AI-Based Pathology for Assessing Sensitivity to Atezolizumab/Bevacizumab in HCC

Get Permission

As reported in The Lancet Oncology, Zeng et al developed an artificial intelligence (AI)-based pathology method that distinguished between better and poorer response to treatment with atezolizumab plus bevacizumab in patients with hepatocellular carcinoma (HCC).

As stated by the investigators: “Clinical benefits of atezolizumab plus bevacizumab … are observed only in a subset of patients with hepatocellular carcinoma, and the development of biomarkers is needed to improve therapeutic strategies. The atezolizumab-bevacizumab response signature (ABRS), assessed by molecular biology profiling techniques, has been shown to be associated with progression-free survival…. The primary objective of our study was to develop an AI model able to estimate ABRS expression directly from histological slides and to evaluate if model predictions were associated with progression-free survival.”

Study Details

In the multicenter retrospective study, the ABRS AI-based predictive model (ABRS-P) was developed using data from the Cancer Genome Atlas. The model was externally validated in samples from a surgical resection series of 225 patients and a biopsy series of 157 patients with HCC.

The predictive value of the model was assessed in a series of biopsy samples from a cohort of 122 patients with HCC treated with atezolizumab/bevacizumab. Spatial transcriptomics were performed in the atezolizumab/bevacizumab cohort, with prediction heatmaps being matched with in situ expression profiles.

Key Findings

In the development series, the mean Pearson’s coefficient between ABRS-P values and ABRS score (mean expression of ABRS genes) across all cross-validations was r = 0.62 (standard deviation = 0.09, mean P < .0001). The best model had a correlation of r =  0.71, 95% confidence interval [CI] = 0.58–0.81, P < .0001).  

Correlations between ABRS-P values and ABRS scores were r = 0.60 (95% CI = 0.51–0.68, P < .0001) in the external surgical resection series and r = 0.53 (95% CI = 0.40–0.63, P < .0001) in the external biopsy series.

In the patients treated with atezolizumab/bevacizumab, median progression-free survival was 12 months (95% CI = 7 months to not reached) among 74 with ABRS-P–high tumors vs 7 months (95% CI = 4–9 months) among 48 with ABRS-P–low tumors (P = .014). Spatial transcriptomics showed significantly higher ABRS score, as well as upregulation of other immune effectors, in tumor areas with high vs low ABRS-P values.

The investigators concluded: “Our study indicates that AI applied on hepatocellular carcinoma digital slides is able to serve as a biomarker for progression-free survival in patients treated with atezolizumab-bevacizumab. This approach could be used in the development of inexpensive and fast biomarkers for targeted therapies. The combination of AI heatmaps with spatial transcriptomics provides insight on the molecular features associated with predictions. This methodology could be applied to other cancers or diseases and improve understanding of the biological mechanisms that drive responses to treatments.”

Julien Calderaro, MD, of the Department of Pathology, Henri Mondor-Albert Chenevier University Hospital, Créteil, France, is the corresponding author for The Lancet Oncology article.

Disclosure: The study was funded by Institut National du Cancer, Fondation ARC, China Scholarship Council, Ligue Contre le Cancer du Val de Marne, Ipsen, Fondation Bristol Myers Squibb Pour la Recherche en Immuno-Oncologie, and others. For full disclosures of the study authors, visit

The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.