Validation of a new prognostic staging system for hepatocellular carcinoma: The JIS score compared with the CLIP score

Authors


Abstract

The Japan Integrated Staging score (JIS score), which combines the Child-Turcotte-Pugh classification and tumor-node-metastasis staging, has been proposed as a better prognostic staging system for hepatocellular carcinoma (HCC) than the Cancer of the Liver Italian Program (CLIP) scoring system. In this study, validation was performed among a larger patient population. A total of 4,525 consecutive patients with HCC who had been diagnosed at five institutions were included. Stratification ability, prognostic predictive power, and reproducibility were analyzed and compared with results from the CLIP scoring system. Only 45% (1,951 of 4,525) of all patients were categorized as early stage HCC according to JIS score (0 or 1), whereas 63% (2,878 of 4,525) of the patients were categorized as having a CLIP score of 0 or 1. Significant differences in survival curves were not observed among CLIP scores 3 to 6. In contrast, survival curves showed significant differences among all the JIS scores. The same JIS scoring subgroups showed a similar prognosis, and good internal reproducibility was observed in each of the institutions. Multivariate analysis of the prognosis in all 4,525 patients proved the JIS score to be the best prognostic factor. Furthermore, the Akaike information criteria proved that the JIS scoring system was statistically a better model for predicting outcome than the CLIP scoring system. In conclusion, the stratification ability and prognostic predictive power of the JIS score were much better than that of the CLIP score and were simple to obtain and remember. (HEPATOLOGY 2004; 40:1396–1405.)

A clinical staging system for cancer patients provides guidance for patient assessment and appropriate therapy. It is useful for the decision to treat patients aggressively while avoiding the overtreatment of patients who would not tolerate therapy or whose life expectancy rules out any chance of success. Clinical staging is also an essential tool for comparison between groups in therapeutic trials and between different studies.

The current classifications most commonly used for hepatocellular carcinoma (HCC) are the Okuda staging system,1 the Child-Turcotte-Pugh staging system,2 tumor node metastasis (TNM) staging,3, 4 and Cancer of the Liver Italian Program (CLIP) score.5 Among these, the CLIP score is currently the most commonly used integrated staging score, including both tumor and liver disease stage.

Although the CLIP score has been well validated in its prognostic value for HCC patients by many authors,6–8 it has some problems and limitations when applied to patients who are diagnosed in the early and asymptomatic stage of the disease.9 First, the CLIP score can discriminate patient populations that score from 0 to 3, but it is not able to discriminate patient groups with a score of 4 to 6. Second, the definition of tumor morphology in the best prognostic subgroup is too advanced, being uninodular with extension of less than 50% of the tumor.10 As a result, the prognosis of their best prognostic subgroup is poor. In other words, this system cannot identify the best prognostic subgroup that would benefit from curative and aggressive treatment.10 Third, nearly 80% of the patient population is classified into a CLIP score of 0 to 2—as has been proved by many studies—which shows poor stratification ability.9

The Japan Integrated Staging (JIS) system, a new system that is based on a combination of the Child-Turcotte-Pugh system and the Liver Cancer Study Group of Japan (LCSGJ)—which is also concordant with TNM classification for HCC by the International Hepato-Pancreato-Biliary Association and the International Union Against Cancer (UICC)11—has recently been proposed in Japan.9 The aim of this multicenter study was to validate the the JIS score in terms of stratification ability, reproducibility, prognostic predictive power, and suitability as a prognostic model compared with the CLIP score.

Abbreviations

JIS, Japan Integrated Staging; HCC, hepatocellular carcinoma; CLIP, Cancer of the Liver Italian Program; TNM, tumor-node-metastasis; LCSGJ, Liver Cancer Study Group of Japan; UICC, International Union Against Cancer; CT, computed tomography; AFP, alpha-fetoprotein.

Patients and Methods

Patients.

