Comparative efficacy of combination regimens based on interventional therapy and immune checkpoint inhibitors (ICIs) in patients with intermediate- and advanced-stage hepatocellular carcinoma: a systematic review, meta-analysis, and network meta-analysis.

Introduction
Hepatocellular carcinoma (HCC) ranks sixth in global cancer incidence and is the third leading cause of cancer-related mortality [1]. Early-stage HCC is usually asymptomatic, which often leads to diagnosis at intermediate or advanced stages and thereby reduces the chances of curative treatment [2]. Currently, the main therapeutic approaches for intermediate- and advanced-stage HCC include transarterial interventional therapies and immune checkpoint inhibitor (ICI)-based regimens. Both modalities, when used alone or in combination, have shown promising outcomes in recent clinical studies.
Transarterial interventional therapy constitutes a cornerstone of treatment for intermediate- and advanced-stage HCC and includes procedures such as transcatheter arterial chemoembolization (TACE), transarterial embolization (TAE), transarterial radioembolization (TARE), and hepatic arterial infusion chemotherapy (HAIC) [3]. These approaches exploit the liver’s dual blood supply: normal hepatic tissue primarily receives blood from the portal vein, whereas HCC lesions are mainly supplied by the hepatic artery. This difference enables selective arterial interventions that can effectively target tumor tissue while sparing normal parenchyma. The efficacy of these treatments has been demonstrated in multiple studies [4, 5]. Meanwhile, the introduction of ICIs has substantially expanded the treatment landscape for advanced HCC. Several pivotal clinical trials have confirmed the efficacy of ICI monotherapy and combination regimens. For example, the CheckMate 9DW trial showed that nivolumab plus ipilimumab as a first-line therapy improved 18-month (34% vs. 18%) and 24-month (28% vs. 12%) progression-free survival (PFS) rates compared with lenvatinib or sorafenib (LEN/SOR). The study also reported a longer PFS2 (median progression-free survival on next-line therapy) for the combination regimen (19.3 vs. 15.4 months; hazard ratio [HR] = 0.70, 95% confidence interval [CI]: 0.58–0.84) [6].
Building on the local control advantage of interventional therapy and the systemic immune activation induced by ICIs, the exploration of combination and multi-drug regimens has naturally emerged as a key research focus. Among dual-combination strategies, TACE plus ICI has been widely investigated and has shown encouraging efficacy and acceptable safety profiles. For instance, the phase II PETAL trial reported a 12-week progression-free survival rate of 93% and a median PFS of 8.95 months (95% CI, 7.30-not reached) for TACE combined with pembrolizumab, with manageable toxicity [7]. Regarding triple-combination approaches, regimens combining TACE with tyrosine-kinase inhibitors (TKIs) or anti-angiogenic antibodies and ICIs [TACE-TKI-ICI/anti-VEGF–ICI] are gaining attention. The phase III EMERALD-1 study (TACE + durvalumab + bevacizumab) demonstrated a significant improvement in PFS compared with TACE alone (HR = 0.77, 95% CI: 0.61–0.98) with a tolerable safety profile [8]. Similarly, the LEAP-012 study showed a significant PFS benefit with TACE + lenvatinib + pembrolizumab compared with TACE alone (HR = 0.66, 95% CI: 0.51–0.84) [9], suggesting a potential new treatment option for intermediate- and advanced-stage HCC.
Despite the growing body of research on combination therapies, several key knowledge gaps remain in the current evidence base. First, there is a lack of systematic evaluation of novel ICIs such as tislelizumab in interventional combination settings, and comparative evidence among different ICIs remains limited. Second, although evidence for TACE-TKI-ICI is increasing, integrated analyses comparing various triple-combination strategies (e.g., different TKI and ICI pairings) are still lacking, and data on HAIC-based triple regimens are scarce. Third, evidence guiding the selection between dual and triple combinations, as well as identifying suitable patient subgroups, is still insufficient, thereby limiting the evidence base for clinical decision-making.
Therefore, this study aims to systematically evaluate the efficacy and safety of combination regimens based on interventional therapy and ICIs in patients with intermediate- and advanced-stage HCC. We further compare the clinical performance of various ICI types and explore the value of emerging triple-combination regimens. By addressing these evidence gaps, this study seeks to provide integrated evidence to inform and refine clinical decision-making for the management of intermediate- and advanced-stage HCC.

