Comparison of radiofrequency ablation and transarterial chemoembolization for hepatocellular carcinoma within the Milan criteria: A propensity score analysis

Authors

  • Chia-Yang Hsu,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    3. Department of Medicine, National Yang-Ming University Hospital, Yilan, Taiwan
    Search for more papers by this author
  • Yi-Hsiang Huang,

    1. Institute of Clinical Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    Search for more papers by this author
  • Yi-You Chiou,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Radiology, Taipei Veterans General Hospital, Taipei, Taiwan
    Search for more papers by this author
  • Chien-Wei Su,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    Search for more papers by this author
  • Han-Chieh Lin,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    Search for more papers by this author
  • Rheun-Chuan Lee,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Radiology, Taipei Veterans General Hospital, Taipei, Taiwan
    Search for more papers by this author
  • Jen-Huey Chiang,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Radiology, Taipei Veterans General Hospital, Taipei, Taiwan
    Search for more papers by this author
  • Teh-Ia Huo,

    Corresponding author
    1. Institute of Pharmacology, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    • Division of Gastroenterology, Department of Medicine, Taipei Veterans General Hospital, 201 Shipai Road, Section 2, Taipei 112, Taiwan
    Search for more papers by this author
    • ph: + 886 2 2871 2121, extension 3055; Fax: + 886 2 2873 9318

  • Fa-Yauh Lee,

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    Search for more papers by this author
  • Shou-Dong Lee

    1. Faculty of Medicine, National Yang-Ming University School of Medicine, Taipei, Taiwan
    2. Departments of Medicine, Taipei, Taiwan
    Search for more papers by this author

  • This study was supported by grant DOH99-TD-C-111-007 from the Center of Excellence for Cancer Research at Taipei Veterans General Hospital (Taipei, Taiwan), by grant V99C1-169 from Taipei Veterans General Hospital (Taipei, Taiwan), and by grant RD-2010-013 from National Yang-Ming University Hospital (Yilan, Taiwan).

Abstract

Radiofrequency ablation (RFA) and transarterial chemoembolization (TACE) are used to treat hepatocellular carcinoma (HCC). This study was designed to compare the long-term survival of HCC patients within the Milan criteria who underwent RFA or TACE. In all, 315 RFA patients and 215 TACE patients with HCC within the Milan criteria were analyzed. Propensity scores were generated to select matched patients. For the propensity model, 101 patients were selected from each arm of the study. Independent prognostic predictors were determined with the Cox proportional hazards model. The long-term survival was significantly better for the RFA group in the univariate survival analysis (P = 0.048). In the Cox model, the following were identified as independent predictors of poor prognosis (TACE was not): age > 69 years (P = 0.026), serum α-fetoprotein level > 20 ng/mL (P = 0.003), ascites (P < 0.001), Eastern Cooperative Oncology Group performance status ≥ 1 (P = 0.004), total tumor volume (TTV) > 8.2 cm3 (P = 0.020), and vascular invasion (P = 0.023). With similar baseline patient characteristics generated in the propensity score model, there was no significant difference in the long-term survival rates of the 2 groups of patients. A subgroup analysis showed that among patients with a TTV < 11 cm3, the RFA group had significantly better long-term survival than the TACE group (P = 0.032). In conclusion, TACE and RFA lead to comparable long-term survival rates for HCC patients within the Milan criteria. Patients with a smaller TTV (<11 cm3) are likely to benefit more from RFA treatment. Further studies are needed to compare RFA and TACE in patients with early-stage cancers. Liver Transpl 17:556–566, 2011. © 2011 AASLD.

Hepatocellular carcinoma (HCC) is a common primary liver cancer in Asia.1 The current therapeutic options for HCC are resection, liver transplantation, percutaneous chemical or thermal ablation, transarterial chemoembolization (TACE), and chemotherapy/targeted therapy. Among a variety of locoregional ablation techniques, radiofrequency ablation (RFA) is considered the treatment of choice for small HCCs.2 For patients with a single tumor ≤ 2 cm, RFA as a treatment strategy has been shown to be equally effective as surgical resection.3, 4 Recently, RFA has been reported to be safe and effective as a first-line treatment for a single HCC up to 5 cm in diameter.5 On the other hand, for HCC patients diagnosed at an intermediate or advanced stage, TACE has been suggested as a useful palliative anticancer treatment for prolonging survival.6

According to the American Association for the Study of Liver Diseases guidelines published in 2005,6 patients assigned to RFA and TACE treatment have discrete prognostic characteristics, such as tumor burden, liver cirrhosis severity, and general performance status. Until now, there have been few studies comparing RFA and TACE as primary anticancer treatments in terms of long-term survival because of different considerations in treatment allocation. To obtain a definite conclusion, a well-designed randomized trial must be conducted. However, this could be difficult because of the present treatment guidelines and ethical concerns.

