Transarterial chemoembolization (TACE) combined with immunotherapy and anti-angiogenic agents in intermediate-stage hepatocellular carcinoma (CHANCE2202): a target trial emulation study

Summary

Background

This study aimed to assess the clinical outcomes of transarterial chemoembolization (TACE) with immune checkpoint inhibitors (ICIs) plus vascular endothelial growth factor (VEGF) inhibitors or tyrosine kinase inhibitors (TKIs) (combination therapy) versus TACE monotherapy as a first-line treatment for intermediate-stage hepatocellular carcinoma (HCC).

Methods

This nationwide, retrospective cohort study employed a target trial emulation framework with a cloning-censoring-weighting approach. Patients with intermediate-stage HCC receiving either combination therapy or TACE monotherapy between January 2018 and December 2022 in China were included. Co-primary outcomes were overall survival (OS) and progression-free survival (PFS) per modified Response Evaluation Criteria in Solid Tumors (mRECIST), assessed using restricted mean survival time (RMST). Hazard ratio (HR) was additionally estimated using Cox proportional hazards models for reference. For both RMST and HR, 95% CIs were obtained by bootstrapping. Secondary outcomes included PFS per RECIST 1.1, objective response rate (ORR) per both mRECIST and RECIST 1.1, and safety. This study is registered at ClinicalTrials.gov (NCT05332496).

Findings

A total of 941 patients were included in the study, with 308 (32.7%) receiving combination therapy, and 633 (67.3%) receiving TACE monotherapy. Median OS was 32.9 with combination therapy versus 23.0 months with TACE monotherapy, with an RMST difference of 9.2 months (95% CI 4.5–14.3, bootstrapped p < 0.001; HR 0.57 [95% CI 0.43–0.70]). Median PFS was 18.0 and 12.9 months in the respective groups, with an RMST difference of 6.7 months (95% CI 3.3–10.7, bootstrapped p = 0.001; HR 0.70 [95% CI 0.58–0.82]). Combination therapy also yielded a higher ORR per mRECIST (60.5% versus 44.3%; p < 0.001). Similar results for PFS and ORR were observed when assessed using RECIST 1.1. Grade ≥3 adverse events occurred in 64 (20.8%) and 43 (6.8%) patients, respectively.

Interpretation

Combining TACE with ICIs and VEGF inhibitors or TKIs was associated with improved OS and PFS than TACE monotherapy, with an acceptable safety profile, supporting its potential as a first-line treatment strategy for intermediate-stage HCC.

Funding

National Natural Science Foundation of China (82130060, 82502493), China Postdoctoral Science Foundation (2025M772071), Jiangsu Provincial Basic Research ProgramNatural Science Foundation-Frontier Leading Technology Basic Research Project (BK20232008), Jiangsu Provincial Medical Innovation Center (CXZX202219), Postdoctoral Fellowship Program of CPSF (GZC20251385), and Natural Science Foundation of Jiangsu Province (BK20251687).

Research in context.

Evidence before this study

Transarterial chemoembolization (TACE) is the first-line recommended treatment for intermediate-stage hepatocellular carcinoma (HCC). TACE combined with immune checkpoint inhibitors (ICIs) plus vascular endothelial growth factor (VEGF) inhibitors or tyrosine kinase inhibitors (TKIs) is expected to be synergistic. We searched PubMed from database inception to October 12, 2025, for relevant articles using the following search terms: “hepatocellular carcinoma” AND “transarterial chemoembolization” AND “immune checkpoint inhibitor” AND (“tyrosine kinase inhibitor” OR “anti-VEGF”), without language restrictions. Several randomized trials have been published. However, no prospective multicenter trial has yet reported overall survival in intermediate-stage disease. In view of this gap, we report findings from a multicenter cohort study using a target trial emulation design.

Added value of this study

Within target trial emulation framework, this nationwide cohort study provides real-world evidence on first-line TACE plus ICIs with VEGF inhibitors or TKIs in intermediate-stage HCC, a setting without prospective multicenter OS reporting to date. The combination was associated with longer survival and higher response than TACE alone, with a manageable safety profile. Findings were consistent across clinically relevant subgroups and remained robust in multiple sensitivity analyses.

Implications of all the available evidence

Phase 3 trials have shown PFS benefit when TACE is combined with ICIs plus VEGF inhibitors or TKIs, but OS data remain pending. Although this nationwide real-world study is the first to report OS benefit of TACE combined with immunotherapy-based regimens in intermediate-stage patients, the evidence is still insufficient to change current clinical practice. Prospective trials are required to confirm the OS benefit.

