinically defined algorithm and agent are always linked. Clin Cancer Res; 23(14); 3585–91. ©2017 AACR.

Relevance of deep learning to facilitate the diagnosis of HER2 status in breast cancer

Tissue biomarker scoring by pathologists is central to defining the appropriate therapy for patients with cancer. Yet, inter-pathologist variability in the interpretation of ambiguous cases can affect diagnostic accuracy. Modern artificial intelligence methods such as deep learning have the potential to supplement pathologist expertise to ensure constant diagnostic accuracy. We developed a computational approach based on deep learning that automatically scores HER2, a biomarker that defines patient eligibility for anti-HER2 targeted therapies in breast cancer. In a cohort of 71 breast tumour resection samples, automated scoring showed a concordance of 83% with a pathologist. The twelve discordant cases were then independently reviewed, leading to a modification of diagnosis from initial pathologist assessment for eight cases. Diagnostic discordance was found to be largely caused by perceptual differences in assessing HER2 expression due to high HER2 staining heterogeneity. This study provides evidence that deep learning aided diagnosis can facilitate clinical decision making in breast cancer by identifying cases at high risk of misdiagnosis.

Abstract LB-094: A comparative study of PD-L1 diagnostic assays and the classification of patients as PD-L1 positive and PD-L1 negative

Background: PD-1/PD-L1 directed antibodies are emerging as effective therapeutics in multiple oncology settings. Keynote 001 and Checkmate 057 have shown more frequent response to PD-1 targeted therapies in NSCLC patients with high tumour PD-L1 expression than patients with low or no PD-L1 expression. Multiple diagnostic PD-L1 tests are available using different antibody clones, different staining protocols and diverse scoring algorithms. It is vital to compare these assays to allow appropriate interpretation of clinical outcomes. Such understanding will promote harmonization of PD-L1 testing in clinical practice. Methods: Approximately 500 tumour biopsy samples from NSCLC patients, including squamous and non-squamous histologies, will be assessed using three leading PD-L1 diagnostics assays. PD-L1 assessment by the Ventana SP263 assay that is currently being used in Durvalumab clinical trials (positivity cut off: ≥25% tumour cells with membrane staining) will be compared with the Dako 28-8 assay (used in the Nivolumab Checkmate 057 trial at the 1%, 5% and 10% tumour membrane positivity cut offs), and the Dako 22C3 assay (used in the Pembrolizumab Keynote 001 trial) at the 1% and 50% cut offs). Results: Preliminary data from 81 non-squamous patients indicated good concordance between the Ventana SP263 and Dako 28-8 assays. Optimal overall percent agreement (OPA) was observed between Dako 28-8 at the 10% cut off and the Ventana SP263 assay (OPA; 96%, Positive percent agreement (PPA); 91%, Negative percent agreement (NPA); 98%), where the Ventana SP263 assay was set as the reference. Data on the full cohort will be presented for all three assays, and a lower 95% confidence interval calculated using the Clopper-Pearson method. Conclusions: This study indicates that the patient population defined by Ventana SP263 at the 25% cut off is similar to that identified by the Dako-28-8 assay at the 10% tumour membrane cut off. This, together with data on the 22C3 assay, will enable cross comparison of studies using different PD-L1 tests, and widen options for harmonization of PD-L1 diagnostic testing. Citation Format: Marianne J. Ratcliffe, Alan Sharpe, Anita Midha, Craig Barker, Paul Scorer, Jill Walker. A comparative study of PD-L1 diagnostic assays and the classification of patients as PD-L1 positive and PD-L1 negative. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr LB-094.

Correction to: Abstracts

Due to an error with the registration system, the following abstract was regrettably omitted from the Poster Sessions. The abstract should have been included as PS-10-021 and displayed on page S166.