The study design and analytical approach are schematically represented in Figure 1 . Initially, 60 patients who were diagnosed with hepatocellular carcinoma (HCC) complicated by hepatic arteriovenous fistula at Shandong Cancer Hospital and Institute from January 2021 to June 2023 were identified for this study. After applying the inclusion and exclusion criteria, 40 HCC patients were ultimately included. The medical record data of the patients were collected through retrospective analysis. This study was approved by the Ethical Review Committee of the Shandong Cancer Hospital and Institute. Informed consent was waived because this study was retrospective. All data were anonymized and handled in strict compliance with confidentiality regulations. This study was conducted in accordance with the Declaration of Helsinki.

The diagnosis of HCC was rigorously confirmed in all cases using either well-established American Association for the Study of Liver Diseases (AASLD) or the European Association for the Study of the Liver (EASL) (12), or through conclusive pathological examination of liver tissue. Hepatic arteriovenous fistulas (HAVF), comprising arterial-portal shunts (APS) and arterial-hepatic venous shunts (AVS), can be diagnosed and classified through DSA or contrast-enhanced CT (13, 14). HAVF evaluation and severity classification were accomplished via DSA imaging, incorporating a standardized grading approach based on existing studies (15–17), as presented, as follows: The severity grading employs a 4-level system: grade 0 indicates negligible or minimal restrictive shunting; grade 1 represents shunting affecting local branches (APS to segmental portal vein, AVS to local hepatic venous branches); grade 2 was assigned to shunt reaching the ipsilateral main trunk (APS to ipsilateral main portal vein, AVS to hepatic segmental vein); grade 3 was defined for cases where the shunt reached the main portal vein presenting with hepatofugal portal venous flow shunting affecting contralateral or major veins (AVS to main hepatic vein or inferior vena cava) (6, 17, 18). Both Grade 2 and 3 were classified as high-flow fistulas.

Patients were carefully selected based on the following inclusion criteria (1): age 18 years or older; (2) preserved liver function with Child-Pugh A-B or ALBI grade 1-2; (3) good performance status (ECOG 0-1); (4) presence of high-flow HAVF (grade 2-3) documented by DSA or contrast-enhanced CT; (5) ineligibility for curative surgical resection or ablative procedures; and (6) expected survival of at least 3 months. Patients were excluded if they had (1) incomplete clinical records; (2) insufficient follow-up (<6 months); (3) prior HAIC or systemic treatment; (4) complete main portal vein occlusion with inadequate collateral flow; or (5) concurrent malignancies.

All patients underwent detailed superior mesenteric and hepatic angiography for comprehensive tumor and HAVF evaluation. The femoral artery was punctured using standard Seldinger technique under local anesthesia, with a 5 French catheter subsequently deployed to analyze hepatic arterial configuration and tumor blood supply patterns. Under Digital Subtraction Angiography (DSA) or Cone Beam CT (CBCT) guidance, the operator performed angiography of the celiac trunk, superior mesenteric artery, and infra-diaphragmatic arteries to comprehensively evaluate tumor vascularity. Using guidewire support, a microcatheter was precisely positioned within the main arterial feeders supplying the tumor. Collateral circulation was selectively embolized beforehand to redirect blood flow and improve therapeutic delivery to the target lesion. Any flow detected in the gastroduodenal artery or remote extrahepatic vessels was effectively blocked using coil embolization. After catheter placement, heparinized saline (10,000 U heparin in 100 mL 0.9% sodium chloride) was administered to prevent intracatheter thrombosis. The treatment schedule utilized modified FOLFOX6, featuring oxaliplatin (85 mg/m² delivered over the first 2 hours of day 1), leucovorin (400 mg/m² given during hours 2-4 of day 1), and fluorouracil (400 mg/m² bolus at hour 4, then 2,400 mg/m² as a 46-hour continuous infusion on days 1-2). HAIC treatments continued every 3 weeks until progression or prohibitive toxicity emerged. Dose reductions were acceptable based on hepatic reserve and drug tolerance, adhering to guidelines from prior studies (7, 19).Throughout treatment, patients underwent close monitoring of vital signs and laboratory parameters, with regular imaging assessments. For all HBV-infected patients, baseline viral load testing was performed, and effective antiviral management was implemented when necessary.

