HAIC-Based versus TACE-Based Triple Therapy with Lenvatinib and PD-1 Inhibitors for Hepatocellular Carcinoma with Portal Vein Tumor Thrombus: A Retrospective Propensity Score-Matched Study

Abstract

Background

The optimal locoregional treatment to combine with lenvatinib and PD-1 inhibitors for hepatocellular carcinoma with portal vein tumor thrombus (PVTT-HCC) remains uncertain. We compared HAIC- and TACE-based triple therapy.

Methods

This retrospective single-center study included 359 consecutive patients with PVTT-HCC treated with HAIC or TACE plus lenvatinib and a PD-1 inhibitor. Propensity score matching (PSM) was performed at a 1:1 ratio. Tumor response was assessed using mRECIST. Overall survival (OS), progression-free survival (PFS), conversion-to-surgery, and adverse events were compared. Time-dependent Cox models accounted for cumulative locoregional treatment sessions.

Results

Among 359 patients, 161 received TACE-based therapy and 198 received HAIC-based therapy. After PSM, 139 matched pairs were analyzed. HAIC was associated with higher objective response rate (66.91% vs 42.45%) and disease control rate (82.73% vs 62.59%) than TACE (both P<0.001). Median OS was 23.7 versus 17.1 months (HR 0.67, 95% CI 0.48–0.93; P=0.018), and median PFS was 7.7 versus 5.4 months (HR 0.70, 95% CI 0.54–0.91; P=0.008), favoring HAIC. After time-dependent adjustment, HAIC remained associated with improved OS (HR 0.61; P=0.003) and PFS (HR 0.73; P=0.011). Conversion-to-surgery was more frequent with HAIC (27.3% vs 17.4%; P=0.027). Overall adverse-event rates were similar (89.90% vs 88.82%; P=0.741), although hypertension was more frequent with HAIC (52.02% vs 38.51%; P=0.011).

Conclusion

In patients with PVTT-HCC receiving lenvatinib and PD-1 inhibitors, HAIC-based triple therapy was associated with improved tumor response and longer OS and PFS than TACE-based therapy, with generally comparable safety. Prospective multicenter studies are warranted.

Introduction

Hepatocellular carcinoma (HCC) is among the most frequently diagnosed cancers worldwide, ranking sixth in global incidence and representing the third leading cause of cancer-related mortality.¹ As the most common histologic subtype of liver cancer,² hepatocellular carcinoma (HCC) often presents with insidious symptoms in the early stage;³ consequently, approximately 70% of patients are diagnosed at an intermediate to advanced stage, by which time curative surgical resection is no longer feasible.⁴ Moreover, HCC is characterized by a strong propensity for vascular invasion, with a particular tendency to infiltrate the portal venous system; consequently, approximately 10%–40% of patients present with portal vein tumor thrombus (PVTT) at the time of diagnosis.⁵ According to the Barcelona Clinic Liver Cancer (BCLC) staging system, systemic therapy—primarily based on targeted agents and immunotherapy—is recommended for patients with hepatocellular carcinoma (HCC) complicated by portal vein tumor thrombus (PVTT).⁶ Nevertheless, the clinical benefit of systemic therapies in HCC with PVTT remains limited, as reflected by a median OS of 9.7–14.2 months and a median PFS of 3.6–7.3 months.⁷ Moreover, the progression and treatment resistance of HCC are driven by complex molecular alterations and tumor-promoting regulatory networks, underscoring the biological heterogeneity of advanced HCC and the need for more effective combined therapeutic strategies.^8,9

In recent years, systemic therapy with molecular targeted agents and immune checkpoint inhibitors has substantially changed the treatment landscape of advanced HCC. Immunotherapy has become an important component of systemic treatment, and emerging studies have continued to explore novel molecular targets and therapeutic strategies for advanced HCC.^10,11 Numerous studies have suggested that, for patients with intermediate-to-advanced HCC receiving systemic therapy, the addition of interventional (locoregional) treatment can increase the objective response rate (ORR) and improve clinical outcomes.^12,13 A multicenter retrospective study reported that triple therapy with transarterial chemoembolization (TACE) plus lenvatinib and a PD-1 inhibitor, when applied to unresectable HCC with main portal vein tumor thrombus (Vp4), achieved a high tumor response rate and relatively favorable survival outcomes.¹⁴ In the JSH guidelines, HAIC is positioned as a key treatment option for HCC with PVTT, particularly in cases with major portal vein invasion (Vp3/Vp4).¹⁵

Given that transarterial chemoembolization (TACE) and hepatic arterial infusion chemotherapy (HAIC), as two distinct interventional modalities, have both demonstrated meaningful therapeutic efficacy in hepatocellular carcinoma (HCC) with portal vein tumor thrombus (PVTT), the optimal choice of locoregional treatment to incorporate into systemic therapy for patients with PVTT-HCC remains controversial. In this study, we retrospectively compared the therapeutic efficacy of TACE plus a PD-1 inhibitor and lenvatinib versus HAIC plus a PD-1 inhibitor and lenvatinib in patients with hepatocellular carcinoma and portal vein tumor thrombus (PVTT-HCC).

Materials and Methods

Patients

This retrospective study included patients with hepatocellular carcinoma (HCC) and portal vein tumor thrombus (PVTT) who received interventional therapy combined with targeted therapy and immunotherapy at Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, from April 2019 to August 2025. Patients were stratified according to the interventional modality into a TACE plus targeted therapy and immunotherapy group (TACE group) and an HAIC plus targeted therapy and immunotherapy group (HAIC group). The requirement for written informed consent was waived by the Institutional Review Board/Ethics Committee because of the retrospective nature of the study. The study protocol was reviewed and approved by the Institutional Review Board/Ethics Committee of Tongji Hospital (Approval No: TJ-IRB202511038).

HCC was diagnosed according to the American Association for the Study of Liver Diseases (AASLD) imaging criteria and/or histopathological confirmation when available.⁶ PVTT was evaluated on contrast-enhanced multiphasic imaging (triphasic CT and/or dynamic contrast-enhanced MRI) and was defined as an enhancing intraluminal soft-tissue mass within the portal venous system, typically showing arterial-phase hyperenhancement with subsequent washout on the portal venous and/or delayed phases. PVTT was further classified according to Cheng’s classification based on the extent of portal vein involvement.¹⁶ Bland portal vein thrombosis was differentiated by the absence of arterial-phase enhancement within the thrombus.

The inclusion criteria were as follows: (1) histologically or clinically confirmed HCC with PVTT; (2) age 18–75 years; (3) PVTT confirmed by contrast-enhanced CT or MRI and classified as Cheng’s type I–IV; (4) Child–Pugh class A or B liver function; (5) an Eastern Cooperative Oncology Group (ECOG) performance status score of 0 or 1; (6) no prior anticancer treatment, including radiotherapy, chemotherapy, or other antineoplastic agents; and (7) receipt of at least one session of TACE or at least two cycles of HAIC. The exclusion criteria were as follows: (1) severe comorbidities or major organ dysfunction (eg, cardiac, pulmonary, renal, or other vital organ insufficiency); (2) concomitant other primary malignancies; (3) incomplete clinical data that were inadequate for the planned analyses; and (4) loss to follow-up, precluding the acquisition of follow-up information.

Treatment Procedures

HAIC was scheduled every 3 weeks. All procedures were performed under digital subtraction angiography (DSA) guidance. Angiography of the celiac trunk and superior mesenteric artery was routinely conducted to delineate tumor arterial supply. Together with preprocedural contrast-enhanced CT, MRI, and/or DSA images, a microcatheter was selectively advanced into the tumor-feeding artery, most commonly the left hepatic artery, right hepatic artery, or proper hepatic artery. If part of the tumor blood supply was suspected to originate outside the celiac trunk system, additional angiographic evaluation of the phrenic and intercostal arteries was undertaken, and lesions supplied by these branches were managed with embolization and/or transarterial chemotherapy as appropriate. After catheter and microcatheter placement, the FOLFOX regimen was administered as follows: oxaliplatin 85 mg/m² (0–3 h on day 1), leucovorin 400 mg/m² (3–4 h on day 1), 5-fluorouracil (5-FU) 400 mg/m² bolus (4–5 h on day 1), and 5-FU 2400 mg/m² continuous infusion over 46 h (days 1–2).