The Milan criteria (a single tumor ≤ 5 cm in diameter or 3 or fewer nodules ≤ 3 cm in diameter) have been recommended as the major reference system for liver transplantation in patients with HCC.7 Patients within the Milan criteria are considered to have been diagnosed with HCC at a relatively early stage. However, transplant candidates far outnumber liver donors, and the long-term survival benefit of TACE in this particular patient group is unclear. In this study, by using a propensity score analysis to generate a matched group of HCC patients,8 we compared the long-term survival of RFA and TACE patients with HCC within the Milan criteria. To delineate the differential survival benefits of RFA and TACE in these patients, we determined prognostic predictors and performed a subgroup analysis.

PATIENTS AND METHODS

Patients

Over an 8-year period between January 2002 and January 2010, the data for 2090 patients with newly diagnosed HCC in our hospital were prospectively collected and retrospectively analyzed. Three hundred twenty HCC patients (15%) underwent RFA as the primary treatment, and 617 HCC patients (30%) underwent TACE as the primary treatment; 315 RFA patients and 215 TACE patients who met the Milan criteria were identified, and they formed the basis of this study (Fig. 1). The rationale for treatment selection was based on the physician's advice or patient's preference after a full explanation and a clinical evaluation at the time of diagnosis. The baseline information, including the patient demographics, the etiology of the underlying liver disease, the number and size of the tumor nodules, the serum biochemistries, the severity and complications of liver cirrhosis, the performance status, the cancer stage, and the treatment modality, were recorded when the diagnosis was established. The survival status of the patients was inspected every 3 months until death or dropout from the follow-up program. This study complied with the standards of the Declaration of Helsinki and with current ethical guidelines.

Figure 1.

Distribution of patients according to the primary anti-HCC treatment in the entire cohort. In all, 320 patients (15%) underwent RFA, and 617 patients (30%) underwent TACE. The Milan criteria were met by 315 patients undergoing RFA and by 215 patients undergoing TACE; 101 pairs of matched patients were selected for the propensity model.

Propensity Score Analysis

To investigate causal relationships between treatments and outcomes in an observational database (instead of a randomized controlled trial), a propensity score study was developed, and an attempt was made to reduce the bias in patient selection.8 In this study, a propensity score analysis was used to build a matched group of patients for the comparison of the long-term survival of RFA and TACE patients. Clinical variables with significant differences between RFA and TACE patients with HCC were included for propensity score generation. Logistic regression was applied to generate a continuous propensity score ranging from 0 to 1. One-to-one matching between patients undergoing RFA or TACE with the nearest neighbor method was used to select patients for the subsequent analysis so that the impact of the different treatments on mortality could be assessed.9

Diagnosis and Definitions

The diagnosis of HCC was based on findings of typical radiological features in at least 2 imaging examinations [sonography, contrast-enhanced dynamic computed tomography (CT), magnetic resonance imaging, and hepatic arterial angiography] or was histologically confirmed by needle biopsy or a single positive imaging technique associated with a serum α-fetoprotein (AFP) level > 400 ng/mL.10 Hepatitis B virus (HBV) infection was recognized as the disease underlying HCC if hepatitis B surface antigen (HBsAg) was detected serologically (radioimmunoassay kits, Abbott Laboratories, Chicago, IL). Patients were considered to be infected with hepatitis C virus (HCV) if they were seropositive for an antibody against hepatitis C virus (anti-HCV) according to a second-generation enzyme immunoassay (Abbott Laboratories). Alcoholic hepatitis was diagnosed in patients who consumed at least 40 g of alcohol daily for 5 years or more.11 The diagnosis of cirrhosis was based on typical sonographic or CT findings characteristic of cirrhosis plus at least 1 of the following: hypersplenism, ascites formation, esophageal or gastric varices, and hepatorenal syndrome.12 The severity of cirrhosis was determined according to the Child-Turcotte-Pugh (CTP) classification. Ascites was identified as free peritoneal fluid recognized by sonography or CT. The performance status was determined with the Eastern Cooperative Oncology Group performance scale, which ranges from 0 (asymptomatic) to 4 (restricted to bed). Vascular invasion was defined as the presence of a thrombus adjacent to the tumor in the portal vein with a vague boundary confirmed by at least 2 imaging modalities.13 None of the patients had received a specific anticancer treatment before the diagnosis. The total tumor volume (TTV), which is the sum of the tumor volume of every nodule, was calculated with the following mathematical equations:

