Hepatocellular carcinoma — clinical trials landscape¶
TL;DR — The HCC trial portfolio has shifted from single-agent kinase inhibition toward immune combinations, perioperative therapy, TACE/systemic integration, personalized radiation, biomarker-guided detection, and cell/vaccine approaches. Landmark completed trials form a clear genealogy—SHARP (NCT00105443), REFLECT (NCT01761266), IMbrave150 (NCT03434379), HIMALAYA (NCT03298451), and CheckMate 9DW (NCT04039607)—but they do not provide head-to-head ranking of current first-line options. Active studies are moving checkpoint therapy before and after surgery, into transplant-adjacent settings, and alongside TARE/TACE. Surveillance trials are testing abbreviated MRI against ultrasound and prospective blood-marker cohorts, while prevention trials finally target cirrhosis/MASLD populations. Trial eligibility remains narrower than real-world HCC because Child-Pugh B/C disease, decompensation, transplant, autoimmunity, and poor performance are commonly excluded.
All 39 registry identifiers below were re-retrieved live from ClinicalTrials.gov API v2 on 2026-08-30. Status is a snapshot, not a permanent attribute.
Landmark completed phase 3 genealogy¶
| Trial | NCT | Comparison | Published anchor | What changed |
|---|---|---|---|---|
| SHARP | NCT00105443 | Sorafenib vs placebo | Llovet 2008, PMID 18650514 | First systemic OS benefit |
| REFLECT | NCT01761266 | Lenvatinib vs sorafenib | Kudo 2018, PMID 29433850 | Non-inferior TKI alternative |
| RESORCE | NCT01774344 | Regorafenib vs placebo after sorafenib | Bruix 2017, PMID 27932229 | Validated second line after tolerated sorafenib |
| IMbrave150 | NCT03434379 | Atezolizumab-bevacizumab vs sorafenib | Finn 2020, PMID 32402160 | IO/VEGF first-line standard |
| HIMALAYA | NCT03298451 | Tremelimumab-durvalumab/durvalumab vs sorafenib | Abou-Alfa 2022, PMID 38319892 | Dual-IO first-line option |
| CheckMate 9DW | NCT04039607 | Nivolumab-ipilimumab vs lenvatinib/sorafenib | Yau 2025, PMID 40349714 | OS 23.7 vs 20.6 months; additional dual-IO first-line option |
| IMbrave050 | NCT04102098 | Adjuvant atezolizumab-bevacizumab vs surveillance | Qin 2023, PMID 37871608; Yopp 2026, PMID 41580093 | Initial RFS signal not sustained; benefit–risk does not support use |
| EMERALD-1 | NCT03778957 | TACE + durvalumab ± bevacizumab | Sangro 2025, PMID 39798579 | Positive PFS for TACE + IO/VEGF |
| LEAP-012 | NCT04246177 | TACE + lenvatinib-pembrolizumab vs TACE control | Kudo 2025, PMID 39798578 | Positive PFS for triplet strategy |
ClinicalTrials.gov listed SHARP, REFLECT, RESORCE, IMbrave150, CheckMate 9DW, IMbrave050, and LEAP-012 as completed on 2026-08-30. HIMALAYA and EMERALD-1 were active, not recruiting, reflecting ongoing follow-up or registry maintenance despite primary publication.
