Hepatocellular carcinoma (HCC) continues to pose a major global health burden and is among the most common malignancies and leading causes of cancer-related death worldwide^[1]. Although vaccination and antiviral treatment have reduced the incidence of hepatitis B virus (HBV)-related HCC in some regions^[2], the burden of HCC associated with metabolic dysfunction-associated steatotic liver disease (MASLD) is increasing^[3]. Many patients are diagnosed after the window for curative resection, or ablation has passed^[4], making locoregional and systemic therapies the therapeutic backbone for many patients.

For decades, transarterial chemoembolization (TACE) has been the standard option for intermediate-stage HCC^[5-7]. However, clinical outcomes after TACE are heterogeneous, and recurrence remains common. Mechanistically, TACE induces tumor ischemia and immunogenic cell death^[8], but may also trigger hypoxia-driven angiogenesis and reinforce an immunosuppressive tumor microenvironment^[9,10], thereby limiting durable disease control. Systemic treatments, including immune checkpoint inhibitors (ICIs) and targeted agents including anti-vascular endothelial growth factor (VEGF) antibody and tyrosine kinase inhibitors (TKIs) - can potentially counteract these adaptive changes after TACE, providing a strong biological rationale for combination strategies^[11]. Accordingly, multiple randomized trials have evaluated TACE in combination with systemic therapy across different sequencing paradigms and endpoint frameworks. Recent results from four key trials, including EMERALD-1^[12], LEAP-012^[13], TALENTACE^[14], and CAP-ACE^[15], have demonstrated significant benefits in progression-free survival (PFS) and objective response rate (ORR), accelerating guideline discussions and clinical adoption in selected settings^[16]. Nevertheless, clinicians now face a more complex decision landscape: overall survival (OS) remains immature or not yet definitively improved in several trials, and substantial heterogeneity persists regarding when to initiate systemic therapy relative to TACE, how to operationalize on-demand TACE in trial protocols, and which endpoints best reflect clinically meaningful benefit in the modern era of effective post-progression therapy.

In contrast to previous reviews that mainly summarized efficacy outcomes, this narrative review interprets available evidence from a trial-design perspective. Relevant literature was identified through searches of PubMed, Web of Science, Embase, and ClinicalTrials.gov up to 2025 using core terms: “HCC”, “TACE”, “systemic therapy”, and “combination therapy”. Randomized trials, landmark studies, and major clinical guidelines were prioritized. Trials were highlighted if they impacted clinical practice or illustrated methodological issues critical to interpreting TACE-based combination therapy. A preliminary version of this study has been published as a preprint^[17].

The development of TACE is rooted in the hemodynamic hallmark of HCC, in which tumor blood supply progressively shifts toward the hepatic artery, whereas the non-tumoral liver parenchyma remains predominantly portal venous^[18]. This physiological contrast enables selective intra-arterial drug delivery and embolization to achieve a dual effect - tumor starvation and cytotoxic exposure - while limiting collateral injury to surrounding liver tissue. Its origins can be traced to the first hepatic artery embolization performed by Goldstein in 1976^[19]. Subsequently, Nakakuma in the 1980s innovatively introduced gelatin sponge and iodized oil as a mitomycin carrier, establishing the classic model of conventional TACE (cTACE)^[20]. Two randomized trials published in 2002 subsequently demonstrated a survival advantage of TACE over supportive care and established its role as standard therapy^[21,22]. Since then, technical evolution has largely pursued three goals: higher precision, improved safety, and more consistent efficacy.

Systemic treatment for HCC has progressed from a period of limited benefit to a rapidly expanding therapeutic field. Table 1 summarizes pivotal phase III first-line systemic therapy trials in advanced HCC. Historically, advanced HCC relied primarily on chemotherapy and targeted agents. A pivotal SHARP trial in 2008 demonstrated that the multi-target TKI sorafenib significantly improved OS (10.7 months vs. 7.9 months; HR = 0.69) compared to placebo, establishing sorafenib as the sole first-line standard for years^[37]. Lenvatinib subsequently demonstrated non-inferiority in OS while improving progression-related endpoints and response rates, expanding first-line options^[38]. The Chinese-developed TKI donafenib also showed superior OS to sorafenib in the ZGDH3 trial (12.1 months vs. 10.3 months; HR = 0.83), with lower drug-related grade ≥ 3 adverse events (AEs). Sorafenib, lenvatinib, and donafenib are all oral multi-target TKIs, yet their mechanisms and clinical profiles differ. Sorafenib primarily inhibits VEGFR-2/3, platelet-derived growth factor receptor (PDGFR)-β, and (rapidly accelerated fibrosarcoma) RAF kinases, establishing its historical standard role by blocking angiogenesis and tumor proliferation. Lenvatinib demonstrates broader and more potent inhibition of the VEGFR family (1-3) and, most distinctively, strongly targets fibroblast growth factor receptors (FGFR) 1-4^[39]. This concurrent blockade of the FGFR pathway is the key mechanism behind its significantly higher ORR observed in the REFLECT trial. Donafenib is a deuterated derivative of sorafenib^[40], sharing a similar core target profile but with optimized pharmacokinetics, which explains the better balance between efficacy and safety.

