Abstract
Objectives This study aimed to evaluate the seventh edition of the American Joint Committee on Cancer (AJCC) tumour–node–metastasis (TNM) staging system and to compare its efficacy with those of the fifth and sixth editions of the AJCC staging system and the TNM staging system defined by the Liver Cancer Study Group of Japan.
Methods Data for 754 patients submitted to hepatectomy for hepatocellular carcinoma (HCC) between 1989 and 2005 were reviewed. Tumour-free survival was estimated using the Kaplan–Meier method and compared between subgroups using the log-rank test. Prognostic factors for tumour-free survival were identified by multivariable analysis. The accuracy of these staging systems was evaluated using the Cox regression model and a refined staging system was developed based on the drawbacks of the respective systems.
Results According to the criteria defined by the seventh AJCC TNM staging system, 5-year survival was 50.6% in patients with T1 tumours, 21.0% in patients with T2 tumours, 14.6% in patients with T3a tumours, 12.1% in patients with T3b tumours, and 12.9% in patients with T4 tumours. There was no survival difference between patients with T3a and T3b tumours (P = 0.073), nor between those with T3b and T4 tumours (P = 0.227). Significant prognostic tumour factors were microvascular invasion, tumour multiplicity, bilobar disease and a tumour size of ≥5.0 cm. The fifth and sixth editions of the AJCC TNM staging system were found to be more accurate in prognosis than the seventh.
Conclusions The seventh edition of the AJCC TNM staging system is able to adequately stratify patients with early HCC only. A refined staging system is therefore proposed.
Introduction
Hepatocellular carcinoma (HCC) is the fifth most common malignancy in the world.1 Each year, 500 000–1 000 000 new cases are reported and around 600 000 patients die of this cancer globally.2,3 Hepatic resection and liver transplantation remain the curative treatments that attain longterm survival. Nonetheless, stringent patient selection criteria and the shortage of liver grafts have restricted the widespread application of liver transplantation to patients with early HCC. By contrast, longterm prognosis after hepatic resection is often influenced by postoperative tumour recurrence.4,5 Tumour characteristics that determine the biological aggressiveness and metastatic potential of the disease are therefore important prognostic markers for longterm survival6–8 and hence important elements in the staging of HCC.
In 2003, a consensus panel jointly organized by the American Hepato-Pancreato-Biliary Association (AHPBA) and the American Joint Committee on Cancer (AJCC) concluded that the two different categories (i.e. clinical and pathological) of staging systems for HCC should embrace the entire spectrum of the disease.9 A pathological staging system for HCC is one that is designated solely for HCC patients undergoing hepatic resection or liver transplantation. To date, there are two widely accepted pathological staging systems for HCC, one developed by the AJCC and the other by the Liver Cancer Study Group of Japan (LCSGJ).10,11 Both of them use T (tumour: tumour size, number and nearby invasion), N (node: regional lymph node involvement) and M (metastasis: distant metastasis) as parameters to stage the disease.
The AJCC has in fact developed several editions of its TNM staging system. A major drawback of the fifth AJCC TNM staging system (Table 1), which is an earlier edition, is that its clinical application is complex and thus it fails to stratify patients accordingly. To address this issue, a simplified sixth edition (Table 2) was released in 2002.12 Both systems use tumour size and number, vascular invasion, perforation of the visceral peritoneum, and invasion into adjacent organs as parameters for T-staging. Recent data have shown that the sixth edition of the TNM staging system also fails to demonstrate accuracy in predicting longterm survival in patients undergoing hepatic resection.13 The seventh AJCC TNM staging system (Table 3), which is the latest edition, was released in 2009. It subclassifies advanced-stage tumours with reference to tumour multifocality and portal vein invasion.14 However, its prognostic accuracy has yet to be validated in prospective clinical studies. By contrast, the LCSGJ TNM staging system (Table 4) has been validated in Japanese15 and in Chinese16 patients and compared with the fifth and sixth editions of the AJCC staging system.16
Table 1.
American Joint Committee on Cancer tumour–node–metastasis (TNM) system for staging of primary liver cancer, fifth edition
| Stage | Tumour | Node | Metastasis |
|---|---|---|---|
| I | T1 | N0 | M0 |
| II | T2 | N0 | M0 |
| IIIa | T3 | N0 | M0 |
| IIIb | T1, T2, or T3 | N1 | M0 |
| IVa | T4 | Any N | M0 |
| IVb | Any T | Any N | M1 |
T1, solitary tumour of ≤2.0 cm without vascular invasion.
T2, solitary tumour of ≤2.0 cm with vascular invasion, or multiple unilobar tumours of ≤2.0 cm without vascular invasion, or solitary tumours of >2.0 cm without vascular invasion.
T3, solitary tumour of >2.0 cm with vascular invasion, or multiple unilobar tumours of ≤2.0 cm with vascular invasion, or multiple unilobar tumours of >2.0 cm with or without vascular invasion.
T4, multiple bilobar tumours, or with invasion of a major branch of the portal or hepatic vein, or with invasion of an adjacent organ other than the gallbladder, or with perforation of the visceral peritoneum.
N1, regional lymph node metastasis.
M1, distant metastasis.
Table 2.
American Joint Committee on Cancer tumour–node–metastasis (TNM) system for staging of primary liver cancer, sixth edition
| Stage | Tumour | Node | Metastasis |
|---|---|---|---|
| I | T1 | N0 | M0 |
| II | T2 | N0 | M0 |
| IIIa | T3 | N0 | M0 |
| IIIb | T4 | N0 | M0 |
| IIIc | Any T | N1 | M0 |
| IV | Any T | Any N | M1 |
T1, single tumour without vascular invasion.
T2, single tumour with vascular invasion, or multiple tumours none of are >5.0 cm.
T3, multiple tumours any of which are >5.0 cm, or involving a major branch of the portal or hepatic vein.
T4, with direct invasion of an adjacent organ other than the gallbladder or with perforation of the visceral peritoneum.
N1, regional lymph node metastasis.
M1, distant metastasis.
Table 3.