Current portfolio by disease phase¶
| Disease phase | Active direction | Scientific question |
|---|---|---|
| Prevention | Aspirin or metabolic intervention in cirrhosis/MASLD | Can HCC incidence be reduced before cancer? |
| Surveillance | Abbreviated MRI, biomarker cohorts, risk modeling | Can early detection improve without excessive harm? |
| Neoadjuvant | IO ± anti-VEGF/TKI before resection | Does pathologic response predict recurrence benefit? |
| Adjuvant | IO/TKI after high-risk cure | Can micrometastatic recurrence be prevented? |
| Transplant bridge/downstage | TARE/TACE ± IO | Can tumor be controlled without rejection risk? |
| Intermediate | TACE/TARE plus systemic therapy | Which tumors need combined local-systemic control? |
| Advanced | New checkpoints, bispecifics, vaccines, cell therapy | Can response deepen beyond current combinations? |
| Resistance | Post-IO TKI, microbiome, adaptive platforms | What reverses primary/acquired resistance? |
Surveillance and prevention trials¶
| NCT | Status at retrieval | Design signal | Why it matters |
|---|---|---|---|
| NCT04965259 | Active, not recruiting | ELEGANCE high-risk prospective early-detection cohort | Longitudinal biospecimens and marker validation |
| NCT05486572 | Recruiting | PREMIUM: ultrasound+AFP versus abbreviated MRI+AFP every six months | 4,700-person RCT powered for HCC mortality through planned 2031 completion |
| NCT06658782 | Recruiting | Abbreviated MRI versus ultrasound surveillance | Direct modality comparison in high-risk people |
| NCT07010588 | Not yet recruiting | Ultrasound versus abbreviated MRI with/without hepatobiliary phase | Compares practical MRI strategies |
| NCT07291141 | Recruiting | MRI versus ultrasound in cirrhosis | Detection plus body composition/decompensation risk |
| NCT07529262 | Not yet recruiting | Phase 3 aspirin prevention in cirrhosis | Hard prevention endpoint rather than association |
| NCT07495995 | Not yet recruiting | Phase 2 SAMe prevention in MASLD cirrhosis | Etiology-specific chemoprevention |
| NCT04172779 | Not yet recruiting | Erlotinib HCC chemoprevention | Repurposed pathway intervention |
The critical design distinction is whether a study measures test accuracy, stage shift, or mortality. Better sensitivity does not guarantee net benefit.
Perioperative and adjuvant wave¶
| NCT | Status | Intervention | Boundary |
|---|---|---|---|
| NCT05440864 | Recruiting | Durvalumab-tremelimumab in resectable HCC | Pathologic response and surgical timing |
| NCT04658147 | Active, not recruiting | Perioperative nivolumab ± relatlimab | Small early-phase feasibility |
| NCT05701488 | Active, not recruiting | SIRT + tremelimumab-durvalumab for resectable HCC | Local radiation plus IO before surgery |
| NCT06089382 | Not yet recruiting | Phase 3 sintilimab-lenvatinib after resection with PVTT | Very-high-risk adjuvant population |
| NCT06498622 | Recruiting | Donafenib-envafolimab after high-risk cure | Adjuvant combination |
| NCT07350824 | Not yet recruiting | MRD prediction and adjuvant PD-1 | Biomarker-guided postoperative treatment |
| NCT06311929 | Recruiting | Phase 4 precision adjuvant therapy | Real-world strategy evaluation |
Perioperative trials face special bias: pathologic response is available only to patients reaching surgery, and immune toxicity can delay or prevent definitive treatment.
Locoregional-systemic integration¶
| NCT | Status | Strategy |
|---|---|---|
| NCT03778957 | Active, not recruiting | EMERALD-1: TACE + durvalumab ± bevacizumab |
| NCT04246177 | Completed | LEAP-012: TACE + lenvatinib-pembrolizumab |
| NCT05063565 | Active, not recruiting | TheraSphere + durvalumab-tremelimumab |
| NCT04605731 | Active, not recruiting | Durvalumab-tremelimumab after radioembolization |
| NCT06999694 | Recruiting | Proton radiotherapy + tremelimumab-durvalumab |
| NCT07059494 | Recruiting | Atezolizumab-bevacizumab + Y-90 for transplant pathway |
| NCT01918683 | Active, not recruiting | TACE ± SBRT as transplant bridge |
Combined trials must report liver decompensation and transplant consequences alongside PFS.
