From September 2014 to February 2019, 294 consecutive patients with a total of 480 HCC lesions, which were treated by RFA or SBRT, were initially selected. All patients were local and Japanese and had been diagnosed with HCC at Yokohama City University Medical Center (YCUMC). HCC was diagnosed according to the Japan Society of Hepatology Guideline ; these lesions all had a typical imaging appearance of HCC and/or HCC was histologically proven by biopsy. All RFA was performed at YCUMC, while SBRT treatment was provided at Ofuna Chuo Hospital. YCUMC had complete medical records for all patients as the first diagnoses of HCC and follow-up of these patients were conducted at YCUMC. The data collection and analysis received institutional review board approval at YCUMC, and written informed consent was waived due to the retrospective observational nature of the study.
The inclusion criteria were patients (1) with tumors diagnosed as multifocal HCC at stage A4 or B1, for which they had initially received treatment; (2) for whom undergoing surgery was found to be unfeasible, difficult, or unsuitable (the reasons for inoperability included a central location, decreased liver function, and other comorbidities); (3) for whom the multifocal lesions were initially treated by either RFA or SBRT; and (4) treated at intervals between the RFA and SBRT of no more than three months.
The exclusion criteria were as follows: (1) patients in BCLC stages 0–A3 (170 patients with 170 lesions); (2) patients in BCLC stages greater than B1 (9 patients with 30 lesions); (3) patients who the hospital lost contact with prior to one year’s follow-up (13 patients with 30 lesions); (4) patients unable to meet the modified Response Evaluation Criteria in Solid Tumors (mRECIST)  for the selection of target lesions (less than two lesions ≥1 cm and failed to show pre-treatment enhancement on arterial phase (AP) of computer tomography (CT) or magnetic resonance imaging (MRI) examination; 37 patients with 79 lesions); (5) patients whose multifocal lesions were initially treated in the same way (i.e., all by RFA or all by SBRT; 46 patients with 123 lesions); and (6) patients with treatment intervals between the RFA and SBRT of more than three months (four patients with nine lesions).
As described by Wang et al.  and Hao et al. , a 480 kHz generator (VIVA RF generator; STARmed, Gyeonggi, Korea) capable of producing a maximum power of 200 W and a specifically sized 17-gauge internally cooled, adjustable RF electrode (Proteus; STARmed, Gyeonggi, Korea) were applied. RFA was performed with the guidance of real-time CT/MRI and US fusion imaging carried out by one of the three senior hepatologists (K.N., K.O., and H.F.), each having more than 20 years of experience in interventional techniques. One to three insertions were performed according to the tumor size and shape, requiring an ablative safety margin of no less than 5 mm around the treated lesions. Post-operative contrast-enhanced ultrasonography was performed to determine the adequacy of ablation. If a residual tumor was detected, additional RFA was performed.
We previously described our SBRT methods in detail . During free breathing of patients, Spiral, 4-phase, multidetector CT and/or dynamic contrast-enhanced MRI were conducted, and followed by fusion with a slow-scan CT scan (6–10 s per slice). The gross tumor volume (GTV), including the enhanced tumor, was delineated with the slow-scan CT images. For the internal target volume, an internal margin (4–6 mm) was created around the clinical target volume (CTV) according to the respiratory movement of the diaphragm observed during fluoroscopy. For the planning target volume (PTV), individualized margins of 2 mm were applied around the internal target volume as a setup margin. Multiarc, dynamic conformal radiation was planned using a radiation treatment planning system (FOCUS XiO, version 4.2.0–4.3.3; Computerized Medical Systems, St Louis, MO, USA) and was performed using X-rays from a 6-MV linear accelerator (Clinac 2100C; Varian Medical Systems Inc., Palo Alto, CA, USA). When the tumors were not near the gastrointestinal tract, SBRT with total doses of 35–40 Gy were delivered in five fractions over five to seven days. The total dose of 35 Gy was administered to patients with Child–Pugh class A or B disease, with > 20% of the normal liver receiving > 20 Gy, and a total dose of 40 Gy in the other patients. For the tumors near the gastrointestinal tract (≤ 3 mm distance), a total dose of 42 Gy was delivered in 14 fractions over 18 days . The treatment was planned to enclose the planning target volume with a 60–80% isodose line of the maximal dose.
Interpretation and assessment of outcome data
The first targets of observation in this study were treatment-related adverse effects, which were evaluated by both laboratory testing (objective findings) and clinically-visible complications (constitutional symptoms) . The laboratory parameters, including serum alanine transaminase (ALT), aspartate transaminase (AST), leukocytes counts, platelets counts, total bilirubin (T-BIL), and albumin (ALB), were recorded. Hematologic toxicity by radiation, a principal cause of acute myelosuppression, was evaluated by the decline in leukocytes and platelets. The albumin–bilirubin (ALBI) score was calculated according to the values of T-BIL and ALB. The ALBI grade, ALT, and AST were used to assess treatment-related side effects on liver function. The time points for laboratory evaluations were one or three days before treatment, one week after completion of later therapy (either RFA or SBRT), and three to five months thereafter. In addition, clinically-visible complications were recorded and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.03. The adverse events grade ≥ 3 were recorded for later analysis.
Tumor response and prognosis
The secondary targets for observation were tumor response and local control. Tumor response was evaluated one year after treatment completion of both RFA and SBRT. According to mRECIST for patients with multifocal HCC lesions, the longest viable tumor diameter of the two selected target lesions was measured . Local control was assessed by follow-up radiologic imaging and defined as no tumor recurrence/progression at the primary site.
The tertiary observation target was the prognosis, including time to progression, one- and two-year PFS, and OS. PFS was defined as the time from the initial treatment (RFA or SBRT in our study) of the target HCC lesion to the first occurrence of disease progression or death from any cause, whichever occurred first. OS was defined as the interval from the initially received therapy (either RFA or SBRT) until the last visit or date of death, regardless of the cause of death. The outcome of this study was death or dropout of the patients. Exposure factors were RFA and SBRT treatment. Potential confounders that influence OS include other chronic diseases (diabetes, hypertension, etc.) that the patient also suffers from. They were calculated from the date of the earlier therapy (either RFA or SBRT). Local control was evaluated on a lesion basis; the tumor response, PFS, and OS were evaluated on a patient rather than lesion basis.
Attention points for imaging evaluation
Contrast-enhanced CT and MRI were used as first-line imaging modalities for evaluating tumor response and recurrence in all patients. Contrast-enhanced ultrasound examination and non-enhanced CT or non-enhanced MRI were used for one patient experiencing renal dysfunction during the follow-up periods. Radiological images were all independently evaluated by two hepatologists (enhanced CT and MRI: M.C. and K.H.; contrast-enhanced ultrasound: H.N. and K.O.) who were unaware of any laboratory test information, clinical history, or our therapy strategy. Any interpretation discrepancies were resolved by consensus with the participation of a third expert hepatologist (K.N.) with 20 years of experience in HCC diagnosis and treatment.
According to type and level of distribution, the continuous variables (lesion size, treatment intervals, PFS, and OS) are presented as means and ranges while the categorical variables (such as treatment modalities and rise or fall of laboratory test values) are described as percentages and frequencies, as appropriate. Cumulative rates of PFS and OS were estimated using the Kaplan–Meier method. Analyses were performed using the statistical software SPSS 24.0 (Inc, Chicago, IL). All analyses are descriptive.