Introduction

Hepatocellular carcinoma (HCC) is one of the most prevalent cancers and a leading cause of cancer-related mortality worldwide.¹ According to the Barcelona Clinic Liver Cancer (BCLC) guidelines, transarterial chemoembolization (TACE) is the first-line recommended treatment for intermediate-stage HCC, and systemic therapy is the standard of care for advanced HCC, which is also advised for intermediate-stage HCC with diffuse, infiltrative, or extensive liver involvement.²

Systemic treatment with immunotherapy-based combinations has dramatically transformed the treatment landscape of HCC. Following the success of a series of phase 3 randomized controlled trials (RCTs),3, 4, 5, 6, 7, 8, 9 a range of immune checkpoint inhibitor (ICI)-based strategies, including combinations with vascular endothelial growth factor (VEGF) inhibitors or tyrosine kinase inhibitors (TKIs), and dual ICI regimens have been established as the standard of care for advanced HCC.² TACE induces necrosis of tumor tissue and releases tumor antigens, potentially enhancing tumor-specific immune responses. VEGF inhibitors or TKIs can inhibit TACE-induced angiogenesis and further restore anti-tumor activity.10, 11, 12 Given limited survival benefits of TACE monotherapy and the strong biological rationale for combination therapy, integrating TACE with ICI-based systemic therapies has emerged as a promising approach for the treatment of HCC.^12,13

Emerging evidence indicates that TACE combined with ICIs plus VEGF inhibitors or TKIs significantly prolongs PFS compared to TACE alone in patients with unresectable HCC. Recently, the EMERALD-1, LEAP-012, CARES-005, and TALENTACE trials reported that TACE with durvalumab plus bevacizumab, TACE with pembrolizumab plus lenvatinib, TACE with camrelizumab plus apatinib, or TACE with atezolizumab plus bevacizumab, respectively, led to superior PFS compared to TACE alone in unresectable HCC.14, 15, 16, 17 However, the OS data from these trials remain immature due to insufficient follow-up duration. Thus, it remains unclear whether the addition of ICI plus targeted therapy to TACE can provide the OS benefit in patients with intermediate-stage HCC.

To address the common methodological biases inherent in real-world observational studies, target trial emulation has been proposed as a rigorous analytical framework that applies key RCT design principles to observational data.^18,19 This approach mimics randomization and preserves the intention-to-treat principle, while reducing selection bias and immortal-time bias. It has been increasingly adopted in comparative effectiveness studies using real-world data.^20,21 Therefore, using a target trial emulation approach, we conducted this nationwide, retrospective cohort study (CHANCE2202) to identify efficacy and safety, with a particular focus on OS data, of TACE combined with ICIs and VEGF inhibitors or TKIs versus TACE monotherapy as first-line treatments for intermediate-stage HCC.

Methods

This multicenter retrospective cohort study was designed to emulate a target trial of first-line TACE with ICIs and VEGF inhibitors or TKI (combination therapy) as compared with TACE monotherapy for the treatment outcomes of intermediate-stage HCC using nationwide observational data in China. The key components of the protocol of a hypothetical RCT and the target trial emulation are summarized in Table S1. This study was conducted in accordance with the Declaration of Helsinki, registered at ClinicalTrials.gov (NCT05332496), and reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement and the Transparent Reporting of Observational Studies Emulating a Target Trial (TARGET) statement.¹⁹

Ethics statement

This study was approved by the ethics committee of Zhongda Hospital, Southeast University (2022ZDSYLL069-P01). The study protocol received approval from the institutional review boards of participating centers. Due to the retrospective nature of the study, written informed consent was waived.

Patients

Patients with intermediate-stage HCC who received TACE with or without ICIs plus VEGF inhibitors or TKIs as first-line treatment at 62 hospitals in China from January 2018 to December 2022 were screened. The data utilized in this study were sourced from the database of nationwide, multicenter CHANCE registers in China, as detailed in previous publications.21, 22, 23

The inclusion criteria were as follows: (1) diagnosis confirmed histologically, cytologically, or clinically following the American Association for the Study of Liver Diseases criteria²⁴; (2) intermediate-stage HCC according to BCLC staging²; (3) age 18 years or older; (4) at least one measurable intrahepatic disease as defined by the modified Response Evaluation Criteria in Solid Tumors (mRECIST), which incorporates changes in the enhancement patterns of target lesions on imaging to better reflect tumor response in HCC.²⁵ The main exclusion criteria included: (1) previous treatment with TACE or systemic agents for HCC; (2) presence of overt hepatic encephalopathy or uncontrollable ascites; (3) incomplete information about outcomes or prognostic variables. Accordingly, all analyses were conducted on a complete-case dataset. The example study design diagram was displayed in Figure S1.