American Joint Committee on Cancer tumour–node–metastasis (TNM) system for staging of primary liver cancer, seventh edition
| Stage | Tumour | Node | Metastasis |
|---|---|---|---|
| I | T1 | N0 | M0 |
| II | T2 | N0 | M0 |
| IIIa | T3a | N0 | M0 |
| IIIb | T3b | N0 | M0 |
| IIIc | T4 | N0 | M0 |
| IVa | Any T | N1 | M0 |
| IVb | Any T | Any N | M1 |
T1, single tumour without vascular invasion.
T2, single tumour with vascular invasion, or multiple tumours, none of >5.0 cm.
T3a, multiple tumours, any of which are >5.0 cm.
T3b, involving a major branch of the portal or hepatic vein.
T4, with direct invasion of an adjacent organ other than the gallbladder, or with perforation of the visceral peritoneum.
N1, regional lymph node metastasis.
M1, distant metastasis.
Table 4.
Liver Cancer Study Group of Japan staging system for primary liver cancer
| Criteria | 1 Solitary tumour |
| 2 Tumour diameter of ≤2.0 cm | |
| 3 No bile duct or vascular invasion1 | |
| T1 All three criteria fulfilled | |
| T2 Two of the three criteria fulfilled | |
| T3 One of the three criteria fulfilled | |
| T4 None of the criteria fulfilled | |
Invasion of the bile duct, portal vein or hepatic vein is determined by macroscopic examination of the resected specimen.
In the present study, the survival outcomes of 754 HCC patients who received curative hepatic resection at a single centre were reviewed. The fifth, sixth and seventh AJCC and the LCSGJ TNM staging systems were evaluated on their respective abilities to predict disease-free survival (DFS). Independent prognostic factors for longterm survival were identified by multivariable analysis and then integrated into a refined staging system. The prognostic ability of the refined staging system was compared with the capabilities of the other staging systems.
Materials and methods
This was a retrospective study that required no institutional approval. From January 1989 to December 2005, 841 patients underwent curative hepatic resection for HCC at the Queen Mary Hospital, Hong Kong. The operative technique for hepatic resection has been reported previously.17 Patients who demonstrated positive resection margins (n = 47) and patients who died postoperatively (n = 40) were excluded from the study. Clinicopathological data for the remaining 754 patients were reviewed. All of these patients had been followed up for ≥60 months at the time of data analysis. Data were censored in December 2011.
For the purpose of this study, major vascular invasion was defined as tumour invasion into the first branch of the portal or first tributary of the hepatic vein. Microvascular invasion (MVI) was defined as the presence of tumour emboli within the tributaries of the hepatic vein or branches of the portal vein. In patients with multiple tumours, MVI referred only to the predominant tumour nodule. Tumour invasion into adjacent organs (except the gallbladder) was confirmed by gross examination of the resected specimens and/or histological examination. Tumour perforation into the peritoneum (i.e. ruptured HCC) was defined as the presence of peritumoral haematoma within the peritoneal cavity. As lymph node metastasis is uncommon in resectable HCC,18–20 routine lymphadenectomy was not performed at the study centre. Data with which to assess the prognostic impact of lymph node metastasis were insufficient and thus N status was not evaluated in the present study.
Follow-up protocol for surveillance of HCC
The follow-up protocol for tumour surveillance included computed tomography (CT) at 1 month after resection to confirm gross tumour clearance and subsequently at every 3 months. In patients with chronic renal impairment or known allergy to i.v. contrast material, magnetic resonance imaging (MRI) was performed instead. The latter was also performed when CT findings were ambiguous in terms of tumour recurrence. Serum liver biochemistry was checked monthly in the first year after resection, and every 3 months thereafter. Tumour recurrence was defined as a new lesion with arterial enhancement and portal venous washout on contrast-enhanced CT or MRI. Biopsy was generally avoided for fear of needle tract seeding of tumour cells. All follow-up data were prospectively recorded and survival data were updated after each follow-up visit. No patient was lost to follow-up.
Statistical analysis
Statistical analysis was performed using spss Version 16.0 (SPSS, Inc., Chicago, IL, USA). Continuous variables were expressed as medians (with interquartile ranges) and compared between groups using the Mann–Whitney U-test. Categorical variables were compared using the chi-squared test or Fisher's exact test as appropriate. Overall survival and DFS were estimated by the Kaplan–Meier method and compared between the various subgroups using the log-rank test. The primary endpoint of the study was DFS defined as the period from the time of hepatic resection to the time of tumour recurrence or death related to HCC or complications of liver failure. Overall survival was defined as the period from the time of hepatic resection to the time of death from any cause. Multivariable analysis using the proportional hazard model was used to identify independent prognostic factors affecting DFS. A P-value of <0.05 was considered to indicate statistical significance. Thereafter, each staging system was entered in turn into Cox regression analysis to evaluate its correlation with DFS. The degree of correlation was expressed in value as −2 log-likelihood. A smaller value indicates a closer correlation between the staging system and the full model. To reaffirm the association of staging system with survival, each staging system was further evaluated and ranked using Kendall's tau rank correlation.
Results
Clinicopathological features of the 754 patients are shown in Table 5. The cohort had a male preponderance. The median age of the patients was 54 years. Most patients were hepatitis B virus (HBV) carriers, but HCC had been detected on screening in just over a third of patients. The majority of patients had Child–Pugh class A cirrhosis. Major hepatic resections were performed as right hepatectomy (40.0%), left hepatectomy (9.5%) or more extensive resection (13.0%). Median perioperative blood loss was 600 ml; 75.4% of patients did not require blood transfusion. At the time of data analysis, 299 patients remained alive on follow-up. The three most common causes of death were terminal malignancy (83.0%), variceal bleeding (5.0%) and liver failure (4.0%).
Table 5.
