A sequential therapy with sorafenib followed by regorafenib against single-line atezolizumab and bevacizumab in advanced hepatocellular carcinoma (HCC): Indirect treatment comparisons (ITC) using the RESORCE study.

Abstract

537 Background: Our objective was to explore ITC of sequential therapy with sorafenib followed by regorafenib (sorego) against atezolizumab+bevacizumab (atezobev) in HCC. Naïvely comparing median Overall Survival (OS) from sorafenib initiation on sorego in the RESORCE trial with atezobev from IMbrave150 is biased by population imbalance and selection of patients who survived on sorafenib in RESORCE. We used unanchored Matching Adjusted Indirect Comparison (MAIC) and novel survivorship bias adjusted anchored ITC (SBITC) to overcome these limitations. Methods: Overall Survival (OS) from sorafenib initiation in RESORCE on sorego were compared with reconstructed Individual Patient Data (IPD) OS based on Kaplan-Meier curves from IMbrave150 on atezobev. In MAIC, the sorego arm of RESORCE was weighted on reported baseline characteristics to match the IMbrave150 atezobev arm. In SBITC, sorafenib IPD from IMbrave150 were sampled with replacement and added to the RESORCE BSC arm to correct for “survivorship bias”. Sample size was sufficient for follow-up prior to RESORCE randomisation to be within 5% of sorafenib follow-up in IMbrave150. Anchored Bucher ITC was used to compare to unadjusted IMbrave150 with bootstrapping to obtain standard errors. Hazard Ratios (HRs) were used to compare OS. Results: MAIC gave an OS HR of 0.75 (95% confidence interval 0.60, 0.93; p-value=0.007). Effective sample size (ESS) was reduced from 379 to 271.75. SBITC gave a HR of 0.63 (0.52, 0.76; p-value <0.001) requiring an average of 83 sorafenib patients to be added from IMbrave150. Conclusions: We proposed MAIC and SBITC to overcome biases in naïve comparisons of OS on sorego with atezobev in HCC. However, MAIC is limited to baseline characteristics in IMbrave150 and does not adjust for “survivorship bias”. SBITC depends on sample size from IMbrave150, and makes no adjustment for imbalance in baseline characteristics. Despite these limitations, our methods were consistent with naïve comparisons and suggested longer OS with sorego compared to atezobev in patients with HCC.