OBJECTIVES: There is scarce methodological guidance to conduct indirect treatment comparisons (ITCs) of treatment sequences. Previous reviews in oncology have highlighted that ITC of sequential therapies usually evaluate each line of therapy independently. In hepatocellular carcinoma (HCC), regorafenib was compared against best supportive care (BSC) on patients who progressed on sorafenib in the RESORCE trial, but effectiveness against other treatments have not been explored. We developed a novel approach to ITC of overall survival (OS) on sequential therapy with sorafenib followed by regorafenib (sorego) against atezolizumab + bevacizumab (atezobev) in HCC, by averaging survival curves across lines of therapy.

METHODS: We first conducted an ITC of first-line treatments for HCC using fractional polynomial network meta-analysis. Then, we fit parametric models to second-line treatments in the RESORCE trial. RESORCE reported the time from initiation of first line sorafenib to randomisation, which we used to estimate time to initiate second line therapy. We then combined curves at first and second-line treatments using a weighted average of the first line and RESORCE curves, with weights defined as the inverse square root of number of patients on each line of therapy.

RESULTS: A lognormal model fit to RESORCE OS was selected by statistical fit and clinical plausibility of extrapolations. The averaged curves showed that the sorego sequence had a marginally better OS. The time-varying hazard ratios (HRs) showed a benefit of sorego against atezobev over time, with a HR (95% confidence interval [CI]) of 0.98 (0.82-1.12) at 12 months, to 0.95 (0.84-1.09) at 24 months, and 0.87 (0.80-0.96) at 36 months. Nevertheless, the results were sensitive to the selection of parametric model for RESORCE OS.

CONCLUSIONS: We proposed a novel parametric ITC methodology to estimate overall efficacy of sequences of treatments. Further applications are needed to test this method in other treatment sequences.