This was a historical cohort study of patients with treatmentnaïve multinodular HCC within the Milan criteria between 2013 and 2022 at Asan Medical Center, Seoul, Republic of Korea. A total of 483 patients were analyzed retrospectively (Fig. 1). Patients who met all the following criteria were included 1) age ≥18 years, 2) diagnosed with HCC within the Milan criteria, 3) two or three tumors, 4) preserved or moderately preserved liver function corresponding to Child-Pugh (CP) class A or B7, and 5) underwent LR or TACE as a primary treatment. Patients who met any of the following criteria were excluded 1) LT, radiofrequency ablation (RFA), LR with RFA, best supportive care, or other primary treatments, 2) HCC beyond the Milan criteria or with extrahepatic metastasis at baseline, 3) severely impaired liver function, classified as CP class B8 or above to minimize the impact of deteriorated liver function on survival outcomes, 4) concurrent non-HCC malignancies, and 5) missing variables or insufficient data. Data on patient characteristics and outcomes were obtained from the electronic medical records of Asan Medical Center. HCC was radiologically or histologically diagnosed according to international guidelines.3,4 The clinical information of the study population, including demographics, liver disease etiologies, and laboratory data, was evaluated at diagnosis. Furthermore, we initially assessed tumor characteristics, such as the number and size of the tumors, presence of macrovascular invasion, and invasion of the bile duct at the time of diagnosis, using dynamic computed tomography (CT) or multiphasic magnetic resonance imaging (MRI) for a baseline study. If a patient underwent LR as the initial treatment for HCC, the type of resection was also investigated. After TACE or LR, the clinical status of all patients was assessed using physical examination and laboratory findings during hospitalization. Patients who underwent LR were evaluated for postoperative complications, such as bleeding, wound complications, and infection, using laboratory findings and CT.

Subsequently, patients in the TACE group underwent an initial disease evaluation using CT 4-6 weeks after TACE initiation. If complete remission (CR) was not achieved and incomplete lipiodol deposition or residual HCC was identified, additional TACE was considered. Patients were followed up every 1-3 months for disease evaluation using laboratory findings and imaging, such as CT or MRI. Patients who underwent LR were monitored at regular intervals, every 1-3 months, for postoperative disease evaluation.

Ethical approval was obtained from the Institutional Review Board (IRB) of Asan Medical Center (IRB No. 2019-0932), and the requirement for informed consent was waived owing to the retrospective nature of the study. Moreover, we followed the guidelines for reporting observational epidemiological studies (Supplementary Table 1).

Among the 483 patients, 107 (22.2%) underwent LR as the primary treatment for HCC. We used the Brisbane 2000 Terminology of Liver Anatomy and Resections to classify the type of LR.9 A major hepatectomy was defined as the resection of three or more contiguous or non-contiguous liver segments. At our center, LRs are performed by highly experienced hepatobiliary surgeons with >10 years of expertise in hepatobiliary surgery and liver transplantation. Anatomical resection is the preferred approach over non-anatomical resection when deemed suitable, with the extent of hepatectomy tailored to the estimated remnant liver volume and functional hepatic reserve. Post-hepatectomy biochemical dysfunction (PHBD) was defined as increased international normalized ratio >1.5 or hyperbilirubinemia (total bilirubin >2.9 mg/dL) on or after postoperative day ≥5.10 Surgery-related complications other than PHBD were defined according to the Clavien-Dindo classification.11

Additionally, the 90-day mortality after index surgery was also investigated. The TACE procedure was performed in our hospital as previously described.12 After selective catheterization of the feeding artery using a microcatheter, either 50 mg of adriamycin or cisplatin (2 mg/kg body weight) was infused into the feeding artery, followed by an infusion of an emulsion of iodized oil and embolization with gelfoam slurry.

The first recurrence after treatment initiation was evaluated. In the LR group, recurrence was defined as radiological detection of tumors after LR, regardless of the location. According to the modified response evaluation criteria in solid tumors (mRECIST), the first recurrence after TACE was defined as viable tumor detection after achieving CR.13 mRECIST measures the size of target lesions enhanced in the arterial phase and washed out in the venous phase. Based on mRECIST, CR was defined as the disappearance of any intratumoral arterial enhancement in all target lesions. If residual tumors were observed in the disease evaluation and no new lesions developed after TACE, we considered them to be in non-CR. Therefore, this case was excluded from the assessment of RFS risk for comparative purposes.