All 4,525 consecutive patients who were diagnosed with HCC before treatment at five institutions from 1990 to 2002 were included in this study. The study protocol was approved by the Institutional Ethics Review Board at each of the institutions. To compare the validity of CLIP score (Table 1), TNM stage by LCSGJ (Table 2), Child-Turcotte-Pugh classification, and JIS score (Fig. 1, Table 3), several data were collected. The JIS score was obtained via the summation of tumor stage score (stage I, 0; stage II, 1; stage III, 2; and stage IV, 3) and Child-Turcotte-Pugh score (Child-Turcotte-Pugh A, 0; Child-Turcotte-Pugh B, 1; Child-Turcotte-Pugh C, 2).9 The maximum diameter of the tumor was defined using B-mode ultrasonography. Vascular invasion was assessed via dynamic computed tomography (CT) and angiography. Lymph node invasion and distant metastases were assessed via routine screening study such as ultrasonography, dynamic CT, and chest X-ray. Bone scintigraphy or brain CT was performed if suggestive symptoms were present. CLIP score was calculated by means of the summation of the points for four variables: Child-Turcotte-Pugh score, tumor morphology, alpha-fetoprotein (AFP) value, and portal venous invasion. Tumor morphology was determined based on CT findings. Serum AFP value was determined using a commercially available enzyme immunoassay kit (Eiken Chemical Co., Tokyo, Japan) uniformly in all five institutions.

Table 1. Definitions of the CLIP Score
VariableScore
012
Child-Turcotte-Pugh stageABC
Tumor morphologyUninodular and extension ≤50%Multinodular and extension ≤50%Massive or extension >50%
AFP (ng/mL)<400≥400
Portal vein thrombosisNoYes
Table 2. Definitions of TNM Stage by the Liver Cancer Study Group of Japan
T factorI. Single II. <2 cm III. No vascular involvement
T1Fulfilling 3 factors
T2Fulfilling 2 factors
T3Fulfilling 1 factor
T4Fulfilling 0 factors
Stage IT1 N0 M0
Stage IIT2 N0 M0
Stage IIIT3 N0 M0
Stage IV-AT4 N0 M0, or any T N1 M0
Stage IV-BAny T N0-1 M1
Figure 1.

Definition of JIS scoring system. The JIS score can be obtained by summing up the TNM stage score (stages I, II, III, and IV are allocated to scores 0, 1, 2, and 3, respectively) and Child-Turcotte-Pugh stage score (Child-Turcotte-Pugh stages A, B, and C are allocated to scores 0, 1, and 2, respectively).

Table 3. Definitions of the JIS Score
VariableScore
0123
Child-Turcotte-Pugh stageABC
TNM stage by LCSGJIIIIIIIV

Patients were diagnosed with HCC based on histological confirmation of tumor tissue taken from resected or biopsied samples in 2,023 cases (Table 4). In the remaining 2,502 patients, diagnosis was made according to the reliable clinical criteria12, 13 fulfilling the following conditions: compatible clinical background (association with liver cirrhosis or viral hepatitis) with typical imaging findings and rising trend of increased HCC-related tumor markers; serum AFP level of more than 400 ng/mL; serum des-γ-carboxyprothrombin level of 100 mAU/mL or more; or AFP-L3 fraction of more than 15%. Typical imaging findings for HCC include: a mosaic pattern with a halo on B-mode ultrasonography; hypervascularity on angiography; and a high-density mass in the arterial phase dynamic CT with a low-density mass in the portal phase dynamic CT using a helical or multidetector raw CT scanner. When typical findings for HCC were not obtained through dynamic CT or angiography, CT during hepatic arteriography and CT during arterial portography or T1- and T2-weighted imaging associated with superparamagnetic iron oxide–enhanced magnetic resonance imaging13 were performed. In cases that did not show typical findings in any of the imaging modalities or no increase of tumor markers, biopsy was performed to confirm the diagnosis of HCC.

Table 4. Characteristics of the 4,525 Patients Included in the Study
VariableValue%
  1. Abbreviations: DCP, des-γ-carboxyprothrombin; HCV, hepatitis C virus; HBV, hepatitis B virus; PEIT, percutaneous ethanol injection therapy; PMCT, percutaneous microwave coagulation therapy; RFA, radiofrequency ablation; TACE, transcatheter arterial chemoembolization.

