This retrospective study was approved by the institutional review board and the need for informed consent was waved. From February 2009 to September 2010, patients who met the following inclusion criteria were assessed and treated with combination of TACE and RFA: 1) 3 or less HCCs, 2) the size of the largest tumor ranging in size from 3 cm to 5 cm; 3) Child-Pugh class A or B; 4) no evidence of vascular invasion or extrahepatic metastases on the CT or MR images; 5) absence of severe coagulopathy (i.e., prothrombin activity < 40% or platelet count < 40000/mL); and 6) tumor located at least 1 cm away from the hilar bile duct. All patients suspected of having HCC underwent a routine physical examination, blood laboratory tests, and dynamic CT and/or MRI. The diagnosis of HCC was based on the American Association for the Study of Liver Diseases guidelines (18) as follows: typical vascular pattern (hypervascular in the arterial phase, and wash-out in the portal/delayed phase) of the liver nodule in at least one of the dynamic CT or MRI. No patient underwent a percutaneous biopsy. The location and size of each patient's HCC was evaluated using the CT or MR images. Tumor size was defined as the maximum diameter measured on the CT or MR images, and the segmental location of the tumor was recorded in the data sheet according to the Couinaud nomenclature.

During the study period, 21 consecutive patients with HCCs were included (Tables 1, 2). Among them, 2 patients had 2 HCC nodules each. Both patients had one tumor larger than 3 cm at the longest diameter and the other smaller than 3 cm, making a total of 23 HCCs. Percutaneous RFA was also performed for these 2 HCCs < 3 cm, but these 2 lesions were not of our interest in this study, and thus excluded from the analysis. Of the 21 patients, 16 patients had no treatment history of HCC, whereas the other 5 patients had a previous treatment history using the following methods: hepatic resection (n = 3) or percutaneous RFA (n = 2). The size of 21 HCCs was 3.6 ± 0.4 cm in the longest diameter with a range of 3-4.5 cm (median: 3.5 cm). Seventeen (81.0%) HCCs were located in the subphrenic region, either directly contacting the diaphragm or within 1 cm from the diaphragm.

For the patients who met the inclusion criteria, planning US was performed by one certificated abdominal radiologist with at least 8 years' experience in liver US (more than 15000 cases) and US-guided interventional procedures (more than 2000 cases) including biopsy, drainage and RFA, at the beginning of this study. Through the planning US, the radiologist searched for the HCC thoroughly and evaluated the expected electrode path, adjacent organ vulnerable to thermal injury, tumor size, lesion conspicuity, and heat-sink effect to determine if percutaneous RFA was feasible. All planning US were performed with a multi-frequency 4C1 convex array probe (Acuson Sequoia 512, Siemens Medical Solutions, Mountain View, CA, USA). At the time of planning US, the radiologist was aware of the patients' clinical information and liver CT and/or MR images. After a careful evaluation of the prior liver CT and/or MR images, the radiologist searched for the HCC thoroughly and evaluated the expected electrode path, adjacent organ vulnerable to thermal injury, tumor size, lesion conspicuity, and heat-sink effect to determine if percutaneous RFA was feasible (19).

All TACE procedures were performed on an inpatient basis by one of 2 interventional radiologists with at least 9 years of experience in interventional radiology at the starting point of this study. The hepatic artery was catheterized after celiac and superior mesenteric arteriography using a 5 Fr catheter (Cook, Bloomington, IN, USA). Depending on the size, location and arterial supply of the tumor, the 3 Fr microcatheter (Microferret; Cook, Bloomington, IN, USA) was advanced toward the tumor-feeding arteries for selective embolization. Therefore, segmental embolization of the tumor supplying artery was performed, sparing the majority of hepatic parenchymal arterial supply, using an emulsion of iodized oil (Lipiodol; Andre Gurbet, Aulnay-sous-Bois, France) and doxorubicin hydrochloride (Adriamycin RDF; Ildong Pharmaceutical, Seoul, Korea). An infusion of this emulsion was performed until arterial flow stasis had been achieved and/or iodized oil was visualized in the portal branches. The dose of embolization agent was determined according to the tumor size, tumor number, feeding vessels and liver function status. This procedure was followed by the embolization of gelatin sponges (1-mm in diameter; Gelfoam; Upjohn, Kalamazoo, MI, USA). After embolization, angiography was performed again to determine the extent of the vascular occlusion and the presence of any residual tumor staining.