Methods
This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [10] (Table S1). To ensure transparency and methodological rigor, the study protocol was prospectively registered with PROSPERO (CRD42024619661).

Literature search and study selection
Studies were considered eligible if they fulfilled the following inclusion criteria: (1) enrolled patients with a pathological or clinical diagnosis of intermediate or advanced-stage HCC; (2) the interventions included at least one study arm with a combination regimen based on interventional therapy and ICIs; (3) the studies included at least one of the following outcomes of interest: PFS, OS, objective remission rate (ORR), or grade ≥ 3 adverse events (AEs); (4) published phase I, II, or III clinical trials or cohort studies. Exclusion criteria comprised: (1) studies involving patients with secondary HCC; (2) clinical studies involving multiple immunological agents and different interventional techniques (e.g., HAIC for some patients and TACE for others); and (3) incomplete or unavailable data on the target outcome.
We conducted a comprehensive search across PubMed, Embase, Cochrane, and Web of Science from database inception through June 2025. The literature search utilized keywords such as “intermediate and advanced-stage hepatocellular carcinoma,” “interventional therapy,” and “immune checkpoint inhibitors” (complete search strategy available in Table S2). Two authors (JTS and YJS) independently screened titles and abstracts and evaluated full-text articles for eligibility, with discrepancies adjudicated by a third author (WJC).

Data extraction and risk of bias assessment
The primary outcome was PFS, while secondary outcomes included OS, ORR, and the rate of grade ≥ 3 AEs. Relevant data were independently extracted from the included studies by two authors (JTS and YJS) and recorded in a pre-designed Excel sheet. The collected data included baseline characteristics, sample size, treatment modality and dose, OS, PFS, ORR, the rate of grade ≥ 3 Aes, and study design.
We used three tools to assess the methodological quality and potential biases in the included studies: the Cochrane risk of bias tool for randomized controlled trials, the methodological index for non-randomized studies (MINORS) for non-randomized single-arm clinical trials, and the Newcastle–Ottawa Quality Assessment Scale (NOS) for cohort studies. The assessments were independently conducted by two authors (JTS and YJS). All disagreements were resolved through discussion and consensus with a third author (WJC).

Data analysis
Statistical analyses were performed in R software (v4.4.1), using the meta package for single-arm meta-analyses (metaprop and metamean functions) and the gemtc and JAGS packages for NMA. For the meta-analysis of survival data, hazard ratios were preferred as the outcome measure. If studies did not report HRs, data were extracted and estimated using methods described by Popat et al. and Jayne et al. [11, 12].
We performed a single-arm meta-analysis to derive pooled estimates of median PFS and OS, and to calculate the pooled rates of objective response, and grade ≥ 3 AEs for each treatment regimen. A network meta-analysis (NMA) was conducted using a Bayesian framework with a random-effects consistency model to estimate relative treatment effects. The analysis employed Markov chain Monte Carlo (MCMC) methods with four independent chains using non-informative priors and normal distributions. The number of simulations was set to 20,000, and the number of iterations was set to 50,000 to obtain the posterior distributions. Trajectory plots and Brooks–Gelman–Rubin diagnostic plots were employed to visually assess the level of model convergence. The effect sizes were expressed in terms of hazard ratios or relative risks, along with their respective 95% confidence intervals. Treatment rankings were determined by the probability of superiority and summarized using the surface under the cumulative ranking (SUCRA) values, ranging from 0 (least effective or least toxic treatment) to 1 (the highest efficacy or toxicity). Consistency for direct and indirect evidence was evaluated through node-splitting test. Study heterogeneity was assessed using the Q test and quantified with the I2 statistic.