Age, yr  
 Median68.0 
 Range26–97 
Sex, n  
 Male3,34173.8
 Female1,18426.1
Modality of diagnosis, n  
 Histology2,02344.7
 Imaging + tumor marker (AFP, DCP, AFP-L3)2,50255.2
Cause of parenchymal disorder, n  
 HCV3,18770.4
 HBV48410.6
 HCV + HBV851.8
 Non-B, non-C4068.9
 Non3638.0
Child-Turcotte-Pugh stage, n  
 A3,13469.2
 B1,01522.4
 C3768.3
TNM stage by LCSGJ, n  
 I76816.9
 II1,71537.9
 III1,51033.3
 IV53211.7
Initial treatment modality, n  
 No1964.3
 Yes  
  Surgery1,17325.9
  Percutaneous ablation therapy85018.7
   PEIT47810.5
   PMCT1062.3
   RFA2665.8
  TACE1,75338.7
  Chemolipiodolization52711.6
  Other260.5

Patients were followed up from 1.2 years to 13 years (median follow-up period: 4.2 years). A total of 99% (4,496 of 4,525) of the patients were followed up at the five institutions.

Survival Curves.

Overall survival was the only end point used in the analysis. Survival was measured from information obtained on the date of the last follow-up. Univariate survival curves were estimated using the Kaplan-Meier method; differences in the survival rates between the groups were compared via the log-rank test. Stratification ability between each curve in the CLIP and JIS scores was compared. Furthermore, the survival of the best prognostic subgroup (i.e., score 0) was compared between CLIP and JIS scores.

Internal Reproducibility Among Institutions.

The total of 4,525 patients consisted of 1,409, 1,360, 774, 641, and 341 from each of the five institutions, respectively. The survival curves based on CLIP score and JIS score were compared at each institution to analyze the internal reproducibility of the results obtained from all of the data.

Comparison of Survival Curves Among Patients With the Same JIS Score.

Survival was compared between the subgroups with the same JIS score. Namely, JIS score 1 consists of two subgroups: Child-Turcotte-Pugh grade A patients with TNM stage II HCC and Child-Turcotte-Pugh grade B patients with TNM stage I HCC. Similarly, the JIS scores 2 and 3 patient subgroups consisted of three subgroups, and the JIS score 4 patient group consisted of two subgroups, respectively (Fig. 1). The differences between survivals were analyzed via the log-rank test.

Prognostic Predictive Power.

The prognostic significance of the JIS score was evaluated by univaliate analysis followed by multivariate prognostic analysis using the stepwise Cox proportional hazard regression model. The following variables were used for the analysis: patient age and sex; Child-Turcotte-Pugh stage and its constitutive variables (albumin, bilirubin, prothrombin time, ascites, and encephalopathy); tumor type; tumor extension; portal vein thrombosis; AFP (ng/mL) and des-γ-carboxyprothrombin levels (mAU/mL) at diagnosis; evidence of distant metastasis at diagnosis; and treatment modality.

Statistical Analysis.

Values were expressed as the median ± SD. Univariate survival curves were estimated using the Kaplan-Meier method; differences in survival rates between groups were compared using the log-rank test. The Bonferroni correction for multiple comparisons was applied. The likelihood ratio test was performed to measure homogeneity—that is, whether or not the difference in survival time is small among patients classified into the same group by that system. The likelihood ratio test can also estimate the monotonicity of gradient; the mean survival time for a group classified as favorable by that system is always longer than the survival times noted in less favorable groups. The Akaike information criteria14 were also used to evaluate the discriminatory ability of the given model. All analyses were performed with SAS statistical software (version 8.2; SAS Institute, Cary, NC) or the SPSS Medical Pack for Windows (version 10.0; SPSS, Inc., Chicago, IL).

Results

Patient Characteristics.

Table 4 shows the characteristics of the 4,525 patients collected from five institutions. The patients were predominantly men (male/female ratio, 4:1), and the median age was 65 years.

Patient distribution according to CLIP score is shown in Table 5. Approximately 63% (2,878 of 4,525) of all patients were categorized into the early stage according to CLIP score (score 0 or 1), whereas only 45% (1,951 of 4,525) of the patients were categorized into the early stage according to JIS score (score 0 or 1). Similarly, only 7% of the patients were categorized into advanced stage according to CLIP scores (scores 4, 5, 6), whereas 24% of the patients were categorized into advanced stage according to JIS scores (scores 3, 4, 5).