The advance of ICIs, mainly including programmed cell death-(ligand) protein 1 [PD-(L)1] inhibitors and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), have fundamentally reshaped the treatment paradigm. Although initial attempts with PD-(L)1 inhibitors in the first-line setting (e.g., nivolumab in CheckMate 459^[41]) failed to meet their primary survival endpoints, combination strategies proved revolutionary. The updated landmark IMbrave150 trial established the combination of the PD-(L)1 inhibitor atezolizumab and the VEGF inhibitor bevacizumab as a new global standard, demonstrating significantly superior OS compared to sorafenib (19.2 months vs. 13.4 months; HR = 0.66)^[42]. The STRIDE regimen [CTLA-4 inhibitor tremelimumab plus PD-(L)1 inhibitor durvalumab] in the HIMALAYA trial also outperformed sorafenib in OS (16.4 months vs. 13.8 months; HR = 0.76)^[43]. The recently reported CheckMate 9DW trial^[44] demonstrated a significant OS improvement for the combination of nivolumab and ipilimumab vs. lenvatinib/sorafenib (23.7 months vs. 20.6 months; HR = 0.79). Together with the results of the HIMALAYA trial, these findings confirm the efficacy of the combination protocol of CTLA-4 inhibitor plus PD-(L)1 inhibitor.

Notably, China-developed combination regimens have also demonstrated equally remarkable efficacy. In the ORIENT-32 trial, sintilimab plus a bevacizumab biosimilar significantly improved OS and PFS compared with sorafenib^[45]. Subsequently, the RATIONALE-301 trial^[46], which compared PD-(L)1 inhibitor tislelizumab with sorafenib, reported a non-inferior OS outcome (15.9 months vs. 14.1 months; HR = 0.85), providing crucial evidence for global research on single-agent immunotherapy. Furthermore, in the CARES-310 trial, camrelizumab plus apatinib achieved a median OS of 23.8 months (vs. 15.2 months for sorafenib, HR = 0.64)^[47]. Recently, the SCT-I10A-C301 trial showed that finotonlimab plus a bevacizumab biosimilar significantly prolonged OS compared with sorafenib (22.1 months vs. 14.2 months; HR = 0.60)^[48]. The HEPATORCH trial demonstrated that toripalimab plus bevacizumab improved OS compared with sorafenib (20.0 months vs. 14.5 months; HR = 0.76)^[49]. Together, these studies have expanded the range of first-line systemic treatment options for HCC, particularly in China, and underscore the growing contribution of regionally developed regimens to the global evidence base.

Since the introduction of sorafenib into clinical practice, continuous efforts have been made to enhance the efficacy of TACE by combining with systemic therapy. The three combination strategies-sequential, interrupted, and continuous- were systematically proposed by Strebel in 2008^[50], established a crucial theoretical framework for subsequent clinical trial designs, as shown in Figure 1. Early randomized trials combining TACE with TKIs or antiangiogenic agents in intermediate-to-advanced HCC yielded largely negative results, whereas later studies demonstrated that trial success is highly sensitive to (i) the timing of systemic exposure around TACE; (ii) progression/response assessment criteria that govern TACE continuation; and (iii) whether TACE is delivered in a protocol-mandated “scheduled” manner or as response-guided “on-demand” treatment.

With immunotherapy plus anti-angiogenic therapy emerging as a dominant systemic backbone, triple therapy integrating TACE with ICI and anti-VEGF/TKI has become a logical next step. Real-world evidence and target-trial emulation studies have reported encouraging disease control^[58,59]. Meanwhile, several randomized trials have provided more rigorous signals of benefit, most consistently in PFS and ORR. However, OS remains immature or difficult to improve definitively in intermediate-stage populations where subsequent effective therapies and crossover can dilute survival differences. Below, we summarized and reflected pivotal randomized trials using a standardized evidence-focused template.

