This study retrospectively evaluated patients with HCC who underwent LR with MVI at the Hepatobiliary Surgery Department of the First Affiliated Hospital of Chongqing Medical University from 1 January 2020 to 30 June 2024. The study was conducted in accordance with the Declaration of Helsinki, and was approved by the institutional ethics committees of our medical center (K2014-039-01). The study was retrospective and no further patient consent was required.

Patients who met the following criteria were enrolled: (1) HCC stage BCLC 0-B prior to surgery; (2) postoperative pathology confirmed hepatocellular carcinoma with MVI; (3) radical liver resection (negative margins confirmed by pathology); (4) without any preoperative anticancer treatments; (5) HAIC treatment only or no treatment after surgery; (6) no history of other malignancies or autoimmune diseases. Exclusion criteria: (1) R1 resection (postoperative pathology suggesting positive margins) or preoperative imaging suggesting extrahepatic metastases; (2) non-HCC confirmed by postoperative pathology; (3) with MVI negative or macrovascular invasion; (4) Preoperative anti-tumor therapies; (5) Post-operative treatments other than HAIC or HAIC in combination with other anti-tumor therapies; (6) Patients who relapsed or died within 60 days of surgery.

All patients underwent routine preoperative examinations including ultrasound, enhanced electronic CT or enhanced magnetic resonance imaging (MRI) to assess tumor diameter, BCLC stage, resectable extent and residual liver volume. In addition, liver function was assessed using the Child-Pugh classification and cirrhosis using ICG15 in all patients. The hepatectomy method contains non-anatomical resection and anatomical resection, and the surgical technique used depends on the location and distribution of the tumor. Anatomical hepatectomy is the complete resection of the segment of liver with the tumor or the segment of liver limited by the branches of the portal vein of the tumor. Non-anatomical hepatectomy is the resection of the tumor and part of the non-tumor liver parenchyma (24, 25). Radical liver resection was defined as the complete removal of all detected tumors without involving any major branch of the portal or hepatic veins, without invasion of adjacent organs and without lymph node or distant metastasis, and tumor-free margins confirmed by histopathology (10). Postoperative adjuvant modalities, including HAIC, are recommended for all patients with MVI. However, patients will decide whether to receive one or more adjuvant treatments, and the number of treatments, based on their medical compliance, economic status or other social factors. Before receiving adjuvant therapy, patients and their families must be fully informed about the relevant treatment modalities and adverse effects, and sign informed consent forms. Meanwhile, all patients with hepatitis B virus need to be treated with antiviral drugs.

PA-HAIC: Patients were re-evaluated approximately 4–6 weeks post-operatively for blood counts, liver and kidney function, alpha-fetoprotein, CT or MRI enhancement of the epigastric region, and were treated with PA-HAIC after assessment of no tumor recurrence and absence of obvious contraindications. In the Seldinger technique, a hepatic artery catheter is placed through the femoral artery into the appropriate hepatic artery and any suspicious tumor staining in the remaining liver is detected by digital subtraction angiography (DSA) or CT angiography. The catheter was left in place and connected to the infusion pump on the ward. The following chemotherapy drugs are pumped continuously: oxaliplatin 85 mg/m² from 0 to 3 h on day 1, leucovorin 400 mg/m² from 3 to 4.5 h on day 1, 5-fluorouracil 400 mg/m² from 4.5 to 6.5 h on day 1, and 5-fluorouracil 2,400 mg/m² over 46 h from days 1 to 3 (26). The patient had almost complete restriction of movement of the right lower limb during the infusion. At the end of the chemotherapy, the catheter was removed and the puncture site bandaged for 8 h to allow movement of the right lower limb, then the blood and liver functions were re-checked and the patient was discharged from hospital if there were no obvious abnormalities. The interval between two cycles of PA-HAIC was set at 4–5 weeks. If patients were unable to tolerate the treatment due to physical frailty, pain, fever, nausea, vomiting, or myelosuppression, the interval could be appropriately extended and/or the drug dosage adjusted. However, the interval was not allowed to exceed 2 months, and the dosage could not be reduced to less than 75% of the standard dose.

All patients were followed during outpatient visits or hospitalizations. Follow-up was conducted every 1–2 months during the first 6 months and every 3–6 months thereafter. Approximately two months after LR, at least two imaging modalities-ultrasound, CT, or MRI-were performed to evaluate tumor recurrence. Patients with confirmed recurrence at this time point were excluded from further analysis, while those without recurrence continued regular follow-up. During subsequent follow-up, patients underwent routine blood tests, liver function tests, AFP, and abdominal ultrasound. If recurrence was suspected, contrast-enhanced CT or MRI was performed for confirmation. Recurrence was defined as any tumor nodule confirmed by at least two imaging modalities or by histopathological biopsy. The study endpoint was recurrence-free survival (RFS), defined as the time from LR to the diagnosis of tumor recurrence. All patients were followed until 31 January 2025 or until loss to follow-up or death.