Resistance and adaptive strategies¶
| NCT | Status | Hypothesis |
|---|---|---|
| NCT04770896 | Active, not recruiting | Atezolizumab plus lenvatinib/sorafenib after atezolizumab-bevacizumab |
| NCT05690048 | Recruiting | Fecal microbiota transfer to overcome atezolizumab-bevacizumab resistance |
| NCT07328009 | Recruiting | Personalized adaptive platform |
| NCT07310173 | Recruiting | AST-3424 in AKR1C3-high HCC |
| NCT06109272 | Recruiting | Livmoniplimab-budigalimab |
| NCT06954467 | Recruiting | JS207 + JS007 first line |
Adaptive platforms can test multiple hypotheses efficiently, but small molecular strata and changing standards complicate control arms.
Cell and vaccine therapies¶
| NCT | Status | Platform | Key concern |
|---|---|---|---|
| NCT06044506 | Recruiting | Autologous natural-killer cells | Persistence and manufacturing |
| NCT07123545 | Recruiting | Autologous neoantigen-specific T cells | Turnaround and personalized product |
| NCT04147078 | Recruiting | Personalized dendritic-cell vaccine after surgery | Antigen selection and recurrence endpoint |
| NCT06676982 | Recruiting | CD19 CAR-T in advanced HCC | Biological rationale and off-tumor effects need clarity |
HCC cell therapy must navigate antigen heterogeneity, an immunosuppressive liver, cirrhosis, and on-target injury to normal tissue.
Trial-population gap¶
Common exclusions create evidence gaps:
- Child-Pugh B/C or active decompensation;
- uncontrolled varices or recent bleeding;
- solid-organ transplant;
- active autoimmune disease;
- severe cardiovascular or renal disease;
- poor ECOG performance;
- mixed HCC-cholangiocarcinoma;
- limited access to trial centres.
These exclusions protect safety and endpoint interpretability but limit real-world transportability.
Endpoints and pitfalls¶
| Endpoint | Strength | Pitfall |
|---|---|---|
| Overall survival | Patient-important and integrative | Subsequent therapy and competing liver death |
| PFS | Earlier readout | Imaging timing and uncertain OS translation |
| Recurrence-free survival | Relevant after cure | Does not distinguish delayed recurrence from cure |
| Objective response | Useful for conversion/downstaging | Criteria and confirmation vary |
| Pathologic response | Direct tissue effect | Only resected patients assessed |
| Time to deterioration/QoL | Patient-centered | Missing data and open-label bias |
| Child-Pugh/ALBI worsening | Liver preservation | Tumor, toxicity, and infection confounded |
Live registry snapshot and denominator discipline¶
| Portfolio | Registered status on 2026-08-30 | Planned enrollment | Interpretation |
|---|---|---|---|
| HIMALAYA (NCT03298451) | Active, not recruiting | 1,324 | Published efficacy does not require the registry to be “completed”; long-term follow-up remains active |
| CheckMate 9DW (NCT04039607) | Completed | 732 | Registry enrollment differs from the 668 randomized patients in the global publication (Yau 2025, PMID 40349714) |
| IMbrave050 (NCT04102098) | Completed | 668 | Updated RFS HR 0.90 (95% CI 0.72–1.12); initial benefit was not sustained (Yopp 2026, PMID 41580093) |
| EMERALD-1 (NCT03778957) | Active, not recruiting | 724 | PFS publication precedes final registry closure and mature OS |
| LEAP-012 (NCT04246177) | Completed | 479 | Completed recruitment/follow-up snapshot; OS interpretation still depends on published maturity |
| MORPHEUS-Liver resistance study (NCT04770896) | Active, not recruiting | 557 | Tests continuation/change of pathway after atezolizumab-bevacizumab |
| ELEGANCE detection cohort (NCT04965259) | Active, not recruiting | 2,002 | Large prospective biomarker cohort, not a randomized mortality trial |
| PREMIUM surveillance RCT (NCT05486572) | Recruiting | 4,700 | HCC-mortality endpoint; estimated completion September 2031 |
| Adaptive platform (NCT07328009) | Recruiting | 350 | Efficiently tests multiple hypotheses but requires contemporaneous controls |
| AspiRe HCC prevention (NCT07529262) | Not yet recruiting | 890 | Phase 3 prevention ambition; no efficacy inference until enrollment and results |
| MRD-guided adjuvant study (NCT07350824) | Not yet recruiting | 276 | Registry concept, not validation that MRD-triggered treatment improves outcomes |
Registry status must not be confused with evidence status. “Recruiting” does not mean promising, “completed” does not mean positive, and a planned sample is not an analyzed sample. Conversely, an active-not-recruiting registry can have a practice-changing primary publication while survival follow-up continues.