The primary outcome of this study was the OS. OS was defined as the duration from the date of LR or TACE to death from any cause, LT, or the last follow-up date, whichever occurred first. We also analyzed the RFS and compared the two treatments. RFS after LR was defined as the duration from LR to recurrence, whereas RFS after TACE was defined as the duration from the achievement of a complete response after TACE to recurrence. OS in the study cohort and RFS in patients who underwent LR and TACE and achieved complete response were analyzed. For patients who underwent LR, postoperative adverse outcomes, which included any liver-related morbidity after LR, including PHBD, ascites, bleeding, wound infection, bile leakage, fluid collection, and 90-day mortality, were also evaluated.

To investigate the possible association between the number of tumors and survival outcomes, we performed a sensitivity analysis based on the number of tumors. The entire patient population was stratified into two subgroups based on the number of tumors at baseline: a subgroup with two tumors and a subgroup with three tumors at baseline. OS and RFS were evaluated and compared between the two treatments in each subgroup.

Regarding baseline characteristics, data were summarized as mean values±standard deviation for continuous variables and numbers with percentages for categorical variables. The t-test or Mann-Whitney U test was used for continuous variables and the chi-square or Fisher’s exact test for categorical variables, as appropriate. OS and RFS rates were estimated using the Kaplan-Meier method. A Cox proportional hazards model was used to identify predictive factors for OS and RFS. We also used propensity score (PS) matching analysis to minimize potential confounding between the two treatment groups in the entire cohort and within subgroups based on number of tumors. Missing values ranging from 2.3% (diabetes) to 0.2% (total cholesterol and alkaline phosphatase) of the baseline laboratory data were imputed using linear interpolation in the MICE package. A 1:2 nearest neighbor matching scheme with a caliper size of 0.2 was used for PS matching. None of the standardized differences between baseline covariates in the matches exceeded 0.2. Statistical analyses were performed using the R software (http://cran.r-project.org/). Statistical significance was set at P<0.05.

The median age of the study population was 68 years, and 85.1% were men. Among the 483 patients, 107 (22.2%) and 376 (77.8%) underwent LR and TACE, respectively, as primary treatments for HCC (Table 1). Chronic hepatitis B was the most common etiology in both groups. Compared with patients in the TACE group, those in the LR group had a significantly lower rate of CP class B7 and significantly higher platelet and serum albumin levels. The LR group had a significantly higher proportion of patients with two tumors than the TACE group.

Of the 107 patients in the LR group, 30 (28.0%) underwent major resection and 77 (72.0%) underwent minor resection. The most common type of major and minor resection were hemihepatectomy and right anterior sectionectomy, respectively.

During a median follow-up of 5.3 years, the OS rates after treatment in the entire cohort at 1, 3, 5, 7, and 10 years were 97.5%, 85.7%, 73.6%, 68.6%, and 60.8%, respectively. Among the 483 included patients, 140 (29.0%) died during the follow-up period and the median OS was not reached.

The OS rates after LR at 1, 3, 5, 7, and 10 years were 100.0%, 97.1%, 90.7%, 85.5%, and 75.8%, respectively. In the TACE group, the OS rates at 1, 3, 5, 7, and 10 years were 96.7%, 82.4%, 68.5%, 63.4%, and 56.4%, respectively (Fig. 2A). The median OS of the LR and TACE groups were also not reached during the follow-up.

During the follow-up period, 76 patients (71.0%) in the LR group experienced recurrence. Among the 376 patients in the TACE group, 237 (63.0%) achieved a CR of the target tumor lesions after TACE. Among them, 190 (80.2%) experienced recurrence after achieving CR (Supplementary Table 2). In the LR group, 75 patients (98.7%) and one patient (1.3%) showed intrahepatic and extrahepatic recurrence (peritoneal seeding), respectively. In the TACE group, the recurrence pattern was mostly intrahepatic (98.4%) and no cases of isolated extrahepatic recurrence were observed. One patient (0.5%) exhibited both intrahepatic and extrahepatic recurrences, whereas two patients (1.1%) died of liver-related causes rather than recurrences.

The RFS rates after LR at 1, 3, 5, 7, and 10 years were 82.0%, 54.0%, 37.8%, 28.3%, and 19.5%, respectively. In the TACE group, the RFS rates at 1, 3, 5, 7, and 10 years were 59.7%, 18.1%, 7.9%, 5.5%, and 4.8%, respectively. The median RFS of the LR group was significantly longer than that of the TACE group (3.4 vs. 1.2 years, P<0.001) (Fig. 3A).

As the treatment for recurrence, TACE was performed in 56 (73.7%) and 132 (69.5%) patients in the LR and TACE groups, respectively (Supplementary Table 2). Local ablation treatments such as RFA and percutaneous ethanol injection therapy were administered to 15 (19.7%) and 34 (16.5%) patients in the LR and TACE groups, respectively.

