Comparison of 7 staging systems for patients with hepatocellular carcinoma undergoing transarterial chemoembolization

Authors


Abstract

BACKGROUND.

Many liver staging systems have been proposed for patients with hepatocellular carcinoma after locoregional therapy; however, controversies persist regarding which system is the best. In this study, the authors compared the performance of 7 staging systems in a cohort of patients with hepatocellular carcinoma who underwent transarterial chemoembolization.

METHODS.

In total, 131 patients with hepatocellular carcinoma who underwent transarterial chemoembolization between August 1998 and February 2005 were included in the study. Demographic, laboratory, and tumor characteristics were determined at diagnosis and before therapy. At the time of censorship, 109 patients had died (83.2%). Predictors of survival were identified by using the Cox proportional hazards model. The likelihood-ratio chi-square statistic and the Akaike Information Criterion were calculated for 7 prognostic systems to evaluate their discriminatory ability. Comparisons of the survival rate between each stage were performed to evaluate the monotonicity of the gradients using Kaplan-Meier estimation and the log-rank test.

RESULTS.

The 5-year survival rate for the entire cohort was 13.6%. The independent predictors of survival were serum albumin level (≤3.4 g/dL), the presence of ascites, serum α-fetoprotein level (>60 ng/mL), and portal or hepatic vein tumor thrombosis (P = .001, P = .001, P = .004, and P = .000, respectively). The Cancer of the Liver Italian Program classification system was superior to the other 6 prognostic systems regarding discriminatory ability and the monotonicity of the gradients.

CONCLUSIONS.

In this comparison of many staging systems, the Cancer of Liver Italian Program system provided the best prognostic stratification for a cohort the patients with hepatocellular carcinoma who underwent transarterial chemoembolization. Cancer 2008. © 2007 American Cancer Society.

Many liver staging systems have been proposed to determine treatment options and to predict the prognosis of patients with hepatocellular carcinoma (HCC).1–7 However, it is neither easy nor feasible to satisfy the aspects of both treatment and prognosis simultaneously by using the available staging systems. Primary staging systems like the Barcelona Clinic Liver Cancer (BCLC) system or the Japanese Integrated Staging (JIS) system have been developed based on databases from surgically oriented centers.5, 6 Unfortunately, surgical indications for HCC usually are very limited, and transcatheter arterial chemoembolization (TACE) has been practiced worldwide as a palliative treatment.8–11 It is not clear whether the systems that are appropriate for surgical group also can work well for predicting the survival of patients who undergo TACE.

Recently, it has been reported that the Child-Pugh nominal classification12 and 2 modified systems from the Cancer of the Liver Italian Program (CLIP)13 and the JIS system14 are effective for predicting the survival of patients who are undergoing locoregional therapy. However, controversies persist regarding which is the most appropriate prognostic system for these patients.

In the current study, we tried to compare 7 liver staging systems to determine the most suitable system for patients with HCC who were undergoing TACE, including the primary staging systems and the newly proposed, modified systems from the JIS and the CLIP.

MATERIALS AND METHODS

Patient Selection

The requirements for entering the study were as follows 1) prothrombin time ratio (ie, normal time divided by patient time) >40%, 2) platelet count >40,000/mm3 (40 × 109/L), 3) newly diagnosed patients with no previous treatment for the HCC, 4) ineligible for surgical resection or transplantation, 5) Child-Pugh class A or B, 6) no evidence of extrahepatic tumor invasion, and 7) the patient agreed to undergo TACE. Surgically ineligible patients were those who were adjudged unresectable by surgeons because of the functional status and volume of the liver or those who refused surgical therapies. From August 1, 1998 to February 28, 2005, in total, 131 patients who had HCC and underwent TACE in our hospital met the inclusion criteria and were evaluated prospectively. All medical records were reviewed retrospectively. For the patients with portal vein thrombosis, TACE was performed taking into consideration the functional status of the liver. All 131 patients were monitored from the time of diagnosis to the date of death or of data collection if they remained alive. All patients had known 2-year survival and long-term survival status to be included in the survival analysis, and no patients were excluded. The study was censored on March 1, 2007. All but 5 patients were men. The patients ranged in age from 43 years to 82 years (mean age ± standard deviation, 64.5 ± 8.4 years). The median number of tumors was 2 (range, 1–18 tumors), and the median greatest tumor dimension was 3.6 cm (range, 1–20 cm). Other characteristics are shown in Table 1. This study complied with the standards of the Declaration of Helsinki and current ethical guidelines.