Treatment

Treatment allocation was determined through a locally multidisciplinary team (MDT)–based decision-making process at each participating center.²⁶ This evaluation comprehensively considered patient-related factors (e.g., age, comorbidities, and ECOG performance status), tumor and disease characteristics (e.g., tumor burden, biomarkers, and liver function), and treatment feasibility (e.g., contraindications, availability, potential efficacy, and financial burden).²⁷ Detailed description of the MDT process and treatment protocol were provided in the Supplementary Material.

The TACE procedures included either conventional TACE (cTACE) or drug-eluting beads TACE (DEB-TACE), which were standardly performed by physicians with over 10 years of experience in interventional radiology, following established guidelines.²⁸ Feeding arteries of the tumors were as selective as possible in order to obtain better treatment efficacy and to reduce treatment-related complications. “On-demand” TACE was repeated based on evidence of viable tumors or intra-hepatic recurrence, as determined by contrast-enhanced computed tomography or magnetic resonance imaging. When residual viable tumors were confirmed or new lesions developed in patients with adequate liver function, repeated TACE was performed.

VEGF inhibitors or TKIs were administered concomitantly with ICIs, and TACE was performed either simultaneously, or within a three-month window before or after ICIs initiation. All agents were administered according to prescribed doses and frequencies (Table S2). Dose reduction due to adverse events (AEs) was permitted for TKIs, but not for ICIs or VEGF inhibitors. Dose interruption due to AEs was permitted for both ICIs and VEGF inhibitors or TKIs. Systemic treatment was continued until either disease progression or the occurrence of unacceptable toxic effects.

Follow-up

The follow-up protocol included regular monitoring of vital signs, clinical symptoms, and treatment-related AEs, and conducting periodic laboratory tests and radiological examinations every 6–9 weeks. Patients underwent contrast-enhanced computed tomography or magnetic resonance imaging scans for tumor response evaluation by two independent radiologists with a minimum of ten years of experience at each center. These radiologists received standardized training to ensure consistent evaluations, conducted either online or offline. Patients continued to be followed routinely until either death or the end of the study (July 30, 2023). For the combination therapy group, T0 is defined as the time when all eligibility criteria are met and TACE, an ICI, or a VEGF inhibitor/TKI is first initiated, whichever occurs earliest. For the TACE monotherapy group, T0 is defined as the time when all eligibility criteria are met and TACE treatment is initiated. For the sensitivity analysis assessing how the definition of T0 impacts the results, Td was defined as the time point at which eligible patients were diagnosed with intermediate-stage HCC.

Outcomes

The co-primary outcomes were OS and PFS according to mRECIST. OS was defined as the duration from T0 to death from any cause. The PFS was determined as the period from T0 to the first documented disease progression or death from any cause. Secondary outcomes included PFS per Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1), objective response rate (ORR) per RECIST 1.1 or mRECIST, and safety profiles. The severity of AEs was classified using the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0.

Target trial emulation and statistical analysis

We adopted the cloning-censoring-weighting approach akin to the intention-to-treat analysis used in RCTs to minimize potential selection bias and immortal time bias.^18,29 First, given that the treatment strategy of interest included a three-month grace period for initiating combination therapy, we applied clones for all eligible patients at T0, the time of the initial dose of first treatment. Second, we censored any clones deviating from their assigned therapies during the grace period. Considering that artificial censoring may introduce potential selection bias, we performed the inverse probability of censoring weighting (IPCW) to re-weight patients remaining in the risk set, compensating for censored patients and ensuring the comparability between the combination therapy group and the TACE monotherapy group over the grace period. Finally, given the emulated trial design with the cloning step and the similarities between arms during the grace period, which led to a violation of the proportional hazards assumption, we utilized the restricted mean survival time (RMST) to compare differences in OS and PFS between groups.³⁰ Hazard ratio (HR) from the Cox proportional hazards model was also calculated for reference. Accounting for the inflated sample size after the clone step, a non-parametric bootstrap was utilized to obtain 95% CI through 1000 resamples. Under the target trial emulation framework, interpreting the results as causal estimates relies on rigorous methodology and key assumptions, including consistency, exchangeability, and positivity. In addition, all approaches relied on the correct model specification for the weights.