Within 14 days after TACE, percutaneous RFA was performed by a radiologist with more than 4 years' experience (> 200 cases) in RFA of HCC at the starting point of this study. The patients were placed in the supine position and treated under local anesthesia or general anesthesia depending on the tumor size and location. When the index tumor was abutting the portal vein, we preferred general anesthesia rather than local anesthesia. RFA was performed in an interventional suite equipped with a flat-panel biplane fluoroscopy/angiography instrument (Allura Xper FD 20/10, Philips Healthcare, Best, The Netherlands). US guidance was used concurrently to determine the safe skin entry site and to enable accurate targeting of the index tumor, as well as monitoring of the ablation process. Seventeen-gauge, internally cooled electrodes with a 3 cm-long exposed tip (Cool-tip; Valleylab, Boulder, CO, USA or Well-point RF Electrode; STARmed, Goyang, Korea) were used. RF energy was delivered using an impedance-based control algorithm of the generator according to the manufacturer's instructions for each device. Artificial ascites was infused, whenever it was required, to improve the sonic window and decrease the degree of collateral thermal injury to the adjacent diaphragm or colon (20, 21). The site of skin puncture was determined based on the findings of both biplane fluoroscopy and US. Therefore, traversal of the critical structures, such as colon, gallbladder, and large vessels, was avoided in all cases. Although the tumor was located in the subphrenic region, a transhepatic approach not traversing the thoracic cavity was possible in almost all cases using the oblique approach.

Real-time biplane fluoroscopy was used to place the radiofrequency electrode through the tumors, if the index tumor was visible on both anteroposterior and lateral projections due to iodized oil retained from prior TACE. The targeting time, which is the time measured from the skin puncture of the electrode to the placement of the electrode through the index tumor, was typically within one minute. To achieve adequate ablative margin (at least 0.5 cm), multiple overlapping ablations were applied as needed, depending on the tumor size, shape, and location. The number of radiofrequency electrodes was dependent on the visibility of the target tumor on fluoroscopy and the location of the tumor. Generally, multiple electrodes (up to 3 electrodes) were used if the lesion conspicuity on fluoroscopy was not radiopaque enough for multiple overlapping treatments. In addition, if the tumor was located close to any critical structures vulnerable to thermal injury, multiple electrodes were used for precise placement of the electrode. When multiple electrodes were used, sequential ablation of each electrode was performed since generator capable of delivering RF energy with switching mode was not available at the time of the study period. Multiple overlapping treatments were possible in most tumors under the real-time guidance of biplane fluoroscopy, because the visibility of iodized oil retained remained almost unchanged after ablation cycles. The ablation time was variable and dependent on the US findings. If a large echogenic zone was obtained in 3-6 min, the tip of the electrode was repositioned under real-time guidance of biplane fluoroscopy. In other words, a short ablation time with a multiple overlapping strategy rather than a standard 12 min single ablation (i.e., 6 min × 2 overlapping ablations rather than 12 minutes × 1 ablation) was adopted because we believe that the former could create a larger ablation zone than the latter. If the index tumor was not clearly demonstrated on the fluoroscopic imaging, US guidance alone was used for the RFA procedure. At the end of the procedure, tract ablation was performed to prevent bleeding or tumor seeding.