Propensity score-matching (PSM) analysis was involved to minimize alternative factors and sampling bias between the two groups. A 1:1 nearest neighbor matching algorithm was used with a caliper width of 0.02. PSM was performed using SPSS 27.0 statistical software (IBM Corp., Armonk, NY, USA).

Statistical analyses were performed with SPSS 27.0. The Shapiro–Wilk test was used to test the normality of continuous variables, and the independent samples t-test was used to detect continuous data that followed the normal distribution, expressed as the mean ± standard deviation. The Mann–Whitney U-test was used to detect continuous data that were not normally distributed, expressed as median (interquartile range, IQR). Categorical data were detected using the chi-squared test, and expressed as numbers (n) and proportions (%). Univariate and multivariate analyses were performed in Cox risk models to identify independent prognostic factors for RFS. Survival analyses were performed using the Kaplan–Meier method, and differences in the survival curves were analyzed using the log-rank test. K-M curves were plotted using R software (version 4.2.1 http://www.r-project.org). P-value <0.05 was considered statistically significant.

A total of 175 patients who met the criteria were included in this study, and the flowchart of patient selection is presented in Figure 1.

The entire patient age range was 56.67 ± 11.27 years, with 156 (88.14%) males and 21 (11.86%) females; 72 patients were treated postoperatively with HAIC, with the majority receiving 1–2 HAIC treatments (83.33%), and 103 patients received no postoperative treatment. In the entire cohort, patients in the PA-HAIC group were younger than those in the LR group (53.58 ± 10.78 years vs. 58.96 ± 11.18 years, p = 0.002), and no differences were seen in other baseline characteristics. After PSM, 67 patients were matched in the PA-HAIC and LR groups, respectively. The age range of all patients was 54.87 ± 10.06 years, with a total of 118 (88.06%) males and 16 (11.94%) females, and most patients in the PA-HAIC group received 1–2 HAIC treatments (82.09%). The baseline characteristics of the patients in the two groups did not differ significantly. A total of 138 cycles of PA-HAIC were administered to 72 patients. No treatment-related deaths or severe adverse events were observed. All adverse events were mild and alleviated with symptomatic treatment. Baseline information for the entire cohort and PSM cohort patients is summarized in Table 1.

The median follow-up time for all patients was 34.00 months (95% CI, 25.98–42.03 months). In the entire cohort, a total of 117 (66.86%) patients relapsed, including a total of 39 (54.67%) patients in the PA-HAIC group and 78 (75.73%) patients in the LR group. The median RFS (mRFS) was 24.00 months (95% CI, 19.90–28.11 months) for all patients, and 33.00 months (95% CI, 29.32–36.68 months) and 15.00 months (95% CI, 11.58–18.51 months) for patients in the PA-HAIC and LR groups, respectively (p < 0.001). The 1-, 2- and 3-year RFS rates were 87.50% (95% CI, 79.86%–95.14%), 75.20% (95% CI, 64.42%–85.98%) and 34.30% (95% CI, 19.01%–49.59%) in the PA-HAIC group and 60.80% (51.20%–70.40%), 31.30% (21.30%–41.30%) and 10.10% (2.65%–17.55%) in the LR group, respectively, with statistically significant differences (p < 0.001) (Figure 2A).

In PSM cohort, a total of 89 (66.42%) patients relapsed, including a total of 38 (56.72%) patients in the PA-HAIC group and 51 (76.12%) patients in the LR group. The mRFS was 27.00 months (95% CI, 22.70–31.30 months) for all patients, and 33.00 months (95% CI, 28.74–37.26 months) and 18.00 months (95% CI, 16.25–19.75 months) for patients in the PA-HAIC and LR groups, respectively (p < 0.001). The 1-, 2- and 3-year RFS rates were 86.50% (95% CI, 78.27%–94.73%), 73.30% (95% CI, 65.07%–81.53%) and 31.10% (95% CI, 22.87%–39.33%) in the PA-HAIC group and 66.10% (57.87%–74.33%), 32.10% (19.95%–44.25%) and 13.50% (3.50%–23.50%) in the LR group, respectively, with statistically significant differences (p < 0.001) (Figure 2B).

Univariate and multivariate Cox regression analyses were performed in both the entire cohort and the PSM cohort in Table 2 and Supplementary Table S1. In the entire cohort, multiple tumor (2.236, 95% CI 1.487, 3.363), and tumor diameter (1.095, 95% CI 1.020, 1.175) were independent risk factors for RFS. In contrast, compared with LR, PA-HAIC (0.369, 95% CI 0.246, 0.553) were identified as independent protective factors for RFS. In the PSM cohort, it was found that age (0.977, 95% CI 0.953, 0.996), multiple tumor (3.002, 95% CI 1.729, 5.211), tumor diameter (1.162, 95% CI 1.055, 1.256), low differentiation (4.064, 95% CI 1.358, 12.164) were the independent risk factors for RFS. PA-HAIC (0.276, 95% CI 0.165, 0.463) were identified as independent protective factors for RFS.