Table 5. Patient Distribution According to the CLIP Score and JIS Score (n = 4,525)
Staging Systemn%
CLIP score  
 01,18126
 11,68737
 292321
 34189
 41744
 5932
 6491
JIS score  
 055212
 11,39931
 21,47133
 375717
 42445
 51022

Survival According to Staging System.

Patient survival according to Child-Turcotte-Pugh stage, TNM stage, CLIP score, and JIS score are shown in Table 6; 10-year survival in the best prognostic subgroup was 21%, 37%, 30%, and 48%, respectively. Stratification ability of the overall survival rate was good according to TNM stage as well as Child-Turcotte-Pugh stage (P < .001) (Fig. 2). However, when accounting for both the TNM stage and Child-Turcotte-Pugh stage, the survival curves differed greatly according to TNM stage, even in the same Child-Turcotte-Pugh stage A patient subgroup (Fig. 3). The same tendency was observed in Child-Turcotte-Pugh stage B or C patient subgroups (data not shown). Conversely, the survival curves differed greatly according to Child-Turcotte-Pugh stage, even in the same TNM stage I patient subgroup (Fig. 4). The same results were observed in TNM stages II-IV patient subgroups (data not shown).

Table 6. Patient Survival by Child-Turcotte-Pugh Stage, TNM Stage by LCSGJ, CLIP Score, and JIS Score
Staging SystemnSurvival (%)
3-year5-year10-yearMedian (mo)
Child-Turcotte-Pugh stage     
 A3,13464462153.6
 B1,0154524031.6
 C376211013.8
TNM stage by LCSGJ     
 I76881633777.5
 II1,71567451553.2
 III1,51046281032.1
 IV53213608.5
CLIP score     
 01,18179643076.6
 11,68764411549
 29234424630
 3418196115.2
 417413905.9
 5934203
 6490002.8
JIS score     
 055287734895.8
 11,39972522061.9
 21,47156331041.5
 37572513319.1
 424413208.2
 51021003
Figure 2.

Kaplan-Meier curves according to (A) tumor-node-metastasis TNM stage by the Liver Cancer Study Group of Japan and (B) Child-Turcotte-Pugh grade (N = 4,525). Statistically significant differences were observed among survival curves of TNM stages and Child-Turcotte-Pugh stages via log-rank test (P < .0001).

Figure 3.

Overall survival rates according to Child-Turcotte-Pugh grade (N = 4,525). Even in the same Child-Turcotte-Pugh grade A patients, the survival rates were greatly different according to the TNM stage (P < .0001).

Figure 4.

Overall survival rates according to tumor-node-metastasis TNM stage (N = 4,525). Even for the same stage I patients, survival rates were very different according to Child-Turcotte-Pugh grade (P < .0001).

Stratification Ability and Internal Reproducibility.

Significant differences in survival curves were observed among CLIP 0, 1, 2, and 3 patient subgroups, whereas no difference was observed among CLIP 3, 4, 5, and 6 patient subgroups (Fig. 5). In contrast, statistically significant differences were observed between survival curves for all JIS scores (scores 0-5) (Fig. 5). The exact same results were obtained in each of the five subgroup institutions.

Figure 5.

Overall survival rate according to CLIP score compared with JIS score (N = 4,525). Stratification ability in the JIS score is much better than in the CLIP score, especially in advanced stage HCC subgroups. Log-rank test with the Bonferroni correction for multiple comparisons was applied. NS, not significant.

Comparison of Survival Curves Between the Same JIS Scores.

JIS score 1, which consisted of two subgroups; scores 2 and 3, which consisted of three subgroups; and score 4, which consisted of two subgroups, showed no statistical differences in the survival rates between the subgroups with the same JIS score (Fig. 6).

Figure 6.

Comparison of survival curves between the subgroups showing same JIS score. Similar survival curves were obtained in the two subgroups in JIS score 1 as well as in JIS score 4. Furthermore, similar survival curves were obtained in the three subgroups in JIS score 2 as well as JIS score 3. No statistical difference was observed between survival curves in the same JIS score subgroup.

Prognostic Factors.