TKIs and PD-1 inhibitors were initiated at least 3 days before or after HAIC and maintained until disease progression or prohibitive toxicity developed. Before treatment, patients were fully briefed on therapeutic efficacy and potential adverse events. PD-1 inhibitors were delivered intravenously every 3 weeks over 30-60 minutes at fixed doses: sintilimab 200 mg, camrelizumab 200 mg, and tislelizumab 200 mg. Tyrosine kinase inhibitors (TKIs), including apatinib (250 mg, once daily), donafenib (200 mg, once daily), and lenvatinib (8 mg, once daily), was administered orally. Treatment was halted in the event of any unacceptable or severe adverse event (AE) (grade 3 or higher AE) or any intolerable grade 2 drug-related AE. Treatment was also discontinued when disease progression was observed.

Multiple clinical outcomes were investigated, including tumor response, HAVF improvement, HAVF recanalization, OS, PFS, and AEs. Based on Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1) and Modified RECIST (mRECIST) standards, systematic tumor response evaluation was performed at 4-8 week intervals using MRI and dynamic contrast-enhanced CT to precisely monitor tumor burden changes and necrotic area distribution. The treatment response represented the optimal response observed throughout the follow-up period, categorized as complete response (CR), partial response (PR), progressive disease (PD), or stable disease (SD).The objective response rate (ORR) was determined as the combined total of CR and PR, while the disease control rate (DCR) was computed as the cumulative count of CR, PR, and SD.PFS was calculated as the period from the start of combination therapy to disease progression or death), and OS was defined as the time between the start of the therapy and death from any cause. Improvement in HAVF was a reduction grade one level, while no improvement or progression of the fistula after treatment was deemed ineffective. Adverse events were documented and evaluated according to the Common Terminology Criteria for Adverse Events version 5.0.

Statistical analyses were performed using SPSS software (version 26.0). Parametric continuous variables were expressed as mean ± standard deviation and evaluated through Student’s t-test, whereas non-parametric continuous variables were reported as medians and assessed via Mann-Whitney U-test. Categorical variables were described as counts and percentages, with group comparisons performed using chi-square tests or Fisher’s exact tests when appropriate. Survival analysis was conducted using the Kaplan-Meier approach, with log-rank testing for intergroup comparisons. Both univariate and multivariate Cox proportional hazards analyses were implemented to determine prognostic indicators. Statistical significance was defined as a two-tailed p-value < 0.05.

By the end of follow-up in August 2024, a total of 60 HCC patients met the inclusion criteria. Among them,10 patients with low-flow hepatic arterioportal shunts, 5 patients did not complete imaging examination and evaluation after treatment, 5 patient was lost to follow-up, all of which were excluded ( Figure 1 ). Finally, 40 HCC patients were included in this study, and their baseline characteristics were summarized in Table 1 . Male patients accounted for 72.5% of the study population. Cirrhosis was present in 62.5% of patients. Patients with Vp3-4 portal vein tumor thrombus accounted for 90%(36/40) of the cohort. Most patients (85.0%) were HBsAg positive. Extrahepatic metastasis was observed in 27.5% of patients. The majority of patients (95.0%) had Barcelona Clinic Liver Cancer (BCLC) stage C disease, while 5.0% had BCLC stage B disease. Regarding hepatic arteriovenous fistula (HAVF), 42.5% of patients were classified as grade 2 and 57.5% as grade 3. For HAVF somatotype, 32.5% of patients had arteriovenous shunt (AVS) and 67.5% had arterioportal shunt (APS).

The follow-up duration reached a median of 13.5 months (3.5-16.4 months). During the study period, 150 HAIC treatment sessions were delivered, averaging 3.75 cycles per patient. According to mRECIST evaluation criteria, patients treated with HAIC+TKI+PD-1 inhibitor combination demonstrated CR in 7.5%, PR in 67.5%, and SD in 17.5% of cases, resulting in an ORR of 75.0% and DCR of 92.5%. When assessed using RECIST v1.1 criteria, the ORR and DCR were 42.5% and 92.5%, respectively ( Table 2 , Figures 2a, b ).

During follow-up,18 OS events occurred, with a median OS of 10.4 months (95% CI: 7.4-13.4; Figure 3a ) and a median PFS of 5.5 months (95% CI: 3.4–6.9; Figure 3b ). Univariate analysis revealed several factors significantly associated with overall survival (OS), including extrahepatic metastasis (HR 0.46, 95% CI 0.23-0.92, p=0.030), BCLC staging (HR 1.98, 95% CI 1.09-3.58, p=0.024), HAVF disappearance (HR 0.31, 95% CI 0.17-0.55, p<0.001), and HAVF grade (HR 2.51, 95% CI 1.40-4.48, p=0.002). In the subsequent multivariate analysis, extrahepatic metastasis (HR 0.34, 95% CI 0.17-0.71, p=0.004), BCLC staging (HR 2.00, 95% CI 1.15-3.48, p=0.015), and HAVF disappearance (HR 0.29, 95% CI 0.16-0.55, p<0.001) remained independently predictive of OS.

