Prognostic and predictive factors from the phase III CELESTIAL trial of cabozantinib (C) versus placebo (P) in previously treated advanced hepatocellular carcinoma (aHCC)

Abstract

Background: C improved overall survival (OS) and progression‐free survival (PFS) vs P in the phase 3 CELESTIAL trial (NCT01908426). Median OSwas 10.2mowithC vs 8.0mo with P (HR 0.76, 95% CI0.63‐0.92; p=0.005), and median PFS was 5.2 movs 1.9mo (HR 0.44, 95% CI 0.36‐0.52; p <0.001) (Abou‐Alfa, NEJM 2018). Here, retrospective univariate (UV) and multivariate (MV) analyses were done to identify prognostic and predictive factors for OS from CELESTIAL. Methods: 707 patients (pts) were randomized 2:1 to receive C (60mgqd) or P. Ptswere Child‐Pugh grade A and ECOG PS < 1 and must have received prior sorafenib and could have received up to 2 prior lines of systemic therapy for HCC. UV and MV analyses of OS were done using the Cox proportional hazard regression model to compare subgroups defined by baseline variables within each treatment group (Bruix, J Hepatol 2017). MV analyses used backward selection with cutoffs of p < 0.1 for inclusion and p < 0.05 for retention of prognostic factors in the final model. Variables with p < 0.05 for treatment‐subgroup interaction were considered possible predictive factors for treatment benefit. Results: Age, sex, race, region, and etiology were not significant covariates in stepwise UV and MV analyses of OS. Nine baseline variables related to disease status, laboratory values, or prior therapy were identified as possible prognostic factors for OS in one or both treatment groups (Table). No covariates were found to be predictive of an OS benefit with C. Conclusions: Exploratory UV and MV analyses identified high baseline values of AFP, alkaline phosphatase, ALBI grade, neutrophil to lymphocyte ratio, and number of disease sites as possible prognostic factors for shorter OS in previously treated aHCC. Disease etiology and demographic factors such as race and region were not found to be prognostic for OS.