Multiple analyses were conducted to support the robustness of the primary results. First, stabilized inverse probability of treatment weighting (sIPTW) was utilized to emulate the randomization and address the imbalance of potential confounders, ensuring the exchangeability and reducing the influence of extreme weights. The directed acyclic graph was created to identify these confounders (Figure S2). Variables exhibiting a standardized mean difference of less than 0.1 were deemed balanced. Next, a landmark analysis was performed with the landmark time set at three months after T0 to further mitigate potential immortal-time bias.³¹ This three-month window corresponded to the grace period allowed for initiating combination therapy, ensuring balanced follow-up and reducing bias related to delayed treatment initiation. Additionally, OS and PFS were recalculated using Td as the follow-up start point, serving as a sensitivity analysis for the definition of T0. Other analytical methods were also employed as sensitivity analyses to address potential immortal-time bias, confounding bias or measurement error. Detailed descriptions of target trial emulation framework and statistical analyses were provided in Supplementary Material (pages 13–17). All analyses were performed using R software (version 4.4.2).

Multiplicity and sample size estimation

Considering the co-primary endpoints for this study, the overall two-sided significance level of 0.05 was split into a two-sided significance of 0.04 and 0.01 for the testing of OS and PFS, respectively. The allocation and recycling of significance level was conducted through a graphical approach.^32,33 No interim analysis was considered.

The ideal target trial sample size was pre-estimated based on literature review using PASS software (version 15.0.5). We calculated that 477 patients (159 for combination therapy, 318 for TACE monotherapy), targeting 274 deaths to achieve 80% power (two-sided α = 0.04) for an OS HR of 0.69 (median 36.0 versus 25.0 months) over an 18-month accrual and 54-month total trial duration, with a 5% annual loss to follow-up. This size also provides 97% power (two-sided α = 0.01) for PFS (HR = 0.62, median 13.0 versus 8.0 months). In our target trial emulation, we exceeded this minimum to include as many patients as possible, aiming to improve generalizability and enhance statistical power.³⁴

Role of the funding source

The funders had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. GJT, ZGR, HDZ, BYZ, JJC, and ZCJ had full access to all data in the study. GJT and ZGR had the final responsibility for the decision to submit the manuscript for publication.

Results

Patients and treatment

A total of 941 patients were included in the study, with 308 (32.7%) receiving combination therapy, and 633 (67.3%) receiving TACE monotherapy. The patient selection flowchart was presented in Fig. 1. All eligible patients meeting the inclusion criteria were included in the trial emulation. After applying sIPTW adjustments, one patient was excluded due to weighting inconsistencies. The distribution of propensity scores of sIPTW showed substantial overlap between the two groups (Figure S3), indicating balanced treatment allocation without extreme imbalances or certain allocation to one treatment group.

Fig. 1.

Flowchart of the study. Abbreviations: HCC, hepatocellular carcinoma; TACE, transarterial chemoembolization; ICIs, immune checkpoint inhibitors; VEGF, vascular endothelial growth factor; TKIs, tyrosine kinase inhibitors.

Before applying sIPTW, the combination therapy group exhibited a higher proportion of HBV infection, greater tumor burden, higher alpha-fetoprotein levels, and a more extensive history of HCC-related treatments. After sIPTW adjustment, the baseline characteristics were well-balanced between the two groups. Detailed baseline characteristics before and after sIPTW adjustment are presented in Table 1. A visualization of the standardized mean differences for each baseline covariate was shown in Figure S4.

Table 1.

Patient baseline characteristics before and after applying stabilized inverse probability of treatment weighting.^a