equation image
equation image

The maximum radius and number of HCC nodules were measured according to CT or magnetic resonance imaging results. For HCC nodules less than 2 cm in diameter without typical dynamic imaging characteristics, liver biopsy was performed before treatment. All RFA and TACE sessions were supervised and reviewed by 3 experienced liver radiologists (Y.Y.C., J.H.C., and R.C.L.) who were blinded to the survival information of the study patients.

Treatment

RFA was performed with the following standard procedure in our hospital. After the skin was cleansed with iodized alcohol, the most appropriate approach for electrode insertion was selected. Under local anesthesia and ultrasound guidance, the RFA procedure was performed with a 17-gauge cooled-tip electrode and the Cool-Tip radiofrequency system (Radionics, Burlington, MA). The ablation was performed in an automatic impedance control mode in which the current output was automatically adjusted. As the RFA energy was being delivered, the size and shape of the hyperechoic area with a 1-cm safety margin were monitored with sonography until full coverage of the targeted lesion was completed. Post-RFA sonography was performed immediately to confirm that there was no definite hemorrhaging or hematoma.

The Seldinger technique of arterial embolization was administered as the standard TACE procedure.14 LCN equipment (GE Medical System, Waukesha, WI) was used for the selection of proper or common hepatic artery angiography. A 4-Fr catheter (Terumo, Tokyo, Japan) was used for femoral artery puncture. HCC nodules were localized with hepatic arteriography and superior mesenteric arterial portovenography. Tumor staining (HCC tumor vascularity) was investigated with 100 to 150 mL of a radiocontrast agent (Telebrix, Laboratoire Guerbet, Aulnay-Sous-Bois, France) injected by a power injector (CT9000 ADV, Liebel-Flarsheim, St. Louis, MO). The infusion of a mixture of 20 to 30 mg of adriamycin (Carlo Erba, Milan, Italy) and 5 to 10 mL of Lipiodol (Laboratoire Guerbet) was performed after the arteries supplying the tumor were catheterized with a 3-Fr catheter superselectively. A sufficient amount of the emulsion and 2- to 3-mm strips of Gelfoam (Upjohn, Kalamazoo, MI) were delivered to the tumoral area until complete flow stagnation was achieved.

After RFA or TACE, laboratory tests (including blood cell counts and serum biochemistry tests) were performed during the hospitalization period. Posttreatment follow-up examinations, which included liver sonography, contrast-enhanced dynamic imaging, and measurements of serum biochemistries and AFP levels, were performed every 3 months. Tumor recurrence was defined as the detection of either a viable tumor (not completely necrotic) after treatment or new lesions that developed elsewhere in the liver during the follow-up. When tumor recurrence was confirmed, RFA or TACE was performed repeatedly if this was indicated.

Statistical Methods

The chi-square test or Fisher's exact test (2-tailed) was used for categorical data comparison, and the Mann-Whitney or Kruskal-Wallis ranked-sum test was used for continuous data. For the analysis of prognostic predictors, continuous variables were split by the median values and were treated as dichotomous covariates. In subgroup analysis, median values were used as the cutoffs to dichotomize continuous variables. The Kaplan-Meier method with log-rank testing was applied to compare the survival distributions. Factors that were significant in the univariate survival analysis were introduced into the multivariate Cox proportional hazards model to determine the adjusted risk ratios and 95% confidence intervals (CIs). All statistical analyses were conducted with SPSS for Windows, version 16 (SPSS, Inc., Chicago, IL). A P value less than 0.05 was considered statistically significant.

Abbreviations:

AFP, α-fetoprotein; anti-HCV, antibody against hepatitis C virus; BCLC, Barcelona Clinic Liver Cancer; CI, confidence interval; CLIP, Cancer of the Liver Italian Program; CT, computed tomography; CTP, Child-Turcotte-Pugh; HBsAg, hepatitis B surface antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; INR, international normalized ratio; NA, not applicable; PAIT, percutaneous acetic acid injection therapy; PEIT, percutaneous ethanol injection therapy; PT, prothrombin time; RFA, radiofrequency ablation; RR, relative risk; SD, standard deviation; TACE, transarterial chemoembolization; TTV, total tumor volume.