Table 1. Pretreatment Baseline Patient and Tumor Characteristics
CharacteristicNo. of patients (%)Median value*Range
  • AST indicates aspartate aminotransferase; ALT, alanine aminotransferase; HBV, hepatitis B virus; HCV, hepatitis C virus.

  • *

    For continuous variables, the cutoff was set at the median value.

Age: <66/≥66 y64/67 (48.9/51.1)66 y44–82 y
Sex: Men/women126/5 (96.2/3.8) 
Underlying causes: HBV/HCV/HBV and HCV/others62/34/3/32 (47.3/26/2.3/24.4) 
Performance status: 0/1 or 266/65 (50.4/49.6) 
Serum AST level: ≤50/>50 U/L74/57 (56.5/43.5)489–191
Serum ALT level: ≤40/>40 U/L66/65 (50.4/49.6)4010–195
Serum albumin level: >3.4/≤3.4 g/dL61/70 (45.8/54.2)3.42–4.5
Serum total bilirubin level: <1/≥1 mg/dL58/73 (44.3/55.7)10.3–2.8
Ascites: No/yes103/28 (78.6/22.4) 
Hepatic encephalopathy: No/yes129/2 (98.5/1.5) 
Child-Pugh class: A/B92/39 (70.2/29.8) 
Portal hypertension: No/yes69/62 (52.7/47.3) 
Platelet count: <100,000/≥100,000 × 109/L71/60 (54.2/45.8)105,00041,000–368,000
Serum α-fetoprotein level: ≤60/>60 ng/mL64/67 (48.9/51.1)60.461.8–242,000
Tumor number: Single/multiple62/69 (47.7/52.3)21–18
Tumor maximal dimension: ≤4/>5 cm98/33 (74.8/25.2)3.6cm1–19
Tumor distribution: Unipolar/bipolar98/33 (74.8/25.2) 
Tumor thrombosis in portal or hepatic vein: No/yes118/13 (90.1/9.9) 
 Main portal trunk1 (0.8)  
 Right or left main portal vein6 (4.6)  
 Second or distal portal branch4 (3.1)  
Lymph node metastasis: No/yes128/3 (97.7/2.3) 

The diagnosis of HCC was verified histologically by percutaneous needle biopsy in 13 patients. For the other patients, the diagnosis was established based on the characteristic radiologic features in ≥2 image examinations, including ultrasound, contrast-enhanced dynamic computed tomography (CT), magnetic resonance imaging, and hepatic angiography (for hypervascular tumors that were observed with ≥2 imaging modalities), or by a single positive imaging technique associated with a serum α-fetoprotein level >400 ng/mL.15 In the majority of patients, the etiology of cirrhosis was chronic viral hepatitis B or hepatitis C (Table 1).

Treatment Procedures

All patients had an enhanced dynamic CT scan obtained within 4 weeks before TACE. Informed consent was obtained from all patients before the procedure. For TACE, embolization through the segmental or subsegmental arterial feeders was performed whenever possible. When the tumors were supplied by multiple segmental arterial feeders or when catheterization of a segmental tumor feeder failed, TACE usually was performed through the right or left hepatic artery. For patients with bilobar disease, we tried to treat all tumors by superselective TACE if possible. If liver function was not good as Child-Pugh class B or C, then larger tumors were treated selectively to preserve liver function.

We administered iodized oil-doxorubicin hydrochloride (Adriamycin; Donga, Seoul, Korea) emulsion into the feeders. The volume of the iodized oil ranged from 3 mL to 12 mL, and the amount of doxorubicin ranged from 20 mg to 100 mg. Once the flow became sluggish, gelatin sponge particles (Gelfoam; Cutanplast, Mascia Brunelli S.p.A., Italy) that were mixed with mitomycin-C (Kyowa; Kyowa Hakko Kogyo, Tokyo, Japan) and contrast material were administered into the feeders until the blood flow stopped completely.

Analysis of the Prognostic Factors for Overall Survival and Image Interpretation

The overall survival rate was compared for 19 possible prognostic factors, including 14 baseline patient factors (patient age, sex, performance status, portal hypertension, presence of ascites, hepatitis B infection, hepatitis C infection, aspartate aminotransferase level, alanine aminotransferase level, albumin level, total bilirubin level, platelet count, prothrombin time, and Child-Pugh class) and 5 baseline tumor factors (serum α-fetoprotein level, tumor location [unilobar or bilobar], multiplicity of tumors, greatest tumor dimension >5cm, and portal or hepatic vein tumor thrombosis).