Clinicopathological features of the 754 patients with hepatocellular carcinoma (HCC) included in the present study
| Age, years, median (range) | 54 (19–82) |
| Gender (male : female) | 608:146 |
| HBsAg positivity, n (%) | 649 (86.1%) |
| HCC detected on screening, n (%) | 219 (29.0%) |
| Child–Pugh class, n (%) | |
| A | 729 (96.7%) |
| B | 25 (3.3%) |
| Total bilirubin, μmol/l, median (range) | 12 (2–61) |
| Albumin, g/l, median (range) | 41 (17–53) |
| Aspartate transaminase, U/l, median (range) | 48 (13–768) |
| Alpha-fetoprotein, ng/ml, median (range) | 128 (1–1 335 900) |
| Platelet count × 109/l, median (range) | 174 (27–686) |
| Indocyanine green retention at 15 min, %, median (range) | 10.4 (1.2–78.0) |
| Tumour size, cm, median (range) | 6 (0.5–28.0) |
| With tumour of ≤5.0 cm, n (%) | 342 (45.4%) |
| With tumour of >5.0 cm, n (%) | 412 (54.6%) |
| With solitary tumour, n (%) | 564 (74.8%) |
| With ≥ 3 tumour nodules, n (%) | 120 (15.9%) |
| With ruptured HCC, n (%) | 52 (6.9%) |
| Major vascular invasion, n (%) | 45 (6.0%) |
| Microvascular invasion, n (%) | 337 (44.7%) |
| Invasion into adjacent organs, n (%) | 45 (6.0%) |
| Tumour differentiation, n (%) | |
| Good | 175 (23.2%) |
| Moderate | 391 (51.9%) |
| Poor | 156 (20.7%) |
| Undifferentiated | 4 (0.5%) |
| Not available | 28 (3.7%) |
| Liver status, n (%) | |
| Normal | 90 (11.9%) |
| Chronic hepatitis | 267 (35.4%) |
| Cirrhosis | 397 (52.7%) |
Table 6 shows the prognostic impact of 21 clinicopathological factors associated with longterm DFS. Patient factors associated with poor prognosis were male gender, serum bilirubin of >12 μmol/l, serum aspartate transaminase (AST) of >48 U/l and serum albumin of ≤41 g/dl. Tumour factors associated with poor prognosis were symptomatic HCC, serum alpha-fetoprotein (AFP) of >400 ng/ml, tumour diameter of >5.0 cm, moderately or poorly differentiated tumour, bilobar tumours, multiple tumour nodules, major vascular invasion, MVI, tumour invasion into adjacent organs, and rupture of HCC. When all of these 14 clinicopathological factors were entered into multivariable analysis, tumour factors found to retain independent prognostic influence were MVI, tumour multiplicity, bilobar tumours, tumour diameter of >5.0 cm and symptomatic presentation (Table 7).
Table 6.
Univariable analysis for disease-free survival (DFS) in hepatocellular carcinoma (HCC)
| Factors | 5-year DFS | P-value |
|---|---|---|
| Age ≤55 years (n = 407) | 34.6% | 0.094 |
| Age >55 years (n = 347) | 27.6% | |
| Male (n = 608) | 29.6% | 0.052 |
| Female (n = 146) | 38.9% | |
| Hepatitis B virus negative (n = 98) | 28.5% | 0.939 |
| Hepatitis B virus positive (n = 649) | 32.2% | |
| Child–Pugh class A (n = 729) | 31.8% | 0.233 |
| Child–Pugh class B (n = 25) | 19.2% | |
| Asymptomatic (n = 320) | 43.4% | <0.0001 |
| Symptomatic (n = 434) | 22.5% | |
| Total bilirubin ≤12 μmol/l (n = 416) | 32.7% | 0.042 |
| Total bilirubin >12 μmol/l (n = 338) | 29.8% | |
| Albumin ≤41 g/dl (n = 439) | 27.7% | 0.032 |
| Albumin >41 g/dl (n = 315) | 36.5% | |
| Aspartate transaminase ≤48 U/l (n = 379) | 40.6% | <0.0001 |
| Aspartate transaminase >48 U/l (n = 375) | 22.1% | |
| Alpha-fetoprotein ≤400 ng/ml (n = 455) | 33.4% | 0.001 |
| Alpha-fetoprotein >400 ng/ml (n = 294) | 27.9% | |
| Platelet count ≤174 × 109/l (n = 378) | 31.7% | 0.989 |
| Platelet count >174 × 109/l (n = 376) | 31.1% | |
| Indocyanine green retention at 15 min ≤10.5% (n = 372) | 34.7% | 0.062 |
| Indocyanine green retention at 15 min >10.5% (n = 360) | 27.2% | |
| With tumour ≤5.0 cm (n = 342) | 42.9% | <0.0001 |
| With tumour >5.0 cm (n = 412) | 21.8% | |
| With tumour ≤2.0 cm (n = 83) | 49.4% | 0.011 |
| With tumour >2.0 cm (n = 671) | 29.2% | |
| With solitary tumour (n = 564) | 37.2% | <0.0001 |
| With multiple tumours (≥2) (n = 190) | 14.2% | |
| Unilobar HCC (n = 657) | 33.6% | <0.0001 |
| Bilobar HCC (n = 97) | 16.5% | |
| No ruptured HCC (n = 702) | 33.0% | <0.0001 |
| With ruptured HCC (n = 52) | 9.6% | |
| No major vascular invasion (n = 709) | 32.7% | <0.0001 |
| With major vascular invasion (n = 45) | 11.1% | |
| No microvascular invasion (n = 417) | 43.6% | <0.0001 |
| With microvascular invasion (n = 337) | 15.6% | |
| No invasion into adjacent organs (n = 709) | 32.3% | 0.006 |
| With invasion into adjacent organs (n = 45) | 17.8% | |
| Resection margin ≤1.0 cm (n = 411) | 30.4% | 0.141 |
| Resection margin >1.0 cm (n = 309) | 33.3% | |
| HCC well differentiated (n = 170) | 37.1% | 0.003 |
| HCC not well differentiated (n = 539) | 29.3% | |
| Liver normal (n = 90) | 27.8% | 0.709 |
| Liver not normal (n = 664) | 31.9% |
Table 7.
Multivariable analysis for disease-free survival in hepatocellular carcinoma (HCC)
| Factors | P-value | HR | 95% CI |
|---|---|---|---|
| Albumin | 0.017 | 0.978 | 0.961–0.996 |
| Platelet count | <0.001 | 0.998 | 0.996–0.999 |
| Tumour >5.0 cm | 0.003 | 1.037 | 1.012–1.061 |
| Bilobar HCC | 0.035 | 1.304 | 1.019–1.670 |
| Symptomatic HCC | 0.004 | 1.355 | 1.099–1.670 |
| Multiple tumours | <0.001 | 1.633 | 1.349–1.976 |
| Microvascular invasion | <0.001 | 1.910 | 1.592–2.291 |
HR, hazard ratio; 95% CI, 95% confidence interval.