EMERALD-1^[12] was the first reported phase III trial evaluating TACE combined with immunotherapy and anti-angiogenic therapy. This multiregional, randomized, double-blind, placebo-controlled study compared TACE plus durvalumab and bevacizumab, TACE plus durvalumab, and TACE plus placebo in patients with unresectable HCC eligible for embolization [Table 2]. The enrolled population was broad and clinically heterogeneous, including patients with Barcelona Clinic Liver Cancer (BCLC) stage A-C disease; approximately 26% had stage A disease, about 98% had Child-Pugh class A liver function, and nearly half were beyond the up-to-seven criteria. This broad eligibility enhanced external validity but also complicated the interpretation of OS by introducing heterogeneity in tumor burden, disease stage, TACE suitability, and subsequent treatment pathways.

The primary endpoint was PFS assessed according to RECIST 1.1. TACE was performed according to local practice, with one to four procedures permitted and all planned sessions required to be completed within 16 weeks after the first procedure. Durvalumab was initiated at least 7 days after the first TACE session, whereas bevacizumab was delayed until at least 14 days after the final session. This design clearly prioritized peri-procedural safety, particularly by avoiding concurrent exposure to bevacizumab during embolization. Mechanistically, however, the potentially long interval between the first embolization and bevacizumab initiation may have reduced coverage of the early post-embolization window characterized by hypoxia and VEGF upregulation^[60,61].

The trial showed that triple therapy significantly prolonged PFS compared with TACE plus placebo (15.0 months vs. 8.2 months, HR = 0.77). In contrast, the TACE plus durvalumab group did not show PFS benefit (10.0 months vs. 8.2 months, HR = 0.94). The finding suggests that bevacizumab is not merely an additional component, but may be essential for effective synergy. Notably, curve separation appeared closer to the period after bevacizumab introduction rather than immediately after embolization, further supporting the contribution of anti-angiogenic therapy. Nevertheless, OS remains immature. Given its exploratory role in triple therapy development, the broad eligibility criteria of EMERALD-1 are understandable and provide useful subgroup signals for future trial design. Current PFS-based subgroup analyses suggest potentially greater benefit in patients with no vascular invasion, multifocal intrahepatic disease, and ECOG PS 0, however, these findings remain hypothesis-generating and require further validation. In addition, RECIST 1.1 may trigger early reassessment when new intrahepatic lesions appear, and treatment discontinuation due to adverse events occurred in approximately 28% of patients in the triple therapy group. Therefore, EMERALD-1 provides proof of concept that adding both durvalumab and bevacizumab can improve early disease control, but future trials must better optimize anti-angiogenic timing, response criteria, toxicity management, and post-progression treatment pathways.

LEAP-012^[13] was a global, randomized, double-blind phase III trial evaluating TACE plus lenvatinib and pembrolizumab vs. TACE plus dual placebo in unresectable, non-metastatic HCC. Unlike EMERALD-1, which enrolled a broader embolization-eligible population, LEAP-012 used stricter selection criteria: patients had Child-Pugh class A liver function, Eastern Cooperative Oncology Group performance status 0-1, liver-confined disease, and tumors technically suitable for embolization. Extreme tumor burden was excluded, including tumors ≥ 10 cm, more than 10 tumors, or tumor occupation of ≥ 50% of the liver. Therefore, LEAP-012 tested triple therapy in a more standardized, TACE-suitable population rather than in all embolization-eligible patients.

The design was also more synchronized with the biological window of TACE. Lenvatinib and pembrolizumab were started before embolization, with the first TACE performed 2-4 weeks after randomization. Lenvatinib was administered at 8 or 12 mg once daily according to body weight, and pembrolizumab was administered at 400 mg every 6 weeks. To improve safety, lenvatinib was interrupted around the time of TACE, and each tumor could receive no more than 2 TACE sessions. This approach reduced procedural heterogeneity and the risk of excessive embolization; however, the fixed per-tumor TACE cap may also have limited effective locoregional exposure in selected patients who could still have benefited from additional selective embolization.