For progression-free survival (PFS), univariate analysis demonstrated significant associations with extrahepatic metastasis (HR 0.52, 95% CI 0.30-0.89, p=0.019), HAVF disappearance (HR 0.68, 95% CI 0.52-0.89, p=0.005), HAVF grade (HR 3.19, 95% CI 1.66-6.11, p<0.001), and PVTT classification (HR 1.03, 95% CI 1.00-1.06, p=0.034). Multivariate Cox regression analysis confirmed extrahepatic metastasis (HR 0.56, 95% CI 0.32-0.99, p=0.045), HAVF disappearance (HR 0.20, 95% CI 0.05-0.83, p=0.027), and HAVF grade (HR 2.66, 95% CI 1.27-5.57, p=0.009) as independent prognostic indicators for PFS. Other clinical parameters including gender, AFP levels, tumor number, ECOG performance status, and Child-Pugh classification showed no statistically significant correlation with either survival endpoint ( Table 3 ).

We analyzed how HAVF affected survival by examining outcomes in patients with and without HAVF during follow-up. Of patients presenting with HAVF, resolution occurred in 29 patients (72.5%), while HAVF persisted in 11 patients (27.5%). Significant survival differences emerged between groups. Patients without HAVF demonstrated superior survival metrics compared to those with HAVF (median OS: 13.0 months [95% CI, 7.9-18.0] vs 4.6 months [1.8-7.4], P=0.005; median PFS: 7.5 months [4.0-11.0] vs 2.1 months [1.3-4.9], P=0.016; Figures 3c, d ). In multivariate analysis, HAVF grade, extrahepatic spread, and HAVF disappearance were found to be independent prognostic markers associated with PFS and OS ( Table 3 ).

After triple therapy, the efficacy rate of HAVF treatment reached 85.0%(34/40)( Table 4 ). The majority of patients demonstrated significant hemodynamic improvement following treatment intervention. Complete disappearance of HAVF was documented in 29 (72.5%) patients during follow-up imaging studies ( Figures 4 , 5 ). These cases showed complete resolution of the abnormal arteriovenous communication with restoration of normal hepatic vascular architecture.

Concurrent reduction in cancer thrombus was observed in 31 of 36 evaluable patients (86.1%), indicating effective management of the underlying malignant process. Tumor reduction was documented in 17 (42.5%) patients, reflecting the dual therapeutic benefit of the intervention on both vascular and oncologic outcomes.

Among patients presenting with both cancer embolus and HAVF, combined improvement was achieved in all 31 cases (100%). Furthermore, in patients with concurrent tumor burden and HAVF, combined improvement was observed in 16 of 17 cases (94.1%), demonstrating the efficacy of treatment in addressing multiple pathologic processes simultaneously.

Most importantly, recanalization events were completely absent among all patients during follow-up (0/40, 0.0%), indicating sustained therapeutic success. This absence of recurrence suggests that the treatment approach provides sustained vascular stabilization without compromise of long-term patency ( Table 4 ).

Treatment-related mortality was not observed in this study. At least one treatment-related adverse event was reported in 35 patients (87.5%) ( Table 5 ). Treatment- and laboratory-related adverse events (AEs) were evaluated according to CTCAE version 5.0. Hypothyroidism emerged as the most common treatment-related AE (52.5%), followed by hypertension (42.5%), diarrhea (37.5%), and fatigue (27.5%). Other notable treatment-related AEs included proteinuria (20.0%), skin pruritus (17.5%), and decreased appetite (17.5%). Among laboratory abnormalities, thrombocytopenia (42.5%) and decreased albumin (40.0%) were most frequently observed, with AST elevation (32.5%), neutropenia (25.0%), ALT elevation (20.0%), and elevated total bilirubin (20.0%) also commonly reported. Grade 3-4 AEs were less frequent, with hypertension and thrombocytopenia (both 12.5%) being the most common severe events. Additional grade 3-4 AEs included hypothyroidism (7.5%), decreased albumin (7.5%), and several events occurring at 5.0% (skin pruritus, diarrhea, and proteinuria). Interestingly, no cases of immunological hepatitis or immune pneumonia were documented throughout the study period.