In each of the four staging systems (the fifth, sixth and seventh editions of the AJCC and the LCSGJ TNM staging systems), DFS diminished progressively with advancing T-stage. With the fifth AJCC TNM system (Table 1), 1-, 3- and 5-year survival rates were 78.5%, 64.6% and 52.3%, respectively, in patients with T1 tumours, 81.8%, 59.9% and 50.3%, respectively, in patients with T2 tumours, 46.0%, 24.2% and 18.1%, respectively, in patients with T3 tumours, and 31.1%, 15.6% and 11.9%, respectively, in patients with T1 tumours (Fig. 1). There were significant differences in DFS between subgroups of patients with T2 and T3 tumours (P < 0.001), and between those with T3 and T4 tumours (P = 0.001). However, the survival difference between patients with T1 and T2 tumours was not significant (P = 0.315).
Figure 1.
Disease-free survival in the study cohort (n = 754) according to the fifth edition of the American Joint Committee on Cancer tumour–node–metastasis (TNM) staging system
With the sixth AJCC TNM staging system (Table 2), 1-, 3- and 5-year survival rates were 81.6%, 60.3% and 50.6%, respectively, in patients with T1 tumours, 50.7%, 27.9% and 21.0%, respectively, in patients with T2 tumours, 28.7%, 18.0% and 13.9%, respectively, in patients with T3 tumours, and 36.6%, 17.2% and 12.9%, respectively, in patients with T4 tumours (Fig. 2). Patients with T1 tumours had a significantly better survival outcome than those with T2 tumours (P < 0.001), and patients with T2 tumours had a significantly better survival outcome than those with T3 tumours (P = 0.001). However, the survival curves for patients with T3 and T4 tumours almost overlap one another (P = 0.899), which suggests that tumours at these two stages entailed similar prognoses.
Figure 2.
Disease-free survival in the study cohort (n = 754) according to the sixth edition of the American Joint Committee on Cancer tumour–node–metastasis (TNM) staging system
As the sixth and seventh (Table 3) AJCC TNM staging systems use the same T-stage classifications for T1 and T2 tumours, longterm survival rates for patients with T1 and T2 tumours remained unchanged between these two editions of the TNM staging system. Major changes occurred with T3 tumours. Rates of 1-, 3- and 5-year survival were 31.5%, 19.1% and 14.6%, respectively, in patients with T3a tumours, 21.2%, 15.2% and 12.1%, respectively, in patients with T3b tumours, and 36.6%, 17.2% and 12.9%, respectively, in patients with T4 tumours (Fig. 3). There was a significant survival difference between patients with T2 and T3a tumours. However, the survival differences among patients with T3a, T3b and T4 tumours were not statistically significant.
Figure 3.
Disease-free survival in the study cohort (n = 754) according to the seventh edition of the American Joint Committee on Cancer tumour–node–metastasis (TNM) staging system
With the Japanese TNM staging system, 1-, 3- and 5-year DFS rates were 75.6%, 62.8% and 50.0%, respectively, in patients with T1 tumours, 67.1%, 43.5% and 36.3%, respectively, in patients with T2 tumours, 33.5%, 18.4% and 12.8%, respectively, in patients with T3 tumours, and 23.1%, 19.2% and 15.2%, respectively, in patients with T4 tumours. Patients with T2 tumours had significantly better survival than those with T3 tumours (P < 0.001). Survival rates in patients with T1 and T2 tumours differed insignificantly (P = 0.330). The survival difference between patients with T3 and T4 tumours was also not significant (P = 0.300).
Proposal of a new tumour staging system
In view of the inadequacy of all current staging systems in stratifying patients with surgically resectable HCC, a refined T-staging system was proposed based on the multivariable analysis of the present study (Table 8). The five tumour-related factors (MVI, tumour number, lobar distribution, tumour size, symptomatic presentation) that had the greatest impact on survival formed the foundation of the refined system. Serum albumin level and platelet count were not included because the majority of patients (>90%) in the present study had Child–Pugh class A cirrhosis only and thus the prognostic influence of their liver function would be limited. By contrast, as MVI was identified as the most powerful prognostic factor for DFS, it was selected as the main stratifier for T-staging. Five-year DFS rates in patients with and without MVI were 44.0% and 15.6%, respectively (P > 0.001).
Table 8.
Proposed refined T-staging system for primary liver cancer
| T1 | Single tumour ≤5.0 cm without microvascular invasion |
| T2 | Single tumour >5.0 cm without microvascular invasion |
| Or single tumour ≤5.0 cm with microvascular invasion | |
| Or unilobar multiple tumours without microvascular invasion | |
| T3 | Single tumour >5.0 cm with microvascular invasion |
| T4 | Unilobar multiple tumours with microvascular invasion |
| Or bilobar tumours | |
| Or tumour invasion into a branch of the portal or hepatic vein | |
| Or tumour invasion into an adjacent organ except the gallbladder or rupture into the peritoneal cavity |
In the absence of MVI, the size of a solitary tumour had a significant correlation with survival (Fig. 4). Five-year DFS according to tumour size was 51.5%, 55.8%, 44.9% and 23.1%, respectively, in patients with tumours of ≤2.0 cm, 2.1–5.0 cm, 5.1–10.0 cm and ≥10.0 cm (P = 0.001). Based on these findings, a solitary tumour without MVI was classified as T1. The next step was to define the most appropriate tumour size to be used as the cut-off limit for stratification. The sensitivity of different tumour sizes from 1.0 cm to 6.0 cm was estimated by the receiver operating characteristic curve (Table 9). The sensitivity of tumour size at 5.0 cm was 66.0%, which was the highest. Hence, a size of 5.0 cm was chosen as the cut-off value for T1 tumours. Five-year survival rates in patients with tumours of ≤5.0 cm without MVI and patients with tumours of >5.0 cm without MVI were 42.9% and 21.8%, respectively (P < 0.001).
Figure 4.