CT examinations were performed with an 8-slice multidetector CT scanner (Lightspeed; GE Medical Systems, Milwaukee, Wis) with 5-mm collimation and a 17.5-mm-per-second table speed or with a single-detector helical scanner (Prospeed Advantage; GE Medical Systems) with 10-mm collimation and a 10-mm-per-second table speed. All patients underwent both nonenhanced and contrast-enhanced 3-phase helical CT imaging 1 month after TACE.

Portal hypertension was defined by the presence of either esophageal varices or splenomegaly with a platelet count <100,000/mL.5 Performance status assessment followed the guidelines of the Eastern Cooperative Oncology Group.16 The number of tumors was determined from the pretreatment CT scan. Tumor size was determined as the greatest dimension of the tumor measured on the pretreatment CT scan. Vascular invasion was assessed by dynamic CT and hepatic angiography. Lymph node invasion and distant metastases were assessed with a routine screening study, such as ultrasonography, dynamic CT, and a chest x-ray. Bone scintigraphy or a brain CT was obtained if suggestive symptoms were present. Abdominal lymph nodes with a smallest dimension ≥10 mm were regarded as metastatic lymph nodes.

A residual, viable tumor was judged to be present when an enhanced portion was observed within or around of the original mass on a 1-month follow-up CT scan. If no definite evidence of residual tumor was noted on the 1-month follow-up CT, then a 3-phase contrast-enhanced CT was obtained at a 3- or 4-month interval thereafter. Local tumor progression was judged present when an enhanced portion was observed within or at the margin of the original mass on the follow-up CT scan. For local tumor progression, an additional TACE or ablative therapy was performed. For small recurrent tumors that measured ≤3 cm in greatest dimension, radiofrequency ablation was considered first if the tumor was not located in a difficult location.

The 7 staging systems that we analyzed for this study included the Child-Pugh nominal classification, the Okuda system, the BCLC system, the CLIP system, the Model for End-stage Liver Disease (MELD)-based modified CLIP system, the JIS system, and the MELD-based modified JIS system (Table 2 and 3).1–6, 12, 13, 17 All staging systems that are known as effective for the prediction of prognosis in patients with HCC who are undergoing locoregional therapy were included in this study. However, prognostic scoring systems in which the number of stratifications is >7, like the Tokyo scoring system or the Chinese University Prognostic Index, were not included.7, 18 Two radiologists with 11 years and 6 years of experience (M.Y.K. and Y.O.P., respectively) interpreted the CT scans and angiographic images independently, and final decisions were reached by consensus.

Table 2. The Cancer of the Liver Italian Program (CLIP) Classification From the CLIP Investigators and a Modified CLIP Staging System Based on the Model for End-stage Liver Disease*
VariableCLIP score
OriginalModified
  • MELD indicates Model for End-stage Liver Disease.

  • *

    Adapted from Huo TI, Huang YH, Lin HC, et al. Am J Gastroenterol. 2006;101:975–982.13

  • MELD score = 10[0.957Ln (Scr) + 0.378 Ln (t-bil) + 1.12Ln (INR) + 0.643 (cause of cirrhosis)], where Ln indicates natural logarithm; Scr, serum creatinine (mg/dL); t-bil, total bilirubin; INR, international normalized ratio. Note that, for disease caused by cirrhosis, a score of 0 was used for alcohol-related liver disease or cholestatic liver disease, and a score of 1 was used for all other causes.

Child-Pugh class
 A00
 B11
 C22
MELD score
 <100
 10–141
 >142
Tumor morphology
 Uninodular and extension ≤50%00
 Multinodular and extension ≤50%11
 Massive or extension >50%22
Serum α-fetoprotein, ng/mL
 <40000
 ≥40011
Portal vein thrombosis
 No00
 Yes11
Table 3. The Japan Integrated Staging System (JIS) and the Model for End-stage Liver Disease-based, Modified JIS System*
VariableJIS score
OriginalModified
  • TNM indicates tumor, lymph node, metastasis classification; MELD, Model for End-stage Liver Disease.

