Advances in the diagnosis and surgical treatment of hepatocellular carcinoma (HCC) have improved the prognosis for patients with HCC who undergo liver resection. The objective of this study was to evaluate prognostic predictors for patients with HCC who underwent liver resection in a Japanese nationwide data base.
In this study, the authors analyzed 12,118 patients with HCC in a Japanese nationwide data base who underwent liver resection between 1990 and 1999 and compared them with a previous analysis of patients between 1982 and 1989. All patients were evaluated for prognostic factors.
During the last decade, the increases in patients who were without hepatitis B virus surface antigen, who had small tumors, and who had portal vein invasion were noted. The 5-year overall survival rates for patients with HCC improved to 50.5%, compared with < 40% in the previous analysis. A multivariate analysis using a stratified Cox proportional hazards model according to associated liver disease indicated that age, degree of liver damage, α-fetoprotein level, maximal tumor dimension, number of tumors, intrahepatic extent of tumor, extrahepatic metastasis, portal vein invasion, hepatic vein invasion, surgical curability, and free surgical margins were independent prognostic predictors for patients with HCC. Operative mortality decreased from 2.3% in 1990–1991 to 0.6% in 1998–1999.
Hepatocellular carcinoma (HCC) is a common malignancy in Japan. Recent progress in the diagnosis and treatment of HCC has improved patient outcomes. With earlier diagnosis and tumors detected at an early stage, there is substantially increased survival, and curative surgical resection possible.1–3 Advances in surgical techniques and perioperative management have led to a decline in the operative morbidity and mortality of patients who have HCC with chronic liver disease. However, long-term survival remains unsatisfactory.
The Liver Cancer Study Group of Japan (LCSGJ) has been conducting a nationwide survey of patients with primary liver carcinoma since 1965 to evaluate epidemiologic and clinical characteristics, histopathologic features, diagnosis, treatment modalities, and outcomes. In 1994, the LCSGJ reported on predictive factors for long-term prognosis after liver resection for patients with HCC between 1982 and 1989 in Japan.2 Since that study was published, new imaging methods and laboratory tools have been introduced for the diagnosis of HCC.4, 5 In addition, it is recognized now that hepatitis C viral infection confers a high risk for HCC.6 In the 1990s, a screening program for HCC was initiated, and several new therapeutic modalities for HCC, such as thermal ablation therapy and locoregional infusion chemotherapy through the hepatic artery, have been applied. Liver transplantation also has become an effective treatment for patients with HCC who have severe liver dysfunction.7 However, new strategies for the treatment of patients with HCC still need to be established.
We reevaluated prognostic predictors for patients with HCC in a large-scale data base, taking into consideration recent advances in the diagnosis and treatment of HCC. In this study, we analyzed a Japanese nationwide data base between 1990 and 1999 to reevaluate prognostic factors for patients with HCC who underwent liver resection and compared the findings with data from our previous report.2
MATERIALS AND METHODS
In the nationwide follow-up survey of primary hepatic carcinoma conducted by the LCSGJ, patients with primary malignant liver tumors who were diagnosed with imaging studies, preoperative clinical data, and/or histopathologic studies at approximately 800 institutions in Japan were registered every 2 years, and registered patients were followed prospectively. In this data base, there were 50,267 patients who were diagnosed with HCC between 1990 and 1999. Among them, we enrolled 12,118 patients with HCC who underwent liver resection between January, 1990 and December, 1999. Follow-up ended on December 31, 1999. The median follow-up was 21.5 months (range, 0.03–119.7 months), the mean patient age was 62.5 years, and the male:female ratio was 3.73:1.0.