Impact of tumour size on disease-free survival in patients with solitary tumours without microvascular invasion
Table 9.
Sensitivity of different tumour sizes used as cut-off values for stratification
| Cut-off valuea | AUC, % | P-value | 95% CI |
|---|---|---|---|
| 1.0 cm | 61.5% | 0.211 | 47.0–76.0% |
| 2.0 cm | 63.6% | <0.001 | 58.0–69.2% |
| 3.0 cm | 64.1% | <0.001 | 59.9–68.3% |
| 3.5 cm | 63.6% | <0.001 | 59.5–67.7% |
| 4.0 cm | 64.3% | <0.001 | 60.3–68.3% |
| 4.5 cm | 64.9% | <0.001 | 61.0–68.8% |
| 5.0 cm | 66.0% | <0.001 | 62.1–69.9% |
| 5.5 cm | 65.4% | <0.001 | 61.5–69.3% |
| 6.0 cm | 65.8% | <0.001 | 61.9–69.8% |
Size of largest tumour.
AUC, area under the curve; 95% CI, 95% confidence interval.
Five-year survival rates in patients with a solitary tumour of ≤5.0 cm with MVI, patients with a solitary tumour of >5.0 cm without MVI, and patients with unilobar multiple tumours without MVI were comparable (30.4%, 38.3% and 28.3%, respectively; P = 0.110). Because of the comparable survival rates, these three groups of tumours were all classified as T2 tumour. As the present analysis showed no significant correlation between tumour size and survival in patients with multiple tumours with (P = 0.651) or without (P = 0.384) MVI, it was considered unnecessary to set a cut-off limit for tumour size for multiple tumours.
The next steps involved the definitions of T3 and T4 tumours. Patients with a solitary tumour of >5.0 cm with MVI had a worse prognosis than those with unilobar multiple tumours without MVI (5-year survival rates: 15.1% and 28.3%, respectively; P = 0.013), but a better prognosis than those with unilobar multiple tumours with MVI (5-year survival rates: 15.1% and 8.4%, respectively; P = 0.022) (Fig. 5). By contrast, patients with unilobar multiple tumours with MVI had a similar survival outcome to those with bilobar tumours (5-year survival rates: 8.4% and 16.5%, respectively; P = 0.055). Hence, a solitary tumour of >5.0 cm with MVI was classified as T3 disease, and unilobar multiple tumours with MVI and bilobar tumours were classified as T4 disease. Also classified as T4 were tumours with pathological features such as tumour rupture, invasion of adjacent organs, or invasion into a branch of the portal or hepatic vein. Advanced HCC as such is often manifested by symptoms (100% of patients with major vascular invasion and 95.7% of patients with tumour rupture or tumour invasion into adjacent organs were symptomatic), and symptomatic presentation was a significant prognostic factor for DFS on multivariable analysis. Hence, tumours with these pathological characteristics were grouped together as representing T4 disease.
Figure 5.
Comparison of survival among patients with a solitary tumour of >5.0 cm with microvascular invasion (A; n = 146), unilobar multiple tumours without microvascular invasion (B; n = 53), and unilobar multiple tumours with microvascular invasion (C; n = 82)
Using the refined T-staging system, 1- and 5-year DFS rates (Fig. 6) were 84.8% and 52.2%, respectively, in patients with T1 tumours, 69.6% and 37.1%, respectively, in patients with T2 tumours, 41.5% and 15.1%, respectively, in patients with T3 tumours, and 29.4% and 11.2%, respectively, in patients with T4 tumours (P < 0.001). The corresponding 1- and 5-year overall survival rates were 99.5% and 82.3%, respectively, in patients with T1 tumours, 93.8% and 62.9%, respectively, in patients with T2 tumours, 82.1% and 38.7%, respectively, in patients with T3 tumours, and 68.2% and 26.5%, respectively, in patients with T4 tumours (P < 0.001).
Figure 6.
Disease-free survival in the study cohort (n = 754) under the refined T-staging system
With respect to ability to predict DFS, the refined T-staging system ranked the highest of all five systems in the Cox regression analysis (−2 log-likelihood: 6700.01; P = 0.013) (Table 10). Its Kendall's tau rank correlation coefficient also ranked the highest [Kendall's tau (τ) coefficient: −0.355; P < 0.001] (Table 11). These results indicate that the refined T-staging system achieved the closest association with DFS in the study cohort of all the staging systems considered.
Table 10.
Correlations of different staging systems with disease-free survival by the Cox regression model
| Staging system | Regression | Remaining variation | |||
|---|---|---|---|---|---|
| d.f. | −2 log-likelihood | d.f. | Change in variation | P-value | |
| Full model | 13 | 6677.61a | |||
| Refined | 3 | 6700.01 | 10 | 22.40 | 0.0132 |
| AJCC, 5th edition | 3 | 6706.79 | 10 | 29.18 | 0.0012 |
| AJCC, 6th edition | 3 | 6718.44 | 10 | 40.83 | <0.0001 |
| AJCC, 7th edition | 4 | 6716.30 | 9 | 38.69 | <0.0001 |
| LCSGJ | 3 | 6776.25 | 10 | 98.64 | <0.0001 |
The lesser discrepancy means a stronger correlation with disease-free survival.
d.f., degrees of freedom; AJCC, American Joint Committee on Cancer; LCSGJ, Liver Cancer Study Group of Japan.
Table 11.
Kendall's tau rank correlation coefficients for all staging systems
| Staging system | Kendall's tau (τ) coefficienta | P-value | Rank |
|---|---|---|---|
| Refined | −0.355 | <0.0001 | 1 |
| AJCC, 5th edition | −0.346 | <0.0001 | 2 |
| AJCC, 6th edition | −0.341 | <0.0001 | 3 |
| AJCC, 7th edition | −0.341 | <0.0001 | 3 |
| LCSGJ | −0.254 | <0.0001 | 4 |
A coefficient value of −1 suggests a strong inverse correlation and a value of 0 suggests no correlation.
AJCC, American Joint Committee on Cancer; LCSGJ, Liver Cancer Study Group of Japan.