Computed tomography was performed immediately after RFA to evaluate the technical success and procedure-related complications. When the CT images depicted the presence of a nonenhancing area surrounding the entire tumor, it was defined as a technical success (10, 22). A second RFA session was undertaken only when nodular enhancement was observed near the accumulated iodized oil at the CT images (10, 22). The patients were then followed up with contrast-enhanced CT, one month later, and then every three months after treatment. The overall follow-up time was defined as the interval between the first RFA and either local tumor progression or the last follow-up visit by May 23, 2012.

The presence of major complications was evaluated based on the previous guideline of standardized terminology (23), which is defined as any event leading to substantial morbidity and disability, increased level of care, lengthened hospital stay, or a required blood transfusion or interventional drainage procedure. All other complications were deemed to be minor.

In addition, therapeutic efficacy of RFA was assessed based on standardization paper of image-guided tumor ablation (23). Technical success was assessed based on immediate post-RFA CT images. Primary technique effectiveness was evaluated by dynamic CT performed one month after the combined TACE and RFA. At dynamic liver CT, local tumor progression was defined as a new enhancing lesion within or adjacent to the ablation site (23). The cumulative rate of local tumor progression was estimated with the use of Kaplan-Meier method. A distant metastasis was defined as a new HCC in the liver distant from the treated nodule or in extrahepatic regions.

Radiofrequency ablation was performed with a 2-week interval after initial TACE. At the starting point of this study, a sufficient time interval between the 2 procedures was preferred due to the safety concerns of the patients. On the other hand, the time interval between the 2 procedures was subsequently reduced to less than a week. The mean time interval between the 2 procedures was 6.7 d, ranging from 1 to 14 d. Artificial ascites was introduced in 17 (81.0%) patients immediately before placing the electrodes to minimize collateral thermal injury and enhance the sonic window. A single electrode was used in 16 (76.2%) patients and 3 electrodes were used in the other 5 (23.8%) patients. After segmental TACE, 20 (95.2%) out of 21 HCCs were visible on fluoroscopy, and were ablated under biplane fluoroscopy guidance assisted by US. The remaining HCC nodule, which was poorly visible on fluoroscopy, was ablated under US guidance alone. Multiple overlapping ablations under US guidance alone were technically difficult due to the echogenic zone induced by prior ablation cycles. A long RFA procedure time (total ablation time: 31.1 ± 9.0 min; range, 17.5-47 min) was required to achieve a sufficient ablative margin using multiple overlapping ablations (the number of overlapping ablations: 5.9 ± 2.0; range: 3-12).

On the CT images, the mean diameters of the ablation zone for the long and short axes induced by RFA were 5.2 ± 0.4 cm and 3.8 ± 0.2 cm, respectively. Technical success was achieved in all 21 HCCs after a single RFA session (Fig. 1). Primary technique effectiveness based on the one-month follow-up CT was also achieved in all 21 HCCs.

There were no treatment-related deaths. There was only one major complication, which was a pneumothorax necessitating chest tube placement for 4 days. This was an unexpected complication due to an unplanned transpulmonary approach. It may have been caused by a transgression of the lower margin of the basal lung by the lesser caudal angled approach, which could have been avoided easily if a steeper approach to transgress the liver, not the lung, had been performed. There were 8 minor complications: post-ablation syndromes (n = 5) and diaphragmatic thermal injury (n = 3).

Local tumor progression was found in 4 (19.0%) of 21 HCCs during the follow-up period (mean follow up: 24.8 ± 9.9 months; range: 6.5-37.3 months). Cumulative rates of local tumor progression were estimated as 9.5% and 19.0% at 1 and 3 years, respectively (Fig. 2). Four HCCs, which were found to have local tumor progression 7, 11, 15, and 16 months after the combined treatment, had an adequate ablative margin after the initial combined treatment. The 4 HCCs with local tumor progression were treated with RFA (n = 3) and TACE (n = 1), respectively. A distant intrahepatic metastasis was found in 11 (52.4%) of 21 patients, which were managed with repeated RFA (n = 2), TACE (n = 3), combined TACE and RFA (n = 5) or radiation therapy (n = 1). None of the patients died during the follow-up period.