In the data base, 21 clinicopathologic and biologic variables were selected from LCSGJ questionnaires that were used to survey patients with HCC. Patients were then stratified by gender, age, history of blood transfusion, hepatitis B virus surface antigen (HBs-Ag) status, hepatitis C virus antibody (HCV-Ab) status, degree of liver damage, preoperative serum α-fetoprotein (AFP) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) levels, maximal tumor dimension, number of tumors, intrahepatic extent of tumor, extrahepatic metastasis (including lymph node metastasis indicated in the preoperative imaging studies or operative findings), growth appearance (expansive growth or infiltrative growth), capsular formation, septum formation, portal vein invasion, hepatic vein invasion, bile duct invasion, surgical curability of liver resection, free surgical margins, and histologic-associated liver disease (normal liver, chronic hepatitis, and cirrhosis). The degree of liver damage was classified as A, B, and C (also called clinical Stages I, II, and III, respectively in the Classification of Primary Liver Cancer by the LCSGJ8) and was defined by preoperative measurements of ascites, serum bilirubin level, serum albumin level, indocyanine green retention rate at 15 minutes, and prothrombin activity (Table 1). Intrahepatic extent of tumor also was defined in the Classification of Primary Liver Cancer by the LCSGJ.8 We defined a free surgical margin as a distance of 1 cm between the cut surface and the tumor edge in the resected specimen. Surgical curability was defined by the LCSGJ as follow: Absolute curative resection included live resection with 1 cm of free surgical margin in patients with Stage I disease; relative curative resection included live resection without 1 cm of free surgical margin but with the excised tumor tissue in patients with Stage I disease or liver resection with 1 cm of free surgical margin in patients with Stage II or III disease (in either instance, no tumor thrombi may remain in the portal vein, hepatic vein, or bile duct in images of the remnant liver); relative noncurative resection, in which all macroscopic tumor tissue is removed; and absolute noncurative resection, which is liver resection with part of the macroscopic tumor tissue remaining. Capsular infiltration was excluded, because this factor was determined only in patients who had HCC with capsular formation; and, in 25% of our patients, capsular formation was absent.
ICGR15 indocyanine green retention rate at 15 minutes.
The severity of each finding is evaluated separately. Degree of liver damage is recorded as A, B, or C, based on the highest grade that contained at least two findings.
Serum bilirubin (mg/dL)
Serum albumin (g/dL)
Prothrombin activity (%)
Overall, cumulative survival rates were obtained using the Kaplan–Meier method. The differences in survival between the groups were compared using the log-rank test. The starting point for calculating survival was the date of surgery, and the endpoint was the date of death. All deaths, including operative deaths, were considered the endpoint. Patients who remained alive on December 31, 1999, were censored. After the univariate analysis of the factors affecting survival, only significant variables except PIVKA-II were used in the multivariate analysis using the stratified Cox proportional hazards model, because data records of PIVKA-II were not complete. In the Cox model, baseline hazards were stratified by underlying liver disease, because the underlying liver disease affected survival in the univariate analysis and crossed the Kaplan–Meier curves. It was expected that this stratified Cox model would be more powerful than the subgroup analysis by underlying liver disease. The data derived from 7056 patients with complete records were used in the multivariate analysis. P values < 0.05 were considered statistically significant. Statistical analysis was carried out using SAS software (version 8.02; SAS Inc., Cary, NC).
The 1-year, 3-year, and 5-year overall survival rates for patients with HCC were 85.3%, 67.0%, and 50.5%, respectively (Fig. 1). The 5-year survival rate was 10% greater than the rate in our previous report, in which the 5-year survival rate was < 40%.2 Categorization of variables, patient ratios, and the 3-year and 5-year survival rates are shown in Table 2. Compared with our previous report, the proportion of patients age > 60 years increased from 46% to 66%. The percentage of patients with positive HBs-Ag status decreased from 25% to 20%; and the percentage of patients with positive HCV-Ab status, which was not examined in the previous study, was > 60%. The development of diagnostic imaging studies made it possible to detect small tumors, so that patients with tumor s that measured ≤ 2 cm in greatest dimension increased from 16% to 21%, and their 5-year survival rate improved from 53% to 66%. This early detection of small HCC tumors led to an increase in patients with AFP levels < 20 ng/mL from 32% to 43%. The percentage of patients who had solitary tumors was similar to the percentage in our previous report; however, their 5-year survival rate increased from 41% to 56%. Improved surgical techniques have made it possible to resect advanced hepatic tumors. The proportion of patients with portal vein invasion who underwent liver resection increased from 15% to 24%. The indications for surgical treatment of advanced HCC have been expanding.
Table 2. Univariate Analysis of Patients with Hepatocellular Carcinoma
No. of patients (%)
Survival rate (%)
PIVKA-II: protein induced by vitamin K absence or antagonist-II; Eg: expansive growth (well demarcated border); Ig: infiltrative growth (poorly demarcated border).
H0: a solitary tumor measuring ≤ 2.0 cm in greatest dimension with no vascular invasion; Hs: tumor(s) limited to 1 subsegment (Couinaud segment); H1: tumor(s) limited to 1 segment; H2: tumor(s) limited to 2 segments; H3: tumor(s) limited to 3 segments; H4: tumor(s) involving > 3 segments.