Validation of the proposed T-staging system
In order to verify the refined T-staging system, a more recent cohort of patients (n = 279) who underwent curative hepatic resection between 1 June 2006 and 31 July 2009 were stratified accordingly. All patients were subject to ≥36 months of postoperative follow-up. As a result, 1- and 5-year DFS rates were 81.5% and 55.1%, respectively, in patients with T1 tumours (n = 22), 67.4% and 38.7%, respectively, in patients with T2 tumours (n = 102), 60.9% and 16.2%, respectively, in patients with T3 tumours (n = 87), and 29.2% and 13.5%, respectively, in patients with T4 tumours (n = 68). Likewise, 1- and 5-year overall survival rates were 98.8% and 82.5%, respectively, in patients with T1 tumours, 89.5% and 61.4%, respectively, in patients with T2 tumours, 95.7% and 49.2%, respectively, in patients with T3 tumours, and 74.2% and 28.0%, respectively, in patients with T4 tumours. There were significant differences in terms of disease-free (P < 0.001) and overall (P < 0.001) survival among the four groups of patients.
Discussion
This study evaluated the prognostic power of four TNM staging systems, the fifth, sixth and seventh editions of the AJCC and the Japanese staging system, and found that none of them were able to adequately stratify the study cohort into various subgroups with clear differences in survival. Indeed, the suboptimal performance of the AJCC TNM staging systems had been reported previously. In 1994, Izumi et al. reported the failure of the fifth edition of the AJCC TNM system to establish prognoses in 104 patients undergoing hepatic resection.21 In 2010, Li et al. reported the inadequacy of the sixth edition of the TNM staging system in stratifying 298 patients subjected to hepatic resection for locally advanced tumours (i.e. T3 and T4).22 In the present analysis, the fifth edition of the AJCC TNM staging system was found to be capable of stratifying patients with advanced-stage tumours, but failed to distinguish between T1 and T2 tumours. By contrast, the sixth and seventh editions were able to stratify patients with T1 and T2 tumours, but failed to show any survival difference between patients with T3 and T4 tumours. In fact, T3 tumours are defined in the same way in the sixth and seventh editions of the AJCC TNM staging system, except that the later edition subclassifies them into T3a (multiple tumours, any of which can be >5.0 cm in size) and T3b (tumour invasion into a major branch of the portal or hepatic vein). Nonetheless, the present analysis showed that survival curves for T3a, T3b and T4 tumours derived from the seventh edition of the TNM staging system approximated one another closely, displaying very slight differences in patient survival among the three groups. Thus, the subclassification of T3 tumours is not justified.
Tumour invasion into a branch of the portal or hepatic vein was found to entail worse prognosis than tumour invasion into adjacent organs or tumour rupture into the peritoneal cavity. In fact, the pronounced prognostic effect of portal vein invasion had been confirmed by various studies.15,22,23 In a retrospective analysis of 13 772 patients submitted to hepatic resection for HCC, portal vein invasion was found to be the most dominant prognostic factor among factors including tumour size, tumour number and degree of tumour cell differentiation.15 A previous report by the present group also highlighted the understating of the prognostic influence of portal vein invasion by TNM staging systems.16 Portal vein invasion is a precursor to intrahepatic metastasis and is a major risk factor for early recurrence after hepatic resection.23,24 By contrast, tumour invasion into adjacent organs or perforation into the peritoneal cavity is likely to manifest as spontaneous rupture. An earlier study by the present group showed that patients with spontaneously ruptured HCC could achieve longterm survival if the initial rupture episode was controlled successfully and followed by a staged hepatic resection.25 As patients with major vascular invasion and patients with tumour rupture or tumour invasion into adjacent organs demonstrate similar survival, the present authors do not support the classifying of tumours with these pathological features into different tumour stages, as in the seventh AJCC system. Nonetheless, for the purpose of counselling patients on their prognosis, it would be appropriate to classify tumours with invasion into adjacent organs or rupture into the peritoneal cavity as T3b, and tumours with portal vein invasion as T4.
This study also evaluated the Japanese TNM staging system, which appears to be simpler to use than the AJCC systems. In the Japanese staging system, points are allocated to elements classifying the T-stage. The elements are tumour size, tumour multiplicity, and major vascular or bile duct invasion. With this staging system, patients with T2 and T3 tumours demonstrated a significant difference in survival, but differences between patients with T1 and T2 tumours, and between patients with T3 and T4 tumours were not significant. Such results may be explained by the relatively small numbers of patients in the T1 and T4 groups in the study. Nonetheless, a major drawback of the Japanese staging system is that it attributes equal prognostic influence to tumour size, tumour number, and vascular or bile duct invasion. The multivariable analysis in the present study illustrated that these factors differ in the degree of their respective prognostic impacts. Another major pitfall of the Japanese TNM staging system is that it bases its tumour classification purely on the macroscopic appearance of tumours and does not describe microscopic features. Such a staging system permits the easy communication of prognostic information with patients before surgery, but delivers a less precise prognosis. The lack of detailed pathological data also undermines the system's appropriateness for use in the design of drug trials for adjuvant therapies.
In view of the insufficiency of the four staging systems in patient stratification and prognosis, the present authors propose a refined T-staging system based on the findings of the current study. An important aspect of the refined system is that it reintroduces a cut-off value for tumour size for solitary tumours. In the fifth edition of the AJCC staging system, a tumour size of 2.0 cm is set as the cut-off value for solitary tumours, as well as multiple tumours. In the sixth and seventh editions, the cut-off value for multiple tumours is shifted from 2.0 cm to 5.0 cm, and there is no cut-off for solitary tumours because the size of such tumours was not considered to have prognostic influence. However, the present study found that the size of solitary tumours had significant prognostic impact, whereas the impact of the size of multiple tumours was insignificant, irrespective of MVI status. Hence, the refined system does not include a cut-value for the size of multiple or bilobar tumours.
Also reintroduced in the refined system is lobar distribution. Again, lobar distribution is recognized in the fifth but discarded in the sixth and seventh editions of the AJCC system because its clinical application is complex. Moreover, lobar distribution has been reported to be irrelevant in prognosis.12 Despite such claims, bilobar involvement in fact implies either multicentric tumour occurrence or intrahepatic metastasis, both of which indicate aggressive tumour behaviour. Hence, the disregard of lobar distribution in T-staging would certainly jeopardize a system's accuracy.