LEAP-012 met its PFS endpoint, with a median of 14.6 months vs. 10.0 months and a hazard ratio of 0.66. The primary endpoint was RECIST 1.1 by blinded central review, modified to allow up to five intrahepatic target lesions and to require new intrahepatic tumors to meet LI-RADS 5 criteria before being considered progression. This design attempted to balance regulatory rigor with the clinical complexity of post-TACE intrahepatic changes. OS remained immature at the published interim analysis, with 24-month survival rates of 75% vs. 69% (HR = 0.80). Grade 3 or worse treatment-related adverse events were frequent, occurring in 71% vs. 32%, indicating that systemic-first triple therapy requires careful toxicity management.

More importantly, LEAP-012 provides useful clues for future population refinement. Subgroup results suggested stronger PFS benefit in patients with BCLC stage B disease, ECOG PS 0, ALBI grade 1, Child-Pugh score 5, alpha-fetoprotein > 400 ng/mL, and medium-to-high tumor burden. However, OS data remain immature, and these subgroup findings should be interpreted as exploratory. Thus, the major lesson of LEAP-012 is not simply that systemic-first triple therapy improves early disease control, but that future trials should prospectively enrich for patients with sufficient risk of TACE failure to justify systemic intensification, while still preserving enough liver function and performance status to tolerate sustained combination therapy.

TALENTACE was a phase III, open-label, randomized trial comparing on-demand TACE plus atezolizumab and bevacizumab with on-demand TACE alone in systemically untreated, unresectable HCC. Its key design feature was population enrichment. Eligible patients had intermediate-to-high tumor burden, defined as maximum tumor diameter plus tumor number ≥ 6; most had BCLC stage B disease, while selected BCLC stage C patients with limited portal vein invasion, mainly Vp1-2, or ECOG PS 1 were also allowed. Thus, TALENTACE focused on a clinically important boundary population: patients who are still technically suitable for TACE, but at sufficient risk of TACE failure to justify systemic intensification.

In the experimental arm, atezolizumab 1,200 mg plus bevacizumab 15 mg/kg every 3 weeks was initiated at least 14 days and within 8 weeks after the first TACE. This TACE-first strategy differs from the systemic-first design of LEAP-012 and preserves real-world retreatment flexibility by avoiding a fixed per-tumor TACE cap. The co-primary endpoints were TACE-PFS and OS. TACE-PFS was defined as time to untreatable progression, TACE failure/refractoriness, or death, making it closer to a treatment-strategy failure endpoint than conventional RECIST progression alone.

At the first interim analysis, TALENTACE met its TACE-PFS endpoint, with a median of 11.30 months vs. 7.03 months (HR = 0.71). RECIST 1.1 PFS also favored the combination arm, 10.32 months vs. 6.37 months (HR = 0.64). However, OS remained immature, with only 38.6% of events reported; median OS was 34.53 months vs. 35.38 months (HR = 0.96). The apparently neutral early survival result may reflect several factors: immature follow-up, unexpectedly long survival in the control arm, subsequent treatment after progression, and the possibility that improved TACE-PFS mainly delays treatment-strategy failure rather than immediately translating into death reduction.

The current subgroup data should be interpreted cautiously. Publicly available reports suggest generally consistent TACE-PFS benefit across predefined subgroups, but detailed subgroup hazard ratios have not yet been fully published. Therefore, TALENTACE should not yet be used to define a definitive biomarker-like beneficiary subgroup. Its deeper implication is conceptual: future trials may need to enrich for patients with both high risk of inadequate TACE control and sufficient liver and performance reserve to tolerate sustained systemic therapy. Toxicity remains a key constraint, as grade 3-4 treatment-related adverse events occurred in 60.8% vs. 40.5%, and treatment withdrawal due to adverse events occurred in 21.1% vs. 2.3%. Thus, TALENTACE represents a shift from asking whether triple therapy works in all embolization-eligible patients to asking whether it works in a risk-enriched population using an endpoint that captures TACE-strategy failure rather than radiographic progression alone.

The CAP-ACE, also reported as CARES-005, was a multicenter, open-label, randomized phase II trial conducted in China that compared TACE plus camrelizumab and rivoceranib/apatinib with TACE alone in patients with unresectable HCC eligible for embolization. Eligible patients had no extrahepatic metastasis, Child-Pugh class A liver function, and ECOG PS 0-1. Most patients had BCLC stage B or C disease.