Characteristics	Before sIPTW				After sIPTWb
	Combination therapy (n = 308)	TACE monotherapy (n = 633)	P value	SMD	Combination therapy (n = 307)	TACE monotherapy (n = 633)	P value	SMD
Median age (years)	58 (51–66)	62 (53–68)	<0.001	0.277	60 (53–67)	61 (52–68)	0.808	0.025
Sex			0.937	0.012			0.952	0.004
Female	51 (16.6)	102 (16.1)			51 (16.6)	104 (16.4)
Male	257 (83.4)	531 (83.9)			256 (83.4)	529 (83.6)
Etiology			<0.001	0.31			0.922	0.008
Hepatitis B virus	267 (86.7)	472 (74.6)			242 (78.8)	497 (78.5)
Others	41 (13.3)	161 (25.4)			65 (21.2)	136 (21.5)
Cirrhosis			0.699	0.032			0.915	0.008
Absent	99 (32.1)	213 (33.6)			101 (32.9)	210 (33.2)
Present	209 (67.9)	420 (66.4)			206 (67.1)	423 (66.8)
Child-Pugh class			0.782	0.026			0.941	0.005
A	267 (86.7)	543 (85.8)			264 (86.0)	545 (86.1)
B	41 (13.3)	90 (14.2)			43 (14.0)	88 (13.9)
ALBI grade			0.967	0.018			0.977	0.015
1	131 (42.5)	274 (43.3)			133 (43.3)	272 (43.0)
2	171 (55.5)	346 (54.7)			168 (54.7)	348 (55.0)
3	6 (1.9)	13 (2.1)			6 (2.0)	12 (1.9)
Six-and-twelve score			0.033	0.18			0.994	0.008
≤6	52 (16.9)	126 (19.9)			59 (19.2)	120 (19.0)
>6 but ≤12	139 (45.1)	320 (50.6)			151 (49.2)	309 (48.8)
>12	117 (38.0)	187 (29.5)			98 (31.9)	204 (32.2)
Alpha-fetoprotein			0.038	0.148			0.911	0.008
≤400 ng/mL	190 (61.7)	435 (68.7)			202 (65.8)	419 (66.2)
>400 ng/mL	118 (38.3)	198 (31.3)			105 (34.2)	214 (33.8)
TACE type			0.254	0.084			0.113	0.116
cTACE	186 (60.4)	408 (64.5)			181 (59.0)	408 (64.5)
DEB-TACE	122 (39.6)	225 (35.5)			127 (41.4)	225 (35.5)
HCC-related treatment history			0.003	0.206			0.979	0.002
Absent	257 (83.4)	572 (90.4)			271 (88.3)	559 (88.3)
Present	51 (16.6)	61 (9.6)			36 (11.7)	74 (11.7)
Surgery	33 (10.7)	40 (6.3)	0.025	0.158	23 (7.5)	48 (7.6)	0.991	0.001
Ablation	22 (7.1)	26 (4.1)	0.068	0.132	16 (5.2)	32 (5.1)	0.917	0.007

Data are median (interquartile range) or n (%).

Abbreviations: sIPTW, stabilized inverse probability of treatment weighting; TACE, transarterial chemoembolization; SMD, standardized mean difference; ALBI, albumin-bilirubin; cTACE, conventional TACE; DEB-TACE, drug-eluting beads TACE; HCC, hepatocellular carcinoma.

^aData are expressed as the number (percentage) of patients, unless otherwise specified. Percentages have been rounded and may not sum to 100%.

^bThe total number of patients in the two groups varies slightly in the post-sIPTW pseudo-data set due to weighting.

During the treatment phase, patients with combination therapy received a median of 3 sessions of TACE (interquartile range [IQR], 2–5), 8 cycles of ICIs (IQR, 3–12), 7 cycles of VEGF inhibitors (IQR, 3–9), and 8.2 months of TKIs (IQR, 4.7–11.8). In contrast, those with TACE monotherapy had a median of 4 sessions of TACE (IQR, 2–7). A temporal trend analysis showed a gradual increase in the use of combination therapy over time (Figure S5). Details on the combination modalities and subsequent therapies are provided in Tables S3 and S4.

Efficacy

At data cutoff, the median follow-up was 25.5 months (95% CI, 23.4–27.2) in the combination therapy group and 25.8 months (95% CI, 22.7–28.0) in the TACE monotherapy group. A total of 101 patients (32.8%) in the combination group and 310 (49.0%) in the monotherapy group died. A total of 175 patients (56.8%) receiving combination therapy and 440 patients (69.5%) receiving TACE monotherapy had progressive disease per mRECIST or had died.

In the clone and IPCW analysis (Fig. 2), the combination therapy group had a median OS of 32.9 months (95% CI 31.2–40.6), compared to 23.0 months (95% CI 21.6–26.1) for the TACE monotherapy group. The RMST difference for OS was 9.2 months (95% CI 4.5–14.3, bootstrapped p < 0.001) with HR as 0.57 (95% CI 0.43–0.70). The combination therapy group had a median PFS of 18.0 months (95% CI 15.5–21.9), compared to 12.9 months (95% CI 11.8–14.3) for the TACE monotherapy group. The RMST difference for PFS was 6.7 months (95% CI 3.3–10.7, bootstrapped p = 0.001) with HR as 0.70 (95% CI 0.58–0.82).

Fig. 2.

Kaplan–Meier curves of overall survival (A), and progression-free survival (B) after cloning-censoring-weighting analysis. Abbreviations: TACE, transarterial chemoembolization; CI, confidence interval; HR, hazard ratio; RMST, restricted mean survival time.