  • *

    Adapted from Huo TI, Lin HC, Hwang YH, et al. Cancer. 2006;107:141–148.12

  • MELD score = 10[0.957Ln (Scr) + 0.378 Ln (t-bil) + 1.12Ln (INR) + 0.643 (cause of cirrhosis)], where Ln indicates natural logarithm; Scr, serum creatinine (mg/dL); t-bil, total bilirubin; INR, international normalized ratio. Note that, for disease caused by cirrhosis, a score of 0 was used for alcohol-related liver disease or cholestatic liver disease, and a score of 1 was used for all other causes.

TNM stage: Liver Cancer Study Group of Japan
 I00
 II11
 III22
 IV33
Child-Pugh class
 A0
 B1
 C2
MELD score
 <100
 10–141
 >142

Statistical Analysis

Overall survival was the only endpoint used in the analysis. Survival was defined as the time elapsed from the date of the initial diagnosis to either the date of death or the date of the last follow-up information. For the 19 factors that potentially predicted the overall survival rate, univariate and multivariate analysis was performed using the Cox proportional hazard model. The assumption for the Cox proportional hazard model was checked for the covariates by log-log plot and statistical test (P < .05). The Cox analysis included the time-dependent covariates in the model.19 Parameters that proved to be significant in univariate analysis were tested subsequently with the multivariate Cox proportional hazard model. The forward conditional selection technique was used for the Cox multivariate analysis. For continuous variables, the cutoff was set at the median value in the univariate and multivariate analyses.

The suggested criteria for assessing the performance of any prognostic system were 1) homogeneity within the same group, 2) discriminatory ability between different groups, and 3) monotonicity of gradients.20 The likelihood ratio chi-square test was used to measure homogeneity and the monotonicity of the gradients and was correlated with a Cox proportional hazard regression model.21, 22 The Akaike information criteria (AIC) value also was calculated for each system to measure its discrimination ability22, 23: the smaller the value of this statistic, the better the model. We also evaluated the independent contribution of the classifications considered in explaining survival by comparing the likelihood-ratio chi-square test and the AIC values related to the full model (with all systems included as covariates) with the same values calculated in a reduced model in which the covariates related to the individual classification were removed 1 at a time.23

For the staging systems that were statistically significant according to the likelihood-ratio chi-square test, Kaplan-Meier estimations and log-rank tests were used to evaluate the statistical difference between each stage. If the number of patients with each prognostic score was ≤5, then statistical differences between adjacent prognostic scores were not calculated.

P values <.05 were considered statistically significant. SAS software (version 8.0; SAS Inc., Cary, NC) or the SPSS software package (version 10.0; SPSS Inc., Chicago, Ill) was used for statistical analyses.

RESULTS

At the time of the current statistical analysis, 22 patients were alive, and 109 patients had died. Among them, 75 patients died with recurrent HCC. In total, 65 patients received multiple sessions of TACE. The median follow-up was 24 months (range, 2–83 months). Among the 19 potential prognostic factors of survival, univariate analysis revealed that older age (≥66 years), performance status, the presence of ascites, Child-Pugh class B, serum albumin level (≤3.4 g/dL), serum α-fetoprotein level (>60 ng/mL), multiplicity of tumors, and portal or hepatic vein tumor thrombosis were associated statistically with overall survival (Table 4). However, other factors, such as portal hypertension or platelet count, were not associated significantly with overall survival rate (Table 4). Multivariate analysis revealed that the serum albumin level (≤3.4 g/dL), the presence of ascites, serum α-fetoprotein level (>60 ng/mL), and portal or hepatic vein tumor thrombosis were the only independent significant factors affecting the overall survival rates (P = .001, P = .001, P = .004, and P = .000, respectively) (Table 5). The hazard-ratio was much higher for venous tumor invasion compared with the other prognostic factors. Older age, poor performance status, Child-Pugh class B, and multiple tumors were not significant factors in the multivariate analysis (P = .107, P = .929, P = .371, and P = .085, respectively). The 1-, 2-, 3-, 4-, and 5-year estimated overall survival rates were 77.1%, 46.6%, 28.1%, 19.1%, and 13.6%, respectively. The median survival was 24 months.

Table 4. Analysis of the Potential Risk Factors for Death in Patients With Hepatocellular Carcinoma Undergoing Transarterial Chemoembolization
Variable*Median survival, wkRegression coefficientSEHR (95% CI)P
  • SE indicates standard error; HR, hazard ratio; 95% CI, 95% confidence interval; HBV, hepatitis B virus; HCV, hepatitis C virus; AST, aspartate aminotransferase; ALT, alanine aminotransferase.

