Regorafenib plus nivolumab in unresectable hepatocellular carcinoma: the phase 2 RENOBATE trial

Anti-PD-1 monotherapy has not improved overall survival in first- or second-line settings for unresectable hepatocellular carcinoma (uHCC) in global phase 3 trials. In contrast, combinations of anti-PD-L1 with anti-VEGF (atezolizumab–bevacizumab) or anti-CTLA-4 (durvalumab–tremelimumab, STRIDE) have shown survival benefits over sorafenib. A VEGFR2 TKI (rivoceranib) plus anti-PD-1 (camrelizumab) also improved PFS and OS versus sorafenib, although other TKI plus ICI combinations (cabozantinib–atezolizumab; lenvatinib–pembrolizumab) have not consistently improved OS. Regorafenib, an MKI with anti-angiogenic and immunomodulatory activity including CSF1R pathway inhibition, may reverse myeloid-driven immunosuppression and potentiate anti-tumor immunity. Given the limited predictive biomarkers for ICI response in uHCC and the mechanistic rationale, the RENOBATE phase 2 trial evaluated front-line regorafenib plus nivolumab in uHCC, with comprehensive exploratory immune profiling to identify correlates of response.

Prior studies established that anti-PD-1 monotherapy (for example, nivolumab in CheckMate 459) did not significantly improve OS versus sorafenib in uHCC. Combination regimens have demonstrated variable success: atezolizumab–bevacizumab (IMbrave150) and durvalumab–tremelimumab (HIMALAYA/STRIDE) improved OS, and rivoceranib–camrelizumab improved both PFS and OS versus sorafenib. Conversely, cabozantinib–atezolizumab (COSMIC-312) improved PFS but not OS, and lenvatinib–pembrolizumab (LEAP-002) failed to improve PFS or OS. These discrepancies suggest that distinct molecular and immunologic effects of combination partners influence efficacy. Regorafenib targets VEGFRs and CSF1R, potentially reprogramming immunosuppressive myeloid cells toward an antitumor phenotype. Biomarker studies in uHCC have implicated immune features in ICI outcomes, but no predictive biomarkers are established. Alterations in WNT/β-catenin (for example, CTNNB1) have been associated with immune exclusion and ICI resistance in some contexts, though effects in ICI plus anti-angiogenic combinations may be attenuated.

Design: Open-label, multicenter, single-arm, phase 2 trial (RENOBATE) at three academic centers in South Korea (Asan Medical Center, Samsung Medical Center, Bundang CHA Hospital). Trial registration: NCT04310709. Ethics approvals obtained; written informed consent from all patients. Participants: 42 patients with unresectable or metastatic HCC, transarterial chemoembolization-unfeasible/refractory BCLC B or BCLC C, no prior systemic therapy, ECOG 0–1, Child-Pugh A, at least one measurable lesion by RECIST 1.1. Key exclusions included other HCC histologies (fibrolamellar, sarcomatoid, combined HCC-CCA), active/untreated CNS metastases, significant autoimmune disease, significant cardiovascular disease, uncontrolled infections, prior checkpoint inhibitors, significant bleeding risk, recent major surgery, and others per protocol. Treatment: Nivolumab 480 mg IV day 1 of each 4-week cycle plus regorafenib 80 mg orally daily on days 1–21 of each 28-day cycle (3-weeks-on/1-week-off). Responses assessed every 8 weeks by RECIST 1.1; safety evaluated on C1D1, C1D15, and each cycle day 1. Regorafenib dose reductions allowed; nivolumab dose reductions not allowed. Endpoints: Primary—investigator-assessed ORR per RECIST 1.1 in the ITT population. Secondary—safety (NCI-CTCAE v5.0), ORR per modified RECIST, PFS, OS. Exploratory—correlative biomarker analyses: ctDNA (Guardant360 CDx), single-cell RNA sequencing (scRNA-seq), TCR repertoire, and multicolor flow cytometry of PBMCs collected serially (baseline C1D1 and on-treatment C1D15, C2D1, C3D1). For exploratory analyses, patients were categorized post hoc as early progressors (PD at first evaluation or progressive increase leading to PD; n=14) and long-term responders (sustained tumor burden decrease ≥10 months; n=15). Statistical plan: Fleming’s single-stage design assuming ORR improvement from 7% (sorafenib) to 25% (regorafenib–nivolumab); alpha 0.05 (two-sided), power 90%; planned n=35; with 15% attrition, target n=42. ITT used for efficacy; all treated for safety. Kaplan–Meier for PFS and OS; log-rank tests for comparisons. Molecular/immune assays: Baseline ctDNA with Guardant360 CDx. scRNA-seq on 10x Genomics Chromium, analyzed with Cell Ranger and Seurat; quality control (mitochondrial gene expression >5%, >4500 genes filtered), batch correction, clustering (SNN/UMAP), subclustering for monocytes and T/NK cells, module scoring (gene sets for inflammatory signatures, regorafenib responsiveness, CSF1R knockout, M1-like signatures, TAM signatures), pseudotime (monocle3), and CellChat interactome analyses. TCR diversity by inverse Simpson index on CD8+ T cells. Flow cytometry for CD8+ T-cell proliferation/activation (Ki-67, PD-1, granzyme B, perforin) and monocyte subsets (classical, intermediate, non-classical). In vitro monocyte polarization assays from healthy donor CD14+ monocytes with IFN-γ or IL-4 ± regorafenib, assessing TNF-α+ CD86+ cells.