Results
From an initial pool of 3917 identified records, title and abstract screening excluded 3813 irrelevant studies. Full-text review of 104 articles yielded 45 eligible studies [8, 9, 13–55]. The final analysis included 2 randomized controlled trials, 28 retrospective cohort studies, and 15 single-arm non-randomized trials (n = 4738) evaluating 14 therapeutic regimens (Fig. S10): HAIC plus TKI plus camrelizumab (HAIC-TKI-Camre), TACE plus bevacizumab plus sintilimab (TACE-Bev-Sinti), TACE plus bevacizumab plus atezolizumab (TACE-Bev-Atez), TACE plus durvalumab (TACE-Durva), TACE plus bevacizumab plus durvalumab (TACE-Bev-Durva), TACE plus camrelizumab (TACE-Camre), TACE plus TKI plus camrelizumab (TACE-TKI-Camre), TACE plus TKI plus pembrolizumab (TACE-TKI-Pembro), TACE plus TKI plus sintilimab (TACE-TKI-Sinti), TACE plus TKI plus tislelizumab (TACE-TKI-Tisle), TACE plus TKI plus toripalimab (TACE-TKI-Tori), HAIC plus TKI plus toripalimab (HAIC-TKI-Tori), HAIC plus TKI plus tislelizumab (HAIC-TKI-Tisle), and TAE plus HAIC plus TKI plus tislelizumab (TAE-HAIC-TKI-Tisle). Most studies enrolled patients with Barcelona Clinic Liver Cancer (BCLC) stage B-C and Child–Pugh class A-B cirrhosis. Comprehensive baseline characteristics are provided in Table S3. All included studies were deemed to have a low to moderate risk of bias, supported by clearly defined criteria, explicit objectives, sufficient follow-up, and comprehensive outcome evaluation (Fig.S1, S2, Table S4).
The single-arm meta-analysis, 31 studies (2921 participants) [8, 9, 16–27, 30–32, 34, 35, 40, 41, 43–47, 49–51, 54, 55] for mPFS, 21 studies (1724 participants) [18, 20–25, 27, 32–35, 40, 41, 44, 47, 49–51, 53, 55] for mOS, 45 studies (3618 participants) [8, 9, 13–55] for ORR, and 24 studies (2384 participants) [8, 9, 13, 14, 16–20, 22–24, 26–30, 32, 34, 35, 39, 43, 48, 54] for grade ≥ 3 AEs rates were included. Among all combination regimens based on interventional therapy and ICIs, the pooled mPFS was 9.56 months (95% CI: 8.61–10.51; I2 = 91.5%), the pooled mOS was 18.20 months (95% CI: 16.27–20.13; I2 = 96.7%), the pooled ORR was 59% (95% CI: 55–64%; I2 = 85.5%), and the pooled incidence of grade ≥ 3 AEs was 37% (95% CI: 29–45%; I2 = 91.5%). The descriptive pooled results for specific regimens revealed distinct profiles. For TACE-TKI-Tisle, the pooled mPFS was 11.7 months (95% CI: 8.02–15.37; I2 = 67.6%), the pooled ORR was 72% (95% CI: 63–80%; I2 = 49.7%), and the pooled incidence of grade ≥ 3 AEs was 24% (95% CI: 15–34%; I2 = 65.7%). The pooled ORR for TACE-TKI-Tori was 69% (95% CI: 60–77%; I2 = 1.8%). For TACE-TKI-Camre, which had relatively comprehensive data, the pooled mPFS was 9.90 months (95% CI: 8.56–11.23; I2 = 85.8%), the pooled mOS was 19.61 months (95% CI: 16.82–22.41; I2 = 91.5%), the pooled ORR was 56% (95% CI: 47–64%; I2 = 86.4%), and the pooled incidence of grade ≥ 3 AEs was 35% (95% CI: 22–50%; I2 = 91.4%). Additionally, for HAIC-TKI-Camre, the pooled incidence of grade ≥ 3 AEs was 78% (95% CI: 71–84%; I2 = 0.0%) and the pooled mPFS was 8.50 months (95% CI: 5.10–11.90; I2 = 80.7%). These pooled findings from single-arm studies serve as preliminary references for considering the efficacy and safety of different combination strategies (Fig. 1, Fig.S3).
The network meta-analysis incorporated 27 studies (n = 3,695) [8, 9, 31–55]. Constraints due to insufficient data limited comparisons to 10 interventions for PFS [8, 9, 32–51, 54, 55] and ORR [8, 9, 32–45, 47–55], 8 for OS [9, 32–51, 54, 55], and 6 for grade ≥ 3 AEs incidence [8, 9, 34, 35, 40, 43, 49, 54] (Fig. S11). TACE-TKI-Tisle and TACE-TKI-Camre demonstrated superior progression-free survival compared to TACE-TKI and TACE alone. TACE-TKI-Tori showed overall survival benefits over TACE-TKI-Camre (HR = 0.43; 95% CI 0.20–0.95) and TACE-TKI-Pembro (HR = 0.32; 95% CI 0.13–0.81). In terms of ORR, TACE-TKI-Tisle and TACE-TKI-Tori showed higher effect estimates compared with Bev-Atez, TKI-Camre, and TACE alone. Safety assessment of grade ≥ 3 AEs revealed 5 interventional comparisons without statistically significant differences (Fig. 2).