RESULTS

Characteristics of All Study Patients

A comparison of the baseline demographics of the RFA and TACE patients within the Milan criteria is shown in Table 1. There were no significant differences between the 2 groups in the following: age; sex; prevalence of HBV, HCV, or alcoholism; severity of liver cirrhosis (CTP classification); ascites formation; performance status; and diabetes mellitus (all P values > 0.05). Patients in the TACE group had significantly larger tumors (P < 0.001), more tumor nodules (P = 0.004), larger TTVs (P < 0.001), and a higher prevalence of vascular invasion (P = 0.001). Patients in this group also had significantly higher serum bilirubin levels (P = 0.012) and a tendency toward higher serum AFP levels (P = 0.074). According to the Barcelona Clinic Liver Cancer (BCLC) and Cancer of the Liver Italian Program (CLIP) staging systems,6, 15 patients undergoing TACE had significantly more advanced cancer than patients undergoing RFA (P < 0.001 and P = 0.001, respectively). The TTV distributions of the study patients according to the number of tumor nodules and the size of the main tumor are depicted in Table 2. Patients with a single tumor nodule > 3 cm in diameter (ie, 3.1-5 cm in diameter) had significantly larger TTVs in comparison with patients with a single tumor nodule ≤ 3 cm in diameter (P < 0.001). In addition, among patients with tumor nodules ≤ 3 cm in diameter, the TTV tended to be larger when the number of tumor nodules increased from 1 to 3 (P < 0.001).

Table 1. Comparison of the Baseline Demographics of Patients Who Underwent RFA or TACE
VariableRFA Patients (n = 315)TACE Patients (n = 215)P Value
Age (years), mean ± SD67 ± 1168 ± 100.172
Male, n (%)205 (65)152 (71)0.176
Positive for HBsAg, n (%)153 (49)89 (41)0.103
Positive for anti-HCV, n (%)138 (44)103 (48)0.352
HBV and HCV, n (%)12 (4)10 (5)0.662
Alcoholism, n (%)30 (10)22 (10)0.788
CTP class, n (%)  0.531
 A263 (83)175 (81) 
 B52 (17)40 (19) 
Ascites, n (%)40 (13)28 (13)0.913
Tumor size > 3 cm, n (%)68 (22)76 (35)<0.001
Tumor number (1/2/3), %81/14/571/17/120.004
TTV (cm3), mean ± SD (median)12 ± 14.4 (6.4)17.9 ± 18.2 (10.3)<0.001
Performance status 0, n (%)262 (83)175 (81)0.597
Vascular invasion, n (%)4 (1)14 (7)0.001
Biochemistries, mean ± SD   
 Albumin (g/dL)3.78 ± 0.63.71 ± 0.50.120
 Bilirubin (mg/dL)1.05 ± 1.01.06 ± 0.60.012
 Creatinine (mg/dL)1.17 ± 1.11.08 ± 0.60.610
 PT INR1.10 ± 0.51.07 ± 0.10.254
 Sodium (mmol/L)139.1 ± 3.5139.8 ± 3.10.048
AFP (ng/mL), mean ± SD (median)293 ± 1277 (19)1629 ± 18,276 (25)0.074
Diabetes mellitus, n (%)81 (26)50 (23)0.519
BCLC (stage 0/A/other), %29/53/1810/62/28<0.001
CLIP score (0/1/other), %61/31/844/43/130.001
Table 2. TTV Distribution According to the Size and Number of HCC Nodules
Main Tumor SizeNumber of Tumor NodulesP Value*
123
  • NOTE: The data are presented as means and SDs (with medians in parentheses).

  • *

    Comparison of the 3 groups of patients with small HCCs (≤3 cm) and 1, 2, or 3 tumor nodules (Kruskal-Wallis test).

  • Comparison of patients with a single, small tumor nodule (≤3 cm) and patients with a single, large tumor nodule (3.1-5 cm; Mann-Whitney test).

