Alessandro Vitale and Fabio Farinati equally contributed to this work.Prognosis and designing a treatment modality in patients with hepatocellular carcinoma (HCC) are extremely complex because, in most cases, this neoplasm is accompanied by the cirrhotic liver and other comorbidities. 1 Thus in such cases, tumour stage, the degree of liver function and general conditions of patients are the major deterministic factors. 2,3 A common and promising approach linking prognostic variables with therapeutic choice is to analyse large patient cohorts and, using treatment selection as the main endpoint, identify the fundamental treatment rules. An example of such a data-based approach is the HCC treatment schedule recently proposed by the Hong Kong Liver Cancer (HKLC) group. In this approach, Yau et al 4 applied an algorithm based on classification and regression tree statistical methodology to link the main treatment decision rules with HKLC tumour stages. The main limitation of this approach, however, was that the
Key points• Using a large consecutive cohort of patients with hepatocellular carcinoma from Italy (n = 4,867), we showed that a clear therapeutic hierarchy exists irrespective of tumour stage, live function, patient general conditions and treatment period.• This therapeutic hierarchy, in order of survival benefit gain, follows liver transplantation, liver resection, ablation, intra-arterial therapies, systemic therapy and best supportive care.• The results of this study suggest that a personalized approach based on the therapeutic hierarchy concept should be pursued in the context of a multidisciplinary evaluation, for each patient with hepatocellular carcinoma, irrespective of the stage of the disease.
AbstractBackground: The Italian Liver Cancer (ITA.LI.CA) prognostic system for patients with hepatocellular carcinoma (HCC) has recently been proposed and validated. We sought to explore the relationship among the ITA.LI.CA prognostic variables (ie tumour stage, functional score based on performance status and Child-Pugh score, and alpha-fetoprotein), treatment selection and survival outcome in HCC patients.
Patients and Methods:We analysed 4,867 consecutive HCC patients undergoing six main treatment strategies (liver transplantation, LT; liver resection, LR; ablation, ABL; intra-arterial therapy, IAT; Sorafenib, SOR; and best supportive care, BSC) and enrolled during 2002-2015 in a multicenter Italian database. In order to control pretreatment imbalances in observed variables, a machine learning methodology was used and inverse probability of treatment weights (IPTW) was calculated. An IPTWadjusted multivariate survival model that included ITA.LI.CA prognostic variables, treatment period and treatment strategy was then developed. The survival benefit of HCC treatments was described as a hazard ratio (95% confidence interval), using BSC as a reference value and as predicted median survival.Results: After the IPTW, the six treatment groups became well balanced for most baseline characteristics. In the IPTW-adjust...