In the sIPTW population, the proportional hazards assumption for OS was not met (p = 0.04). The median OS was 32.8 months (95% CI 31.1–40.6) with combination therapy and 23.3 months (95% CI 21.7–26.1) with TACE monotherapy (adjusted HR 0.56 [95% CI 0.44–0.71]; log-rank p < 0.001; Fig. 3A). Additionally, the RMST difference for OS was 9.3 months (95% CI 4.6–14.1, p < 0.001). The PFS per mRECIST was significantly longer with combination therapy than with TACE monotherapy (median, 18.7 months [95% CI 16.2–23.4] versus 12.9 months [95% CI 11.8–14.3]; adjusted HR 0.66, 95% CI 0.55–0.79; log-rank p < 0.001; Fig. 3B). The PFS per RECIST 1.1 was also significantly longer with combination therapy (median, 18.3 months [95% CI 14.6–21.9] versus 12.7 months [95% CI 11.7–14.2]; adjusted HR 0.73, 95% CI 0.60–0.87; log-rank p = 0.002); Figure S6). The confirmed ORR was 60.5% (95% CI 54.5–65.8) with combination therapy versus 44.3% (95% CI 40.3–48.2) with TACE monotherapy per mRECIST (p < 0.001), and 52.3% (95% CI 46.4–57.8) versus 37.0% (95% CI 33.0–40.7) per RECIST 1.1 (p < 0.001), respectively. Both the landmark analysis and the alternative Td analysis supported these findings (Figures S7 and S8).

Fig. 3.

Kaplan–Meier curves of overall survival (A), and progression-free survival (B) after sIPTW analysis. Abbreviations: sIPTW, stabilized inverse probability of treatment weighting; TACE, transarterial chemoembolization; CI, confidence interval; HR, hazard ratio; RMST, restricted mean survival time.

The subgroup analysis demonstrated that OS and PFS were more favorable with combination therapy than with TACE monotherapy across most clinically relevant subgroups (Fig. 4). Additionally, the overall findings remained consistent across multiple sensitivity analyses (Figures S9–S14).

Fig. 4.

Forest plot of overall survival (A) and progression-free survival assessed (B). Abbreviations: TACE, transarterial chemoembolization; HR, hazard ratio; CI, confidence interval; IPCW, inverse probability of censoring weight; NA, not applicable; sIPTW, stabilized inverse probability of treatment weighting; ALBI, albumin-bilirubin; cTACE, conventional TACE; DEB-TACE, drug-eluting beads TACE. ∗P value was estimated using the bootstrap resampling method. ^†P value was calculated using the log-rank test.

Safety

AEs of any grade were reported in 208 patients (67.5%) receiving combination therapy and 299 (47.2%) receiving TACE monotherapy (Table 2 and Table S5). Grade ≥3 AEs were observed in 64 patients (20.8%) in the combination group and 43 (6.8%) in the monotherapy group. No grade 5 AEs were reported in either group. In the combination therapy group, AEs leading to discontinuation of ICIs occurred in 21 patients (6.8%), and discontinuation of VEGF inhibitors or TKIs occurred in 47 patients (15.3%). Fifteen patients (4.9%) experienced dose interruption of ICIs due to AEs. The interruption of VEGF inhibitors or dose reduction or interruption of TKIs due to AEs occurred in 37 patients (12.0%).

Table 2.

Adverse events after treatment in original population.