TACE was initiated within 7 days after diagnosis and was performed via femoral arterial access under local anesthesia. Under digital subtraction angiography guidance, a microcatheter was advanced superselectively into the segmental or subsegmental tumor-feeding arteries. An emulsion of iodized oil mixed with pirarubicin was then infused into the target arteries, with the pirarubicin dose individualized according to baseline liver function, tumor burden (size/volume), tumor vascularity, and body surface area. Thereafter, all identified tumor-feeding arteries were embolized superselectively using gelatin sponge particles until a marked reduction in arterial flow was achieved. The volume of emulsion administered was determined by tumor volume. When an artery–portal vein fistula was detected angiographically, it was occluded before chemoembolization using coil embolization. Repeat TACE sessions were scheduled based on comprehensive reassessment with contrast-enhanced CT/MRI, serum tumor markers, and liver function tests. Lenvatinib and anti–PD-1 antibody therapy were withheld for 3 days before and after each TACE session. Patients with untreated chronic hepatitis B received prophylactic antiviral therapy (entecavir or tenofovir) throughout the anticancer treatment period. In cases of disease refractoriness or unacceptable treatment-related toxicity, subsequent management was determined through a multidisciplinary team approach, with escalation to second-line systemic therapy and/or conversion to HAIC or radiotherapy when appropriate.

Lenvatinib was initiated after locoregional treatment, within 3–7 days after TACE and within 3 days after HAIC, and was continued thereafter according to patient tolerance and liver function status. Lenvatinib was administered orally at 8 mg once daily for patients weighing ≤60 kg and 12 mg once daily for those weighing >60 kg. Dose interruption, reduction, or discontinuation was permitted for clinically significant adverse events or hepatic decompensation at the discretion of the treating physicians. Anti–PD-1 therapy consisted of tislelizumab, sintilimab, or pembrolizumab. PD-1 inhibitors were initiated on day 3 after TACE or HAIC. Each agent was administered intravenously at a fixed dose of 200 mg every 3 weeks (q3w). The specific PD-1 antibody was selected at the discretion of the treating physicians based on patient condition and drug availability, and treatment was continued according to tolerance and liver function status.

HAIC was delayed, withheld, or discontinued according to predefined clinical and laboratory criteria. Specifically, HAIC was withheld in patients with clinically significant ascites requiring intensified diuretic therapy or therapeutic paracentesis, uncontrolled infection, Eastern Cooperative Oncology Group performance status deterioration to ≥2, neutrophil count <1.5 × 10⁹/L, platelet count <50 × 10⁹/L, hemoglobin <80 g/L, persistent grade ≥2 nausea/vomiting or abdominal pain despite supportive treatment, or any grade ≥3 gastrointestinal toxicity according to CTCAE version 5.0. Treatment was resumed only after recovery to an acceptable clinical condition and after reassessment by the multidisciplinary team. Elevations in liver enzymes or bilirubin alone were not considered absolute contraindications unless accompanied by clinical hepatic decompensation. Similar clinical and laboratory criteria were applied to TACE. In addition, TACE was withheld in the presence of hepatic decompensation, clinically significant hyperbilirubinemia, severe coagulopathy, renal dysfunction, or other conditions considered to increase the risk of embolization-related hepatic ischemia.

Criteria for Conversion Surgery

The feasibility of conversion surgery was reassessed during follow-up by a multidisciplinary team consisting of hepatobiliary surgeons, interventional radiologists, medical oncologists, and radiologists. Surgical resection was considered when the following criteria were met: (1) radiological tumor response or durable disease control according to mRECIST; (2) regression or stabilization of PVTT allowing technically feasible macroscopic complete resection; (3) absence of uncontrolled extrahepatic disease; (4) preserved liver function, defined as Child–Pugh class A or well-compensated Child–Pugh class B without refractory ascites, hepatic encephalopathy, or uncontrolled hyperbilirubinemia; (5) ECOG performance status of 0–1; and (6) adequate future liver remnant based on preoperative imaging and institutional surgical criteria. The final decision for surgery was made after multidisciplinary discussion and patient consent.

Follow-up

Clinical and radiological data were retrospectively collected from the electronic medical record and follow-up system. The collected variables included baseline demographic and clinical characteristics (age, sex, body mass index [BMI], smoking status, and alcohol consumption), hepatitis B virus (HBV)–related information (HBV infection status and serum HBV DNA levels), liver function and hepatic reserve assessments (Child–Pugh class and albumin–bilirubin [ALBI] grade), serum tumor biomarkers (alpha-fetoprotein [AFP] and protein induced by vitamin K absence or antagonist-II [PIVKA-II]), and tumor-related characteristics (tumor number, maximum tumor diameter, and the number of locoregional interventional procedures). Imaging assessments were performed every 6 weeks and independently reviewed by two radiologists according to modified Response Evaluation Criteria in Solid Tumors (mRECIST). Tumor response was categorized as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD). The objective response rate (ORR) was defined as the proportion of patients achieving CR or PR, and the disease control rate (DCR) was defined as the proportion achieving CR, PR, or SD. AFP response without normalization was assessed at the first follow-up, approximately 6 weeks after treatment initiation, among patients with a baseline AFP level ≥400 ng/mL and was defined as a decrease of at least 20% from baseline.¹⁷ Patients whose AFP levels normalized were not included in this category. Progression-free survival (PFS) was calculated from the initiation of locoregional treatment (TACE or HAIC) to radiologic progression or death from any cause, and overall survival (OS) was calculated from treatment initiation to death from any cause. Adverse events were assessed and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), version 5.0.

Adverse events were retrospectively reviewed from inpatient medical records, laboratory test results, nursing records, medication records, and discharge summaries. Laboratory abnormalities, including liver function impairment, hematologic toxicity, and coagulation-related abnormalities, were graded according to the corresponding CTCAE v5.0 criteria.¹⁸ Clinical adverse events, including fever, abdominal pain, nausea/vomiting, hypertension, hand-foot syndrome, and gastrointestinal bleeding, were also graded using CTCAE v5.0. For patients with repeated events, the highest grade observed during the treatment period was recorded.

Pathological Response Assessment

Pathological response was evaluated in patients who underwent conversion surgery. Resected specimens were reviewed by experienced pathologists using hematoxylin and eosin-stained sections. The percentage of residual viable tumor was estimated within the treated tumor bed. Pathological complete response (pCR) was defined as 0% residual viable tumor. For categorical reporting, major pathological response (MPR) was defined as >0% and ≤10% residual viable tumor because pCR was reported separately. Partial pathological response was defined as >10% and ≤50% residual viable tumor, whereas no pathological response was defined as >50% residual viable tumor.^19,20 Microvascular invasion was assessed according to routine pathological criteria and categorized as M0, M1, or M2.

Statistical Analysis

All statistical analyses, except for the time-dependent Cox regression analyses, were performed using SPSS software (version 26.0). Time-dependent Cox regression analyses were conducted using R software (version 4.5.3) with the survival package. Continuous variables were assessed for distributional characteristics before analysis. Normally distributed continuous variables are presented as mean ± standard deviation (SD) and were compared using the independent-samples t test, whereas non-normally distributed variables are presented as median and interquartile range (IQR) and were compared using the Mann–Whitney U-test. Categorical variables are presented as number (percentage) and were compared using the chi-square test or Fisher’s exact test, as appropriate.

Propensity score matching (PSM) was performed to minimize selection bias and balance baseline characteristics between the TACE and HAIC groups. Propensity scores were estimated for each patient using a logistic regression model including age, sex, body mass index (BMI), HBsAg status, HBV-DNA level, Child–Pugh class, ALBI grade, AFP level, DCP/PIVKA-II level, PVTT classification, hepatic vein invasion, tumor number, maximum tumor size, and extrahepatic metastasis. Patients in the two groups were then matched 1:1 using nearest-neighbor matching without replacement, with a caliper width of 0.2 of the standard deviation of the logit of the propensity score, as recommended in previous methodological studies of propensity score matching. This caliper was selected to reduce the likelihood of poor matches, improve covariate balance, and minimize residual confounding while preserving an adequate matched sample size.^21,22 Because the number of locoregional treatment sessions was accumulated during follow-up and differed between the TACE-based and HAIC-based groups, additional Cox proportional hazards models with time-dependent covariates were performed. Follow-up time for each patient was divided into multiple intervals according to the timing of each locoregional treatment session. The cumulative number of locoregional treatment sessions was updated at the beginning of each interval and modeled as a time-dependent covariate. Treatment group, defined as HAIC-based versus TACE-based therapy, was included as a fixed covariate. The models were adjusted for clinically relevant baseline prognostic factors, including PVTT classification, Child-Pugh class, tumor number, extrahepatic metastasis, and AFP level.

