This retrospective study was approved by the institutional review boards of the Second Xiangya Hospital and the Affiliated Cancer Hospital of Xiangya School of Medicine and was performed in accordance with the Declaration of Helsinki. The need for written informed consent was waived by the institutional review board due to the retrospective nature of the present study.

Between June 2014 and September 2020, 438 consecutive patients with HCC, who underwent thermal ablation as an initial therapy at two institutions, were selected for this study. Patients meeting the following criteria were included: those with (a) a preserved liver function (Child-Pugh A or B); (b) an Eastern Cooperative Oncology Group performance status 0; (c) at least one target lesion for measurement; and (d) the absence of macrovascular invasion and extrahepatic metastasis. The exclusion criteria were as follows: patients who (a) did not follow-up (n=69); (b) had more than five HCC lesions (n=47); (c) had infiltrative HCC (n=38). Finally, 284 patients were included in the present study. Among those 284 patients, 69 were from institution A and 215 were from institution B. The flowchart of the study population is shown in Figure 1 .

Thermal ablation was performed under computed tomography (CT) or ultrasound guidance using a microwave ablation (MWA) (KY-2000, Kangyou Medical Instrument Co. Ltd., China) or a radiofrequency ablation (RFA) system (CELON Power System, Olympus Corporation, Japan) by four board-certified senior interventional radiologists. The microwave ablation was set at 60 to 80 W for 3 to 20 min. The radiofrequency ablation was set at 30 to 80 W for 10 to 40 min. For tumors larger than 3.0 cm, an overlapping ablation was performed. Contrast-enhanced CT immediately after thermal ablation was routinely performed to assure a minimum safety margin of 0.5 cm when possible. For multifocal lesions, a maximum of three lesions was ablated in a single session, and the remaining were ablated within one week after the first session.

Contrasted-enhanced CT or magnetic resonance (MR) was routinely performed approximately four to six weeks after the first attempt thermal ablation. The treatment response was classified into complete response and incomplete response according to the modified Response Evaluation Criteria in Solid Tumors (mRECIST). An incomplete response was defined as residual viable tumor foci at the edge of the ablative zone while a complete response was defined as an ablated lesion with no viable tumor. Patients with multiple tumors were classified into the complete response group only if all the tumors were ablated completely in the first attempt itself; otherwise, they were categorized into the incomplete response group.

In patients with an incomplete response after a first attempt ablation, the feasibility of repeated thermal ablation was assessed. Usually, the lesion was considered infeasible to ablation when it was in an unfavorable location or had poor conspicuity on unenhanced CT or ultrasound. When considered feasible, the patients were treated with repeated thermal ablation. When considered infeasible, TACE was performed as an alternative therapeutic option by several board-certified senior interventional radiologists. The femoral artery was routinely catheterized. Selective hepatic arteriography was performed and a 2.2-F coaxial microcatheter (Carnelian, Tokai Medical Products, Japan) was placed in the tumor feeders. Chemoembolization was performed using a super-selective approach by slowly injecting an oil-epirubicin emulsion. The gel-foam slurry was used to embolize the proximal tumor feeders. The oil-epirubicin emulsion was prepared by mixing iodized oil (up to 15 mL; Lipiodol, Guerbet Asia Pacific Ltd, Hong Kong, China) with distilled water containing 50 mg to 120 mg of epirubicin dissolved in a ratio of 3:1.

After the procedure, the patients were followed up at 1, 3, 6, and 12 months in the first year and every 6 months subsequently if no recurrence was detected. At each follow-up visit, contrast-enhanced CT/MRI of the liver was performed to evaluate the radiological response according to the mRECIST. Local tumor progression (LTP), intrahepatic distal recurrence (IDR), and extrahepatic recurrence (ER) were recorded. If tumor progression occurred, the choice of treatment for recurrent HCC was dependent on the site of the tumor, liver function, and the general condition of the patient.

The primary outcome was progression-free survival (PFS), which was defined as the time from the first thermal ablation until the date of radiological progression or the last follow-up (April 30, 2021). Radiological progression was defined as patients with LTP, IDR, and ER. The secondary outcome was overall survival (OS), which was defined as the time interval between the first ablation and death from any cause or the last follow-up (April 30, 2021).

Data on the potential risk factors for PFS and OS were collected, including the first attempt ablation (complete/incomplete), Child-Pugh class (A/B), age (≤56 years/>56 years), gender (male/female), number of tumors (single/multiple), modality of guidance for ablation (CT/ultrasound), type of thermal ablation (MWA/RFA), tumor size (≤3 cm/>3 cm), etiology of hepatitis (none/hepatitis B virus/hepatitis C virus/alcoholic), cirrhosis (presence/absence), level of alpha-fetoprotein (AFP; ≤400 ng/mL/>400 ng/mL), and level of the systemic immune inflammation index (SII; ≤330/>330). The SII was calculated as (N*P)/L, where “N” represents the neutrophil count, “P”, the platelet count, and “L”, the lymphocyte count; a cut-off of 330 for SII was used as previously described (11). In the present cohort, patients were stratified based on the median age (56 years).