SUCRA analysis suggested that TACE-TKI-ICI generally obtained higher efficacy rankings than other strategies. For PFS, TACE-TKI-Tisle and TACE-TKI-Tori had the highest SUCRA values (0.896 and 0.831, respectively). Bayesian ranking probability analysis showed that TACE-TKI-Tisle had a 32% probability of being the best regimen and a 31% probability of being the second-best, whereas TACE-TKI-Tori had a 35% probability of being ranked first and a 19% probability of being ranked second. Regarding OS, TACE-TKI-Tori had the highest rank (SUCRA = 0.981) and an 85% probability of being the top-performing regimen. For ORR, both TACE-TKI-Tisle and TACE-TKI-Tori also had high SUCRA values (0.892 and 0.871, respectively). In contrast, regimens combining TKIs with ICIs without interventional therapy (e.g., TKI-Sinti and TKI-Camre) were associated with lower SUCRA values across all efficacy outcomes, with ranking probabilities suggesting they were more likely to be ranked lower. Safety analysis showed a different pattern. TACE-TKI-Pembro (SUCRA = 0.694) and TACE-TKI-Camre (SUCRA = 0.760) had the highest SUCRA values for adverse events, indicating a potentially higher risk of serious toxicity. Notably, TACE-TKI-Tisle, which showed favorable efficacy rankings, was in the middle range for safety (SUCRA = 0.426) (Fig. S12, Table S5).
In the subgroup network meta-analysis, most pairwise comparisons did not reach statistical significance; however, trends favoring the triple combination of TACE-TKI-ICI over dual or monotherapy approaches were observed in several subgroups, including those defined by Child–Pugh A, HBV positivity, ≤ 3 Tumors, prior treatment status, and TACE type. Notably, in treatment-naïve patients, TACE-TKI-Tori was associated with a greater survival benefit (SUCRA = 0.981) than TACE-TKI-Camre (HR = 2.28, 95% CI: 1.14–4.89) and TACE-Bev-Atez (HR = 2.96, 95% CI: 1.22–7.63), suggesting its potential suitability for the first-line setting. The SUCRA ranking analysis offered a comparative perspective on the relative efficacy of different strategies. Among the comparable regimens, TACE-TKI-Tisle and TACE-TKI-Tori tended to rank higher for PFS and OS across most subgroups. However, some subgroup analyses suggested potential variations in treatment rankings. For PFS, TACE-TKI-Sinti ranked higher than TACE-TKI-Tisle in HBV-negative and Child–Pugh B populations, whereas TACE-TKI-Tori was ranked higher than TACE-TKI-Tisle in treatment-naïve patients. For OS, TACE-TKI-Sinti showed a superior ranking to TACE-TKI-Tisle in patients with ECOG 0, while TACE-TKI-Camre demonstrated a better ranking in those with metastatic disease or a history of prior treatment (Fig.S5–S8).

Discussion
In recent years, treatment strategies for intermediate- and advanced-stage HCC have evolved substantially. The integration of interventional therapies and ICIs has emerged as a major research focus, with multiple pivotal studies confirming the clinical benefits of this combination approach [8, 9]. Collectively, these findings support the concept of comprehensive multimodal treatment for intermediate and advanced-stage HCC and suggest new avenues for optimizing therapeutic strategies in clinical practice.