A key feature of CAP-ACE was its crossover design. Patients initially assigned to TACE alone were allowed to cross over to receive camrelizumab plus rivoceranib/apatinib after progression. This design was clinically and ethically reasonable and also resembles real-world practice, in which patients are often switched to combined systemic therapy after TACE failure. However, this inevitably reduced the ability of the trial to detect an OS difference because the control group could later receive the active systemic regimen. The primary endpoint was RECICL-assessed PFS, which is particularly relevant in TACE-based trials because it better captures viable tumor response after embolization than size-based criteria alone. CAP-ACE showed a strong PFS advantage: median PFS was 10.8 months vs. 3.2 months (HR = 0.34), and time to unTACEable progression was also prolonged, 13.7 months vs. 3.9 months, suggesting that early systemic intensification can delay the point at which a TACE-based strategy becomes ineffective. Subgroup analyses showed RECICL-PFS benefit in both BCLC A/B and BCLC C disease, with hazard ratios of 0.34 and 0.39, respectively; however, OS benefit was not confirmed. Safety is another important limitation. Grade 3 or higher treatment-related adverse events were more frequent with the combination, including aspartate aminotransferase elevation, alanine aminotransferase elevation, hypertension, and platelet count decrease. These toxicities were described as manageable, but they highlight the recurring challenge in TACE-based triple therapy: stronger early disease control may come at the cost of greater hepatic and systemic toxicity.

Overall, CAP-ACE reinforces two important concepts. First, RECICL-based PFS and time to unTACEable progression may better capture the benefit of delaying TACE-strategy failure than conventional radiographic endpoints alone. Second, when crossover is allowed, OS becomes strongly influenced by post-progression treatment pathways and may underestimate the value of early combination therapy. Thus, CAP-ACE provides strong randomized evidence that early TACE plus camrelizumab and rivoceranib/apatinib improves disease control and delays unTACEable progression, but it should not be overinterpreted as definitive proof of OS benefit.

The central challenge for future TACE-based combination trials is no longer whether systemic therapy can enhance early tumor control, but how to design treatment strategies that preserve enough effective TACE and systemic exposure to convert early disease control into durable survival benefit. Biologically, this PFS-OS dissociation may reflect the fact that triple therapy primarily enhances early intrahepatic tumor control rather than fully modifying the later determinants of survival in HCC. TACE can reduce viable tumor burden and promote antigen release, while anti-VEGF/TKI therapy may suppress post-embolization angiogenic rebound and render the tumor microenvironment more permissive to ICI activity^[8-10]. These mechanisms are well suited to improve radiographic response and delay progression. However, OS is additionally influenced by hepatic functional reserve, cirrhosis-related competing mortality, cumulative liver injury after repeated TACE, treatment-related toxicities and interruptions, and effective post-progression therapies or crossover in control arms. Therefore, PFS may serve as a more sensitive signal of biological synergy, whereas OS requires sustained tumor control together with preservation of liver function. This discrepancy underscores several critical areas that require attention in future research.

First, the optimal timing and sequencing of TACE combined with systemic therapies must be directly compared. Current studies utilize heterogeneous strategies, including systemic-first and TACE-first, without definitive comparative evidence. Prospective trials specifically designed to test the impact of different sequencing approaches - while controlling for TACE technique, retreatment protocols, and continuity of systemic therapies - are essential to determine whether sequencing significantly influences synergy and long-term outcomes.

Second, the endpoints used in these trials need to be modernized to better capture clinically meaningful benefits. In settings where active subsequent therapies and crossover are common, OS may be diluted, making it statistically challenging to measure. Therefore, composite endpoints incorporating time to untreatable progression, liver function preservation, conversion to curative therapy rates, and patient-reported outcomes should be considered, as they more accurately reflect the true clinical benefit. When using TACE-specific endpoints, simplicity and clinical interpretability should be prioritized to facilitate broad adoption and reproducibility.

Third, TACE operational standardization should align more closely with real-world practice. A persistent gap exists between the on-demand TACE approach used in clinical practice and the protocol-driven session limits seen in trials. Future studies should incorporate adaptive, response-guided TACE algorithms, considering imaging response, liver reserve, and performance status, rather than relying on fixed session caps that may under-treat or over-treat certain patient subgroups.

Future progress hinges on large-scale, intelligently designed trials that simultaneously address these intertwined questions, including timing, endpoint selection, and TACE standardization. The goal is to move beyond demonstrating PFS superiority and to reliably deliver the OS benefit that defines meaningful therapeutic advancement for patients with intermediate-to-advanced HCC.