Among the significant variables identified via univariate analysis, five variables were identified as independent prognostic variables: JIS score, ICGR15, AFP value, des-γ-carboxyprothrombin value, and portal invasion. As shown in Table 7, each JIS score was shown to be an independent prognostic factor among several factors via multivariate analysis. The relative risk of each JIS score showed a high value ranging from 1.587 to 3.178 compared with other independent variables, even though the JIS score had five independent strata. In addition, as in Table 8, the independent homogenizing ability and stratification value of the CLIP score and JIS score by means of the likelihood ratio test within a Cox proportional hazard regression model showed the JIS score to have a higher χ2 (1238.053) than the CLIP score (χ2 =1062.092). Furthermore, a lower Akaike information criteria value was shown in the JIS score (33642.365) than in the CLIP score (33822.326), proving that the JIS score is a better prognostic model.

Table 7. Prognostic Variables by Using Cox Proportional Hazard Model
VariablesRelative Risk95% CIP Value
  1. Abbreviations: ICG R15, indocyanine green retention test at 15 minutes; DCP, des-γ-carboxyprothrombin.

JIS (0 → 1)1.8581.264–2.732.0016
JIS (1 → 2)1.6821.365–2.072<.0001
JIS (2 → 3)1.8841.542–2.303<.0001
JIS (3 → 4)1.5871.176–2.143<.0001
JIS (4 → 5)3.1781.919–5.262<.0001
ICG R15 (>15%)1.2961.093–1.537.0029
AFP (>400 ng/mL)1.6201.376–1.907<.0001
DCP (>200 mAU/mL)1.3671.164–1.604.0001
Portal invasion (+)1.5591.246–1.951.0001
Table 8. Evaluation of Prognostic Stratification and Homogeneity in CLIP Score and JIS Score (n = 4,525)
ModelLikelihood Ratio (χ2)Akaike Information Criteria
  1. Note. Regarding discriminatory ability, homogeneity, and monotonicity of gradients, the model with the higher χ2 by the likelihood ratio test was considered the better model. Furthermore, the lower value of Akaike information criteria is considered the better model for discriminatory ability.

CLIP score1062.09233822.326
JIS score1238.05333642.365

Discussion

In HCC, the prognostic assessment and choice of treatment options are important but extremely complicated, unlike carcinomas arising in other organs: Prognosis of HCC depends not only on the grade of cancer (i.e., tumor staging), but also on the grade of residual liver function (i.e., liver disease stage).15–18

To reliably estimate prognosis in patients with HCC, both liver function and tumor-related factors should be considered; however, staging systems seldom include both. In fact, the well-known UICC and American Joint Committee on Cancer staging (TNM) criteria do not define the relative prognostic weight of variables in terms of residual liver function.3, 4 An accurate prognostic staging system is important for patients with HCC in order to provide adequate therapeutic guidance, avoiding overtreatment in intolerant patients or those whose life expectancy rules out any chance of success. Finally, it is important to obtain better patient stratification when planning controlled trials on the locoregional treatment of HCC or examining the results of open trials reporting on new treatment methods.

Several retrospective studies have described the clinical course of HCC patients, but most of them involve large series of individuals that were diagnosed at advanced stages more than 10 years ago,19–31 when regular screening was uncommon and imaging assessment was less accurate. During the last 10 years, the implementation of screening programs in high-risk populations has led to an increased detection of early and asymptomatic tumors.32 Thus, an integrated staging system for HCC in terms of both tumor stage and liver disease stage should be established to assign the best treatment options to the best patient groups, determine the prognosis, compare the treatment efficacy of several treatment options, and clarify geographical and racial differences of HCC itself.

The Okuda classification,1 a classification that includes both tumor and liver function factors, has been widely accepted in Western countries in the past. However, it is outdated as it does not include important tumor factors, such as the unifocal, multifocal, or diffuse nature of the tumor, whether there is vascular invasion, or whether the tumor is less than 2 cm in diameter—all of which have prognostic significance in early stages of HCC.