• Population: 42 enrolled/treated; median age 61 years (40–79); 73.8% male; BCLC C in 88.1%; HBV etiology 71.4%; prior TACE 83.3%.
• Primary endpoint: ORR (RECIST 1.1) 31.0% (CR 2.4% [1/42], PR 28.6% [12/42]); SD 50.0% (21/42); PD 14.3% (6/42); not assessable 4.8% (2/42). Met predefined threshold (≥25%).
• Secondary efficacy: ORR by modified RECIST 33.3% (CR 4.8% [2/42], PR 28.6% [12/42]). With median follow-up 11.1 months (95% CI 6.11–14.0): median PFS 7.38 months (95% CI 4.12–13.0); 1-year PFS rate 37.8%. 1-year OS rate 80.5% (95% CI 63.0–90.3%); median OS not reached. Median duration of response (RECIST 1.1) 10.3 months (95% CI 8.2–13.9).
• Safety: Any-grade AEs in 92.9% (39/42); grade 3 AEs in 23.8% (10/42); no grade 4 AEs or treatment-related deaths. Common AEs: palmar-plantar erythrodysesthesia 38.1% (16/42), alopecia 26.2% (11/42), skin rash 23.8% (10/42), fatigue 23.8% (10/42), AST increased 23.8% (10/42; grade 3 in 4.8% [2/42]). Discontinuation of one agent due to AEs in 9.5% (4/42) (regorafenib in 3; nivolumab in 1); no discontinuation of both agents. Regorafenib dose reductions in 19.0% (8/42). Median adherence to any study agent 7.36 months; to full combination 5.61 months.
• ctDNA: Most frequent mutations—TP53 69%, CTNNB1 26%. Altered WNT/β-catenin pathway genes were not associated with inferior PFS/OS.
• scRNA-seq/immune correlates: Long-term responders exhibited increased CD8+ TCR repertoire diversity (inverse Simpson index) and increased frequencies of effector and MKI67+ proliferating CD8+ T cells during treatment; enrichment of inflammatory and nivolumab-responsiveness gene signatures in MKI67+ CD8+ T cells; upregulation of cytotoxicity genes at C1D15 (validated by increased Ki-67+ and Granzyme B+Perforin+ CD8+ T cells by flow cytometry). Monocyte analyses revealed trajectories toward M1-like activation versus M2-skewed non-classical monocytes; regorafenib–nivolumab increased the proportion of M1-branch cells and M1-like gene signatures only in long-term responders, with increased ratios of classical to non-classical monocytes (flow cytometry). Classical monocytes in long-term responders showed enrichment for regorafenib-responsive and CSF1R-knockout signatures; IFN-γ-related signaling from proliferating CD8+ T cells to classical monocytes was evident, with enrichment of IFN-γ response and antigen processing/presentation signatures. In vitro, regorafenib promoted M1-directed polarization (TNF-α+CD86+) of classical monocytes. Early progressors showed upregulation of TMEM176A/B in classical monocytes and enrichment of TAM-related signatures; long-term responders showed increased NLRP3 and IL18 expression and NLRP3 inflammasome pathway signature, not observed in early progressors.
• External anti-PD-1 monotherapy cohort: Cytotoxic features increased among CXCR3+ CD8+ T cells in responders, but monocyte regorafenib/CSF1R signatures were not enriched post-therapy, supporting a regorafenib-specific myeloid modulation with the combination.