In the univariate analysis, there were significant differences in survival among the groups stratified by age, degree of liver damage, serum level of AFP and PIVKA-II levels, maximal tumor dimension, number of tumors, intrahepatic extent of tumors, extrahepatic metastasis, growth appearance, capsular formation, septum formation, portal vein invasion, hepatic vein invasion, bile duct invasion, surgical curability, free surgical margin, and associated liver disease (Table 2). Compared with our previous report, the survival of patients with negative HBs-Ag status, including patients with HCV-Ab, was no better than the survival of patients with negative HBs-Ag status. There also were no significant differences in survival between patients with negative and positive HCV-Ab status.
The multivariate analysis using a Cox proportional hazards model stratified by associated liver disease indicated that age, degree of liver damage, AFP level, maximal tumor dimension, number of tumors, intrahepatic extent of tumors, extrahepatic metastasis, portal vein invasion, hepatic vein invasion, surgical curability, and free surgical margin were independent prognostic predictors for patients with HCC (Table 3). During the last decade, operative mortality was 2.3% in 1990–1991, 2.0% in 1992–1993, 1.4% in 1994–1995, 1.5% in 1996–1997, and 0.6% in 1998–1999.
Table 3. Multivariate Analysis Using the Stratified Cox Proportional Hazard Model by Associated Liver Disease
Surgical curability (absolute noncurative vs. others)
Surgical free margin (positive vs. negative)
The previous LCSGJ report analyzed predictive factors for prognosis of approximately 5800 patients who underwent liver resection for HCC between 1982 and 1989.2 In the current study, we analyzed a large cohort of > 12,000 patients who underwent liver resection for HCC between 1990 and 1999 in a nationwide survey of primary hepatic cancer in Japan. The development of new diagnostic techniques, such as dynamic computed tomography and magnetic resonance imaging, has lead to an increase in patients who had negative HBs-Ag status with low AFP levels, small tumors, and portal vein invasion compared with our previous report.2 The number of patients doubled during the last decade, and the patients' profiles have changed. Advances in therapeutic techniques and perioperative patient care have improved long-term outcomes after radical resection for HCC. In the current study, we analyzed patients with HCC in Japan in the 1990s, during which the 5-year survival rate improved to 50.5% compared with < 40% in the 1980s.2 The overall cumulative survival rate was better than the rates reported in series from Asian and Western countries.9, 10 Operative mortality also improved to < 1.0% in 1998–1999, compared with > 3.0% in 1982–1987.2 Therefore, prognostic factors for patients with HCC should be reevaluated.
In our previous study, the univariate analysis showed significant differences for 13 of 14 factors; and multivariate analysis using the Cox proportional hazards model showed that 3 tumor factors (tumor size, number of tumors, portal vein invasion), 3 clinical factors (age, AFP level, and associated liver disease), and 1 operative factor (surgical curability) were independent predictors of long-term prognosis for patients with HCC. In the univariate analysis for the current study, 18 of 21 factors showed significant differences; and the multivariate analysis, which was stratified by associated liver disease, found that 6 tumor factors (tumor size, number of tumors, intrahepatic extent of tumor, extrahepatic metastasis, portal vein invasion, and hepatic vein invasion), 3 clinical factors (age, degree of liver damage, and AFP level), and 2 operative factors (surgical curability and free surgical margin) were independent prognostic factors for overall survival. Thus, tumor size, number of tumors, and portal vein invasion are well known prognostic factors after resection in patients with HCC.
With regard to tumor size, some reports have shown that patients with tumors that measured ≤ 5 cm in greatest dimension had a better prognosis compared with patients who had tumors > 5 cm. A new International Union Against Cancer (UICC) TNM classification system makes use of 5 cm as a tumor cut-off size.11 However, establishing a screening program for HCC in patients who are at high risk for hepatitis virus infection will increase the diagnoses of small sized HCC tumors. In the current study, the percentage of patients with tumors that measured ≤ 2 cm increased to 21% of all patients, compared with 16% of all patients in our previous report. Three-fourths of patients had tumors that measured ≤ 5 cm. Patients who had tumors that measured ≤ 2 cm had a significantly better prognosis compared with patients who had tumors that measured 2–5 cm. Therefore, in the TNM classification system proposed by the LCSGJ, ≤ 2 cm is the tumor cut-off size.12
Hepatic vein invasion was not evaluated as a prognostic factor in the previous study,2 because only a few patients had hepatic vein invasion. In this large-scale study, however, hepatic vein invasion was an independent prognostic factor, along with portal vein invasion, although the rate of hepatic vein invasion was only 9.0%. Conversely, bile duct invasion was not a predictor in the multivariate analysis, although there were significant differences in survival in the univariate analysis between patients with and without bile duct invasion. Bile duct invasion occurred along with vascular invasion in most patients, which may explain these findings.13
Most patients with HCC have HBV or HCV infection. The long-term survival of patients with HCC who have different hepatitis viral infections has been controversial.14, 15 In the current study, there was no significant difference in survival stratified by either hepatitis B or hepatitis C serology; although, in the previous report, patients who had negative HBsAg status had a better prognosis compared with patients who had positive HBsAg status. Conversely, associated liver disease was an important prognostic factor. Patients who had normal livers and chronic hepatitis had a better prognosis compared with patients who had cirrhosis, although there were no significant differences between patients with normal livers and patients with chronic hepatitis. To select adequate therapeutic options for patients with HCC, prognosis should be assessed at the time of preoperative clinical assessment. The degree of liver damage classification, defined by preoperative clinical data similar to the Child–Pugh classification system, was a significant predictor. The degree of liver damage classified by preoperative clinical data was more useful than histologic evaluation of associated liver disease, not only to estimate hepatic functional impairment for determining the appropriate surgical procedure but also to predict patient prognosis.