Although the present refined T-staging system was subsequently validated in a more recent cohort of patients, it is subject to a major drawback attributable to the fact that it was derived from a single-centre experience using survival data pooled from a patient population that was predominantly affected by HBV-related HCC. The prognosis in HBV-related HCC is somewhat different from that of hepatitis C virus (HCV)-related HCC because there are differences in the respective disease aetiologies.26,27 Whether such a system is also applicable in areas in which HCV infection is endemic should be clarified in future studies.
A comparison of the present findings with those of two landmark studies on pathological tumour staging12,15 shows that the relatively low proportion of patients (11.1%) with very early-stage HCC (i.e. tumour size of ≤2.0 cm) in the present study may have affected the prognostic impact of different tumour sizes. Such discrepancies in the distribution of tumour sizes between the present study and other studies might be attributable to the differences in the tumour sizes adopted as cut-off values in the respective staging systems.
One of the merits of the present study is its long follow-up period. All of the patients included had been followed for ≥60 months at the time of data analysis, which allowed sufficient time to monitor the behaviour of tumours in each T-stage. Another merit of the study refers to its use of a mono-centre experience in the surgical management of HCC patients. The sixth edition of the AJCC TNM staging system was developed using survival data pooled from four different centres. A big disadvantage of a multicentre study is that it is difficult to standardize data for the operative techniques used in hepatic resection among different surgeons at the various centres. Variations in operative approach may make a significant difference to longterm survival.28 For instance, excessive intraoperative bleeding calls for blood transfusion, which in itself promotes tumour recurrence after hepatic resection and impairs longterm survival.29,30 Use of a standardized operative approach is important if survival outcomes are not to be affected by surgical confounding factors and such an approach can be achieved more readily by a single team of surgeons.17
By contrast with the studies by Vauthey et al.12 and Minagawa et al.,15 DFS was selected as the primary endpoint for survival analysis in the present study. Survival after tumour recurrence is influenced by factors such as remnant liver function, tumour characteristics and subsequent treatment strategy. To date, a widely accepted treatment algorithm for tumour recurrence after hepatic resection has yet to be developed and not all treatments (i.e. salvage transplantation, repeated resection, local ablation) are available at every centre. Hence, using overall survival as the primary endpoint may not enable a fair and equal comparison of survival outcomes among different centres or regions. However, the T-staging system proposed in the present study demonstrated a clear stratification of patients in terms of not only DFS, but also of overall survival, indicating that it may be a clinically practicable staging system.
Conclusions
The performance of the AJCC TNM staging system with respect to the prognosis of patients submitted to hepatic resection for HCC remains subject to limitations. A refined T-staging system for resectable HCC is therefore proposed and appears to be able to achieve satisfactory patient stratification. Nonetheless, further studies are necessary to validate its applicability to other cohorts of patients.
Conflicts of interest
None declared.
References
- Bosch FX, Ribes J, Borras J. Epidemiology of primary liver cancer. Semin Liver Dis. 1999;19:271–285. doi: 10.1055/s-2007-1007117. [DOI] [PubMed] [Google Scholar]
- Sherman M. Hepatocellular carcinoma: epidemiology, risk factors, and screening. Semin Liver Dis. 2005;25:143–154. doi: 10.1055/s-2005-871194. [DOI] [PubMed] [Google Scholar]
- Nordenstedt H, White DL, El-Serag HB. The changing pattern of epidemiology in hepatocellular carcinoma. Dig Liver Dis. 2010;42(Suppl. 3):206–214. doi: 10.1016/S1590-8658(10)60507-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Poon RT, Ng IO, Fan ST, Lai EC, Lo CM, Liu CL, et al. Clinicopathologic features of longterm survivors and disease-free survivors after resection of hepatocellular carcinoma: a study of a prospective cohort. J Clin Oncol. 2001;19:3037–3044. doi: 10.1200/JCO.2001.19.12.3037. [DOI] [PubMed] [Google Scholar]
- Kaibori M, Ishizaki M, Saito T, Matsui K, Kwon AH, Kamiyama Y. Risk factors and outcome of early recurrence after resection of small hepatocellular carcinomas. Am J Surg. 2009;198:39–45. doi: 10.1016/j.amjsurg.2008.07.051. [DOI] [PubMed] [Google Scholar]
- Ishizawa T, Hasegawa K, Aoki T, Takahashi M, Inoue Y, Sano K, et al. Neither multiple tumours nor portal hypertension are surgical contraindications for hepatocellular carcinoma. Gastroenterology. 2008;134:1908–1916. doi: 10.1053/j.gastro.2008.02.091. [DOI] [PubMed] [Google Scholar]
- Taketomi A, Toshima T, Kitagawa D, Motomura T, Takeishi K, Mano Y, et al. Predictors of extrahepatic recurrence after curative hepatectomy for hepatocellular carcinoma. Ann Surg Oncol. 2010;17:2740–2746. doi: 10.1245/s10434-010-1076-2. [DOI] [PubMed] [Google Scholar]
