Requirements for study participation were as follows: (a) an adult patient with hepatic cirrhosis and either a single 5 cm in diameter HCC or smaller, or as many as three 3 cm each in diameter or smaller HCCs, (b) absence of vascular invasion or extrahepatic metastases, (c) sharp definition from the surrounding liver parenchyma without evidence of adjacent satellite nodules, (d) hepatic cirrhosis classified as Child-Pugh class A or B, (e) prothrombin time ratio (i.e., normal time divided by patient's time) greater than 40%, (f) platelet count higher than 40,000 per cubic millimeter (40-10⁹ /L), (g) no previous treatment for HCC, (h) ineligibility for surgical resection or transplantation, (i) patient agreement to have TACE, (j) no residual enhanced area within or around the tumors on the one-month follow-up CT imaging. The reason for the confined inclusion criteria to nodular small or intermediate tumors was to facilitate comparison with the clinical results from other treatment modalities such as radiofrequency (RF) ablation. From July 1998 to November 2005, 372 patients with HCC had been referred to our center for TACE. Among them, 74 tumors in 59 patients were included in our study. The median follow-up period was 17 months (4-77 months). All but two patients were male. Their age ranged from 44 to 78 years (mean±SD: 62.5±9.5 years). The tumor size ranged from 1.0 cm to 4.5 cm (mean±SD: 2.4±0.9 cm). Additional patient characteristics are shown in Table 1.

For nine patients, the diagnosis of HCC was proved histopathologically. For the other patients, the diagnosis was established on the basis of characteristic imaging findings on the three-phase helical computed tomography (CT) and conventional angiography and/or the presence of elevated tumor marker levels in serum (alpha-fetoprotein level > 200 ng/mL) (12). In the vast majority of patients, the etiology of cirrhosis was chronic viral hepatitis B or C (Table 1).

All patients had enhanced dynamic CT within four weeks prior to TACE. All the TACE procedures were performed by interventional radiologists with experience of over five and three years. Hepatic angiography was performed using 5-Fr angiographic catheters, followed by superselection of segmental arterial feeders using a microcatheter. Then we administered an iodized oildoxorubicin hydrochloride (Adriamycin; Kyowa Hakko Kogyo, Tokyo, Japan) emulsion into the feeders. The volume of iodized oil ranged from 3 to 10 ml, and the amount of doxorubin ranged from 20 to 70 mg. Once the flow became sluggish, gelatin sponge particles (Gelfoam; Upjohn, Kalamazoo, MI) that were mixed with mitomycin-C (Kyowa Hakko Kogyo, Tokyo, Japan) and contrast material (Iopromide; Schering, Berlin, Germany) were administered into the feeders until blood flow stopped completely. While performing the segmental TACE, attempts were made to completely occlude the arterial feeder. A small amount of saline solution was then injected slowly to confirm the complete occlusion of the segmental arterial feeder. If contrast media retention was partially washed out after the saline injection, additional gelatin sponge particles were infused until complete stasis of flow was obtained.

Among the 74 tumors, three were supplied by both segmental arteries on selective angiography using microcatheters. In these circumstances, both segmental feeding arteries were embolized.

Local recurrence rate was compared by 12 possible prognostic factors: patient age, hepatitis C infection, modified Child-Pugh classification, number of tumors, size of tumor nodule, serum alpha-fetoprotein level, serum albumin level, platelet count, homogeneity of iodized oil accumulation within the nodule, tumor location in segmental border zone, tumor location in subcapsular area, and contact of tumor with adjacent vessels.

The number of tumors was determined by pre-embolization CT. Tumor size was determined as the maximal diameter of the nodule measured on the pre-embolization CT. The CT examinations were performed with an 8-slice multidetector CT scanner (Lightspeed; GE Medical Systems, Milwaukee, WI) with 5-mm collimation and 17.5-mm/sec table speed, or a single-detector helical scanner (Prospeed Advantage; GE Medical Systems, Milwaukee, WI) with 10-mm collimation and a 10-mm/sec table speed.

The border zone between hepatic segments was determined by tracing the portal venous tree during the portal phase of the spiral CT scan (13, 14). The segmental border zone was defined as an area without traceable portal veins between hepatic segments on the CT scan. Segmental border zone lesions were defined as lesions crossing an imaginary border between hepatic segments.

All patients underwent both non-enhanced and contrast-enhanced three-phase helical CT four weeks after the TACE. The pattern of iodized oil accumulation in the masses was evaluated with the four-week follow-up CT scan. When the tumor nodules showed compact iodized-oil accumulation without any defect, they were classified as a "homogeneous" pattern; otherwise they were classified as an "inhomogenous" pattern.

When the tumor nodules showed sharp definition from the surrounding liver parenchyma, without evidence of adjacent satellite nodules, they were classified as "nodular" tumors. Otherwise, they were classified "nonnodular" tumors.

Tumor location was classified as either subcapsular (abutting the hepatic capsule) or nonsubcapsular. The tumors were also classified into two groups of contacting or non-contacting tumors according to their position relative to adjacent visible (> 1-mm diameter) blood vessels on the basis of whether part of the tumor was attached to the vessel or not (15). These two factors were known as potential predictors of local recurrence after RF ablation (15-18).

Residual viable tumor was determined to be present when an enhanced portion was observed within or around the original mass on the one-month follow-up CT scan. If no definite evidence of residual tumor was noted on this one-month follow-up CT, then a three-phase contrast-enhanced CT was performed at 3- or 4-month intervals thereafter. Local tumor recurrence was determined to be present when iodized oil from the lesion disappeared, or when an enhanced portion was seen within or at the margin of the original mass on the next follow-up CT scan, after the first one-month follow-up CT scan. For recurrent tumors, additional therapies such as TACE or RF ablation were performed.

Two abdominal radiologists with five and four years of experience interpreted the CT images including segmental zonal anatomy independently; they were blinded to whether the tumor showed local recurrence on the follow-up CT images. Final decisions were reached by consensus.

For the 12 potential prognostic factors of local tumor recurrence, univariate and multivariate analyses were performed using the Cox proportional hazard model. Parameters that proved to be significant with the univariate analysis were subsequently tested with the multivariate Cox proportional hazard model.

Specifically, nodular tumors were divided into four groups as IA, IB, IIA, or IIB according to whether they showed homogeneous (Group I) or inhomogeneous (Group II) iodized oil accumulation on the one-month follow-up CT imaging, after segmental TACE, or whether they were located within the liver segment (Group A) or within the segmental border zone (Group B).

Tumor characteristics such as tumor size, iodized oil uptake pattern on the one-month follow-up CT imaging as well as laboratory data were compared between Group IA and the other three groups combined as well as between each group. Comparison of tumor characteristics and laboratory data between the groups was performed using the chi-square test. The local recurrence rate was compared between the groups using the Kaplan-Meier method and the log rank test.

P-values less than 0.05 were considered statistically significant. The SPSS software package (Version 10.0; SPSS Inc., Chicago, IL) was used for statistical analysis.