Serum levels of the tumor markers AFP and PIVKA-II were associated with significant differences in survival. Furthermore, AFP was an independent prognostic factor. We reported previously that AFP and PIVKA II were indicators of a poor prognosis in patients with HCC.2, 3 The Cancer of the Liver Italian Program investigators also reported that AFP was an independent prognostic factor.16 Koike et al. reported that the PIVKA-II (des-γ-carboxy prothrombin) level was the most useful predisposing clinical parameter for the development of portal vein invasion.17 However, PIVKA-II was recognized as a useful tumor marker for HCC only in the late 1990s, and it was not examined in 30% of patients in the current study. Therefore, PIVKA-II was not included in our multivariate analysis. Recently, some reports have shown that Lens culinaris agglutinin-reactive AFP is another useful predictor for HCC.18 These three markers should be evaluated further to determine which tumor markers will be useful in the near future for clinical screening and for determining prognosis in patients with HCC.
Surgical curability was an important prognostic factor. Patients who underwent absolute noncurative resection in which residual tumor remained had a significantly worse survival compared with patients who underwent other types of surgical resection. Llovet et al. reported that the median survival of patients with unresectable HCC who were managed with systematic treatment was 17 months; and their 1-year, 2-year, and 3-year survival rates were 54%, 40%, and 28%, respectively.19 Those results were similar to the survival results for patients in the current study who underwent absolute noncurative resection. Cytoreduction surgery for HCC does not contribute to improved outcome. Free surgical margin was another important prognostic factor. To avoid tumor recurrence, it is important to perform liver resection with adequate free margins, because microsatellite nodules and histologic venous permeation have been found in adjacent, apparently noncancerous liver.20
Overall, within 5 years after they underwent liver resection, 80% of patients developed recurrent disease, and the most common cause of postoperative death was HCC, either due to tumor recurrence or due to multicentric carcinogenesis in the remnant liver. Local ablation therapy and transcatheter arterial chemoembolization in the treatment of recurrent tumors have contributed to improvements in the prognosis for patients with HCC. Liver transplantation is another surgical modality for patients with HCC who have severely impaired liver function, although few Japanese patients underwent liver transplantation in the 1990s. New postoperative adjuvant therapies with interferon may be crucial in reducing the rate of recurrence, especially recurrent multicentric carcinogenesis.21 However, it remains unknown whether this treatment improves survival.
There are some differences between the recent UICC11 and LCSGJ TNM staging systems.12 The major differences between these two TNM staging systems are the tumor cut-off size discussed above and the extent of vascular invasion. Some studies reported that microscopic vascular invasion reflected on prognosis after resection.22, 23 For patients with HCC, several effective, nonsurgical modalities were applied, such as transcatheter arterial chemoembolization, local ablation therapy, etc. Therefore, it is important that patients with HCC select adequate therapeutic options based on a reliable prognostic preoperative assessment using imaging studies and clinical data, and not based on histopathologic reviews of resected specimens. The results of the current study, in which we reevaluated prognostic factors for patients who underwent liver resection using a recent, large-scale data base, will provide useful information with which to evaluate these TNM staging systems. Recently, new staging systems for HCC reflecting tumor status and liver functional status also have been proposed by several groups, such as the Cancer of the Liver Italian Program score,16 the Barcelona Clinic Liver Cancer stage,24 and the Japan Integrating Staging score.25 It will be necessary in the future to establish a common international staging system to guide discussions of treatment for patients with HCC.