Overall survival (OS) was defined as the time from treatment initiation to death from any cause, and patients alive at the last follow-up were censored. Progression-free survival (PFS) was defined as the time from treatment initiation to radiologic disease progression or death from any cause, whichever occurred first. Median follow-up was estimated using the reverse Kaplan–Meier method rather than by directly calculating the median observed time. In this approach, death events were treated as censored observations and censored observations were treated as events, allowing estimation of the potential follow-up duration among the study cohort. Survival curves were estimated using the Kaplan–Meier method and compared using the Log rank test; Kaplan–Meier curves were plotted using GraphPad Prism. Univariable and multivariable prognostic analyses were performed using the Cox proportional hazards regression models, with results reported as hazard ratios (HRs) and 95% confidence intervals (CIs).

Subgroup analyses were performed in the overall cohort before propensity score matching. Treatment-by-subgroup interaction tests were conducted by including interaction terms between treatment modality and each prespecified subgroup variable in Cox proportional hazards regression models. P values for interaction were reported to assess potential effect modification. Because multiple subgroup analyses were performed, these analyses were considered exploratory and hypothesis-generating. No formal multiplicity adjustment, such as Bonferroni correction, was applied to the subgroup analyses. A two-sided P value <0.05 (α=0.05) was considered statistically significant.

Results

Basic Patient Information

Between April 2019 and August 2025, a total of 359 consecutive patients who received first-line triple therapy were included in this study. Among these patients, 161 received TACE treatment, and 198 received HAIC treatment. The mean age was 52.25 ± 11.1 years, and 89.7% (322/359) were male. The median maximum tumor diameter was 9.3 cm; extrahepatic metastasis and Cheng type III–IV PVTT were present in 25.9% (93/359) and 32.0% (115/359) of patients, respectively. Hepatitis B virus infection was the predominant etiology (83.8%), and all patients with chronic HBV infection received antiviral therapy with entecavir or tenofovir.

Overall, 139 patients in the TACE-based triple therapy group and 139 patients in the HAIC-based triple therapy group were included in the analysis after 1:1 PSM, while patients who did not meet the matching criteria were excluded. Characteristics including age, sex, BMI, etiology, HBV-DNA, PVTT stage, tumor number, largest tumor size, AFP, DCP, Child-Pugh class, ALBI grade, hepatic vein invasion and metastasis were matched and shown in Table 1. There were no significant differences in the baseline characteristics between the two groups (P > 0.05). In addition, HAPVF was assessed based on preprocedural DSA reports. Among the overall cohort, HAPVF was identified in 66 patients (18.38%). The prevalence of HAPVF was 16.77% (27/161) in the TACE group and 19.70% (39/198) in the HAIC group, with no significant difference between groups (χ² = 0.51, P = 0.476). The number of locoregional treatment sessions was not normally distributed in either group (both P < 0.001, Lilliefors normality test) and was therefore summarized as median (IQR). The median number of locoregional treatment sessions was 2 (IQR, 1–4) in the TACE group and 4 (IQR, 3–5) in the HAIC group. The corresponding mean values were 2.69 ± 1.74 and 3.98 ± 2.04, respectively.

Table 1.

Baseline Characteristics Before and After PSM

Characteristics	Before PSM				After PSM
	TACE+L+P,n=161	HAIC+L +P,n=198	χ2	P value	TACE+L+P,n=139	HAIC+L+P,n=139	χ2	P value
Sex
Men	147(91.3%)	175(88.38%)	0.82	0.365	127(91.37%)	123(88.49%)	0.64	0.425
Women	14(8.7%)	23(11.62%)			12 (8.63%)	16(11.51%)
Age, years
<65	137(85.09%)	177 (89.39%)	1.5	0.221	118 (84.89%)	121 (87.05%)	0.27	0.604
≥65	24(14.91%)	21 (10.61%)			21 (15.11%)	18 (12.95%)
BMI
<25	133 (82.61%)	156 (78.79%)	0.83	0.363	111 (79.86%)	112 (80.58%)	0.02	0.88
≥25	28 (17.39%)	42 (21.21%)			28 (20.14%)	27 (19.42%)
HBsAg
Negative	22 (13.66%)	36 (18.18%)	1.34	0.247	22 (15.83%)	20 (14.39%)	0.11	0.738
Positive	139 (86.34%)	162 (81.82%)			117 (84.17%)	119 (85.61%)
HBV-DNA
<2000 IU/mL	103 (63.98%)	128 (64.65%)	0.02	0.895	88 (63.31%)	87 (62.59%)	0.02	0.901
≥2000 IU/mL	58 (36.02%)	70 (35.35%)			51 (36.69%)	52 (37.41%)
Child-Pugh class
A	126 (78.26%)	165 (83.33%)	1.49	0.223	112 (80.58%)	111 (79.86%)	0.02	0.88
B	35 (21.74%)	33 (16.67%)			27 (19.42%)	28 (20.14%)
ALBI Grade
1	107 (66.46%)	130 (65.66%)	0.03	0.873	95 (68.35%)	92 (66.19%)	0.15	0.701
2	54 (33.54%)	68 (34.34%)			44 (31.65%)	47 (33.81%)
AFP,ng/mL
<400	72 (44.72%)	83 (41.92%)	0.28	0.594	64 (46.04%)	61 (43.88%)	0.13	0.718
≥400	89 (55.28%)	115 (58.08%)			75 (53.96%)	78 (56.12%)
PIVKA-II, ng/mL
<40	15 (9.32%)	20 (10.10%)	0.06	0.803	14 (10.07%)	12 (8.63%)	0.17	0.68
≥40	146 (90.68%)	178 (89.90%)			125 (89.93%)	127 (91.37%)
Portal vein invasion (Cheng classification)
I	54 (33.54%)	48 (24.24%)	8.17	0.043	45 (32.37%)	44 (31.65%)	0.67	0.88
II	67 (41.61%)	75 (37.88%)			57 (41.01%)	57 (41.01%)
III	23 (14.29%)	38 (19.19%)			21 (15.11%)	25 (17.99%)
IV	17 (10.56%)	37 (18.69%)			16 (11.51%)	13 (9.35%)
Hepatic vein invasion, n (%)
No	144 (89.44%)	162 (81.82%)	4.1	0.043	123 (88.49%)	122 (87.77%)	0.03	0.853
Yes	17 (10.56%)	36 (18.18%)			16 (11.51%)	17 (12.23%)
Number of tumors
<4	86 (53.42%)	112 (56.57%)	0.36	0.551	79 (56.83%)	78 (56.12%)	0.01	0.904
≥4	75 (46.58%)	86 (43.43%)			60 (43.17%)	61 (43.88%)
Maximum tumor size
<10cm	98 (60.87%)	115 (58.08%)	0.29	0.593	87 (62.59%)	82 (58.99%)	0.38	0.539
≥10cm	63 (39.13%)	83 (41.92%)			52 (37.41%)	57 (41.01%)
Extrahepatic metastasis
No	116 (72.05%)	150 (75.76%)	0.64	0.425	101 (72.66%)	102 (73.38%)	0.02	0.893
Yes	45 (27.95%)	48 (24.24%)			38 (27.34%)	37 (26.62%)

Tumor Response

When tumor response was assessed using mRECIST before PSM, the numbers of patients achieving CR, PR, SD, and PD were 3 (1.86%), 65 (40.37%), 33 (20.50%), and 60 (37.27%) in the TACE group, respectively, and 4 (2.02%), 118 (59.60%), 35 (17.68%), and 41 (20.71%) in the HAIC group, respectively. The ORR was 42.24% versus 61.62% (P<0.001), and the DCR was 62.73% versus 79.29% (P<0.001) in the TACE and HAIC groups, respectively (Table 2).

Table 2.