≤3 cm5.4 ± 4.3 cm3 (4.2 cm3)10.4 ± 8.9 cm3 (8.1 cm3)11.6 ± 15 cm3 (8.1 cm3)<0.001
3.1-5 cm35.8 ± 15.8 cm3 (33.5 cm3)NANA 
P Value<0.001NANA 

Survival Analysis of All Study Patients

A comparison of the long-term survival of the RFA and TACE groups is presented in Fig. 2. During a mean follow-up period of 26 ± 17 months, 69 patients undergoing RFA (22%) and 67 patients undergoing TACE (31%) died. The estimated 1-, 3-, and 5-year survival rates of RFA and TACE patients were 93% and 89%, 72% and 63%, and 55% and 43%, respectively (P = 0.048). Factors that could be linked to survival (age; gender; etiology of liver disease; CTP classification; number, size, and total volume of tumor nodules; serum biochemistries; diabetes mellitus; performance status; vascular invasion; and treatment modalities) were investigated in the survival analysis. The prognostic factors predicting an increased risk of mortality included the following: age > 69 years (P = 0.035), HCV infection (P = 0.031), ascites (P < 0.001), CTP class B (P < 0.001), TTV > 8.2 cm3 (P = 0.012), serum bilirubin level > 0.8 mg/dL (P = 0.002), serum albumin level < 3.8 g/dL (P < 0.001), serum AFP level > 20 ng/mL (P = 0.014), performance status ≥ 1 (P < 0.001), TACE treatment (P = 0.048), vascular invasion (P = 0.021), and prothrombin time (PT) international normalized ratio (INR) > 1 (P = 0.01). As shown in Table 3, 6 factors were identified as independent predictors of poor prognosis in the adjusted Cox proportional hazards model (TACE was not): serum AFP level > 20 ng/mL [relative risk (RR) = 1.69, 95% CI = 1.20-2.39, P = 0.003], age > 69 years (RR = 1.49, 95% CI = 1.04-2.08, P = 0.026), ascites (RR = 3.13, 95% CI = 1.90-5.15, P < 0.001), performance status ≥ 1 (RR = 1.82, 95% CI = 1.16-2.87, P = 0.004), TTV > 8.2 cm3 (RR = 1.51, 95% CI = 1.07-2.39, P = 0.020), and vascular invasion (RR = 2.37, 95% CI = 1.13-4.98, P = 0.023).

Figure 2.

Comparison of the long-term survival distributions of HCC patients within the Milan criteria according to RFA and TACE. Significantly better survival was found for HCC patients undergoing RFA.

Table 3. Independent Prognostic Predictors for All Patients and for Patients Selected for the Propensity Score Analysis in the Cox Proportional Hazards Model
 RR95% CIP Value
All patients (n = 530)   
 AFP > 20 ng/mL1.691.20-2.390.003
 Age > 69 years old1.491.04-2.080.026
 Ascites3.131.90-5.15<0.001
 Performance status ≥ 11.821.16-2.870.004
 TTV > 8.2 cm31.511.07-2.390.020
 Vascular invasion2.371.13-4.980.023
Patients in the propensity model (n = 202)   
 Performance status ≥ 13.502.15-5.68<0.001
 Vascular invasion3.411.45-8.000.005

Characteristics of the Patients Selected for the Propensity Model

Variables that showed significant baseline differences between RFA and TACE patients were selected for the propensity score model. These variables included the following: the number, size, and total volume of the tumor nodules; vascular invasion; and the serum bilirubin and sodium levels. One hundred one patients from both the RFA treatment arm and the TACE treatment arm were selected (Table 4 and Fig. 1). There were no significant baseline differences in age, sex, etiology of liver disease, tumor-related parameters, severity of cirrhosis, performance status, vascular invasion, or cancer staging systems between these 2 groups of patients. A matched study population was constructed to compare the long-term survival of RFA and TACE patients.