  • *

    For continuous variables, the cut-off was set at the median value.

  • Statistically significant (P < .05).

Age: <66/≥66 y80/116−0.5100.1960.600 (0.409–0.881).009
Sex: Women/men107/101−0.0640.5900.938 (0.295–2.979).914
Performance status: 0/1 or 2133/791.0470.3652.849 (1.394–5.823).004
HBV infection: No/yes110/920.3760.1961.457 (0.993–2.139).055
HCV infection: No/yes92/130−0.2940.2180.745 (0.486–1.142).176
AST: ≤50/>50 U/L103/1010.0970.1951.102 (0.752–1.615).617
ALT: ≤40/>40 U/L95/1040.0480.1941.049 (0.718–1.534).804
Albumin: >3.4/≤3.4 g/dL130/760.6330.1991.884 (1.276–2.780).001
Total bilirubin: ≤1/>1 mg/dL104/1010.1180.1951.126 (0.767–1.651).545
Platelet count: ≥100,000/<100,000 × 109/L109/1000.0030.1971.003 (0.682–1.476).986
Prothrombin time: ≥80%/<80%111/920.3130.1971.367 (0.929–2.012).113
Presence of ascites: No/yes115/750.690.2301.994 (1.269–3.132).003
Child-Pugh class: A/B115/731.290.3813.631 (1.721–7.662).006
Portal hypertension: No/yes116/920.3840.1971.468 (0.998–2.159).051
Serum α-fetoprotein: ≤60/>60 ng/mL137/750.7190.2002.052 (1.386–3.039).000
Tumor location: Unipolar/bilobar106/800.3210.2201.378 (0.895–2.122).146
Tumor multiplicity: No/yes144/900.5400.1981.715 (1.164–2.527).006
Maximal tumor size: ≤5/>5 cm104/790.2380.2181.269 (0.828–1.945).275
Venous invasion: No/yes110/391.7230.3175.599 (3.010–10.41).000
Table 5. Prognostic Factors That Were Independently Significant by Multivariate Cox Analysis
VariableRegression coefficientSEHR (95% CI)P
  • SE indicates standard error; HR, hazard ratio; 95% CI, 95% confidence interval.

  • *

    Statistically significant (P < .05).

Albumin: >3.4/≤3.4 g/dL0.6440.2011.904 (1.284–2.823).001*
Ascites: No/yes0.7620.2362.143 (1.349–3.404).001*
Serum α-fetoprotein >60 ng/mL: No/yes0.5970.2091.817 (1.206–2.738).004*
Venous tumor invasion: No/yes1.6800.3345.366 (2.791–10.319).000*

The univariate and multivariate likelihood ratio chi-square test and the AIC test are described in Table 6. The CLIP system had the highest chi-square score in the univariate and multivariate likelihood-ratio chi-square tests. It also had the lowest AIC value in the univariate AIC test and the highest value in the multivariate AIC test (full model with the CLIP system removed). Therefore, the CLIP system was the best system with regard to the univariate and multivariate likelihood-ratio chi-square and AIC tests. The modified CLIP system and the modified JIS system were the next best systems after the CLIP system with regard to these 2 tests.

Table 6. Comparison of the Discrimination Ability Between the 7 Liver Staging Systems as Determined by the Likelihood-ratio Chi-square Test and the Akaike Information Criterion Test*
ModelLog likelihoodLikelihood-ratio Chi-square test (P)Akaike information criterion
  • C-P indicates Child-Pugh; BCLC, Barcelona Clinic Liver Cancer staging classification; MELD, Model for End-stage Liver Disease scoring system; CLIP, Cancer of the Liver Italian Program classification; Mod CLIP, the MELD-based, modified CLIP staging system; JIS: Japanese Integrated Staging System; Mod JIS, the MELD-based, modified JIS system.

  • *

    The CLIP system showed the best results.

  • Statistically significant (P < .05).

