Digital pathology superior to manual SP263 PD-L1 TC scoring for stage II-IIIA high PD-L1 expression NSCLC

Prof Martin Reck - German Center of Lung Research, Grosshansdorf, Germany

I also had the pleasure to present an exploratory analysis of the IMpower010 trial. This was a trial investigating adjuvant atezolizumab in patients with resected early-stage non-small cell lung cancer after adjuvant chemotherapy. We have seen an improvement in disease free survival in patients with PD-L1 expressing tumours and also an impact on overall survival in particular when we look at the patients with high PD-L1 expression on the tumour cells.

In this analysis we looked at a new method to assess the PD-L1 expression on the tumour cells by an AI-based algorithm which was used in an experimental part of the study. We had two groups of patients, we had first the analysis of patients with a standard assessment of PD-L1 by the pathologist and then we had a second readout of the samples by the AI-based algorithm.

When we look at the data we had two very important results. So, number one, we identified more samples with high PD-L1 expression using the AI-based algorithm compared to the conventional pathology. Number two, when we looked at the signal of efficacy related to conventional pathology or related to AI, this was completely comparable. So the signal of efficacy was there in both cohorts.

For the practice this is of importance because in Europe we do have the label of using atezolizumab as an adjuvant treatment only in patients with high PD-L1 expression. This AI-based algorithm might help us to identify more patients with a high PD-L1 expression on the tumour cells which might benefit from this treatment.