Thanks for your submission. We'll get back to you soon!

PathAI and Bristol Myers Squibb find Positive Association of Digital PD-L1 Expression with Outcomes in nivolumab-treated Patients in an Exploratory analysis

Digital analysis shows higher sensitivity for PD-L1 expression and accurate detection of low-level PD-L1 expression in patients compared to manual reading

June 25, 2020

For more information:

Isabella Canuso

Boston, Massachusetts - – PathAI, a global provider of artificial intelligence (AI)-powered technology for use in pathology research, announced that the results of an exploratory analysis of digital scoring of PD-L1 expression to assess response in patients treated with Bristol Myers Squibb’s PD-1 inhibitor, nivolumab was presented as a poster #2017 at the 2020 American Association for Cancer Research (AACR) Annual Meeting. AACR was held from June 22 to June 24.

The key finding of this study shows that PathAI’s digital quantification of PD-L1 expression in samples across multiple tumor types identified more patients that were PD-L1 positive, while maintaining the same response rate as the PD-L1 positive population identified using the manual PD-L1 analysis. This retrospective analysis of melanoma (MEL) and urothelial carcinoma (UC) clinical trial samples compared both the prevalence and response characteristics of the manual PD-L1 read-outs to the AI-derived quantifications. The analysis showed the digital analysis has higher sensitivity for PD-L1 expression and accurate detection of low-level PD-L1 expression in patients compared to manual reading. In three retrospective analyses, patients who were PD-L1 negative by manual assessment, were shown to have improved outcomes among nivolumab treated patients when identified as positive by digital analysis. In all cohorts, as previously reported, anti-tumor activity was observed regardless of PD-L1 status.

Training a classifier for PD-L1 expression in multiple tumor types

In this study, digital scoring was conducted by PathAI’s research platform using an algorithm trained on annotations collected from a network of board-certified pathologists. PathAI collected tens of thousands of examples to train a classifier that automatically evaluates PD-L1 expression on tissue stained using the Dako PD-L1 IHC 28-8 pharmDX assay. Once trained, the models were applied to two Bristol Myers Squibb sponsored melanoma clinical trial sets (CheckMate067 and CheckMate238), and one clinical trial of advanced UC (CheckMate275). The AI-derived PD-L1 quantification of these samples showed high correlation to manual PD-L1 expression scoring results.

Positive association of digital PD-L1 expression in multiple tumor types with outcomes in nivolumab-treated patient

The response of the PD-L1 positive patients to the nivolumab therapy regimens, was assessed using RECIST overall response criteria. Results suggest that despite showing increased prevalence of PD-L1 positivity across multiple cut-off points (PD-L1+ TC ≥ 1% and ≥ 5%) by the digital predictions, the response rates were indistinguishable between digital and manual methods. Additionally, patients identified by digital but not manual assessment showed improved clinical benefit compared with patients identified as TC PD-L1 negative by both manual and digital methods. PathAI’s research platform and PD-L1 algorithm are not for use in diagnostic procedures.

“This poster shows the potential of an AI-powered PD-L1 test to provide improved patient selection over traditional manual scoring. Such a finding could have important future consequences for cancer patients.” said PathAI co-founder and Chief Executive Officer Andy Beck, M.D., Ph.D.

About PathAI

PathAI is a leading provider of AI-powered research tools and services for pathology. PathAI’s platform promises substantial improvements to the accuracy of diagnosis and the efficacy of treatment of diseases like cancer, leveraging modern approaches in machine and deep learning. Based in Boston, PathAI works with leading life sciences companies and researchers to advance precision medicine. To learn more, visit