Adverse events	Any grade			Grade ≥3
	Combination therapy (n = 308)	TACE monotherapy (n = 633)	P value	Combination therapy (n = 308)	TACE monotherapy (n = 633)	P value
Any adverse event	208 (67.5)	299 (47.2)	<0.001	64 (20.8)	43 (6.8)	<0.001
Abdominal pain	85 (27.6)	161 (25.4)	0.48	13 (4.2)	21 (3.3)	0.46
Pyrexia	68 (22.1)	96 (15.2)	0.01	6 (1.9)	8 (1.3)	0.40
Increased AST	59 (19.2)	77 (12.2)	0.01	6 (1.9)	8 (1.3)	0.40
Nausea	53 (17.2)	79 (12.5)	0.06	2 (0.6)	3 (0.5)	0.67
Increased ALT	47 (15.3)	71 (11.2)	0.09	9 (2.9)	9 (1.4)	0.13
Hypertension	42 (13.6)	0	<0.001	16 (5.2)	0	<0.001
Vomiting	37 (12.0)	58 (9.2)	0.20	2 (0.6)	1 (0.2)	0.25
Fatigue	31 (10.1)	32 (5.1)	0.01	4 (1.3)	0	0.01
Elevated bilirubin	31 (10.1)	45 (7.1)	0.13	3 (1.0)	1 (0.2)	0.11
HFSR	29 (9.4)	0	<0.001	7 (2.3)	0	<0.001
Rash	27 (8.8)	0	<0.001	2 (0.6)	0	0.11
Diarrhea	25 (8.1)	0	<0.001	4 (1.3)	0	0.01
Proteinuria	18 (5.8)	0	<0.001	3 (1.0)	0	0.04
RCCEP	18 (5.8)	0	<0.001	2 (0.6)	0	0.11
Hypothyroidism	17 (5.5)	0	<0.001	1 (0.3)	0	0.33
Anorexia	16 (5.2)	12 (1.9)	0.01	2 (0.6)	1 (0.2)	0.25
Thrombocytopenia	15 (4.9)	6 (0.9)	<0.001	3 (1.0)	0	0.04
Pruritus	12 (3.9)	0	<0.001	2 (0.6)	0	0.11
Ascites	9 (2.9)	4 (0.6)	0.01	2 (0.6)	0	0.11
Gastrointestinal hemorrhage	8 (2.6)	0	<0.001	1 (0.3)	0	0.33
Decreased WBC	8 (2.6)	4 (0.6)	0.02	2 (0.6)	0	0.11
Abdominal distension	6 (1.9)	11 (1.7)	0.80	1 (0.3)	0	0.33
Pneumonitis	4 (1.3)	0	0.01	2 (0.6)	0	0.11
Weight decrease	3 (1.0)	0	0.04	0	0	NA
Alopecia	3 (1.0)	0	0.04	0	0	NA
Constipation	2 (0.6)	0	0.11	0	0	NA
Infusion reaction	2 (0.6)	0	0.11	0	0	NA
Epistaxis	1 (0.3)	0	0.33	0	0	NA

Abbreviations: TACE, transarterial chemoembolization; AST, aspartate aminotransferase; ALT, alanine aminotransferase; HFSR, hand foot skin reaction; RCCEP, reactive cutaneous capillary endothelial proliferation; WBC, white blood cell.

Discussion

This study offers a comprehensive multicenter evaluation of clinical outcomes of combining TACE with ICIs and VEGF inhibitors or TKIs as a first-line treatment for intermediate-stage HCC in a real-world setting. The findings suggest that TACE combined with ICIs and VEGF inhibitors or TKIs significantly improves clinical outcomes for patients with intermediate-stage HCC. Specifically, this combination therapy yielded an RMST difference for OS of 9.2 months and for PFS (per mRECIST) by 6.7 months. It also reduced the risk of death by 43% and the risk of disease progression or death by 30%, compared to the TACE monotherapy, while achieving a significantly higher ORR with acceptable safety profiles. Notably, employing a target trial emulation framework, which aims to mitigate confounding, selection, and immortal time bias, enhanced the credibility of these findings. Subgroup analyses showed generally consistent survival benefits across clinical subgroups, and multiple sensitivity analyses affirmed the robustness of these results.

The findings of our study are consistent with and extend those from other studies (Table S6). In the EMERALD-1 trial, TACE combined with durvalumab and bevacizumab significantly prolonged PFS compared to TACE alone (median, 15.0 versus 8.2 months; p = 0.03) in patients with unresectable HCC.¹⁴ Similarly, the LEAP-012 trial demonstrated a comparable PFS improvement with TACE in combination with pembrolizumab and lenvatinib (median, 14.6 versus 10.0 months; p = 0.002).¹⁵ Notably, 57.4% of patients in EMERALD-1 and 57.0% in LEAP-012 were classified as intermediate-stage and received the respective TACE-based combination therapy, showing PFS improvement with HRs of 0.71 (95% CI 0.52–0.95) and 0.57 (0.41–0.77), respectively. Although theses phase 3 trials have not conclusively proved that whether this TACE-based combination strategy can improve OS due to immature follow-up, this target trial emulation study has shown that this combination can extend OS in patients with intermediate-stage HCC with real-world evidence.

Currently, the optimal timing and sequencing of TACE combined with immunotherapy-based systemic therapies remain uncertain, highlighting the need for future investigation.¹² In the EMERALD-1 trial, TACE was administered in 1–4 cycles over the first 16 weeks, with durvalumab initiated at least 7 days after the initial TACE procedure, and bevacizumab added to durvalumab at least 14 days after the final TACE procedure.¹⁴ In the LEAP-012 trial, the first TACE was administered 2–4 weeks after the start of systemic therapy with pembrolizumab and lenvatinib, limiting TACE to two sessions per lesion.¹⁵ In both CARES-005 and the TALENTACE trial, TACE was performed on-demand.^16,17 In TALENTACE, patients received TACE followed by atezolizumab-bevacizumab within 14 days to 8 weeks after TACE, whereas in CARES-005, camrelizumab-rivoceranib was initiated within 2 weeks after the first TACE procedure. In our study, TACE was repeated in an on-demand mode by a multidisciplinary team, with a 3-month grace period allowed for completing combination therapy. Patients receiving combination therapy underwent a median of 3 TACE sessions, while those in the TACE monotherapy group had a median of 4 sessions. This treatment pattern more closely reflects real-world clinical practice, in which therapeutic decisions are typically individualized based on patient response and multidisciplinary assessment.

