FDA analysis of treatment efficacy based on etiology of hepatocellular carcinoma

4117 Background: Retrospective analyses of trial-level data report that etiology of hepatocellular carcinoma (HCC) (e.g., viral versus [vs.] non-viral) may influence efficacy of systemic therapies used in advanced HCC. FDA conducted an analysis of 9 randomized controlled trials (RCTs; n=5,417) to assess treatment effects based on drug class, treatment line, and HCC etiology. Methods: FDA conducted a retrospective, exploratory analysis of 3 RCTs comparing first line (1L) treatment with anti-programmed cell death (ligand)-1-containing regimens (αPD-[L]1) to single-agent anti-vascular endothelial growth factor therapy (αVEGF), and 6 RCTs comparing second line (2L) treatment consisting of αPD-(L)1 or αVEGF to placebo. For each line of treatment (1L or 2L), meta-analyses of overall survival (OS) were performed for the overall treatment effect and by disease etiology, viral (hepatitis B [HBV], or hepatitis C [HCV]) vs. non-viral. Subgroup analyses were performed to investigate the confounding effect of subsequent therapy (yes vs. no) and region (Asian vs. non-Asian) on etiology. Trial-level hazard ratios (HRs) and 95% confidence intervals (CI) were mostly calculated using patient-level data submitted to FDA. Pooled HRs and CIs were calculated using a random-effects model. Results: While 1L αPD-(L)1 improved OS compared to 1L αVEGF in the overall population (HR [95% CI]: 0.8 [0.73, 0.88]), treatment effects were comparable for viral (HR [95% CI]: 0.75 [0.6, 0.92]; n=1,435) and non-viral (HR [95% CI]: 0.83 [0.72, 0.95]; n=979) subgroup analyses, and not influenced by subsequent therapy or region. Analyses based on specific viral infection (i.e., HBV or HCV), showed that improvement in OS for αPD-(L)1 vs. αVEGF appeared larger for HBV (HR [95% CI]: 0.7 [0.59, 0.82]; n=833) compared to HCV (HR [95% CI]: 0.83 [0.6, 1.16]; n=602). Treatment effects were also comparable in 2L for different etiologies for both αPD-(L)1 (HR [95% CI] for viral: 0.76 [0.62, 0.93]; n=529 and non-viral: 0.85 [0.65, 1.11]; n=337) and αVEGF (HR [95% CI] for viral: 0.79 [0.69, 0.9]; n=1,288 and non-viral: 0.74 [0.63, 0.88]; n=849), and not influenced by subsequent therapy. Nonetheless, compared to placebo, αPD-(L)1 had a larger treatment effect with HBV (HR [95% CI]: 0.69 [0.5, 0.94]; n=459) compared to non-viral etiologies. Conclusions: The results did not demonstrate that etiology of HCC was predictive of OS for both treatment classes based on retrospective, exploratory meta-analyses of 9 RCTs; however, the treatment effects may be more pronounced in patients with HBV who received αPD-(L)1 as 1L or 2L. Uncertainty remains regarding subgroup treatment effects given limited number of trials in the analyses (mostly leading to approvals), overlapping CIs, number of patients in some subgroups, retrospective nature, and exclusion of trials with inadequate data regarding etiology of HCC.