What earlier trial generations teach¶
A systematic review of randomized HCC trials from 2002–2020 showed that positive evidence accumulated unevenly across disease stages and that many apparently plausible interventions failed at phase 3 (Haber 2021, PMID 34126063). STORM is the canonical adjuvant warning: sorafenib activity in advanced disease did not prevent recurrence after cure, and molecular analyses found prognostic signals without a validated predictive subgroup that rescued the negative trial (Pinyol 2019, PMID 30108162). By contrast, adjuvant cytokine-induced killer cells showed recurrence benefit in a randomized setting but did not become a globally standardized platform because manufacturing, replication, and evolving comparators matter (Lee 2015, PMID 25747273).
Conversion and neoadjuvant studies introduce denominator risk. A single-arm lenvatinib-plus-PD-1 phase 2 can report resection among responders, but the meaningful denominator is everyone who started conversion therapy and the endpoints include progression that closes surgery, operative delay, pathologic response, recurrence, and OS (Zhang 2023, PMID 37730273). In neoadjuvant cabozantinib-nivolumab, 5 of 15 patients had a pathologic response; spatial analysis revealed different resistance and recurrence programs, but 15 patients cannot validate a surrogate endpoint (Zhang 2023, PMID 37723590).
Early-phase innovation: signal versus proof¶
| Platform | Early evidence | Missing proof |
|---|---|---|
| TACE plus lenvatinib | Prospective multicentre single-arm phase 2 in selected Child-Pugh A intermediate HCC | Randomized attribution and OS (Kudo 2024, PMID 38344448) |
| TACE then pembrolizumab | PETAL phase 1b: 15 patients; no dose-limiting synergistic toxicity, treatment-related events in 93% | Comparative efficacy and decompensation risk (Pinato 2024, PMID 38578610) |
| Personalized neoantigen vaccine plus pembrolizumab | Phase 1/2 demonstrated feasibility and immune/clinical signals | Manufacturing scalability, randomized benefit, and biomarker-defined responders (Yarchoan 2024, PMID 38584166) |
| GPC3 CAR-T | Phase-1 trials established feasibility and antitumor activity | Persistence, antigen escape, on-target liver injury, and controlled efficacy (Shi 2020, PMID 32371538) |
| Perioperative checkpoint therapy | Multiple early trials and pathologic responses | Validated relationship between pathologic response, recurrence, OS, and transplant safety (Llovet 2024, PMID 38424197) |
Highest-value portfolio gaps¶
- Head-to-head first-line IO combinations.
- Randomized sequencing after IO-combination failure.
- Child-Pugh B-specific trials with liver and cancer endpoints.
- Completion of a modern surveillance mortality trial outside HBV; PREMIUM is recruiting but has no outcome data (NCT05486572).
- Risk-enriched non-cirrhotic MASLD surveillance.
- Prospective biomarker-treatment interaction.
- Transplant-safe perioperative IO strategies.
- Patient-reported outcomes that separate tumor and cirrhosis burden.
Open questions¶
- Will PFS-positive TACE combinations improve OS and preserve liver function?
- Which MRD assay is ready to randomize adjuvant treatment (NCT07350824)?