Intermediate-stage HCC encompasses a heterogeneous patient population, particularly regarding tumor burden, leading to considerable variation in clinical outcomes.² In this study, approximately 80% of patients had high tumor burdens overall, while the combination therapy group had a higher tumor burden than the TACE monotherapy group, which may reflect real-world clinical conditions. Subgroup analyses indicated that among patients with low tumor burden, there was no difference in OS or PFS benefits between the two groups. Those with high tumor burden (six-and-twelve score >6 but ≤12, or >12) derived significant OS and PFS benefits from combination therapy (Fig. 4). These findings suggest that TACE combined with ICIs plus VEGF inhibitors or TKIs may be preferable for patients with high tumor burden in clinical practice. Regrettably, the LEAP-012 trial excluded patients with high tumor burden (diameter >10 cm or number >10), preventing the assessment of survival outcomes in this subgroup. Given the heterogeneity of intermediate-stage HCC, identifying the subgroup most likely to benefit from TACE-based combination strategies remains an important direction for future research.

While the incidence of AEs was higher in the combination therapy group, the safety profile was manageable, with only a small percentage of patients discontinuing ICIs (6.8%) or VEGF inhibitors or TKIs (15.3%) due to AEs, and experiencing grade 3 or higher AEs (20.8%). This aligns with previous studies, where combination therapies involving TACE combined with systemic agents showed increased toxicity but remained within manageable limits.^21,22,35,36 It would be interesting to explore the impact of AEs on treatment efficacy in future studies.³⁷ Importantly, the enhanced efficacy of the combination therapy does not come with a prohibitive increase in severe AEs, making it a viable option for patients.

Our study has several limitations. Firstly, although the target trial emulation framework and multiple sensitivity analyses were employed to mitigate selection and immortal time biases, the potential for residual biases and unmeasured confounders remains a consideration due to the retrospective nature of the study. In particular, socioeconomic factors such as treatment affordability and insurance coverage may have influenced treatment selection but were not captured in our models. Additionally, potential between-center variability could not be formally assessed, although all participating centers were tertiary hospitals where clinical practice generally followed standardized national guidelines. Nonetheless, the E-value analysis suggested that such confounding is unlikely to fully explain the observed survival benefit. Secondly, the use of multiple drugs and their various combinations introduce heterogeneity to some extent, potentially affecting the generalizability of our results. Drawing from the concept of “umbrella trials”, we pooled various ICIs and TKIs with similar targets to evaluate broader treatment strategies for HCC. These drugs were approved and available in China, with favorable therapeutic effect in HCC confirmed through several RCTs.^4,6,38 Finally, the study population consisted predominantly of Chinese patients with HBV-related HCC, and the response to immunotherapy in different etiologic populations possibly varied to some extent,¹³ which may limit the applicability of the findings to other populations and etiologies.

In conclusion, this nationwide target trial emulation study suggests that adding ICIs and VEGF inhibitors or TKIs to TACE as a first-line treatment for intermediate-stage HCC can improve PFS, OS, and ORR, while maintaining a manageable safety profile. With longer follow-up, our data may further clarify the magnitude of survival benefit associated with these combinations. Future results from ongoing phase 3 trials, once mature, will ultimately provide more definitive evidence.

Contributors

GJT, ZGR, HDZ, YZ contributed to conception and design. YZ, JWZ, JJC, and ZCJ conducted statistical analysis. All authors and the CHANCE2202 investigators contributed to acquisition, analysis, or interpretation of data. GJT, ZGR, HDZ, JJC, and ZCJ accessed and verified the data. JJC, ZCJ, BYZ and HDZ drafted the manuscript, and all authors critically reviewed it for important intellectual content. GJT and ZGR provided supervision. All authors reviewed and approved the final version of the manuscript. GJT, ZGR, HDZ, BYZ, JJC, and ZCJ had full access to all data in the study. GJT and ZGR had final responsibility for the decision to submit for publication.

Data sharing statement

Data are not available for public sharing owing to privacy, ethical, and legal considerations. Summary statistical data may be obtained from the corresponding author upon reasonable request and with approval from the Chinese multicenter registry.

Declaration of interests

All authors declare no competing interests.