Atezolizumab plus bevacizumab (A/B) is a first-line therapy for unresectable hepatocellular carcinoma (HCC). Only a small proportion of patients respond to treatment. This study integrated radiomic and clinical data derived from routine pre-treatment imaging to predict outcomes after immunotherapy. A total of 152 patients from two international centres receiving A/B were retrospectively reviewed. Deep learning autosegmentation generated whole liver masks from pre-treatment CTs. Radiomic features combined with clinical variables were used to predict 12-month mortality post A/B. Radiomic and integrated radiomic-clinical models were developed using seven machine learning models in combination with 13 feature selection techniques in the Imperial College London (ICL) cohort. K-means clustering identified high- and low-risk groups and predicted overall survival (OS), progression-free survival (PFS) and response. Model performance was assessed in the independent Assistance Publique-Hôpitaux de Paris (AP-HP) cohort. The integrated radiomic-clinical model outperformed BCLC stage (AUC 0.61, p <0.001) and ALBI grade (AUC 0.48, p <0.001) in the ICL (AUC 0.89, 95% CI 0.75-0.99) and AP-HP (AUC 0.75, 95% CI 0.64-0.85) cohorts. Integrated model-stratified high-risk patients had significantly shorter median OS (ICL: 5.6 months vs. 28.2 months; p <0.001; AP-HP: 5.8 months vs. 15.7 months; p <0.001) and PFS (ICL: 2.4 months vs. 14.6 months; p <0.001; AP-HP: 2.1 months vs. 6.1 months; p = 0.046). Low-risk patients had significantly higher immune checkpoint inhibitor response rates compared to high-risk patients (35.6% vs. 21.4%; p = 0.038). In multivariable analysis, radiomic group was the strongest predictor of OS (hazard ratio 3.22, 95% CI 1.99-5.20; p <0.001) and PFS (hazard ratio 1.82, 95% CI 1.18-2.80; p = 0.010). Radiomic-based models predict survival outcomes and response to immunotherapy in patients with advanced HCC. Deep learning in combination with machine learning can stratify patients and allows for precision treatment strategies. Prognostic markers predicting survival and response to immunotherapy in hepatocellular carcinoma are lacking. This study used deep learning and machine learning to develop and validate an integrated radiomic-clinical model which can predict survival and response to atezolizumab plus bevacizumab from pre-treatment imaging. Radiomic-based machine learning models can risk-stratify patients with advanced HCC receiving atezolizumab plus bevacizumab.

Objective: Circulating tumour DNA (ctDNA) is a promising non-invasive biomarker in cancer. We aim to assess the dynamic of ctDNA in patients with hepatocellular carcinoma (HCC). Design: We analysed 772 plasmas from 173 patients with HCC collected at the time of diagnosis or treatment (n=502), 24 hours after locoregional treatment (n=154) and during follow-up (n=116). For controls, 56 plasmas from patients with chronic liver disease without HCC were analysed. All samples were analysed for cell free DNA (cfDNA) concentration, and for mutations in TERT promoter, CTNNB1 , TP53 , PIK3CA and NFE2L2 by sequencing and droplet-based digital PCR. Results were compared with 232 corresponding tumour samples. Results: In patients with active HCC, 40.2% of the ctDNA was mutated vs 14.6% in patients with inactive HCC and 1.8% in controls (p<0.001). In active HCC, we identified 27.5% of mutations in TERT promoter, 21.3% in TP53 , 13.1% in CTNNB1 , 0.4% in PIK3CA and 0.2% in NFE2L2, most of the times similar to those identified in the corresponding tumour. CtDNA mutation rate increased with advanced tumour stages (p<0.001). In 103 patients treated by percutaneous ablation, the presence and number of mutations in the ctDNA before treatment were associated with higher risk of death (p=0.001) and recurrence (p<0.001). Interestingly, cfDNA concentration and detectable mutations increased 24 hours after a locoregional treatment. Among 356 plasmas collected in 53 patients treated by systemic treatments, we detected mutations at baseline in 60.4% of the cases. In patients treated by atezolizumab-bevacizumab, persistence of mutation in ctDNA was associated with radiological progression (63.6% vs 36.4% for disappearance, p=0.019). In two patients progressing under systemic treatments, we detected the occurrence of mutations in CTNNB1 in the plasma that was subclonal in the tumour for one patient and not detectable in the tumour for the other one. Conclusion: ctDNA offers dynamic information reflecting tumour biology. It represents a non-invasive tool useful to guide HCC clinical management.

Hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA) are the most frequent primary liver cancers, accounting for approximately 80% and 15%, respectively. HCC carcinogenesis occurs mostly in cirrhosis and is a complex multi‐step process, from precancerous lesions (low‐grade and high‐grade dysplastic nodules) to progressed HCC. During the different stages of liver carcinogenesis, there is an accumulation of pathological, genetic and epigenetic changes leading to initiation, malignant transformation and finally tumour progression. In contrast, a small subset of HCC occurs in normal liver from the transformation of hepatocellular adenoma (HCA), a benign hepatocellular tumour. The recent molecular classification enables to stratify HCAs according to their risk of complication, in particular malignant transformation, associated with mutations in exon 3 of the catenin beta 1 (CTNNB1) gene. Cholangiocarcinoma (CCA) derives from the multistep malignant transformation of preneoplastic lesions, like biliary intraepithelial neoplasia (BilIN) and intraductal papillary neoplasm of the bile duct (IPNB), for which a pre‐operative diagnosis remains difficult. Different genetic alterations are involved in BilIN and IPNB progression, leading to the development of tubular or intestinal adenocarcinoma. The aims of this review are to describe the main clinical and molecular features of preneoplastic lesions leading to the development of HCC and CCA, their implications in clinical practice and the perspectives for future research.