Categorical variables were compared using the Chi-square test or Fisher’s exact test, as appropriate. Continuous variables were compared using the Mann-Whitney U test or t-test, as appropriate. To diminish potential confounding and selection bias in the two groups, propensity score-matching (PSM) was applied. PSM is a statistical method which can reduce the bias due to confounding variables by accounting and matching the covariates that may affect the treatment outcome. Each group was matched according to the generated PSM using a caliper width of 0.1. In patients with multiple lesions, the characteristics of the largest lesion were recorded for analysis. The potential risk factors for PFS and OS were compared using a log-rank test before and after PSM. Cox proportional hazard ratios (HRs) were calculated to identify the independent risk factors for PFS and OS. All potential variables were included in the multivariate Cox regression analysis. Statistical analysis was carried out using a statistical software (Statistical Package for the Social Sciences version 24, International Business Machines Corporation, USA) or R software (version 4.0.2, http://www.R-project.org). P <0.05 was considered to be statistically significant.

The entire study population included 238 males and 46 females, with a mean age of 56.0 ± 11.0 years, ranging from 27 to 81 years. There were 240 patients with hepatitis B, 14 with hepatitis C, and 4 with alcoholic hepatitis. The diagnosis of HCC was based on either histological evidence (n=41) or radiological evaluation with reference to the 2018 version of Liver Imaging Reporting and Data System (n=243). There were 244 patients with Child-Pugh A and 40 patients with Child-Pugh B. In the entire study population, 48 patients were in the very early stage (Barcelona Clinic Liver Cancer [BCLC]-0), 190 were in the early stage (BCLC-A), and 46 were in the intermediate stage (BCLC-B). The baseline characteristics of the patients from the two institutions are shown in Table 1 .

There were 236 patients in the first attempt complete response group and 48 in the incomplete response group before PSM. The PFS (19.7 months, 95% CI: 15.1-24.3 months vs. 6.3 months, 95% CI: 3.0-9.6 months; P<0.001) and OS (P=0.044) were significantly higher in the complete response group than in the incomplete response group. After a one-to-one PSM analysis, there were 46 pairs of matched patients. The baseline characteristics of the patients before and after PSM are presented in Table 2 . The PFS was significantly higher in the first attempt complete response group than in the incomplete response group (22.9 months, 95% CI: 15.8-30.0 months vs. 6.3 months, 95% CI: 3.0-9.6 months; P<0.001). However, there was no difference in the OS between the two groups after PSM analysis (P=0.296). Figure 2 shows the survival curves of PFS and OS of the first attempt complete response and incomplete response groups before and after PSM.

After a first attempt incomplete ablation, 29 patients received repeated ablation while 19 received TACE. Figures 3 and 4 respectively show the typical imaging of patients who received repeated ablation and TACE. In some patients, repeated ablation was not feasible due to the unfavorable location of the lesions (n=13) or poor lesion conspicuity on unenhanced CT or ultrasound (n=6). There were no significant differences in PFS (P=0.437) and OS (P=0.178) between patients with repeated ablation and TACE after a first attempt incomplete ablation. Figure 5 shows the survival curves of PFS and OS of the repeated ablation and TACE groups after a first attempt incomplete ablation.

The univariate analysis for PFS showed that a first attempt incomplete response (P<0.001) was associated with tumor progression. However, between the two groups, no differences in PFS were noted in age (P=0.914), gender (P=0.624), the number of tumors (P=0.507), modality of guidance for ablation (P=0.361), thermal ablation type (P=0.244), tumor size (P=0.560), the etiologies of hepatitis (P=0.834), presence or absence of cirrhosis (P=0.670), AFP level (P=0.721), Child-Pugh score (P=0.071) or SII level (P=0.438). Multivariate Cox proportional hazards regression analysis showed that first attempt incomplete response (P<0.001; HR, 2.537; 95% CI, 1.620-3.975) and Child-Pugh class B (P=0.017; HR, 1.905; 95% CI, 1.122-3.234) were independent risk factors for tumor progression. The details of univariate and multivariate Cox regression analysis for PFS are presented in Table 3 .

The univariate analysis for OS showed that a higher AFP level (P=0.019) and a first attempt incomplete response (P=0.048) were associated with poor OS. However, there were no differences between the two groups in age (P=0.512), gender (P=0.816), the number of tumors (P=0.228), modality of guidance for ablation (P=0.335), thermal ablation type (P=0.242), tumor size (P=0.256), the etiologies of hepatitis (P=0.737), presence or absence of cirrhosis (P=0.839), Child-Pugh score (P=0.089), or SII level (P=0.964). Multivariate Cox proportional hazards regression analysis showed that a higher AFP level (P=0.011; HR: 2.497; 95% CI: 1.236-5.042) and Child-Pugh class B (P=0.026; HR: 2.799; 95% CI: 1.131-6.924) were independent risk factors for poor OS. The details of univariate and multivariate Cox regression analysis for OS are presented in Table 4 .