In multivariate analysis, patients in the TACE group had significantly worse survival outcomes than those in the resection group (adjusted hazard ratio [aHR], 1.81; 95% confidence interval [CI], 1.08-3.04; P=0.03) (Table 2). Old age (aHR, 1.06), increased total bilirubin levels (aHR, 1.27), and chronic hepatitis C (aHR, 2.12) were associated with poor survival outcomes. Increasing size of the largest tumor (aHR, 1.43) was also significantly associated with worse survival in the multivariate analysis.

When analyzing the 344 patients who underwent LR or achieved CR after TACE, TACE was associated with a lower RFS (aHR, 2.41; 95% CI, 1.78-3.25; P<0.001) in the multivariate analysis (Supplementary Table 3). Additionally, tumor size of the largest tumor was significantly associated with lower RFS (aHR, 1.44) in the multivariate analysis. Furthermore, three tumors (aHR, 1.27) showed a trend toward a lower RFS than two tumors, although the result was not statistically significance.

Among the 107 patients in the LR group, ascites developed in 69 patients (64.5%) and was the most common perioperative morbidity. PHBD developed in 11 patients (10.2%) (four and seven who underwent major and minor resections, respectively). Table 3 presents the details of the morbidity after LR. Surgery-related morbidities occurred in 15 patients (13.9%) . The most common surgery-related morbidity was pleural effusion, followed by wound complications, bile leakage, and fluid collection. Among patients who developed postoperative pleural effusion, 5.6% required percutaneous drainage during their hospital stay. Most patients with wound infections required wound repair (n=3, 2.8%). All patients who experienced bile leakage underwent endoscopic intervention. None of the patients experienced life-threatening complications to the extent that necessitated intensive care management, corresponding to grade 4 of the Clavien-Dindo classification of surgical complications. Moreover, none of the patients died within 1 year after LR.

To minimize potential confounding factors, we generated a PS-matched cohort in a 1:2 ratio. Finally, 97 patients in the LR group were matched with 171 patients in the TACE group. Baseline characteristics in this PS-matched cohort between the two groups did not significantly differ, showing a standardized mean difference of <0.2 in all matched variables (Table 1).

In the PS-matched cohort, the TACE group showed significantly worse OS than the LR group (P=0.008) (Fig. 2B). OS rates in the LR group at 1, 3, 5, 7, and 10 years were 100.0%, 96.8%, 89.7%, 85.4%, and 74.5%, respectively, whereas the OS rates of the TACE group at 1, 3, 5, 7, and 10 years were 98.2%, 91.4%, 74.1%, 67.5%, and 59.9%, respectively.

The TACE group demonstrated a significantly poorer RFS outcome than the LR group (P<0.001) (Fig. 3B). The RFS of the LR group at 1, 3, 5, 7, and 10 years were 80.1%, 53.8%, 35.9%, 27.0%, and 24.8%, respectively. In the TACE group, RFS at 1, 3, 5, 7, and 10 years were 60.7%, 19.3%, 8.8%, 5.9%, and 5.9%, respectively.

Among 349 patients with two tumors, 87 and 262 underwent LR and TACE, respectively. Patients who underwent LR in this subgroup showed a significantly longer OS than those in the TACE group (P<0.001) (Fig. 4A). Regarding recurrence, the LR group was significantly associated with better RFS than the TACE group (P<0.001) (Fig. 4B).

Of the 134 patients with three tumors, 20 and 114 underwent LR and TACE, respectively. The OS of the LR group was better than that of the TACE group (P=0.11) (Fig. 4C). The LR group had a significantly better RFS than the TACE group (P<0.001) (Fig. 4D).

When conducting PS-matched analysis with 1:2 ratio for OS evaluation in each subgroup, in the subgroup with two tumors, 70 patients from the LR group were matched with 118 patients from the TACE group. In the subgroup with three tumors, 12 patients from the LR group were matched with 22 patients from the TACE group. However, when performing PS-matching for RFS evaluation, in the subgroup with two tumors, 65 patients from the LR group were matched with 90 patients from the TACE group. In the subgroup with three tumors, 12 patients from the LR group were matched with 17 patients from the TACE group (Supplementary Tables 4, 5).

After PS-matched analysis in each subgroup, the subgroup with two tumors showed that the LR group had longer OS and RFS rates than the TACE group (Supplementary Fig. 1A, B). However, no significant differences were observed in OS between the two treatment groups, whereas the LR group showed better RFS than the TACE group in the subgroup with three tumors (Supplementary Fig. 1C, D).