Tumor Responses Evaluated by mRECIST Before and After PSM

Tumor response	Before PSM
	Overall, N=359	TACE+L +P	HAIC+L +P	χ2	P-value
Objective response rate	190 (52.92%)	68 (42.24%)	122 (61.62%)	13.39	<0.001
Disease control rate	258 (71.87%)	101 (62.73%)	157 (79.29%)	12.04	<0.001
Best response
Complete response	7 (1.95%)	3 (1.86%)	4 (2.02%)	15.48	0.001
Partial response	183 (50.97%)	65 (40.37%)	118 (59.60%)
Stable disease	68 (18.94%)	33 (20.50%)	35 (17.68%)
Progressive disease	101 (28.13%)	60 (37.27%)	41 (20.71%)
Tumor response	After PSM
	Overall, N=278	TACE+L +P	HAIC+L +P	χ2	P-value
Objective response rate	152 (54.68%)	59 (42.45%)	93 (66.91%)	16.78	<0.001
Disease control rate	202 (72.66%)	87 (62.59%)	115 (82.73%)	14.20	<0.001
Best response
Complete response	7 (2.52%)	3 (2.16%)	4 (2.88%)	18.69	<0.001
Partial response	145 (52.16%)	56 (40.29%)	89 (64.03%)
Stable disease	50 (17.99%)	28 (20.14%)	22 (15.83%)
Progressive disease	76 (27.34%)	52 (37.41%)	24 (17.27%)

Note: ORR was defined as CR plus PR, and DCR was defined as CR plus PR plus SD.

When tumor response was assessed using mRECIST after 1:1 PSM, the numbers of patients achieving CR, PR, SD, and PD were 3 (2.16%), 56 (40.29%), 28 (20.14%), and 52 (37.41%) in the TACE group, respectively, and 4 (2.88%), 89 (64.03%), 22 (15.83%), and 24 (17.27%) in the HAIC group, respectively. The ORR was 42.45% versus 66.91% (P<0.001), and the DCR was 62.59% versus 82.73% (P<0.001) in the TACE and HAIC groups, respectively (Table 2).

AFP Response

To further evaluate dynamic biomarker response, AFP changes from baseline to the first follow-up assessment were analyzed among patients with baseline AFP ≥400 ng/mL in the propensity score-matched cohort. The proportion of patients with baseline AFP ≥400 ng/mL was comparable between the TACE and HAIC groups (75/139 [53.96%] vs, 78/139 [56.12%], P=0.718). Baseline AFP levels were also similar between the two groups (median, 16,306.0 [IQR, 1,691.0–59,561.0] vs 15,355.5 [IQR, 1,471.2–52,466.0] ng/mL, P=0.978). At the first follow-up assessment, AFP levels decreased in both groups, with a numerically lower median AFP level in the HAIC group than in the TACE group (2,536.6 [IQR, 345.2–15,836.3] vs 3,695.7 [IQR, 559.0–21,125.9] ng/mL, P=0.412). AFP response without normalization was observed in 41 of 75 patients (54.7%) in the TACE group and 49 of 78 patients (62.8%) in the HAIC group (P=0.306). These findings suggested a numerically higher AFP response rate in the HAIC group, although the difference did not reach statistical significance. Detailed results are shown in Table 3.

Table 3.

AFP Response Among Patients with Baseline AFP ≥400 ng/mL After PSM

Variable	TACE Group	HAIC Group	P value
Patients with baseline AFP ≥400 ng/mL, n (%)	75/139 (53.96%)	78/139 (56.12%)	0.718
Baseline AFP, median (IQR), ng/mL	16,306.0 (1,691.0–59,561.0)	15,355.5 (1,471.2–52,466.0)	0.978
AFP at first follow-up, median (IQR), ng/mL	3,695.7 (559.0–21,125.9)	2,536.6 (345.2–15,836.3)	0.412
AFP response without normalization, n (%)	41/75 (54.7%)	49/78 (62.8%)	0.306

Notes: Data are presented as median (interquartile range) or number (percentage). The analysis was restricted to patients with baseline AFP ≥400 ng/mL. AFP response without normalization was defined as a reduction of ≥20% from baseline at the first follow-up assessment approximately 6 weeks after treatment initiation, with the follow-up AFP level remaining ≥20 ng/mL. Patients achieving AFP normalization were not included in this category. Continuous variables were compared using the Mann–Whitney U-test, and categorical variables were compared using the Pearson χ²-test.

Survival Benefits

Data were collected through August 30, 2025, and the median follow-up was 25.8 months (estimated using the reverse Kaplan–Meier method). Before PSM, the median overall survival (mOS) was 22.2 months (95% CI, 19.6–24.8) in the HAIC+L+P group and 16.0 months (95% CI, 13.6–18.3) in the TACE+L+P group, with a statistically significant difference between groups (P = 0.003). The risk of death was lower in the HAIC group than in the TACE group (HR, 0.64; 95% CI, 0.48–0.86; P = 0.003) (Figure 1).

Figure 1.

Kaplan–Meier survival analyses before and after propensity score matching. (A) Overall survival before propensity score matching. (B) Progression-free survival before propensity score matching. (C) Overall survival after propensity score matching. (D) Progression-free survival after propensity score matching. Numbers of patients at risk at prespecified time points are shown below each Kaplan–Meier curve.

After 1:1 propensity score matching, 139 matched pairs were included in the survival analysis. In the PSM-matched cohort, the median OS was 17.1 months (95% CI, 14.3–19.8) in the TACE group and 23.7 months (95% CI, 20.9–26.4) in the HAIC group (HR, 0.67; 95% CI, 0.48–0.93; P = 0.018, Log rank test). Similarly, the median PFS was 5.4 months (95% CI, 4.16–6.6) versus 7.7 months (95% CI, 6.9–8.5) (HR, 0.70; 95% CI, 0.54–0.91; P = 0.008) (Figure 1).

Post-Progression Treatments

To evaluate whether post-progression treatments may have influenced the observed overall survival difference, subsequent treatment patterns after disease progression were further analyzed in the propensity score-matched cohort. Among 139 patients in the TACE group, 48 (34.5%) received subsequent systemic therapy, 14 (10.1%) received radiotherapy, 22 (15.8%) received additional locoregional therapy, 35 (25.2%) received best supportive care, and 20 (14.4%) had no tumor progression during follow-up. Among 139 patients in the HAIC group, 58 (41.7%) received subsequent systemic therapy, 15 (10.8%) received radiotherapy, 18 (12.9%) received additional locoregional therapy, 34 (24.5%) received best supportive care, and 14 (10.1%) had no tumor progression during follow-up. The overall distribution of post-progression treatment patterns did not differ significantly between the two groups (χ² = 2.45, P = 0.653). The detailed post-progression treatment patterns are summarized in Table 4.

Table 4.

Post-Progression Treatment Patterns in the Propensity Score-Matched Cohort

Post-Progression Treatment	TACE Group (n=139)	HAIC Group (n=139)	χ2	P value
Subsequent systemic therapy	48 (34.5%)	58 (41.7%)	2.45	0.653
Radiotherapy	14 (10.1%)	15 (10.8%)
Additional locoregional therapy	22 (15.8%)	18 (12.9%)
Best supportive care	35 (25.2%)	34 (24.5%)
No tumor progression	20 (14.4%)	14 (10.1%)

Prognostic Factors Associated with OS and PFS

Univariable and multivariable Cox proportional hazards regression analyses for overall survival (OS) and progression-free survival (PFS) were performed in the overall (unmatched) cohort before propensity score matching, and the results are summarized in Table 5 and Table 6. In the multivariable model for OS, treatment (HAIC vs TACE; HR, 0.53; 95% CI, 0.39–0.71; P < 0.001), PVTT classification (III–IV vs I–II; HR, 1.53; 95% CI, 1.11–2.12; P = 0.010), Child–Pugh class (B vs A; HR, 1.60; 95% CI, 1.11–2.29; P = 0.011), tumor number (≥4 vs <4; HR, 1.53; 95% CI, 1.14–2.05; P = 0.005), extrahepatic metastasis (yes/no; HR, 1.56; 95% CI, 1.11–2.20; P = 0.010), and AFP (≥400 vs <400 ng/mL; HR, 1.39; 95% CI, 1.02–1.88; P = 0.035) were identified as independent prognostic factors. For PFS, the multivariable analysis demonstrated that treatment (HAIC vs TACE; HR, 0.657; 95% CI, 0.521–0.829; P = 0.001), Child–Pugh class (B vs A; HR, 1.71; 95% CI, 1.28–2.27; P = 0.001), tumor number (≥4 vs <4; HR, 1.31; 95% CI, 1.04–1.65; P = 0.022), and AFP (≥400 vs <400 ng/mL; HR, 1.55; 95% CI, 1.21–1.98; P = 0.001) remained independently associated with PFS, whereas extrahepatic metastasis showed a borderline association (HR, 1.28; 95% CI, 0.98–1.67; P = 0.066).