Table 4. Comparison of the Baseline Demographics of Patients Undergoing RFA or TACE in the Propensity Score Model
VariableRFA Patients (n = 101)TACE Patients (n = 101)P Value
Age (years), mean ± SD66 ± 1168 ± 110.169
Male, n (%)59 (58)67 (66)0.245
Positive for HBsAg, n (%)46 (46)40 (40)0.393
Positive for anti-HCV, n (%)52 (52)50 (50)0.778
HBV and HCV, n (%)7 (7)6 (6)1
Alcoholism, n (%)13 (13)12 (12)0.831
CTP class A, n (%)74 (73)73 (72)0.874
Ascites, n (%)26 (26)22 (22)0.508
Tumor size > 3 cm, n (%)36 (36)32 (32)0.551
Tumor number (1/2/3), %63/27/1073/15/120.115
TTV (cm3), mean ± SD (median)16.6 ± 1.8 (11.5)17.2 ± 1.9 (10.3)0.935
Performance status = 0, n (%)69 (68)76 (75)0.274
Vascular invasion, n (%)3 (3)6 (6)0.498
Biochemistries, mean ± SD   
 Albumin (g/dL)3.65 ± 0.63.67 ± 0.50.924
 Bilirubin (mg/dL)1.25 ± 1.31.23 ± 0.80.078
 Creatinine (mg/dL)1.15 ± 1.01.11 ± 0.90.914
 PT INR1.10 ± 0.21.08 ± 0.10.950
 Sodium (mmol/L)139.5 ± 3.6139.8 ± 3.70.567
AFP (ng/mL), mean ± SD (median)320 ± 1073 (22)3175 ± 26,701 (25)0.784
Diabetes mellitus, n (%)27 (27)23 (23)0.514
BCLC stage (0/A/other), %15/50/3513/54/330.837
CLIP score (0/1/other), %35/47/1840/47/130.514

Survival Analysis of the Patients Selected for the Propensity Model

In the propensity model, 28 patients in the RFA arm subsequently underwent TACE at some point during the follow-up period. In all, 229 and 44 sessions of RFA and TACE, respectively, were performed for patients in the RFA group. On the other hand, for patients in the TACE arm, 14 subjects subsequently underwent RFA at some point during the follow-up period; in all, 251 and 24 sessions of TACE and RFA, respectively, were performed. None of the 202 patients subsequently underwent liver transplantation during the follow-up period. A comparison of the long-term survival of the 2 groups of patients in the propensity model is shown in Fig. 3. There was no significant difference in the survival of RFA and TACE patients. During a mean follow-up period of 24 ± 17 months, 32 patients undergoing RFA (32%) and 37 patients (37%) undergoing TACE died. The estimated 1-, 3- and 5-year survival rates of RFA and TACE patients with HCC were 85% and 86%, 60% and 55%, and 41% and 36%, respectively (P = 0.476). Potential prognostic predictors were investigated. Four predictors associated with an increased risk of mortality were identified in the univariate survival analysis: ascites (P < 0.001), CTP class B (P < 0.001), performance status ≥ 1 (P < 0.001), and vascular invasion (P = 0.015). In the adjusted Cox proportional hazards model, performance status ≥ 1 (RR = 3.50, 95% CI = 2.15-5.68, P < 0.001) and vascular invasion (RR = 3.41, 95% CI = 1.45-8.00, P = 0.005) were identified as independent predictors of poor prognosis (Table 3).

Figure 3.

Comparison of the long-term survival distributions of HCC patients within the Milan criteria according to RFA and TACE. The patients were selected for the propensity score model with the one-to-one matching method. There were no significant survival differences between these 2 groups of patients.

In the subgroup analysis within the propensity model, patients were stratified according to the main tumor size (≤3 or 3.1-5 cm), the multiplicity of tumor nodules, the TTV (median level = 11 cm3), and the CTP classification. As illustrated in Figs. 4 to 6, there were no significant differences in the long-term survival of patients undergoing RFA or TACE in the different subgroups stratified by the tumor size, the number of HCC nodules (single or multiple), or the CTP classification (all P values > 0.05). However, for patients with a TTV < 11 cm3, RFA provided a significantly better survival benefit than TACE (Fig. 7; P = 0.032).On the other hand, there was no significant survival difference between the RFA and TACE groups among patients with a TTV > 11 cm3 (P = 0.412).

Figure 4.

Subgroup analysis of the overall survival of HCC patients undergoing RFA or TACE according to the main tumor size (≤3 or 3.1-5 cm). There were no significant differences in the long-term survival of the patients undergoing RFA or TACE in the 2 groups.

Figure 5.

Subgroup analysis of the overall survival of HCC patients undergoing RFA or TACE according to the number of tumor nodules (single or multiple). There were no significant differences in the long-term survival of patients undergoing RFA or TACE in the 2 groups.

Figure 6.

Subgroup analysis of the overall survival of HCC patients undergoing RFA or TACE according to the CTP classification (class A or B). There were no significant differences in the long-term survival of patients undergoing RFA or TACE in the 2 groups.

Figure 7.

