Lombardi, P.; Ramon-Gil, E.; Brunetti, L.; Manfredi, G. F.; Merces, G.; Fulgenzi, C. A. M.; DAlessio, A.; Torkpour, A.; Celsa, C.; Stefanini, B.; Yang, H.; Crowley, F.; Marron, T. U.; Saeed, A.; Pinter, M.; Scheiner, B.; Huang, Y.-H.; Lee, P.-C.; Nishida, N.; Po-Ting Lin, R.; Dalbeni, A.; Vivaldi, C.; Masi, G.; Rohlen, N.; von Felden, J.; Kaseb, A.; Galle, P. R.; Kudo, M.; Hsu, W.-F.; Rimassa, L.; Parisi, A.; Kelley, R. K.; Toyoda, H.; Pirisi, M.; Jabar, R. F.; Rakaee, M.; Cabibbo, G.; Camma, C.; Piscaglia, F.; Hwang, S.; Shin, D. J.; Li, M.; Daniele, G.; Mann, D.; Chon, H. J.; Leslie, J.; Pin

Background: Despite improved outcomes with atezolizumab plus bevacizumab (A+B) in hepatocellular carcinoma (HCC), primary refractoriness (PRef), characterised by early progression or short-lived disease stabilisation following treatment, remains a significant and poorly understood clinical challenge. Methods: We analysed 1296 patients with HCC and Child-Pugh A liver cirrhosis treated with frontline A+B (AB-real) and validated findings in 645 trial participants recruited to IMbrave150 and GO30140. PRef was defined by Society for the Immunotherapy of Cancer (SITC) criteria as progressive disease in the first 6 months after treatment initiation. Patients who achieved complete response, partial response or stable disease for [&ge;] 6 months were classified as responders. We performed a multi-parametric analysis of pre-treatment tumour tissue including machine learning-based quantification of tumour-infiltrating lymphocytes, imaging mass cytometry and RNA sequencing (RNAseq) to evaluate differences in the tumour microenvironment (TME) of PRef versus responding patients. We employed conditional inference tree analyses to provide a hierarchical organisation of determinants of PRef. Results: Among 677 AB-real and 378 trial patients evaluable by SITC criteria, PRef identified inferior median OS in comparison with responding patients (ABreal: 7.3 vs. 31.5 months, HR 3.7, 95%CI 2.8-8.5, p<0.001; Trials: 10.8 vs. NR, HR 4.6, 95%CI 3.3-6.3, p<0.001). PRef patients exhibited higher baseline systemic inflammation (neutrophil-to-lymphocyte ratio, NLR [&ge;] 3), a distinctively immunosuppressive TME enriched in CD163+ tumour-associated macrophages and a higher Treg/Teff ratio. RNAseq of tumour tissue demonstrated lower intrinsic immunogenicity in PRef samples, characterised by repressed IFN{gamma} and Teff signatures, with elevated myeloid infiltration. Conditional inference tree analysis identified IFN{gamma} signature downregulation combined with NLR [&ge;] 3 as the strongest contributor of PRef. Conclusions: PRef to A+B identifies a distinct biological entity characterised by unopposed systemic inflammation, myeloid cell infiltration and T-cell depletion. Targeting myeloid-mediated immunosuppression, particularly in patients with low IFN{gamma} signature expression and elevated NLR might enhance responsiveness to A+B.