Other than the Okuda classification, new prognostic staging systems including both the tumor and liver disease stage have been studied. The Barcelona Clinic Liver Cancer system is one such system.33 This classification includes prognostic factors, which were shown to be significant via multivariate analysis. Barcelona Clinic Liver Cancer staging includes the following factors: performance status, single or multifocal tumor, vascular invasion, portal hypertension, Okuda stage, and Child-Turcotte-Pugh classification. It is based on the possibility of curative interventions33, 34; consequently, it provides not only a classification with prognostic relevance but also indications that are useful for treatment selection.11, 34, 35

Another prognostic staging system was proposed by the Cancer of the Liver Italian Program group.5 The CLIP score was derived from a retrospective evaluation of 435 Italian patients with HCC diagnosed from 1990 to 1992. It includes four variables (see Table 1). The CLIP score was successful in discriminating patients with HCC: the higher the score, the poorer the prognosis. The CLIP score, calculated at the time of HCC diagnosis, can be used to inform the patient properly and to decide on the treatment strategy.5 After the CLIP score was established by the retrospective analysis in 1998, it was validated in 2000 through an independent group of patients prospectively enrolled between 1995 and 1997 in the CLIP multicenter randomized clinical trial for 196 patients with HCC.36 They concluded that the CLIP scoring system is useful in treatment planning by improving the baseline prognostic evaluation of patients with HCC, and could be used in prospective therapeutic trials as a stratification variable, reducing the variability of results owing to patient selection. The CLIP score was further validated in internal control series37 as well as in separate series of Italian,6 Canadian,7 and Japanese patients with HCC.8

Although the CLIP score has been well validated in predicting the prognosis of Western and Canadian patients with HCC, the criteria of tumor morphology in the CLIP score are too broad to apply to patients with HCC in countries like Japan, where many early stage HCCs—such as solitary nodules less than 2 cm in diameter without vascular invasion—can be detected based on the established screening system for HCC. The definition of tumor morphology in the best prognostic subgroup (namely CLIP score 0) is a uninodular tumor with an extension of less than 50%, which is a fairly large tumor. Advanced tumor morphology in the best prognostic subgroup is the biggest problem of this system, which is difficult to accept in countries where early detection of small HCCs is fairly common. Therefore, this system is probably unsuitable for the current population of HCC patients, many of whom are diagnosed early at an asymptomatic stage of disease. Second, although most publications show that CLIP scores 0, 1, 2, and 3 patient populations are well discriminated from each other, there are no significant differences between CLIP scores 4, 5, and 6 patient populations as reported by Colombo et al.10 and Llovet et al.38 Furthermore, Llovet et al. also noted that because the CLIP score provides almost every treatment option for all subgroups, it is not useful for treatment decisions.38

In their previous study of prognostic factors in HCC, the CLIP groups stratified the patients into seven groups according to prognostic indicators—though in actuality they evaluated only six groups, because scores 5 and 6 were placed in the same group.5 Furthermore, in the subsequent validation study they reduced the strata to five: CLIP scores 0, 1, 2, 3, and 4-6 (though it was in fact four, considering that CLIP 2 and 3 have almost identical survival figures).37 The exact same findings were observed in different series in Italian,6 Canadian,7 and Japanese8 reports as well. Similarly, in the present series of 4,525 patients with HCC, stratification ability of the CLIP score was fairly good in scores 0-3; however, scores 3-6 patients were not well stratified as reported earlier. Furthermore, the ratio of patients categorized into CLIP scores 0-2 was quite high: 73% by a CLIP group,3 83% (83 of 100) by an Italian report,6 68% (175 of 257) by a Canadian report,7 74% (487 of 662) by a Japanese report,8 and 84% (3,791 of 4,525) in the current series of Japanese patients. This poor stratification and deviation of most HCC patients to CLIP scores 0, 1, and 2 is extremely unfavorable and may be the second biggest drawback of the CLIP scoring system. The lack of predictive power of this type of prognostic classification was stressed in a consensus conference on prognostic studies in hepatology.39 Third, the outcome of their best prognostic subgroup (score 0) exhibits an extremely low survival rate regardless of any treatment applied. Similarly, Bruix and Llovet40 noted that the CLIP system lacks the accuracy to identify early stage patients and distinguish them from intermediate cases. A more sensitive staging system to identify the best prognostic subgroup is needed.