Subgroup analysis of the overall survival of HCC patients undergoing RFA or TACE according to TTV (<11 or >11 cm3). RFA provided significantly better long-term survival than TACE for HCC patients with a TTV < 11 cm3. There were no significant differences in the long-term survival of patients undergoing RFA or TACE with a TTV > 11 cm3.

Analysis of the Cumulative Recurrence of Patients Selected for the Propensity Model

A comparison of the cumulative tumor recurrence in the TACE and RFA groups is presented in Fig. 8. Patients undergoing TACE in the propensity model had a significantly higher cumulative recurrence rate than patients undergoing RFA (P = 0.023). The subgroup analysis showed that among patients with smaller TTVs (<11 cm3), RFA led to significantly less cumulative recurrence than TACE (P = 0.015; Fig. 9). There was no significant difference in the rates of tumor recurrence between the RFA and TACE groups among patients with a TTV > 11 cm3 in the propensity model (P = 0.425).

Figure 8.

Comparison of the cumulative tumor recurrence of HCC patients within the Milan criteria according to RFA and TACE. The patients were selected for the propensity score model. The TACE group had a significantly higher rate of cumulative tumor recurrence than the RFA group.

Figure 9.

Subgroup analysis of the cumulative tumor recurrence of HCC patients undergoing RFA or TACE according to TTV (<11 or >11 cm3). RFA led to significantly less cumulative tumor recurrence than TACE among HCC patients with a TTV < 11 cm3. There were no significant differences in the tumor recurrence of patients undergoing RFA or TACE with a TTV > 11 cm3.

DISCUSSION

For patients with unresectable HCC within the Milan criteria, liver transplantation and percutaneous ablation therapies are generally considered the principal choices for a cure. However, as a curative treatment for small HCCs, liver transplantation is largely limited by the donor organ shortage.16 Although RFA is often considered the treatment of choice for small HCCs, a substantial number of HCC patients have undergone TACE as the first-line treatment for tumor control and survival prolongation.17 In this study, we investigated the survival benefits of RFA and TACE with a propensity score model. We found that for HCC patients who met the Milan criteria, TACE provided survival benefits similar to those of RFA. Moreover, in comparison with the number and size of tumor nodules, TTV demonstrated a better discriminative ability to identify a subset of patients who may benefit more from RFA treatment.

The survival of patients with HCC is highly associated with the tumor burden, the liver functional reserve, and the therapeutic strategy. There has been intense debate about the selection of treatments that are more beneficial to HCC patients who have different baseline characteristics. A randomized controlled trial is usually the standard method for comparing RFA and TACE in HCC patients who meet the Milan criteria. However, because of the current American Association for the Study of Liver Diseases guidelines, which suggest that BCLC stage A patients should consider RFA rather than TACE as the first-line treatment,6 it is quite difficult to conduct a head-to-head comparison in a randomized fashion. With a large patient cohort, a propensity score analysis is a useful strategy for minimizing selection bias when we are interpreting the impact of different treatments on survival in an observational study.8, 9 Our results provide important clinical information that may help researchers to improve the design of future clinical trials.

Encouraging results have been found for RFA in HCC patients with a single tumor less than 4 cm in diameter.18, 19 Our results show that for HCC patients within the Milan criteria, RFA led to better long-term survival than TACE in the univariate analysis (Fig. 2); however, RFA was not an independent favorable prognostic factor in the adjusted Cox model, and this suggests that other factors may play more critical roles in determining long-term survival. Importantly, the propensity score analysis consistently showed that RFA was not superior to TACE in terms of survival prolongation. A probable reason for this similar influence on survival is that RFA has a less satisfactory effect on medium tumors (3.1-5 cm in diameter) and multiple tumors. Studies have shown that HCC patients with larger tumors have lower rates of complete ablation and shortened survival.19-21 In addition, multinodular tumors also may diminish the efficacy of RFA and lead to shortened survival for HCC patients.5, 22

Interestingly, in the subgroup analysis of the propensity score model, HCC patients undergoing RFA with a TTV < 11 cm3 (equivalent to a single nodule 2.8 cm in diameter) instead of a main tumor ≤ 3 cm in diameter or a single HCC nodule had significantly better long-term survival than those undergoing TACE (Fig. 4, 5, 7). This finding not only suggests that HCC patients with a smaller tumor burden could be better candidates for RFA but also implies that TTV might be a better parameter for describing the HCC tumor burden than the number or size of tumor nodules. For HCC patients awaiting liver transplantation, the calculated TTV has been proposed to be a better parameter than the current standard (the number and size of tumor nodules in the Milan criteria).23 Our recent study has shown that TTV is a more reliable prognostic predictor in a newly proposed staging system for HCC.24 Therefore, our findings may help in identifying proper candidates for RFA treatment.25, 26