- Can microbiome intervention reverse IO resistance reproducibly (NCT05690048)?
- Will abbreviated MRI improve HCC mortality and net benefit, not merely detection, in PREMIUM and the companion modality trials (NCT05486572; NCT06658782; NCT07010588)?
- How can decompensated patients be studied safely without excluding the population most unlike pivotal trials?
Related pages¶
- Systemic therapy — interprets completed phase 3 trials.
- Locoregional therapy — expands TACE/TARE combinations.
- Surveillance and early detection — explains detection endpoints.
- Biomarkers — evaluates selection and MRD.
- Red flags and safety concerns — covers eligibility and toxicity.
References¶
- Llovet JM, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med. 2008;359:378-390. PMID 18650514
- Kudo M, et al. Lenvatinib versus sorafenib in unresectable HCC. Lancet. 2018;391:1163-1173. PMID 29433850
- Bruix J, et al. Regorafenib after sorafenib (RESORCE). Lancet. 2017;389:56-66. PMID 27932229
- Finn RS, et al. Atezolizumab plus bevacizumab in unresectable HCC. N Engl J Med. 2020;382:1894-1905. PMID 32402160
- Abou-Alfa GK, et al. Tremelimumab plus durvalumab in unresectable HCC. NEJM Evid. 2022;1:EVIDoa2100070. PMID 38319892
- Yau T, et al. Nivolumab plus ipilimumab versus lenvatinib or sorafenib as first-line treatment for unresectable hepatocellular carcinoma (CheckMate 9DW). Lancet. 2025;405:1851-1864. PMID 40349714
- Qin S, et al. IMbrave050. Lancet. 2023;402:1835-1847. PMID 37871608
- Sangro B, et al. EMERALD-1. Lancet. 2025;405:216-232. PMID 39798579
- Kudo M, et al. LEAP-012. Lancet. 2025;405:203-215. PMID 39798578
- Haber PK, et al. Evidence-based management of HCC: systematic review of randomized trials, 2002–2020. Gastroenterology. 2021;161:879-898. PMID 34126063
- Pinyol R, et al. Molecular predictors and prognostic factors in the phase 3 STORM adjuvant trial. Gut. 2019;68:1065-1075. PMID 30108162
- Lee JH, et al. Adjuvant immunotherapy with autologous cytokine-induced killer cells for HCC. Gastroenterology. 2015;148:1383-1391.e6. PMID 25747273
- Zhang W, et al. Lenvatinib plus anti-PD-1 antibodies as conversion therapy: phase 2 trial. J Immunother Cancer. 2023;11. PMID 37730273
- Kudo M, et al. TACE plus lenvatinib in intermediate HCC: TACTICS-L phase 2 trial. Liver Cancer. 2024;13:99-112. PMID 38344448
- Pinato DJ, et al. Pembrolizumab following TACE: PETAL phase 1b study. Clin Cancer Res. 2024;30:2433-2443. PMID 38578610
- Llovet JM, et al. Adjuvant and neoadjuvant immunotherapies in HCC. Nat Rev Clin Oncol. 2024;21:294-311. PMID 38424197
- Zhang S, et al. Spatial transcriptomics of neoadjuvant cabozantinib-nivolumab identifies resistance and recurrence mechanisms. Genome Med. 2023;15:72. PMID 37723590
- Yarchoan M, et al. Personalized neoantigen vaccine and pembrolizumab in advanced HCC: phase 1/2 trial. Nat Med. 2024;30:1044-1053. PMID 38584166
- Shi D, et al. Glypican-3 CAR-T therapy for advanced HCC: phase 1 trials. Clin Cancer Res. 2020;26:3979-3989. PMID 32371538
- Yopp A, et al. Updated data from IMbrave050: adjuvant atezolizumab plus bevacizumab for high-risk hepatocellular carcinoma. J Hepatol. 2026;84:1102-1111. PMID 41580093