Table 5.

Univariable and Multivariable Analysis for Predictors of OS

Variables	Univariable Analysis		Multivariable Analysis
	HR (95% CI)	P	HR (95% CI)	P
Age (≥65 vs <65)	0.92(0.61–1.39)	0.695	–	–
Sex (Male vs Female)	1.06(0.68–1.68)	0.79	–	–
Treatment (HAIC vs TACE)	0.65(0.49–0.87)	0.004	0.53(0.39–0.71)	<0.001
PVTT classification (III–IV vs I–II)	1.42(1.04–1.93)	0.001	1.53(1.11–2.12)	0.010
Hepatic vein invasion (Yes vs No)	0.99(0.64–1.52)	0.948
HBsAg (Negative vs Positive)	1.27(0.83–1.93)	0.718	–	–
HBV-DNA (≥2000 vs <2000)	0.86(0.63–1.61)	0.322		–
ECOG-PS (1 vs 0)	1.02(0.69–1.50)	0.94	–	–
Child-Pugh (B vs A)	1.64(1.15–2.34)	0.007	1.60(1.11–2.29)	0.011
Tumor number (≥4 vs <4)	1.48(1.11–1.98)	0.007	1.53(1.14–2.05)	0.005
Tumor size, cm (≥10 vs <10)	1.17(0.87–1.56)	0.307	–	–
Extrahepatic metastasis (Yes vs No)	1.49(1.07–2.08)	0.019	1.56(1.11–2.20)	0.010
AFP, ng/mL (≥400 vs <400)	1.38(1.03–1.84)	0.029	1.39(1.02–1.88)	0.035
PIVKA-II, ng/mL (≥40 vs <40)	0.86(0.55–1.36)	0.521	–	–

Table 6.

Univariable and Multivariable Analysis for Predictors of PFS

Variables	Univariate Analysis		Multivariate Analysis
	HR (95% CI)	P	HR (95% CI)	P
Age (≥65 vs <65)	0.87(0.62–1.22)	0.407	–	–
Sex (Male vs Female)	1.20(0.84–1.73)	0.323	–	–
Treatment (HAIC vs TACE)	0.70(0.56–0.88)	0.002	0.657(0.521–0.829)	0.001
PVTT classification (III–IV vs I–II)	1.15(0.90–1.46)	0.266
Hepatic vein invasion (Yes vs No)	1.28(0.93–1.76)	0.128
HBsAg (Negative vs Positive)	0.85(0.63–1.16)	0.316	–	–
HBV-DNA (≥2000 vs <2000)	1.03(0.82–1.30)	0.799	–	–
ECOG-PS (1 vs 0)	0.92(0.69–1.23)	0.583	–	–
Child-Pugh (B vs A)	1.61(1.22–2.14)	0.001	1.71(1.28–2.27)	0.001
Tumor number (≥4 vs <4)	1.27(1.01–1.60)	0.038	1.31(1.04–1.65)	0.022
Tumor size, cm (≥10 vs <10)	1.30(1.03–1.63)	0.026	1.13(0.89–1.43)	0.315
Extrahepatic metastasis (Yes vs No)	1.35(1.04–1.75)	0.023	1.28(0.98–1.67)	0.066
AFP, ng/mL (≥400 vs <400)	1.53(1.22–1.93)	0.001	1.55(1.21–1.98)	0.001
PIVKA II, ng/mL (≥40 vs <40)	0.95(0.65–1.40)	0.798	–	–

Given that patients in the HAIC-based group received more locoregional treatment sessions than those in the TACE-based group, additional Cox regression analyses with time-dependent covariates were performed to account for dynamic differences in treatment exposure during follow-up. The results of the time-dependent Cox analyses are summarized in Table 7. After incorporating the cumulative number of locoregional treatment sessions as a time-dependent covariate, HAIC-based therapy remained independently associated with improved OS compared with TACE-based therapy (HR = 0.61, 95% CI: 0.44–0.84, P = 0.003). The cumulative number of locoregional treatment sessions was also independently associated with a reduced risk of death (HR = 0.82, 95% CI: 0.72–0.94, P = 0.004). Similar findings were observed for PFS. HAIC-based therapy remained independently associated with improved PFS after adjustment for cumulative locoregional treatment sessions as a time-dependent covariate (HR = 0.73, 95% CI: 0.57–0.93, P = 0.011). In addition, a greater cumulative number of locoregional treatment sessions was associated with a lower risk of progression or death (HR = 0.86, 95% CI: 0.77–0.96, P = 0.009).

Table 7.

Time-Dependent Cox Regression Analyses for OS and PFS

Variables	OS HR	95% CI	P value	PFS HR	95% CI	P value
Treatment, HAIC vs TACE	0.61	0.44–0.84	0.003	0.73	0.57–0.93	0.011
Cumulative locoregional treatment sessions, time-dependent	0.82	0.72–0.94	0.004	0.86	0.77–0.96	0.009
PVTT classification, III–IV vs I–II	1.46	1.05–2.03	0.025	1.18	0.92–1.52	0.191
Child-Pugh class, B vs A	1.55	1.07–2.25	0.021	1.63	1.21–2.20	0.001
Tumor number, ≥4 vs <4	1.47	1.08–1.99	0.014	1.28	1.01–1.63	0.043
Extrahepatic metastasis, yes vs no	1.48	1.04–2.10	0.029	1.23	0.94–1.61	0.129
AFP, ≥400 vs <400 ng/mL	1.34	0.98–1.82	0.067	1.49	1.16–1.92	0.002

Notes: Cumulative locoregional treatment sessions were modeled as a time-dependent covariate, and the corresponding HR represents the change in hazard associated with each additional treatment session.

Subgroup Analysis

Subgroup analyses for both OS and PFS were performed in the overall cohort before propensity score matching to preserve sample size within each stratum, and formal treatment-by-subgroup interaction tests were conducted using Cox proportional hazards models (Figure 2). Overall, HAIC-based therapy was associated with lower risks of death and disease progression than TACE-based therapy across most prespecified subgroups.

Figure 2.

Subgroup analyses of overall survival and progression-free survival in the overall cohort before propensity score matching. (A) Overall survival. (B) Progression-free survival. Hazard ratios compare HAIC-based with TACE-based triple therapy. P values for interaction were obtained from treatment-by-subgroup interaction tests.

In patients with advanced PVTT, the point estimates suggested a numerically greater association between HAIC-based therapy and improved survival. However, the treatment-by-PVTT classification interaction did not reach statistical significance for either OS or PFS (P for interaction = 0.057 and 0.064, respectively). Therefore, the apparent greater benefit in patients with advanced PVTT should be interpreted as exploratory rather than confirmatory evidence of treatment-effect modification.

Similar favorable point estimates for HAIC were observed in several clinically relevant subgroups, including patients with elevated AFP, patients with ≥4 tumors, larger tumor size, poorer liver reserve, and extrahepatic metastasis. Nevertheless, because these subgroup analyses were exploratory and multiple interaction tests were performed, the results should be interpreted cautiously and considered hypothesis-generating. Across the remaining subgroups, hazard ratios generally favored HAIC-based therapy, although statistical significance was not consistently observed in smaller strata.

Outcomes and Pathological Findings After Successful Conversion Therapy

We evaluated the conversion-to-surgery rates in the TACE and HAIC groups. The conversion rates were 17.4% (28/161) and 27.3% (54/198), respectively. Chi-square analysis showed that the conversion rate was significantly higher in the HAIC group than in the TACE group (χ² = 4.92, P = 0.027). Among the 82 patients who successfully underwent conversion surgery, no significant differences were observed between the TACE and HAIC groups in terms of pathological response (P = 0.446), microvascular invasion (P = 0.608), or perioperative blood transfusion (P = 1.000). Although a higher proportion of patients in the HAIC group achieved pathological complete response (31.48% vs, 17.86%), the difference did not reach statistical significance. Operative time and intraoperative blood loss were also comparable between the two groups (all P > 0.05). No grade 3 or 4 adverse events were observed during the perioperative period.

