Anchored matching-adjusted indirect comparison (MAIC) of regorafenib (REG) versus cabozantinib (CAB) in advanced hepatocellular carcinoma (HCC).

e16188 Background: REG and CAB, two therapies for second-line (2L) advanced HCC, have been evaluated vs. placebo (PBO) in randomized phase III trials, RESORCE (NCT01774344) and CELESTIAL (NCT01908426), respectively. In the absence of head-to-head studies, indirect treatment comparison (ITC) is needed to compare the efficacy and safety of REG vs. CAB. Anchored MAIC is an ITC method adjusting for cross-trial heterogeneity using PBO as the common comparator or anchor. Methods: Patient-level data from RESORCE were available to match the inclusion criteria and published baseline characteristics of CELESTIAL. 2L patients with prior sorafenib exposure of >3 months were included in the analyses. Overall survival (OS), progression-free survival (PFS), and Grade 3/4 treatment-emerging adverse events (TEAEs) occurring in > 10% of patients were compared. Matching was conducted for the following covariates, which were either imbalanced across trials or likely treatment effect modifiers (EMs): gender, region, Eastern Cooperative Oncology Group (ECOG), prior sorafenib duration, extrahepatic spread, macrovascular invasion, hepatitis B (HBV), hepatitis C (HCV), alcohol use, and alpha-fetoprotein (AFP) ≥400 ng/mL. Anchored MAIC were used to compare the relative effect of each treatment. To account for violation of the proportional hazards assumption, in addition to hazard ratios (HRs), restricted mean survival time (RMST) was estimated to quantify differences in OS and PFS. For safety, the relative effects were expressed as risk differences (RDs) (REG – CAB). Results: After matching, the effective sample sizes of REG and its PBO arm were 214 and 82, respectively. For OS, REG showed favorable although not statistically significant HR, and 30% longer mean survival with an RMST ratio approaching statistical significance. For PFS, there was no numerical difference in HR and no clinically meaningful difference based on RMST analyses (Table). REG showed lower incidence of diarrhea [RD of -7.1% (NE, NE)] and fatigue [RD of -6.3% (-14.6, 2.0)]. Conclusions: In this anchored MAIC, REG showed favorable OS and safety profile, especially for diarrhea and fatigue. No clinically meaningful difference was observed for PFS. The analysis could be limited by reduced sample size after matching and unobserved or missing EMs that could not be adjusted.[Table: see text]