Univariate model
 Okuda433.27313.560 (.000)868.546
 C-P score430.60718.893 (.000)869.213
 BCLC436.8146.477 (.011)877.628
 CLIP421.00338.101 (.000)850.005
 Mod CLIP426.60326.901 (.000)863.205
 JIS423.25133.604 (.000)854.502
 Mod JIS429.24121.625 (.000)866.481
Multivariate model
 Full model409.534867.068
 Removing Okuda411.8324.594 (.032)869.663
 Removing C-P score414.78610.502 (.033)869.571
 Removing BCLC409.940.811 (.667)863.88
 Removing CLIP421.54624.023 (.000)883.092
 Removing Mod CLIP418.97618.883 (.002)875.952
 Removing JIS414.63110.193 (.037)869.262
 Removing Mod JIS417.62516.181 (.003)875.25

The median survival for each disease stage is listed in Table 7 for the 6 prognostic systems that were statistically significant according to the likelihood-ratio chi-square test. In the modified CLIP, JIS, and modified JIS systems and in the C-P score, there was substantial overlap of the survival periods, and only the CLIP system presented monotonously decreasing survival curves. Therefore, the CLIP system had the best results with regard to the monotonicity of gradients.

Table 7. Comparison of 6 Staging Systems That Were Significant in Likelihood-ratio Chi-square Tests With Regard to Median Survival and the Confidence Interval for Each Prognostic Score Using Kaplan-Meier Estimation
Staging systemNo. of patients (No. of deaths)Median survival, wk95% CI, wk
  1. 95% CI indicates 95% confidence interval; CLIP, Cancer of the Liver Italian Program classification; Mod CLIP, the Model for End-stage Liver Disease (MELD)-based, modified CLIP classification; JIS, Japanese Integrated System; Mod JIS, the MELD-based, modified JIS; C-P, Child-Pugh class.

CLIP
 Score 034 (21)164122–206
 Score 155 (47)10480–128
 Score 228 (27)8262–102
 Score 313 (13)4525–65
 Score 41(1)9
Mod CLIP
 Score 026 (15)162104–220
 Score 154 (45)11678–154
 Score 226 (24)7058–82
 Score 314 (14)7241–103
 Score 49 (9)4741–53
 Score 52 (2)9
JIS
 Score 06 (3)14934–264
 Score 144 (30)175141–209
 Score 250 (46)8063–97
 Score 327 (26)7240–104
 Score 44 (4)130–32
Mod JIS
 Score 05 (3)10336–170
 Score 146 (31)153112–194
 Score 251 (46)9471–117
 Score 322 (22)5437–71
 Score 47 (7)9247–137
C-P
 Score 545 (33)153117–189
 Score 647 (40)9563–127
 Score 724 (21)10064–136
 Score 812 (12)360–95
 Score 93 (3)450–96
Okuda
 Stage I86 (67)13097–163
 Stage II45 (42)7360–86

Survival curves according to the 6 staging systems that were statistically significant according to the likelihood-ratio chi-square tests are depicted in Figures 1 through 6. Only the CLIP system and the Okuda system had no overlapping survival curves. For the CLIP system, there were statistically significant differences between all adjacent prognostic scores (P<.05 for all), except between scores of 3 and 4 (Fig. 1). There was only 1 patient who had a CLIP score of 4, and that patient died 9 weeks after the initial diagnosis of HCC. No other staging system demonstrated a statistical difference between all adjacent prognostic scores (Figs. 2–6). Therefore, among the 6 staging systems, the CLIP system had the best results with regard to the monotonicity of gradients.

Figure 1.

Survival curves for patients with hepatocellular carcinoma who underwent transarterial chemoembolization stratified according to the Cancer of the Liver Italian Program staging system: score 0 versus 1, P = .023; score 1 vs 2, P = .027; score 2 vs 3, P = .001. All differences between scores were statistically significant.

Figure 2.

Survival curves for patients with hepatocellular carcinoma who underwent transarterial chemoembolization stratified according to the modified Cancer of the Liver Italian Program staging system based on the Model for End-stage Liver Disease: score 0 versus 1, P = .063; score 1 versus 2, P = .098; score 2 versus 3, P = .428; score 3 versus 4, P = .174. No statistical differences were noted between any scores.

Figure 3.

Survival curves for patients with hepatocellular carcinoma who underwent transarterial chemoembolization stratified according to the Japan Integrated Staging System: score 0 versus 1, P = .618; score 1 versus 2, P = .002; score 2 versus 3, P = .024, score 3 versus 4, P = .008. The difference between scores of 0 and 1 was not statistically significant.

Figure 4.

Survival curves for patients with hepatocellular carcinoma who underwent transarterial chemoembolization stratified according to the modified Japanese Integrated Staging System based on the Model for End-stage Liver Disease: score 0 versus 1, P = .447; score 1 versus 2, P = .027; score 2 versus 3, P = .001, score 3 versus 4, P = .701. The difference was significant only between scores of 1 and 2 and between scores of 2 and 3.