The RENOBATE trial demonstrated clinically meaningful activity of front-line regorafenib plus nivolumab in uHCC, meeting the ORR primary endpoint and yielding a median PFS of 7.38 months and 1-year OS of 80.5%. Efficacy was comparable to other successful ICI plus anti-angiogenic combinations and superior to historical ICI monotherapy in uHCC, suggesting additive or synergistic effects. Safety was favorable, with fewer grade 3–4 events than reported for some other TKI–ICI regimens, potentially due to the reduced regorafenib dose. Exploratory immune profiling linked clinical benefit to diversification and proliferation of CD8+ T cells and to M1-directed polarization of monocytes, with evidence for IFN-γ-mediated crosstalk between proliferating CD8+ T cells and classical monocytes. Enrichment of regorafenib-responsive and CSF1R-knockout gene signatures in classical monocytes among long-term responders supports a regorafenib-driven myeloid reprogramming that may augment anti-tumor immunity in combination with PD-1 blockade. Conversely, early progression was associated with TMEM176A/B upregulation and immunosuppressive TAM-like features, potentially attenuating inflammasome activity and contributing to resistance. ctDNA analyses did not show detrimental associations for WNT/β-catenin pathway alterations in this combination setting, aligning with reports from other ICI–anti-angiogenic regimens.

Regorafenib plus nivolumab is an active and well-tolerated first-line regimen for unresectable HCC, achieving an ORR of 31%, median PFS of 7.38 months, and high 1-year OS. Translational analyses suggest that therapeutic benefit is associated with CD8+ T-cell proliferation/diversification and regorafenib-facilitated M1-directed monocyte polarization via CSF1R pathway modulation and IFN-γ signaling. These findings provide biological rationale for further clinical development of regorafenib–anti-PD-1/PD-L1 combinations in HCC and highlight candidate immune biomarkers (for example, MKI67+ CD8+ T-cell expansion, M1 polarization signatures, TMEM176A/B expression) for response and resistance. Future randomized trials and larger, prospectively designed biomarker studies are warranted to validate these observations and to optimize patient selection and dosing strategies.

• Single-arm design without a nivolumab or regorafenib monotherapy control limits causal attribution to the combination and prevents direct quantification of regorafenib’s additive effect.
• Small sample size and single-country, multicenter study may limit generalizability.
• Exploratory biomarker analyses were post hoc with limited sample numbers for scRNA-seq and flow cytometry subsets; subgroup definitions were determined after primary efficacy analysis.
• Lack of tumor tissue-based correlative analyses; peripheral blood may not fully reflect intratumoral immune dynamics.
• The regorafenib-responsive gene sets may not fully capture regorafenib’s effects in HCC patients, and comparisons to external anti-PD-1 monotherapy cohorts occurred in differing clinical contexts and timepoints.
• Differential mechanisms versus other VEGF-targeted agents were not delineated and may underlie varied efficacy across TKI–ICI combinations.