In this study, the 3- and 5-year survival rates for 315 patients undergoing RFA were 72% and 55%, respectively. These rates are consistent with those of a large cohort study from Japan: for 319 patients who underwent RFA as the primary treatment for HCC, the survival rates at 3 and 5 years were 78% and 54%, respectively.27 In comparison with patients in another study from France,5 the patients treated with RFA in our study showed slightly superior long-term survival (60% and 40% for the French patients at 3 and 5 years, respectively). Therefore, the similar long-term survival rates of the RFA and TACE patients in this study should not be attributed to relatively poor outcomes in the RFA group. Alternatively, TACE is recommended for patients who cannot undergo curative treatments such as hepatic resection, liver transplantation, and percutaneous ablation therapy because these curative treatments have been proposed to be better treatment options for BCLC stage A patients. A few studies have investigated the clinical outcomes of HCC patients who met the Milan criteria and underwent TACE as the first-line treatment. In a nationwide survey from Japan, for tumor-node-metastasis stage I patients treated with TACE, the 3- and 5-year survival rates were 72% and 47%, respectively.28, 29 Additionally, tumor-node-metastasis stage I HCC patients with an adequate liver reserve demonstrated an excellent 5-year survival rate of 52%. Increasing evidence supports the idea that with careful patient selection, TACE could provide survival benefits similar to those of RFA for patients with early-stage HCC.

The application of liver transplantation to patients with small HCCs is often hampered by the lack of donor organs.30 In our hospital, most patients who meet the Milan criteria still undergo RFA or TACE as the primary anticancer treatment; this is distinctly different from the practice of Western society.17, 31 Our findings suggest that both RFA and TACE are feasible treatment options for HCC within the Milan criteria.

Poor performance status and older age were unfavorable prognostic predictors in the survival analysis for all patients (Table 3). Similar results have been reported by other study groups, and this tendency could be explained by a correlation between age, performance status, and an increased prevalence of comorbidities.9, 32 In agreement with previous studies, we found that a higher serum AFP level (>20 ng/mL), ascites, a higher tumor burden (TTV > 8.2 cm3), and vascular invasion were significantly associated with increased mortality among HCC patients.33, 34 In patients selected for the propensity model, only performance status and vascular invasion were identified as independent predictors of poor prognosis. Notably, the prevalence of vascular invasion has been reported to be as high as 59% in surgically resected specimens.35 Although vascular invasion is a common event and indicates a relatively advanced stage of HCC, it is not a contraindication to TACE or RFA. Taken together, our results suggest that in HCC patients with smaller tumor burdens, a detrimental general condition and aggressive tumor behavior may play pivotal roles in outcome predictions.

Although patients in the TACE group had a higher tumor recurrence rate in the propensity model, there was no significant difference in overall survival between the 2 groups. Notably, in the subgroup analysis, RFA was related to less tumor recurrence only in patients with smaller TTVs (<11 cm3). Similar cumulative recurrence rates in patients with larger tumor burdens could have contributed to the comparable long-term survival rates of the 2 groups in the propensity model.

This study has a few potential limitations. First, the initial and follow-up treatment modalities were different in some patients and depended on the tumor progression status and the severity of liver cirrhosis during the follow-up period. Second, in this single-center study, the HBV rate was higher than published rates from Western countries and Japan, so external validation from independent study groups is needed. Third, because no HCC patient undergoing RFA or TACE subsequently underwent liver transplantation as a salvage treatment, the generalizability of this study may be limited, and our results may not be readily applicable to centers with a high volume of liver transplants.

In conclusion, our results indicate that TACE is an effective treatment alternative, and TACE and RFA provide similar long-term survival benefits to HCC patients within the Milan criteria. HCC patients with smaller TTVs (<11 cm3) are likely to benefit more from RFA therapy. Further studies are urgently needed to compare the therapeutic efficacy of RFA and TACE in patients who have small HCCs or belong to BCLC stage A.

Acknowledgements

The authors thank Ms. Ya-Ju Tsai for her help with the data collection and manuscript preparation.

Ancillary