TNM staging and liver damage grade have been commonly used in Japan based on the general rules of the LCSGJ.41 The TNM staging manual of the LCSGJ has recently been revised as the 4th edition.42, 43 This staging system was validated as working very well44 and has therefore been proposed as a new concordant TNM classification of primary liver cancer worldwide by the International Hepato-Pancreato-Biliary Association45 and the UICC.11

Although Japan's staging method including TNM stage and liver damage stage32, 42 is extremely precise and superb, it has not widely been used worldwide until now. This may be because TNM staging includes four grades (stages I-IV), and the liver disease stage, or liver damage classification,42 consists of three grades (liver damage grades A, B, and C). These classifications are extremely useful to precisely determine residual liver function and cancer spread separately. However, this independent classification is not useful for determining prognosis or the suitability of a given therapeutic strategy because it has 12 subgroups (3 liver disease stages × 4 tumor stages) (Fig. 1). If the staging of the LCSGJ is used for the purpose of comparison between therapeutic options, the patient subgroups must be separately evaluated in 12 columns (i.e., 3 liver disease stages and 4 tumor stages), resulting in several survival curves in each group.32, 46

Obviously, because this 12 subgroup classification is extremely precise, detailed analysis of large numbers of patients is possible. However, those 12 classifications are not fully integrated into clinical practice because they do not always reflect prognostic significance in addition to their complexity. This intricate property of LCSGJ staging makes it difficult to use in the overall comparison of different therapies, such as local ablation therapy or resection. These issues are the intrinsic limitations of the staging system based on the TNM stage, not only by the LCSGJ but also by the UICC and the American Joint Committee on Cancer.

The JIS score was proposed by Kudo et al.9 and is regarded as one of the most frequently used staging systems.47 The JIS score is obtained by simply adding both scores for the TNM stage by LCSGJ (or UICC stage) and Child-Turcotte-Pugh stage. Scores for the TNM stage can be easily obtained by allocating stages I, II, III, and IV to scores 0, 1, 2, and 3, respectively. Scores for Child-Turcotte-Pugh stage can be similarly obtained by allocating Child-Turcotte-Pugh A, B, and C to scores 0, 1, and 2, respectively. Fairly good stratification was obtained from scores 0 to 5 by JIS score compared with CLIP score, which cannot discriminate advanced stage HCC (CLIP scores 4-6). Furthermore, patient distribution in each JIS score was much better than that in the corresponding CLIP scores. As a consequence of these favorable characteristics, the long-term survival rate was the best prognostic subgroup in the JIS score (i.e., JIS score 0 was much better than that of CLIP score 0; see Table 6). Furthermore, the worst prognostic subgroup (JIS score 5) was accurately identified by this staging system. This was not possible in the CLIP score as described in the literature as well as the current study (Fig. 5). This feature is extremely important, because this patient subgroup is a group in whom any treatment except transplantation should not be performed because they cannot tolerate any treatment. This accurate identification of the worst prognostic subgroup can be a standardized measure to compare the results of liver transplantation with the best supportive therapy or to compare the therapeutic results among several transplantation centers.

Most importantly, the features mentioned above were obtained not only by whole patient analysis, but also by each institutional analysis. In other words, the internal reproducibility of the results obtained by the whole patient population was clearly shown in this study, proving that the JIS score is a reliable and reproducible prognostic staging system. This favorable feature was further validated via multivariate analysis using the Cox proportional hazard model. The results of likelihood ratio and Akaike information criteria also confirmed the fact that the JIS score was a better staging model than the CLIP score in this patient population. Finally, the fact that patient subgroups with the same JIS scores show a homogeneous background with regard to survival further supports the validity of this system.

In conclusion, the JIS system showed a better stratification ability than the CLIP scoring system. The JIS score can correctly identify the patient subgroup from early, intermediate, advanced, and end-stage HCC patient subgroups, which greatly benefits clinical hepatology. The ultimate condition for the primary staging system for HCC is clinical staging, which can be applied to all patients from early stage to end stage11; in light of this, the JIS system would be of great value as a prognostic staging system. It simply consists of two widely accepted staging systems: TNM tumor staging and the Child-Turcotte-Pugh staging system. Because the TNM stage initially proposed by the LCSGJ is becoming an international concordant TNM stage as proposed by the International Hepato-Pancreato-Biliary Association and the UICC,11 we hope the JIS system will come to be used worldwide in the clinical practice of hepatology. Furthermore, it may be used as an international standard scale when comparing the treatment results among several modalities, institutions, or countries, or when testing the effectiveness of a new treatment method for HCC.

Ancillary