Based on the patients' risk factors for recurrence, patients were classified into combined MVI alone (MVI), MVI + tumor diameter ≥5 cm (MVID), MVI + multiple tumor (MVIN), MVI + tumor diameter ≥5 cm + multiple tumor (MVID + N), and MVI + poor differentiation. Given that the number of patients with MVI + poor differentiation was small in the entire cohort and PSM cohort (5 patients in the PA-HAIC group and 10 in the LR group in the entire cohort, and 3 patients in the PA-HAIC group and 4 patients in the LR group in the PSM cohort), the analysis may be somewhat biased and was therefore not analyzed.

In the entire cohort, although patients with MVI in the PA-HAIC group had a longer mRFS than those in the LR group, there was no statistical difference (PA-HAIC group: 33.00 months, 95% CI 28.67–37.33 months vs. LR group: 27.00 months, 95% CI 22.66–31.34 months, p = 0.200). Considering that the number of patients who relapsed as a percentage of the total subgroup at the follow-up endpoint was lower in the PA-HAIC group than in the LR group (50.00% vs. 67.74%), this finding was required further investigation. For MVID, MVIN and MVID + N, the mRFS of patients in the PA-HAIC group were significantly better than those of the LR group (MVID: 36.00 months, 95% CI 32.51–39.49 months vs. 15.00 months, 95% CI 9.12–20.88 months; MVIN: 38.00 months, 95% CI 21.39–54.61 months vs. 18.00 months, 95% CI 1.99–34.01 months; MVID + N: 24.00 months, 95% CI 5.78–42.22 months vs. 4.00 months, 95% CI 0–8.85 months, respectively). In the PSM cohort, there was a consistent conclusion to be drawn. All of the above data are presented in Figure 3 and Supplementary Figure S1.

It was also found that in the LR groups, the mRFS of patients with MVI, MVI + D/MVI + N and MVID + N progressively decreased and were statistically different. However, no statistical difference was found between patients with MVID and patients with MVIN (Supplementary Table S2). In the PA-HAIC group, there were no statistical differences in mRFS between patients with MVI, MVID, MVIN, and MVID + N (the entire cohort: 33.00 months, 95% CI 28.67–37.33 months; 36.00 months, 95% CI 32.51–39.49 months; 35.00 months, 95% CI 19.97–50.03 months; 24.00 months, 5.78–42.22 months; the PSM cohort: 36.00 months, 95% CI 30.79–41.21 months; 33.00 months, 95% CI 27.98–38.02 months; 32.00 months, 95% CI 24.49–39.51 months; 16.00 months, 95% CI 3.20–28.80 months, respectively), as shown in Figure 4 and Supplementary Figure S2.

Patients were divided into ≤55 and >55 years groups based on the median age of the PSM cohort. The study found significant difference in mRFS between patients aged ≤55 and >55 years in the LR group (13.00 months, 95% CI 6.48–19.52 months vs. 27.00 months, 95% CI 18.12–35.88 months, p = 0.020). In patients aged ≤55 years, PA-HAIC significantly improved patient mRFS (PA-HAIC group: 32.00 months, 95% CI: 27.61–36.39 months vs. LR group: 13.00 months, 95% CI: 6.48–19.52 months, p < 0.001), while in patients aged >55 years, mRFS in the PA-HAIC group was better than in the LR group, but not statistically different (37.00 months, 95% CI: 29.00–45.00 months vs. 27.00 months, 95% CI: 18.12–35.88 months, p = 0.126). However, the results may have been influenced by the lower recurrence rate of patients in the PA-HAIC group compared to the LR group (48.30% vs. 66.70%). In addition, there was no significant difference in mRFS between patients aged ≤55 and >55 years in the PA-HAIC group (32.00 months, 95% CI: 27.61–36.39 months vs. 37.00 months, 95% CI: 29.00–45.01 months, p = 0.130) (Figure 5).

Among the 72 patients who underwent PA-HAIC treatment, 22 patients received one treatment, 38 patients received two treatments, and 12 patients received three or more treatments. The study discovered that patients who received two PA-HAIC treatments had significantly better mRFS than those who received only one PA-HAIC treatment (36.00 months, 95% CI 28.26–43.74 months vs. 31.00 months, 95% CI 21.34–40.66 months, P = 0.045). Although superior to those who received one treatment, the mRFS of patients who received three or more treatments did not show significant difference (35.50 months, 95% CI 31.22–39.68 months vs. 31.00 months, 95% CI 21.34–40.66 months, p = 0.092). Furthermore, patients who underwent three or more treatments exhibited a similar RFS to those who received two treatments (mRFS: 35.50 months, 95% CI 31.22–39.68 months vs. 36.00 months, 95% CI 28.26–43.74 months, p = 0.707). It is important to note that the limited number of patients (only 12 patients) who received three or more treatments may have influenced the results to some degree (Figure 6).

By the end of follow-up, 37 patients in the PA-HAIC group had relapsed and 1 patient had died before relapse; in the LR group, 78 patients had relapsed and no patient had died before relapse. Three patients died after relapse in the PA-HAIC group and five patients died after relapse in the LR group. Patients' treatment patterns after relapse are shown in the Table 3.