This multicenter retrospective study included five tertiary referral hospitals and was conducted in accordance with the Declaration of Helsinki. Institutional review board approvals were obtained from the participating hospitals, and the requirement for written informed consent was waived because of the retrospective study design.

Consecutive patients with unresectable very early-stage or early-stage HCC who underwent MWA as primary therapy between August 2013 and September 2020 were included. The inclusion criteria were as follows: (1) a single tumor of 50 mm or less, or 2–3 tumors that were each smaller than 30 mm (considering the indication of MWA), (2) ECOG performance status (PS) of 0, (3) albumin-bilirubin (ALBI) score of I or II without moderate to massive ascites, and (4) absence of macrovascular invasion or extrahepatic metastasis. The exclusion criteria were as follows: (1) absence of baseline CT imaging 1 month prior to MWA, (2) poor imaging quality, (3) absence of baseline laboratory information within a week prior to MWA, and (4) absence of follow-up data.

The CT examination was performed on four different machines using the following parameters: (1) Aquilion One scanner (Canon Medical Systems): tube voltage, 120 kVp; tube current, auto; rotation time, 0.5 s; matrix, 512 × 512; field of view, 400 × 400 mm; and slice thickness, 5 mm; (2) Somatom Definition As+ scanner (Siemens): tube voltage, 120 kVp; tube current, auto; rotation time, 0.5 s; matrix, 256 × 256; field of view, 452 × 452 mm; and slice thickness, 5 mm; and (3) Somatom Force scanner (Siemens): tube voltage, 110 kVp; tube current, auto; rotation time, 0.5 s; matrix, 512 × 512; field of view, 400 × 400 mm; and slice thickness, 5 mm; (4) Aquilion Prime scanner (Canon Medical Systems): tube voltage, 120 kVp; tube current, 300 mAs; rotation time, 0.35 s; matrix, 256 × 256; field of view, 400 × 400 mm; and slice thickness, 4 mm.

Data on clinical variables were collected, including age, sex, etiology of the underlying liver disease, and liver cirrhosis. Imaging parameters included the number of tumors, tumor size, shape of the tumor, mosaic architecture, rim arterial phase hyperenhancement (APHE), and satellite lesions. Laboratory parameters included neutrophil count, lymphocyte count, platelet count, serum albumin, total bilirubin, and alpha-fetoprotein (AFP).

The CT images were independently reviewed by two board-certified radiologists with 13 and 18 years of experience in abdominal imaging, and the corresponding CT imaging features were recorded accordingly. For patients with multiple tumors, imaging features of the largest tumor were recorded. In cases of disagreement between the two radiologists, a final decision was made by consensus. The SMARS score was calculated as described previously, which included five parameters, such as Shape of tumor, Mosaic architecture, AFP level, Rim APHE, and Satellite lesion [5]. Briefly, 0.767 × Shape of tumor + 1.196 × Mosaic architecture + 0.881 × AFP level + 2.506 × Rim APHE + 1.178 × Satellite lesion − 8.811. A cutoff value of −0.49 was used to identify predicted proliferative and nonproliferative HCCs.

The treatment approach was discussed by a tumor board that included surgeons, interventional radiologists, oncologists, diagnostic radiologists, and hepatologists. Clinicians discussed the treatment recommendations with the patients, and a final decision was made by consensus. The MWA procedures (KY-2000, Jiangsu Kangyou Medical Instrument; ECO-100AI10, Nanjing ECO Medical Technology) were performed by several board-certified senior interventional radiologists. Under CT guidance, the antenna was inserted percutaneously into the tumor. An overlapping technique was used for tumors larger than 30 mm. In patients with multiple tumors, ablation was performed for all tumors in a single session. The MWA was set at 60–140 W, and the ablation time was 3–25 min. Intraprocedural contrast-enhanced CT was performed to determine the safety margins. The technical success of ablation was defined as the complete ablation of the tumor with a safety margin of at least 0.5 cm on CT images.

Patients were observed for 2–3 months after MWA and at least every 6 months thereafter. Contrast-enhanced CT or magnetic resonance imaging, and determination of serum AFP levels were routinely performed to monitor recurrence. Follow-up was performed via telephone interviews (March 2024) or during the last visit to the hospital if a telephone interview was unavailable. The primary endpoint was OS, defined as the time interval between the date of MWA and the date of death or last follow-up. The secondary endpoint was RFS, defined as the time interval between the date of MWA and the date of recurrence or last follow-up.

To compare the therapeutic outcomes of predicted proliferative and nonproliferative HCCs in the subgroups, patients were further divided into groups according to tumor size: smaller than 30 mm and 30–50 mm.

Continuous variables are presented as the mean ± standard deviation (SD) or the median with interquartile range (IQR). Categorical variables are presented as numbers with percentages. Categorical variables were compared using the χ2 test or Fisher’s exact test, as appropriate. Continuous variables were compared using the Mann‒Whitney U-test or t-test, as appropriate. The inter-reader agreement between the two radiologists regarding the CT imaging features in the SMARS score was calculated using the kappa coefficient. Based on the SMARS scores assigned to each patient, the entire study population was divided into predicted proliferative and nonproliferative HCCs, respectively. Propensity score matching (PSM) analysis was applied for baseline characteristics with statistical significance between the two groups to reduce potential confounding and selection biases. The optimal caliper for PSM was set to 0.1. The RFS and OS were compared between the two groups using the log-rank test before and after matching. Statistical analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing; http://www.R-project.org), and a two-sided p < 0.05 denoted statistical significance.