The pathological response and perioperative outcomes after conversion surgery are summarized in Table 8.

Table 8.

Pathological Response and Perioperative Outcomes After Conversion Surgery in the TACE and HAIC Groups

Variables	Overall (n=82)	TACE Group (n=28)	HAIC Group (n=54)	Statistic	P value
Pathological response, n (%)
No response	6(7.32%)	3(10.71%)	3 (5.56%)	χ2=2.665	0.446
Partial response	33(40.24%)	11(39.29%)	22 (40.74%)
MPR	21(25.61%)	9(32.14%)	12(22.22%)
pCR	22(26.83%)	5(17.86%)	17 (31.48%)
Microvascular invasion (MVI), n (%)
M0	53 (64.63%)	17 (60.71%)	36 (66.67%)	χ2=0.99	0.608
M1	16 (19.51%)	5 (17.86%)	11 (20.37%)
M2	13 (15.85%)	6 (21.43%)	7 (12.96%)
Perioperative blood transfusion, n (%)
No	70(85.37%)	24(85.71%)	46(85.19%)	χ2=0.001	1
Yes	12(14.63%)	4(14.29%)	8(14.81%)
Operative time (min), mean ± SD	248.71 ± 92.89	256.68 ± 95.28	244.57 ± 92.26	t=0.56	0.579
Intraoperative blood loss (mL), mean ± SD	366.10 ± 267.99	409.29 ± 254.53	343.7 ± 274.33	t=1.05	0.296

Notes: pCR was defined as 0% residual viable tumor. For categorical reporting, MPR was defined as >0% and ≤10% residual viable tumor, partial response as >10% and ≤50% residual viable tumor, and no response as >50% residual viable tumor.M0 indicated no microvascular invasion; M1 and M2 were classified according to the number and distance of microvascular invasion foci from the tumor.

Abbreviations: pCR, pathological complete response; MPR, major pathological response; MVI, microvascular invasion.

Adverse Events

During treatment, treatment delay occurred in 30 patients (18.63%) in the TACE group and 34 patients (17.17%) in the HAIC group, with no significant difference between groups (χ² = 0.13, P = 0.719). Treatment discontinuation occurred in 19 patients (11.80%) and 18 patients (9.09%) in the TACE and HAIC groups, respectively, and the difference was also not statistically significant (χ² = 0.71, P = 0.401). The main reasons for treatment delay or discontinuation included ascites, hematologic toxicity, gastrointestinal intolerance, liver function deterioration, infection, and patient refusal. These findings suggested that treatment feasibility and tolerability were generally comparable between the two groups.

The incidence of adverse events was comparable between the TACE plus lenvatinib plus PD-1 inhibitor group and the HAIC plus lenvatinib plus PD-1 inhibitor group. The overall rate of any adverse events was 88.82% in the TACE group and 89.90% in the HAIC group, with no significant difference between the two groups (P = 0.741). Likewise, no significant between-group differences were observed in abdominal pain (57.14% vs, 58.59%, P = 0.783), fever (40.37% vs, 42.93%, P = 0.625), nausea and vomiting (42.86% vs, 41.92%, P = 0.858), ALT and AST level increase (55.28% vs, 47.47%, P = 0.141), hypothyroidism (16.15% vs, 13.64%, P = 0.504), weight loss (25.47% vs, 21.21%, P = 0.342), proteinuria (21.74% vs, 19.70%, P = 0.634), or hand-foot syndrome (24.22% vs, 23.23%, P = 0.826). However, hypertension occurred more frequently in the HAIC group than in the TACE group (52.02% vs, 38.51%, P = 0.011). These findings suggest that the safety profiles of the two treatment regimens were generally comparable, except for a higher incidence of hypertension in the HAIC group. The treatment-related adverse events are summarized in Table 9.

Table 9.

Treatment-Related Adverse Events

Adverse Events, n (%)	TACE+Lenvatinib+PD-1 (n=161)	HAIC+Lenvatinib+PD-1 (n=198)	χ2 value	P value
Any adverse events	143(88.82%)	178(89.90%)	0.11	0.741
Abdominal pain	92(57.14%)	116(58.59%)	0.08	0.783
Fever	65(40.37%)	85(42.93%)	0.24	0.625
Nausea and vomiting	69(42.86%)	83(41.92%)	0.03	0.858
ALT and AST level increase	89(55.28%)	94(47.47%)	2.16	0.141
Hypothyroidism	26(16.15%)	27(13.64%)	0.45	0.504
Hypertension	62(38.51%)	103(52.02%)	6.53	0.011
Weight loss	41(25.47%)	42(21.21%)	0.9	0.342
Proteinuria	35(21.74%)	39(19.70%)	0.23	0.634
Hand-foot syndrome	39(24.22%)	46(23.23%)	0.05	0.826

Discussion

PVTT is a well-recognized adverse prognostic factor in HCC and is associated with markedly poor survival, with a median survival time of only 2.7–4 months.²³ Although targeted therapy combined with immunotherapy is currently recommended as a first-line treatment for PVTT-HCC,^24,25 locoregional therapies may provide additional clinical benefit.²⁶ TACE has long been used in unresectable HCC and has shown survival benefit in selected patients,^27,28 while HAIC has also demonstrated promising efficacy, particularly in advanced HCC with PVTT.^23,29 These findings suggest that combining locoregional treatment with systemic therapy may be a rational strategy for PVTT-HCC. Therefore, this study was conducted to compare the efficacy of TACE plus targeted therapy and immunotherapy with that of HAIC plus targeted therapy and immunotherapy in patients with PVTT-HCC.

After propensity score matching (PSM), Kaplan–Meier analyses were performed to compare overall survival (OS) and progression-free survival (PFS) between the two groups. The HAIC plus targeted–immunotherapy group achieved significantly longer median OS (23.7 months) and median PFS (7.7 months) than the TACE plus targeted–immunotherapy group (OS, 17.1 months; PFS, 5.4 months). Several factors may account for these differences. First, prior studies have suggested that, for huge HCC (≥10 cm), HAIC can yield a higher objective response rate than TACE.³⁰ In addition, a retrospective study reported that HAIC was associated with significantly prolonged OS and PFS compared with TACE in patients with Vp3 or Vp4 PVTT-HCC.³¹ Collectively, these findings are consistent with our results and lend support to the potential advantage of HAIC-based targeted–immunotherapy in this high-risk population.

Second, a more favorable hepatic safety profile with HAIC may represent another important explanation for the OS benefit. In PVTT-HCC, portal venous inflow is compromised, and the embolic effect of TACE may further reduce perfusion to non-tumorous liver parenchyma, thereby increasing the risk of severe hepatic ischemia and even acute liver failure. By contrast, HAIC delivers chemotherapy via continuous arterial infusion without direct arterial embolization, which may lessen injury to the residual functional liver and better preserve hepatic reserve, ultimately translating into improved survival. Notably, liver function status is a critical determinant of prognosis in HCC: even under the same first-line regimen (eg, sorafenib), median survival differs markedly across Child–Pugh classes (A, 13.6 months; B, 5.2 months; C, 2.7 months).³² Although direct evidence demonstrating less hepatic impairment with HAIC than with TACE in PVTT-HCC remains limited, we observed lower rates of AST/ALT elevation in the HAIC-based group than in the TACE-based group, suggesting that HAIC may, to some extent, mitigate liver injury. This observation is in line with a previous comparative study,³¹ which reported a higher incidence of grade ≥3 hepatic toxicity in the TACE group and one death attributable to embolization-related acute liver failure. Nevertheless, further well-designed prospective studies are warranted to validate safety differences between HAIC and TACE in PVTT-HCC and to clarify whether these approaches differ meaningfully in their impact on hepatic function.

