Expanded biomarker analysis of an international, open-label phase 2 study of regorafenib plus pembrolizumab in patients with advanced hepatocellular carcinoma (HCC) previously treated with immune checkpoint inhibitors (ICI).

635 Background: The optimal treatment for patients (pts) with advanced HCC who progress on an ICI-containing regimen has not been defined. In pts who progressed on one prior ICI regimen, regorafenib plus pembrolizumab (rego + pembro) showed modest anti-tumor activity (primary completion analysis: overall objective response rate 7.4%, all partial responses [PR]) and an expected safety profile (NCT04696055; El-Khoueiry, ASCO 2024). We conducted additional exploratory analyses of tumor markers to gain insights into the clinical results. Methods: Overall, 95 adults with unresectable advanced HCC (C-P Class A, BCLC stage B/C, ECOG PS 0/1) received oral regorafenib 90 mg once daily 3 weeks on/1 week off plus pembrolizumab 400 mg i.v. every 6 weeks. Prior treatment: Cohort 1 = atezolizumab + bevacizumab (n=68); Cohort 2 = any other ICI alone or in combination (n=27). Biomarker analyses in subsets of pts explored potential correlations with clinical benefit (defined as tumor response or stable disease [SD] ≥ 4 months). Immune cell subtypes in blood were analyzed by flow cytometry. Tumor immune markers were assessed by immunohistochemistry. Gene expression profiling in tumor samples was by RNAseq. Results: Twenty-one pts had baseline samples for RNAseq, 7 of whom had clinical benefit (n= 3 Cohort 1 [including 1 PR], n=4 Cohort 2). Pts with clinical benefit had differential gene expression at baseline including enriched expression of genes associated with Notch, Hedgehog, fibroblast, TGF-β, and insulin-like growth factor binding protein signaling; in contrast, there was higher expression of genes associated with myeloid derived suppressor cells in pts without clinical benefit. Of note, tumor samples from pts in this study had increased Tregs and T-cell exhaustion signatures compared with pts treated with rego + pembro in a first-line study (NCT03347292).Analyses of available paired tumor samples from 4 patients (n=2 SD, n=2 progressive disease) found reductions in activated fibroblast, TGF-β, and collagen gene expression signatures from baseline to Week 6 Day 1 of treatment. Conclusions: Rego + pembro had modest activity in pts who progressed on one prior ICI-based regimen. Exploratory biomarker analyses suggest different gene expression in pts with or without clinical benefit. Clinical benefit was associated with enriched expression of genes in Notch, Hedgehog, fibroblast, and TGF-β pathways. Pre- and post-treatment biopsies showed reductions in fibroblast, TGF-β, and collagen gene expression, which are known features of HCC. In the context of the limitations of small sample size and absence of a control arm, these findings are hypothesis generating and consistent with HCC biology and regorafenib mechanism of action. Clinical trial information: NCT04696055 .