In this meta-analysis, we performed a multidimensional evaluation of published regimens combining interventional therapies with ICIs. Overall, the TACE-TKI-ICI regimen showed better performance across multiple efficacy endpoints. Among these, TACE-TKI-Tisle showed robust and consistent efficacy data. Meanwhile, TACE-TKI-Tori not only exhibited a stronger OS signal but also achieved high SUCRA rankings for PFS and ORR. The enhanced efficacy of triple-agent regimens likely results from synergistic mechanisms. Previous studies indicate that TACE is associated with lower intra-tumor depletion of effector (CD8+/PD-1+) and regulatory (CD4+FOXP3+) T cells, which may convert the tumor microenvironment from immunosuppressive to immunopermissive, thereby enhancing the response to immunotherapies [56]. TKIs, through their anti-proliferative and anti-angiogenic effects, mitigate the hypoxia-driven neovascularization induced by TACE or HAIC, further limiting HCC progression [57]. Additionally, molecular distinctions among ICIs may also contribute to efficacy differences. Tislelizumab binds the CC′ loop of PD-1, exhibiting slower dissociation and higher affinity, while toripalimab targets the FG loop, blocking classical PD-1 signaling and enhancing memory CD8+ T-cell responses and cytotoxic activity [58, 59]. These structural and functional variations may underlie the distinct efficacy profiles observed across populations and tumor microenvironments, providing a biologically plausible rationale for the outcomes of TACE-TKI-ICI strategies.
While combination therapy offers promising efficacy, its safety profile requires careful consideration. The single-arm meta-analysis revealed a pooled incidence of grade ≥ 3 AEs of 37% (95% CI: 29–45%) for combination regimens based on interventional therapy and ICIs. The network meta-analysis further confirmed, based on SUCRA rankings, that combination regimens generally carry a higher risk of severe AEs. Notably, TACE-TKI-Tisle achieved a favorable balance between efficacy and safety, whereas safety data for TACE-TKI-Tori remain limited. These findings highlight the importance of individualized risk–benefit assessment and toxicity monitoring when selecting combination regimens for clinical use.
Subgroup analyses further revealed potential efficacy heterogeneity. Across most subgroups, TACE-TKI-ICI consistently showed better efficacy; however, efficacy rankings varied in specific populations. For example, in HBV-negative or Child–Pugh B patients, TACE-TKI-Sinti ranked higher than TACE-TKI-Tisle, suggesting that liver function and viral infection status may modulate ICI responsiveness. Previous evidence indicates that HBV infection can reshape the immune microenvironment and influence immunotherapy outcomes [60], while patients with Child–Pugh B cirrhosis experience reduced immunotherapy efficacy and increased mortality risk [61, 62]. These findings underscore the impact of hepatic reserve, viral etiology, and prior treatment on treatment efficacy. These factors not only reflect the impact of patient conditions on immune modulation but may also represent a potential source of heterogeneity in the pooled results. Clinically, this highlights the need for personalized therapy based on functional status and disease stage. Identifying optimal regimens for specific subgroups through stratified analyses will be key to advancing precision medicine in HCC.
This study has several limitations. First, most included studies were retrospective cohort studies or non-randomized single-arm trials. Although approximately half of the retrospective cohorts used propensity score matching (PSM) to reduce intergroup differences, and all single-arm meta-analyses included prospective studies, inherent heterogeneity could not be completely eliminated. Second, for some treatment regimens, the number of eligible studies and sample sizes remained limited. To enhance robustness, we conducted both single-arm and network meta-analyses, which showed generally consistent results. Third, limited data on TKI subtype and hepatitis C virus (HCV) infection precluded further subgroup analysis, and results pertaining to these factors should be interpreted cautiously. Fourth, discrepancies between PFS and OS were observed, which is a common finding in oncology meta-analyses. These differences may be attributable to variations in subsequent therapies (summarized in Table S8), limited OS follow-up, or differences in the mechanisms of action among various ICIs. Incomplete reporting of subsequent treatments and survival outcomes in some studies further warrants cautious interpretation of OS results. Considering these limitations, PFS was adopted as the primary endpoint, as it is less influenced by post-progression therapies and better reflects direct treatment efficacy.
In conclusion, this study provides a comprehensive assessment of the efficacy and safety of combination regimens based on interventional therapy and ICIs in intermediate- and advanced-stage HCC. The TACE-TKI-ICI regimen demonstrated promising efficacy, with TACE-TKI-Tisle showing favorable outcomes across multiple endpoints and an acceptable safety profile. TACE-TKI-Tori exhibited strong OS benefits but had limited safety data. Given the uncertainty in relative efficacy and generalizability, these findings are exploratory but provide useful guidance for future research. Prospective randomized trials are warranted to confirm optimal regimens and define patient populations most likely to benefit.

Supplementary Information
Below is the link to the electronic supplementary material.