- Sasaki Y, Yamada T, Tanaka H, Ohigashi H, Eguchi H, Yano M, et al. Risk of recurrence in a longterm follow-up after surgery in 417 patients with hepatitis B- or hepatitis C-related hepatocellular carcinoma. Ann Surg. 2006;244:771–780. doi: 10.1097/01.sla.0000225126.56483.b3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Henderson JM, Sherman M, Tavill A, Abecassis M, Chejfec G, Gramlich T. AHPBA/AJCC consensus conference on staging of hepatocellular carcinoma: consensus statement. HPB. 2003;5:243–250. doi: 10.1080/13651820310015833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Japan Liver Cancer Study Group. General Rules for the Clinical and Pathological Study of Primary Liver Cancer. 4th Japanese edn. Tokyo: Kanehara; 2000. [DOI] [PubMed] [Google Scholar]
- Japan Liver Cancer Study Group. General Rules for the Clinical and Pathological Study of Primary Liver Cancer. 2nd English edn. Tokyo: Kanehara; 2003. [Google Scholar]
- Vauthey JN, Lauwers GY, Esnaola NF, Do KA, Belghiti J, Mirza N, et al. Simplified staging for hepatocellular carcinoma. J Clin Oncol. 2002;20:1527–1536. doi: 10.1200/JCO.2002.20.6.1527. [DOI] [PubMed] [Google Scholar]
- Nathan H, Mentha G, Marques HP, Capussotti L, Majno P, Aldrighetti L, et al. Comparative performances of staging systems for early hepatocellular carcinoma. HPB. 2009;11:382–390. doi: 10.1111/j.1477-2574.2009.00070.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Edge SB. AJCC Cancer Staging Manual. 7th edn. New York, NY: Springer; 2007. Liver; pp. 201–210. [Google Scholar]
- Minagawa M, Ikai I, Matsuyama Y, Yamaoka Y, Makuuchi M. Staging of hepatocellular carcinoma: assessment of the Japanese TNM and AJCC/UICC TNM systems in a cohort of 13 772 patients in Japan. Ann Surg. 2009;245:909–922. doi: 10.1097/01.sla.0000254368.65878.da. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Poon RT, Fan ST. Evaluation of the new AJCC/UICC staging system for hepatocellular carcinoma after hepatic resection in Chinese patients. Surg Oncol Clin N Am. 2003;12:35–50. doi: 10.1016/s1055-3207(02)00086-8. [DOI] [PubMed] [Google Scholar]
- Fan ST, Mau Lo C, Poon RT, Yeung C, Leung Liu C, Yuen WK, et al. Continuous improvement of survival outcomes of resection of hepatocellular carcinoma: a 20-year experience. Ann Surg. 2011;253:745–758. doi: 10.1097/SLA.0b013e3182111195. [DOI] [PubMed] [Google Scholar]
- Kobayashi S, Takahashi S, Kato Y, Gotohda N, Nakagohri T, Konishi M, et al. Surgical treatment of lymph node metastases from hepatocellular carcinoma. J Hepatobiliary Pancreat Sci. 2011;18:559–566. doi: 10.1007/s00534-011-0372-y. [DOI] [PubMed] [Google Scholar]
- Sun HC, Zhuang PY, Qin LX, Ye QH, Wang L, Ren N, et al. Incidence and prognostic values of lymph node metastasis in operable hepatocellular carcinoma and evaluation of routine complete lymphadenectomy. J Surg Oncol. 2007;96:37–45. doi: 10.1002/jso.20772. [DOI] [PubMed] [Google Scholar]
- Ercolani G, Grazi GL, Ravaioli M, Grigioni WF, Cescon M, Gardini A, et al. The role of lymphadenectomy for liver tumours: further considerations on the appropriateness of treatment strategy. Ann Surg. 2004;239:202–209. doi: 10.1097/01.sla.0000109154.00020.e0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Izumi R, Shimizu K, Ii T, Yagi M, Matsui O, Nonomura A, et al. Prognostic factors of hepatocellular carcinoma in patients undergoing hepatic resection. Gastroenterology. 1994;106:720–727. doi: 10.1016/0016-5085(94)90707-2. [DOI] [PubMed] [Google Scholar]
- Li B, Yuan Y, Chen G, He L, Zhang Y, Li J, et al. Application of tumour–node–metastasis staging 2002 version in locally advanced hepatocellular carcinoma: is it predictive of surgical outcome? BMC Cancer. 2010;10:535. doi: 10.1186/1471-2407-10-535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Choi KK, Kim SH, Choi SB, Lim JH, Choi GH, Choi JS, et al. Portal venous invasion: the single most independent risk factor for immediate postoperative recurrence of hepatocellular carcinoma. J Gastroenterol Hepatol. 2011;26:1646–1651. doi: 10.1111/j.1440-1746.2011.06780.x. [DOI] [PubMed] [Google Scholar]
- Poon RT, Fan ST, Ng IO, Lo CM, Liu CL, Wong J. Different risk factors and prognosis for early and late intrahepatic recurrence after resection of hepatocellular carcinoma. Cancer. 2000;89:500–507. [PubMed] [Google Scholar]
- Liu CL, Fan ST, Lo CM, Tso WK, Poon RT, Lam CM, et al. Management of spontaneous rupture of hepatocellular carcinoma: single-centre experience. J Clin Oncol. 2001;19:3725–3732. doi: 10.1200/JCO.2001.19.17.3725. [DOI] [PubMed] [Google Scholar]
- Huang YH, Wu JC, Chen CH, Chang TT, Lee PC, Chau GY, et al. Comparison of recurrence after hepatic resection in patients with hepatitis B versus hepatitis C-related small hepatocellular carcinoma in hepatitis B virus endemic area. Liver Int. 2005;25:236–241. doi: 10.1111/j.1478-3231.2005.01081.x. [DOI] [PubMed] [Google Scholar]
- Kao WY, Su CW, Chau GY, Lui WY, Wu CW, Wu JC. A comparison of prognosis between patients with hepatitis B and C virus-related hepatocellular carcinoma undergoing resection surgery. World J Surg. 2011;35:858–867. doi: 10.1007/s00268-010-0928-z. [DOI] [PubMed] [Google Scholar]
- Tanaka S, Mogushi K, Yasen M, Noguchi N, Kudo A, Kurokawa T, et al. Surgical contribution to recurrence-free survival in patients with macrovascular-invasion-negative hepatocellular carcinoma. J Am Coll Surg. 2009;208:368–374. doi: 10.1016/j.jamcollsurg.2008.10.031. [DOI] [PubMed] [Google Scholar]
- Makino Y, Yamanoi A, Kimoto T, El-Assal ON, Kohno H, Nagasue N. The influence of perioperative blood transfusion on intrahepatic recurrence after curative resection of hepatocellular carcinoma. Am J Gastroenterol. 2000;95:1294–1300. doi: 10.1111/j.1572-0241.2000.02028.x. [DOI] [PubMed] [Google Scholar]
- Katz SC, Shia J, Liau KH, Gonen M, Ruo L, Jarnagin WR, et al. Operative blood loss independently predicts recurrence and survival after resection of hepatocellular carcinoma. Ann Surg. 2009;249:617–623. doi: 10.1097/SLA.0b013e31819ed22f. [DOI] [PubMed] [Google Scholar]