Figure 5.

Survival curves for patients with hepatocellular carcinoma who underwent transarterial chemoembolization stratified according to the Child-Pugh nominal classification: score 5 versus 6, P = .017; score 6 versus 7, P = .898; score 7 versus 8, P = .003; score 8 versus 9, P = .531. The difference was significant only between scores of 7 and 8.

Figure 6.

Survival curves for patients with hepatocellular carcinoma who underwent transarterial chemoembolization stratified according to the Okuda classification: stage I versus II, P = .000. The difference was statistically significant.

DISCUSSION

Many comparative studies have been reported to date on the prognostic stratification of patients with HCC patients who undergo TACE; however, controversy remains regarding the best stratification system.12–14, 24, 25 The current study met all 3 of the following criteria to draw a reliable conclusion regarding this controversy: 1) All of the widely used primary staging systems were included in the analysis, such as CLIP, BCLC, and JIS; 2) the 2 modified systems and the Child-Pugh nominal classification that were reported as suitable for predicting the survival of patients undergoing locoregional therapies were included in the analysis; 3) a comprehensive statistical analysis was performed to reveal both the discriminatory ability and the monotonicity of gradients. Contrary to the previous studies, this study clearly revealed that, for the population of HCC patients undergoing TACE, the CLIP system was the best with regard to predicting survival.

In the current study, the CLIP system produced the best results with regard to discrimination ability and the monotonicity of gradients. Toyoda et al. reported that, before 1990, the CLIP system proved to be more suitable than the BCLC system or the JIS system, but not after 1990, when early detection and early treatment for HCC became more common.26 In the current study, surgical patients were excluded, and only the patients who underwent TACE were included in the analysis; it is plausible that the CLIP system may have been more suitable than the BCLC system or the JIS system for this particular group of patients. These results further support the conclusion of the Consensus Conference on Staging of HCC sponsored by the American Hepato-Pancreato-Biliary Association and the American Joint Committee on Cancer in 2003. The consensus recommendation was to use the CLIP system to stratify the prognosis and to guide treatment recommendations for nonsurgical patients.27 Although the BCLC system or JIS system may be more suitable as a choice for surgical indication, those systems may be less suitable than the CLIP system for predicting the survival of patients who receive nonsurgical treatments, as demonstrated in the current study. In the current study, the factors that were significant in multivariate analysis were similar to the factors that constitute the CLIP score, and they also were similar to the factors determined in previous studies of TACE.28, 29 The most powerful prognostic factor of poor survival was vascular invasion in this study, and similar results were reported in patients who underwent liver resection or transplantation.1, 30–32

Although the BCLC system was validated as the most suitable prognostic system for patients with HCC in Italy and the United States, it was not considered a suitable prognostic system in the Eastern countries until now.33, 34 Furthermore, in the current study, surgical patients were excluded from the analysis, which may be an important reason that the BCLC system was inferior to the JIS system or the CLIP system in predicting the survival of patients with HCC.

The MELD-based, modified JIS system and modified CLIP system reportedly were more suitable predictive models for patients with HCC after locoregional therapy compared with the original JIS and CLIP systems.12, 13 However, in the current study, these modified systems produced less favorable results compared with the original systems. The clinical usefulness of the 2 modified systems needs to be verified further, because they are more complex than the original systems.

The Child-Pugh nominal classification reportedly was the best predictive model for the survival of patients after TACE according to Georgiades et al.14 However, in their study, the CLIP system; the MELD-based, modified CLIP system; and the MELD-based, modified JIS system were not included in the analysis. Furthermore, only the error reduction rates were compared in their study.

The current study had the following limitations: First, pathologic confirmation was not obtained routinely for the staging of patients. Second, this study was performed in a hepatitis B-endemic area. Although the CLIP system originally was developed from a Western area, the results of this study probably can be applied to countries that have a high prevalence of hepatitis C-related HCC. Third, in the future, further modification of the CLIP system may be possible, because vascular invasion was a much more potent prognostic factor than the other factors in our multivariate analysis.

In conclusion, the CLIP system was the best prognostic system for predicting the survival of patients with HCC who were undergoing TACE among the 7 systems analyzed. We hope that further validation studies will be performed and that the survival prediction model for patients who undergo TACE can be standardized in both Western and Eastern countries.

Ancillary