Importantly, the comparative efficacy of TACE-based and HAIC-based interventional strategies remains controversial, and not all published studies have consistently demonstrated a clear survival superiority of HAIC.³³ Some comparative studies have reported comparable outcomes between TACE- and HAIC-based approaches,³⁴ particularly after adjustment for baseline characteristics or in selected patient populations. These discrepant findings may be attributable to differences in PVTT extent, intrahepatic tumor burden, liver function reserve, embolization selectivity, HAIC regimens, systemic treatment combinations, subsequent therapies, and the number and timing of locoregional treatment sessions. Therefore, although our results suggest that HAIC-based triple therapy was associated with improved survival outcomes in this cohort, they should not be interpreted as definitive evidence of superiority across all PVTT-HCC populations. Further prospective multicenter studies are needed to define which patients are most likely to benefit from HAIC-based rather than TACE-based combination strategies.

We performed multivariable Cox regression analysis in the overall cohort of 359 patients with PVTT-HCC. The results demonstrated that interventional modality, PVTT classification, Child–Pugh class, tumor number, extrahepatic metastasis, and AFP level were independently associated with overall survival (OS). Although targeted therapy and immunotherapy have rapidly evolved in the management of advanced HCC in recent years, our findings suggest that the choice of interventional strategy remains a critical determinant of prognosis in patients with PVTT-HCC. As highlighted in a comprehensive review by Llovet et al,³⁵ locoregional therapy continues to play a central role across different stages of HCC management. With the growing body of randomized controlled trials evaluating combinations of locoregional and systemic therapies, integration of immunotherapy with interventional approaches may represent an emerging standard strategy for advanced HCC. In particular, optimization of locoregional treatment strategies appears to remain a key factor influencing survival outcomes in PVTT-HCC. The imbalance in the number of locoregional treatment sessions between groups is an important issue when comparing TACE-based and HAIC-based treatment strategies. Patients in the HAIC-based group received more locoregional treatment sessions, which may reflect differences in treatment feasibility, liver function reserve, disease control, and physician decision-making. If the total number of sessions is analyzed as a fixed covariate, immortal time bias may be introduced because patients must survive long enough to receive subsequent treatments. Therefore, we incorporated cumulative locoregional treatment sessions as a time-dependent covariate. This analysis showed that although a greater number of locoregional treatment sessions was associated with improved OS and PFS, HAIC-based therapy remained independently associated with better survival outcomes after accounting for this dynamic treatment exposure. These results suggest that the survival advantage of HAIC-based therapy was not solely explained by greater treatment intensity.

PVTT classification was also identified as an important prognostic factor. A previous prognostic modeling study³⁶ demonstrated that patients with main trunk (Vp4) PVTT had significantly worse survival than those with branch-type (Vp1–Vp3) PVTT. Similarly, a multicenter cohort study³⁷ reported that main trunk PVTT was an independent adverse prognostic factor in HCC. These findings are consistent with the results of the present study.

Elevated AFP levels were likewise identified as unfavorable factors for both OS and PFS. AFP ≥400 ng/mL is widely used for diagnostic stratification in HCC and has been associated with more aggressive tumor behavior.³⁸ In our cohort, the proportion of patients with AFP ≥400 ng/mL reached 58%, likely reflecting the advanced disease stage and frequent vascular invasion in this population.³⁹ Overall, high AFP levels may indicate more aggressive tumor biology, more advanced PVTT classification, and poorer survival outcomes.

Transarterial chemoembolization (TACE) is performed by selectively delivering a mixture of chemotherapeutic agents and embolic materials into the hepatic artery, thereby exerting dual antitumor effects. First, it enables high intratumoral drug exposure, which may enhance local efficacy while limiting systemic toxicity. Second, embolization of tumor-feeding arterial branches reduces tumor perfusion, leading to hypoxia and ischemic necrosis. When combined with targeted therapy and immunotherapy, TACE-induced tumor necrosis may facilitate antigen release and thereby augment the immune response mediated by PD-1 blockade.⁴⁰ Under physiological conditions, the liver has a dual blood supply from the hepatic artery and portal vein; normal hepatic parenchyma is predominantly supplied by the portal vein, whereas HCC largely depends on arterial perfusion. This provides the biological rationale for TACE to selectively target tumor tissue while relatively preserving non-tumorous liver function. However, in patients with PVTT-HCC, portal venous flow is already compromised by tumor thrombus, and additional arterial embolization may further aggravate hepatic hypoperfusion, increasing the risk of severe hepatic ischemia and even acute liver failure. Moreover, the efficacy of TACE in PVTT-HCC may be limited by the presence of hepatic arterioportal venous fistula (HAPVF), which can divert embolic materials from the hepatic artery directly into the portal vein during the procedure, thereby attenuating the intended embolization effect.⁴¹

Hepatic arterial infusion chemotherapy (HAIC) has been incorporated into treatment algorithms for advanced HCC in several Asian countries, including Korea and Japan, and is regarded as an important locoregional approach. Compared with systemic chemotherapy, HAIC is theoretically advantageous in several respects. Continuous intra-arterial infusion delivers high concentrations of cytotoxic agents directly to intrahepatic tumors, potentially treating not only radiographically apparent lesions but also occult intrahepatic micrometastases or lesions with atypical arterial supply.^42,43 In addition, owing to hepatic first-pass metabolism, a substantial fraction of the infused agents is metabolized within the liver before entering the systemic circulation, which may reduce systemic exposure and treatment-related adverse events.^44,45 Taken together, these mechanistic considerations suggest that HAIC may be particularly suitable for patients with PVTT-HCC, in whom preservation of hepatic perfusion and functional reserve is critical. In the context of targeted–immunotherapy, more effective intrahepatic cytoreduction with less compromise of liver function could translate into improved long-term outcomes.^46,47 Consistent with this rationale, our PSM-adjusted analyses showed longer OS and PFS in the HAIC-based group compared with the TACE-based group.

From an immunological perspective, the interaction between locoregional therapy and immune checkpoint blockade may also contribute to the observed clinical benefit. A recent review emphasized that locoregional therapies can enhance the efficacy of immune checkpoint inhibitors by promoting tumor antigen release, increasing immunogenicity, and remodeling the tumor immune microenvironment.⁴⁸ TACE-induced tumor necrosis may facilitate the release of tumor-associated antigens and promote immune activation, thereby providing a rationale for combination with PD-1 blockade. In addition, TACE has been reported to reduce exhausted CD8+PD-1+ effector cells and regulatory T cells, potentially converting an immunosuppressive microenvironment into a more immunosupportive state. However, TACE-induced embolization may also aggravate tumor hypoxia, which can upregulate HIF-1α and PD-L1 expression and promote immunosuppressive signaling. In contrast, HAIC delivers repeated high-concentration chemotherapy through the hepatic artery without direct arterial embolization. This approach may induce tumor cytoreduction and chemotherapy-associated immunogenic cell death while better preserving hepatic perfusion and avoiding excessive hypoxia-driven immunosuppression. Therefore, HAIC-based therapy may provide a distinct immunological context for PD-1 blockade in PVTT-HCC, characterized by effective intrahepatic tumor control, preserved liver function, and potentially more sustained exposure to lenvatinib and immunotherapy. Nevertheless, because immune-cell infiltration, antigen presentation, PD-L1 expression, T-cell exhaustion, and macrophage polarization were not directly evaluated in the present study, these mechanisms remain speculative and should be investigated in future translational studies using paired tumor samples, multiplex immunofluorescence, spatial transcriptomics, or single-cell sequencing.

Our subgroup analyses provide exploratory insights into patient selection for interventional strategies combined with targeted–immunotherapy. Because these analyses were performed in the overall cohort before propensity score matching to preserve sample size within each stratum, the findings should be interpreted cautiously. Although hazard ratios generally favored HAIC-based therapy across most prespecified subgroups, the point estimates suggesting numerically greater benefits in patients with high-risk tumor burden, including advanced PVTT, AFP ≥400 ng/mL, multiple tumors, larger tumors, ALBI grade 2, and extrahepatic metastasis, should not be regarded as definitive evidence of treatment-effect modification. In particular, formal interaction testing showed that the treatment-by-PVTT classification interaction did not reach statistical significance for either OS or PFS, although a borderline trend was observed. From a biological perspective, a potential advantage of HAIC in patients with high intrahepatic tumor burden remains plausible, as HAIC can deliver sustained intra-arterial chemotherapy exposure and achieve effective intrahepatic cytoreduction without further compromising portal venous inflow, which is already impaired in PVTT-HCC. Preservation of hepatic reserve may also be important for maintaining multimodality therapy, particularly in patients with poorer liver reserve or extensive intrahepatic disease.^49,50 Nevertheless, because of residual confounding, limited sample sizes within individual strata, and the increased risk of type I error from multiple subgroup comparisons without multiplicity adjustment, these subgroup findings should be considered hypothesis-generating and require prospective validation.

In the present cohort, the overall safety profiles of TACE plus lenvatinib plus a PD-1 inhibitor and HAIC plus lenvatinib plus a PD-1 inhibitor were largely comparable. The incidence of any-grade adverse events was similar between groups (88.82% vs, 89.90%, P = 0.741), and no significant between-group differences were observed in common events including abdominal pain, fever, nausea/vomiting, transaminase elevation (ALT/AST), hypothyroidism, weight loss, proteinuria, or hand–foot syndrome. Notably, however, hypertension occurred more frequently in the HAIC group than in the TACE group (52.02% vs, 38.51%, P = 0.011), highlighting the need for vigilant blood pressure monitoring and management in patients receiving HAIC-based targeted–immunotherapy. This finding should be interpreted carefully because hypertension is a well-recognized adverse event associated with lenvatinib and VEGF/VEGFR pathway inhibition rather than a direct toxicity of HAIC itself. Several factors may explain the higher incidence of hypertension observed in the HAIC group. First, lenvatinib was routinely interrupted before and after each TACE session, whereas patients receiving HAIC may have received more continuous systemic therapy. Second, the longer disease control observed in the HAIC group may have resulted in a longer duration of lenvatinib exposure, thereby increasing the likelihood of hypertension being detected during follow-up. Third, although propensity score matching balanced major baseline tumor- and liver function-related variables, residual imbalances in baseline cardiovascular risk factors, pre-existing hypertension, antihypertensive medication use, or cumulative lenvatinib dose intensity could not be completely excluded. Therefore, the higher incidence of hypertension in the HAIC group should be regarded as a clinically meaningful safety signal requiring careful blood pressure monitoring and timely antihypertensive management. To date, high-quality evidence directly comparing the safety of TACE- versus HAIC-based regimens in combination with targeted therapy and immunotherapy remains limited. Nevertheless, a previous meta-analysis comparing HAIC and TACE in unresectable HCC reported higher rates of adverse events such as transaminase elevation, fever, and hyperbilirubinemia in the TACE group.⁵¹ While our findings are directionally consistent in that HAIC-based therapy did not show an overall worse tolerability profile, the retrospective design and the potential for under-reporting or heterogeneity in adverse-event documentation warrant cautious interpretation. Larger prospective studies are needed to further clarify the long-term safety and tolerability of these multimodality strategies.

Several limitations of this study should be acknowledged. First, this was a retrospective single-center study and was therefore subject to inherent selection bias. In addition, although we clarified the clinical and laboratory criteria for treatment delay, withholding, or discontinuation, the selection of locoregional treatment modality was ultimately influenced by physician judgment, patient tolerance, and multidisciplinary evaluation. Therefore, residual treatment-selection bias could not be completely avoided. Although propensity score matching was performed to reduce baseline imbalances between the two groups, residual confounding from unmeasured variables could not be completely excluded.

Second, the subgroup analyses were exploratory in nature, and formal interaction testing was performed. Although several nominally significant treatment-by-subgroup interactions were observed for OS, these findings were not consistently reproduced for PFS and were not adjusted for multiple comparisons. Therefore, the apparent differences in treatment effects across subgroups should be interpreted cautiously and considered hypothesis-generating rather than confirmatory.

Third, post-progression treatments were not prospectively standardized and may have influenced overall survival. In the propensity score-matched cohort, we additionally compared subsequent systemic therapy, radiotherapy, additional locoregional therapy, best supportive care, and the proportion of patients without tumor progression between the two groups. The distribution of post-progression treatment patterns was comparable between the TACE and HAIC groups. Nevertheless, because the choice of subsequent therapy was determined by tumor progression pattern, liver function, patient preference, physician judgment, and treatment availability, residual confounding related to post-progression management could not be completely excluded.

Fourth, adverse events were assessed retrospectively from medical records, which may have resulted in under-reporting or inconsistencies in toxicity assessment.

Fifth, the number of locoregional treatment sessions differed between the two groups and may have influenced treatment outcomes. This difference was likely related, at least in part, to the distinct treatment strategies themselves, as the number of treatment sessions was determined during follow-up rather than being a baseline characteristic. Therefore, this variable was not included in the propensity score matching model. Although locoregional treatment sessions were modeled as a time-dependent covariate, residual confounding related to treatment selection could not be fully excluded.

Sixth, although eligibility for conversion surgery was assessed according to predefined surgical feasibility criteria and multidisciplinary evaluation, the conversion-to-surgery endpoint remained susceptible to selection bias because patients with better tumor response, preserved liver function, and better overall condition were more likely to undergo surgery. Therefore, the higher conversion rate observed in the HAIC group should be interpreted cautiously.

Seventh, this study did not include paired tumor samples or translational analyses of the tumor immune microenvironment. Consequently, the potential differential effects of HAIC and TACE on antigen release, immune-cell infiltration, PD-L1 expression, T-cell exhaustion, regulatory T cells, myeloid-derived suppressor cells, and tumor-associated macrophage polarization could not be directly evaluated.

Prospective multicenter studies with standardized treatment and surgical assessment protocols are warranted to validate these findings and refine patient selection for different locoregional treatment strategies.

Conclusions

In this retrospective single-center study, HAIC plus lenvatinib and PD-1 inhibitors was associated with better OS and PFS than TACE plus lenvatinib and PD-1 inhibitors in patients with HCC and PVTT. The potential benefit of HAIC-based triple therapy appeared more evident in clinically high-risk subgroups; however, subgroup findings should be interpreted as exploratory. Prospective multicenter studies are needed to confirm the optimal locoregional treatment strategy to combine with systemic therapy in this population.

Funding Statement

This study was supported by the Multicenter Major Project of the Research Fund for Hepatobiliary and Pancreatic Malignancies of the Hubei Chen Xiaoping Science and Technology Development Foundation (Grant No. CXPJJH11800001-2018104). The funder had no role in the study design, data collection, data analysis or interpretation, manuscript preparation, or the decision to submit the manuscript for publication.

Abbreviations

AFP, alpha-fetoprotein; ALBI, albumin–bilirubin; DCR, disease control rate; HAIC, hepatic arterial infusion chemotherapy; HCC, hepatocellular carcinoma; ORR, objective response rate; OS, overall survival; PD-1, programmed cell death protein 1; PFS, progression-free survival; PSM, propensity score matching; PVTT, portal vein tumor thrombus; TACE, transarterial chemoembolization.

Data Sharing Statement

The original clinical data generated and/or analyzed during the current study are stored in secure institutional data systems at Tongji Hospital, including the YiduCloud platform and the hospital cloud desktop system. Owing to patient privacy protection requirements, institutional data-security policies, and ethical restrictions, the complete dataset cannot be publicly deposited or exported from these systems. De-identified data that do not contain identifiable patient information may be made available by the corresponding author upon reasonable request, subject to institutional and ethical approval.

Ethics Approval and Informed Consent

This retrospective study was conducted in accordance with the Declaration of Helsinki. The study protocol was reviewed and approved by the Institutional Review Board/Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (Approval No. TJ-IRB202511038). The requirement for written informed consent was waived by the Institutional Review Board/Ethics Committee because of the retrospective nature of the study.

Author Contributions

Jun Yu contributed to the conception and design of the study, data acquisition and curation, statistical analysis and interpretation, preparation of figures and tables, and drafting and revision of the manuscript. Yiran Xu and Xuehan Shen contributed to clinical data collection, patient follow-up, and data verification. Jinpeng Wang and Tianyin Shao contributed to data acquisition, literature review, and interpretation of the clinical findings. Yu Wu and Hanlong Hu contributed to the review of imaging and clinical data, data validation, and revision of the manuscript. Qi Cheng contributed to the study design, methodological guidance, interpretation of the results, and critical revision of the manuscript. Zhiwei Zhang conceived and supervised the study, provided overall scientific and clinical guidance, acquired funding, and critically revised the manuscript. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests.