Abstract
This study compared the prognostic significance of staging between the American Joint Committee on Cancer 8th edition Tumor, Node, Metastasis (TNM) staging system and the Barcelona Clinic Liver Cancer (BCLC) classification in patients with hepatocellular carcinoma (HCC). The study population comprised patients with liver cancer registered in the Taiwan Cancer Database from 2007 to 2013 and was followed up until December 31, 2016. The study included patients with HCC, with known staging in both TNM and BCLC systems, and with follow‐up >1 month. Primary endpoint was overall survival. Univariate and multivariate Cox proportional hazards model were constructed to investigate the significance of staging by two systems. Goodness‐of‐fit of model was evaluated via Akaike's information criterion (AIC), the lower the better. Among 73,136 patients with newly diagnosed liver cancer, a total of 37,062 patients with HCC (25.6% underwent surgery) were eligible. The mean age and overall survival of this cohort were 63.9 years and 27.2%, respectively. Overall survivals for stages I, II, III, and IV (the TNM system) were 54.5%, 34.9%, 10.3%, and 6.4%, respectively. Overall survivals for stages A, B, C, and D (the BCLC classification) were 54.5%, 29.2%, 9.8%, and 4.0%, respectively. The median follow‐up time was 59.4 months. Multivariate Cox proportional hazards model revealed that both systems predicted overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate well. Values of ΔAIC of the BCLC classification and the TNM system were lower for the surgery group and nonsurgery group, respectively. The TNM system (8th edition) predicted long‐term outcome better than the BCLC classification in patients with HCC. But in patients treated initially with surgery, the BCLC classification outperformed the 8th edition of the TNM system.
Implications for Practice
This work demonstrates that the Tumor, Node, Metastasis (TNM) system (8th edition) and the Barcelona Clinic Liver Cancer (BCLC) classification both predict long‐term outcome significantly in patients with hepatocellular carcinoma but that the TNM system (8th edition) predicts long‐term outcome better than the BCLC classification. For patients treated initially with surgery, BCLC classification outperforms in 8th edition TNM system in predicting long‐term outcome.
Keywords: Hepatocellular carcinoma, Barcelona Clinic Liver Cancer classification, TNM staging system, Overall survival
Short abstract
This article compares long‐term outcome predictability between two frequently used staging systems, the American Joint Commission on Cancer (AJCC) 8th edition TNM staging system and the Barcelona Clinic Liver Cancer (BCLC) staging classification, in patients with hepatocellular carcinoma.
Introduction
Hepatocellular carcinoma (HCC) is the most common primary liver cancer and the third or fourth leading cause of cancer‐related death worldwide [1, 2, 3]. Especially in Asia, high incidence of HCC leads to a heavy disease burden and workforce loss for several countries. Although surgery (liver resection and transplantation) offers a chance of cure, many (up to 85%) patients with HCC fail to undergo surgery as an initial treatment for multiple reasons (disease severity, tumor location, comorbidities, patient's choice, etc.) [4]. Most HCC is preceded by liver cirrhosis, complicating and influencing the potential treatment modality and subsequent survival outcomes [3]. As a result, to design a staging system for HCC to correctly predict long‐term outcome of individual patients may be a difficult task.
An ideal staging system should be simple, easy, and quick to be determined as soon as possible after diagnosis; it should also provide information on prognosis and guide therapeutic decisions [5]. The most frequently used staging system for malignant disease including HCC is the American Joint Committee on Cancer (AJCC) and the Union for International Cancer Control TNM staging system, which uses a scheme of tumor (T), node (N), and metastasis (M) categories to describe and stratify the severity of the disease [6]. The TNM staging system has been updated several times (historically 5–7 years a cycle) in response to newly acquired clinical and pathological data and is one of the most widely accepted systems. In October 2016, the AJCC published the 8th edition of the AJCC/TNM cancer staging system, which replaced the 7th edition and took effect after January 1, 2018 [7]. It has been revised with particular focus on international expertise from Asia, where the incidence of hepatobiliary cancer is high [8].
The European Association for the Study of the Liver and American Association for the Study of Liver Diseases recommended another system to stage patients with HCC. The Barcelona Clinic Liver Cancer (BCLC) staging classification was first introduced in 1999 and comprises four stages that select the best candidates for the best therapies available [9]. This model is derived from the results of a study of the outcomes of radical therapy and/or the natural history of patients with untreated HCC [9]. It includes variables related to tumor stage, liver functional status, physical status, and cancer‐related symptoms and links the four stages described with a treatment strategy [4]. Currently, it remains one of the most widely used and validated prognostic tools.
Although there are over a dozen staging systems proposed for patients with HCC, no consensus has been made on the best classification available [5, 10, 11, 12, 13]. Survival after diagnosis with HCC is due primarily to three related factors: cancer biology, delivery of an optimal cancer‐directed therapy, and the patient's underlying health and liver function [6]. No single staging system can incorporate all the nuances of treatment, given the complex and heterogeneous nature of HCC and various factors that play a role in decision making [14]. External prognostic validation is an important step in staging system establishment because prediction models almost always perform better on the development cohort [15], especially after the recent enactment of the 8th edition of the AJCC/TNM cancer staging system. Therefore, a comparative study between frequently used staging systems in patients with HCC allows assessment of model robustness in a new patient cohort with measured and unmeasured differences from the development cohort.
The aim of this study was to compare the long‐term outcome predictabilities among two frequently used staging systems (the 8th TNM staging system and the BCLC classification) for patients with newly diagnosed HCC in a population‐based cancer registry.
Materials and Methods
Study Population
This study examined patients with HCC from a national prospective database, the Taiwan Cancer Registry (TCR). Candidates were patients with International Classification of Diseases for Oncology, third revision (ICD‐O‐3) C22 (liver cancer) from January 2007 to December 2013. The TCR was established by the government to provide individual information (cancer staging, treatment profile, and follow‐up status of individual patients) for public interest and academic research [16, 17]. Participation of hospitals in this registry is almost compulsory; currently it collects information about 87.8% of patients with newly diagnosed HCC nationwide.
Analysis Cohort
Patients with age more than 85 years, pathology report not indicating HCC (ICD‐O‐3 histology code 817), and survival less than 1 month were excluded. Because our focus was staging systems (supplemental online Table 1), patients whose stage was unknown in either the TNM or BCLC system were also excluded.
Independent Variables
Relevant demographic, clinical, pathological, and therapeutic data of patients with HCC were retrieved from the TCR. Possible prognostic variables included patient characteristics (age, gender, Charlson comorbidity index), disease characteristics (tumor size, tumor number, histology grade of cell differentiation, vascular invasion, cirrhosis) and therapeutic factors [surgery, resection margin, transarterial chemoembolization (TAE), and nucleoside analog therapy]. The use of nucleoside analogs indicated that patients took at least lamivudine, entecavir, tenofovir, telbivudine, or adefovir dipivoxil [18]. Comorbidities included myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, peptic ulcer disease, mild liver disease, diabetes without complication, diabetes with chronic complications, hemiplegia (or paraplegia), moderate or severe renal disease, cancer without metastasis, moderate or severe liver disease, metastatic solid tumor, and AIDS [19]. The information of comorbidity (Charlson comorbidity index, Deyo version) and nucleoside analog usage (oral antiviral therapy) required database combining with the National Health Insurance Research Database (NHIRD) to complete. The NHIRD, from the National Health Insurance Administration, offers claim data of patients. Because the insurance system in Taiwan is a one‐payer system, this database covered health care billing services of all patients. Therefore, data of comorbidity and medication can be accurately retrieved.
Dependent Variable
The primary endpoint was overall survival, referring to the percentage of patients who were still alive for a certain period of time after the diagnosis of HCC. It reflects a composite clinical endpoint of mutual influence of cirrhosis, underlying liver disease, other comorbidities, and active malignancy. The other endpoints included cancer‐specific survival, disease‐free survival, and local recurrence‐free rate, which refer to the corresponding percentage of patients who were still event‐free for a certain period of time after the diagnosis of HCC. Information about these outcomes was further validated by linking to the 2002–2016 Cause of Death Data (from the Ministry of Health and Welfare), a third database. Date and cause of death were also retrieved if death occurred. The institutional review board of Taipei City Hospital approved this study and waived informed consent (TCHIRB‐10802005‐W).
Statistical Analysis
SAS statistical software, version 9.2 (SAS Institute Inc., Cary, NC) and SPSS 21.0 (SPSS Inc., Chicago, IL) were used for data merging and statistical analyses. Demographic, clinical, pathological, and therapeutic variables were reported as percentages or means ± SD. Kaplan‐Meier plots and log‐rank test were performed to calculate survival estimates and significance regarding two different staging systems. Univariate and multivariate Cox proportional hazards models were conducted to assess the independent prognostic value of stages defined by the TNM staging system and the BCLC classification. All estimates in multivariate analysis were derived by controlling for age, sex, tumor cell differentiation, cirrhosis, comorbidity, hepatitis virus infection, TAE, and oral antiviral therapy. Hazard ratio (HR) with 95% confidence interval (CI) of estimate was reported. A value of p ≤ .05 was considered to be statistically significant. We used Akaike's information criterion (AIC) to represent the goodness‐of‐fit of models in the multivariate survival analysis. The AIC method gives directions to select the best model from a set of models by computer calculations as follows: AIC = 2k − 2 ln(L), where k refers to the number of parameters and L refers to the maximized value of the likelihood function for the estimated model [16, 20]. The homogeneity (differences in outcomes among patients of the same stage) was assessed by “−2 log‐likelihood,” which could be calculated by Cox proportional hazards models [21]. Because the smallest value of AIC indicates the best model fit, we report ΔAIC (distance between the best model and each other model). The model with zero ΔAIC reflects the best model fit and vice versa.
Results
Initially, we identified 73,136 patients who were newly diagnosed with liver cancer between January 2007 and December 2013. There were 67,624 patients who had pathology report indicating HCC (Fig. 1). Finally, a total of 37,062 patients with known tumor stage (both the 8th TNM system and the BCLC classification) and follow‐up for more than 1 month were eligible for this study. The mean age of this cohort was 63.9 years (SD 11.7 years). The demographic, clinical, and interventional characteristics are summarized in Table 1. Most HCC occurred in age between 55 and 75 years (57.5%, 21,319/37,062) and in male patients (71.5%, 26,513/37,062). Tumor sizes that were smaller than 5 cm and larger than 10 cm accounted for 58.0% (21,487/37,062) and 14.3% (5,294/37,062) of patients, respectively. Twenty‐five point six percent (9,501/37,062) of patients underwent liver surgery as initial intervention for HCC. The median time between initial diagnosis and initial surgery was 0.9 month for the surgery group. In the studied cohort, 66.9% (24,783/37,062) and 70.4% (26,078/37,062) of patients had solitary tumor and cirrhotic liver disease. Many patients (43.6%, 16,169/37,062) had Charlson comorbidity index greater than 4. Chronic infections with hepatitis B virus (HBV) and hepatitis C virus (HCV) were found in 41.5% and 27.9% of the cohort, respectively. The remaining 15.4% of patients had both positive HBV and HCV infection. Additional therapy included TAE (50.9%, 18,872/37,062) and oral antiviral therapy for HBV infection (22.7%, 8,420/37,062).
Figure 1.
Flow chart of patient enrollment in this study. Abbreviations: BCLC, Barcelona Clinic Liver Cancer; HCC, hepatocellular carcinoma; TNM, Tumor, Node, Metastasis.
Table 1.
Baseline characteristics including the TNM and the BCLC classifications in patients with hepatocellular carcinoma (2007–2013, n = 37,062)
| Variables | n (%) |
|---|---|
| Overall | 37,062 (100) |
| Age group, years | |
| <45.0 | 2,388 (6.4) |
| 45.0–54.9 | 6,053 (16.3) |
| 55.0–64.9 | 10,653 (28.7) |
| 65.0–74.9 | 10,666 (28.8) |
| >75.0 | 7,302 (19.7) |
| Gender: male | 26,513 (71.5) |
| Cell grade | |
| Well/moderately differentiated | 10,585 (28.6) |
| Poorly differentiated or undifferentiated | 3,611 (9.7) |
| Unknown | 22,866 (61.7) |
| Primary tumor size | |
| <5.0 cm | 21,487 (58.0) |
| 5.1–10.0 cm | 7,993 (21.6) |
| >10.0 cm | 5,294 (14.3) |
| Unknown | 2,288 (6.2) |
| Liver resection as initial treatment | |
| No | 27,561 (74.4) |
| PEI/PAI | 923 (2.5) |
| RFA | 4,818 (13.0) |
| Others (cryotherapy, laser) or nil | 21,820 (58.9) |
| Yes | 9,501 (25.6) |
| Liver resection (wedge or 1 or 2 segments) | 6,421 (17.3) |
| Liver resection (3 segments) | 878 (2.4) |
| Left (or extended) hepatectomy | 615 (1.7) |
| Right (or extended) hepatectomy | 1,147 (3.1) |
| Liver transplantation | 301 (0.8) |
| Lobectomy, unspecified | 139 (0.4) |
| Resection margin | |
| Positive | 430 (1.2) |
| Negative | 9,054 (24.4) |
| Unknown | 27,578 (74.4) |
| Comorbidities | |
| <1 | 11,323 (30.6) |
| 2, 3, 4 | 9,561 (25.8) |
| >4 | 16,169 (43.6) |
| Unknown | 9 (0) |
| History of cirrhosis: Yes | 26,078 (70.4) |
| Number of tumors | |
| Single | 24,783 (66.9) |
| Multiple | 11,455 (30.9) |
| Unknown | 824 (2.2) |
| Vascular involvement | |
| No | 24,808 (66.9) |
| Yes | 12,145 (32.8) |
| Unknown | 109 (0.3) |
| Hepatitis virus infection | |
| History of HBV infection | 15,394 (41.5) |
| History of HCV infection | 10,354 (27.9) |
| History of HBV and HCV infection | 5,703 (15.4) |
| No history of HBV/HCV infection | 5,611 (15.1) |
| Body mass index | |
| <30.0 kg/m2 | 11,339 (30.6) |
| >30.0 kg/m2 | 8,275 (22.3) |
| Unknown | 17,448 (47.1) |
| Smoking status | |
| Smoker | 12,252 (33.1) |
| Nonsmoker | 8,589 (23.2) |
| Unknown | 16,221 (43.8) |
| Drinking habit | |
| No alcohol consumption | 13,168 (35.5) |
| Habitual drink | 2,750 (7.4) |
| Quit or rare | 4,908 (13.2) |
| Unknown | 16,236 (43.9) |
| Perioperative α‐fetoprotein level | |
| <400 ng/mL | 13,899 (37.5) |
| 401–6,000 ng/mL | 3,140 (8.5) |
| >6,000 ng/mL | 2,226 (6.0) |
| Unknown | 17,797 (48) |
| TAE: Yes | 18,872 (50.9) |
| Nucleos(t)ide analog user: Yes | 8,420 (22.7) |
| BCLC classification | |
| Stage A + 0 | 16,420 (44.3) |
| Stage B | 8,617 (23.3) |
| Stage C | 9,184 (24.8) |
| Stage D | 2,841 (7.7) |
| TNM, 8th edition | |
| Stage I + 0 | 15,694 (42.3) |
| Stage II | 7,851 (21.2) |
| Stage III | 8,374 (22.6) |
| Stage IV | 5,143 (13.9) |
| History of fulminant hepatitis/liver failure: Yes | 525 (1.4) |
| Sorafenib: Yes | 2,213 (6.0) |
| Statin usage (days/year) | |
| Nil | 36,267 (97.9) |
| >180 | 26 (0.1) |
| 1–180 | 769 (2.1) |
| Jaundice: Jaundice at (before) initial diagnosis | 98 (0.3) |
| Overall survival: Yes | 12,489 (33.7) |
| Cancer‐specific survival: Yes | 18,537 (50) |
| Disease‐free survival: Yes | 10,079 (27.2) |
| Local recurrence: Yes | 24,765 (66.8) |
| Recurrence status | |
| No | 12,297 (33.2) |
| Local recurrence | 24,742 (66.8) |
| Distant metastasis | 23 (0.1) |
Abbreviations: BCLC, Barcelona Clinic Liver Cancer; HBV, hepatitis B virus; HCV, hepatitis C virus; PEI/PAI, percutaneous ethanol or acetic acid injection; RFA, radiofrequency ablation; TAE, transarterial chemoembolization; TNM, Tumor, Node, Metastasis (8th edition).
Stages I, II, III, and IV (by the 8th TNM system) accounted for 42.3%, 21.2%, 22.6%, and 13.9% of the cohort, respectively. Stages 0/A, B, C, and D (by the BCLC classification) accounted for 44.3%, 23.3%, 24.8%, and 7.7% of the cohort, respectively. The correlation coefficient between the 8th TNM system and the BCLC classification was 0.71. At the end of study, 66.3% (24,573/37,062) of the cohort died, and 50.0% (18,525/37,062) of the cohort died because of HCC or HCC‐related causes (supplemental online Table 2). Only 27.2% of the patients (10,079/37,062) had lived free of HCC‐related condition. The median follow‐up time was 59.4 months (range 36–120 months; 62.4 ± 18.8 months).
Figure 2 displays Kaplan‐Meier survival plots according to stages I, II, III, and IV by the 8th TNM system and stages 0/A, B, C, and D by the BCLC classification. Both cancer staging systems could predict distinct overall survival and cancer‐specific survival (all p < .001 by log‐rank test between two individual stages). The overall 3‐, 5‐, 7‐year survival rates for stages I and A were 70.5%, 56.1%, 45.7% and 71.7%, 56.4%, 45.3%, respectively. The overall 3‐, 5‐, 7‐year survival rates for stages II and B were 52.5%, 35.3%, 25.8% and 42.6%, 29.5%, 22.3%, respectively. The overall 3‐, 5‐, 7‐year survival rates for stages III and C were 15.9%, 10.7%, 7.8% and 14.6%, 9.6%, 6.8%, respectively. The overall 3‐, 5‐, 7‐year survival rates for stages IV and D were 9.0%, 6.2%, 4.5% and 6.6%, 3.8%, 3.3%, respectively. Table 2 summarizes the results of univariate Cox proportional hazards model with respect to overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate. No matter which staging system was employed, individual stage could predict outcomes superbly regarding overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate. With reference to stage I/A, each stage of the BCLC classification had higher HR (both overall survival and cancer‐specific survival) than that of their counterparts of the 8th TNM system. These prognostic findings still kept in the results of multivariate Cox proportional hazards model (Table 3) except that the HR of stage IV for cancer‐specific survival (HR, 8.83; 95% CI, 8.44–9.24) was higher than that of stage D (HR, 8.79; 95% CI, 8.33–9.28). The prognostic predictability of individual stage by the 8th TNM system and the BCLC classification remained strongly significant for overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate when the cohort was divided into surgery group or nonsurgery group except for stage D in the surgery group. In the surgery group, stage D did not show significant disease‐free survival (HR, 1.23; 95% CI, 0.80–1.90; p = .347) and local recurrence‐free rate (HR, 1.57; 95% CI, 0.94–2.62; p = .085) with reference to stage A.
Figure 2.
Kaplan‐Meier curves comparing the 8th edition of the Tumor, Node, Metastasis (TNM) system and the Barcelona Clinic Liver Cancer (BCLC) classification in patients with hepatocellular carcinoma. (A): Overall survival of the TNM system. (B): Cancer‐specific survival of the TNM system. (C): Overall survival of the BCLC classification. (D): Cancer‐specific survival of the BCLC classification.
Table 2.
Univariate Cox regression of overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate regarding the TNM staging system (8th edition) and the BCLC classification in patients with hepatocellular carcinoma (n = 37,062)
| Variables | Overall survival | Cancer‐specific survival | Disease‐free survival | Local recurrence‐free rate | ||||
|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | p value | HR (95% CI) | p value | HR (95% CI) | P value | HR (95% CI) | p value | |
| BCLC classification (stage A + 0) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage B | 2.21 (2.14–2.29) | <.001 | 2.55 (2.45–2.66) | <.001 | 2.10 (2.04–2.17) | <.001 | 2.40 (2.32–2.49) | <.001 |
| Stage C | 5.61 (5.44–5.80) | <.001 | 6.97 (6.72–7.24) | <.001 | 4.60 (4.46–4.74) | <.001 | 5.81 (5.62–6.00) | <.001 |
| Stage D | 9.76 (9.33–10.21) | <.001 | 11.25 (10.67–11.85) | <.001 | 7.14 (6.83–7.46) | <.001 | 9.50 (9.08–9.94) | <.001 |
| TNM, 8th edition (stage I) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage II | 1.72 (1.66–1.79) | <.001 | 1.86 (1.78–1.94) | <.001 | 1.81 (1.75–1.87) | <.001 | 2.01 (1.94–2.09) | <.001 |
| Stage III | 5.02 (4.85–5.18) | <.001 | 6.27 (6.03–6.51) | <.001 | 4.40 (4.26–4.54) | <.001 | 5.42 (5.24–5.60) | <.001 |
| Stage IV | 7.53 (7.25–7.82) | <.001 | 9.80 (9.39–10.24) | <.001 | 6.21 (5.99–6.44) | <.001 | 8.03 (7.73–8.34) | <.001 |
Abbreviations: BCLC, Barcelona Clinic Liver Cancer; CI, confidence interval (characters in the parentheses indicate reference group); HR, hazard ratio; TNM, Tumor, Node, Metastasis.
Table 3.
Multivariate Cox regression for prognostic factors of overall, cancer‐specific, disease‐free, and local recurrence‐free survival regarding the TNM staging system (8th edition) and the BCLC classification in patients with hepatocellular carcinoma
| Multivariate | Overall survival a | Cancer‐specific survival a | Disease‐free survival a | Local recurrence‐free a | ||||
|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | p value | HR (95% CI) | p value | HR (95% CI) | p value | HR (95% CI) | p value | |
| Overall (n = 37,062) | ||||||||
| BCLC classification (stage A + 0) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage B | 2.17 (2.09–2.24) | <.001 | 2.47 (2.37–2.58) | <.001 | 1.95 (1.89–2.02) | <.001 | 2.26 (2.18–2.34) | <.001 |
| Stage C | 5.01 (4.85–5.18) | <.001 | 6.19 (5.95–6.43) | <.001 | 3.97 (3.85–4.10) | <.001 | 4.97 (4.81–5.14) | <.001 |
| Stage D | 7.37 (7.03–7.72) | <.001 | 8.79 (8.33–9.28) | <.001 | 5.70 (5.44–5.96) | <.001 | 7.19 (6.86–7.54) | <.001 |
| TNM, 8th edition (stage I) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage II | 1.54 (1.49–1.60) | <.001 | 1.65 (1.57–1.72) | <.001 | 1.55 (1.50–1.61) | <.001 | 1.73 (1.67–1.79) | <.001 |
| Stage III | 4.63 (4.47–4.79) | <.001 | 5.71 (5.49–5.95) | <.001 | 3.81 (3.68–3.94) | <.001 | 4.70 (4.54–4.87) | <.001 |
| Stage IV | 6.82 (6.56–7.09) | <.001 | 8.83 (8.44–9.24) | <.001 | 5.34 (5.14–5.54) | <.001 | 6.78 (6.52–7.05) | <.001 |
| Initial treatment: nonsurgery (n = 27,561) | ||||||||
| BCLC classification (stage A + 0) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage B | 2.38 (2.29–2.47) | <.001 | 2.72 (2.59–2.85) | <.001 | 2.13 (2.05–2.21) | <.001 | 2.45 (2.36–2.55) | <.001 |
| Stage C | 5.08 (4.9–5.27) | <.001 | 6.27 (6.01–6.54) | <.001 | 4.04 (3.91–4.18) | <.001 | 4.89 (4.72–5.07) | <.001 |
| Stage D | 7.36 (7.01–7.73) | <.001 | 8.73 (8.26–9.24) | <.001 | 5.82 (5.56–6.10) | <.001 | 7.12 (6.78–7.47) | <.001 |
| TNM, 8th edition (stage I) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage II | 1.58 (1.52–1.64) | <.001 | 1.70 (1.62–1.79) | <.001 | 1.57 (1.51–1.63) | <.001 | 1.71 (1.65–1.78) | <.001 |
| Stage III | 4.88 (4.70–5.07) | <.001 | 6.11 (5.85–6.40) | <.001 | 3.92 (3.78–4.06) | <.001 | 4.75 (4.58–4.93) | <.001 |
| Stage IV | 7.04 (6.75–7.34) | <.001 | 9.28 (8.83–9.74) | <.001 | 5.49 (5.27–5.72) | <.001 | 6.75 (6.47–7.03) | <.001 |
| Initial treatment: surgery (n = 9,501) | ||||||||
| BCLC classification (stage A + 0) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage B | 1.90 (1.75–2.06) | <.001 | 2.17 (1.97–2.38) | <.001 | 1.88 (1.74–2.03) | <.001 | 2.20 (2.00–2.41) | <.001 |
| Stage C | 3.10 (2.80–3.42) | <.001 | 3.67 (3.27–4.12) | <.001 | 2.94 (2.67–3.24) | <.001 | 4.05 (3.63–4.52) | <.001 |
| Stage D | 2.36 (1.70–3.28) | <.001 | 2.15 (1.37–3.36) | .001 | 1.23 (0.80–1.90) | .347 | 1.57 (0.94–2.62) | .085 |
| TNM, 8th edition (stage I) | <.001 | <.001 | <.001 | <.001 | ||||
| Stage II | 1.33 (1.21–1.46) | <.001 | 1.35 (1.21–1.51) | <.001 | 1.38 (1.26–1.51) | <.001 | 1.54 (1.39–1.72) | <.001 |
| Stage III | 2.54 (2.31–2.79) | <.001 | 2.83 (2.54–3.16) | <.001 | 2.67 (2.44–2.93) | <.001 | 3.22 (2.90–3.57) | <.001 |
| Stage IV | 3.23 (2.82–3.69) | <.001 | 3.58 (3.07–4.17) | <.001 | 2.86 (2.51–3.25) | <.001 | 4.07 (3.53–4.69) | <.001 |
Estimates derived from controlling for age at diagnosis, gender, tumor cell grade, comorbidity index at diagnosis, cirrhosis, hepatitis B or C virus infection, oral antiviral medication, and transarterial chemoembolization.
Abbreviations: BCLC, Barcelona Clinic Liver Cancer; CI, confidence interval (characters in the parentheses indicate reference group); HR, hazard ratio; TNM, Tumor, Node, Metastasis.
Table 4 compares the ΔAIC of the 8th TNM system and the BCLC classification with respect to overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate. As a whole, the AIC values for the 8th TNM system were lower (= 0) than those for the BCLC classification. This indicated that the 8th TNM system presented better goodness‐of‐fit for all kinds of survival. For patients in the nonsurgery group, adoption of the 8th TNM system might fit more precisely for multivariate Cox proportional hazards model than the BCLC classification, especially in cancer‐specific survival (ΔAIC = 1,081.639). On the contrary, for patients in the surgery group, adoption of the BCLC classification might fit more precisely for multivariate Cox proportional hazards model than the 8th TNM system, even in patients with cirrhosis (supplemental online Table 3).
Table 4.
Multivariate comparison of delta AIC (ΔAIC: BCLC vs. TNM) values from the TNM staging system (8th edition) and the BCLC classification for the best‐fit model of survival in patients with hepatocellular carcinoma
| Variables | Overall (n = 37,062) | Initial treatment | ||||
|---|---|---|---|---|---|---|
| Nonsurgery (n = 27,561) | Surgery (n = 9,501) | |||||
| −2LLL | ΔAIC | −2LLL | ΔAIC | −2LLL | ΔAIC | |
| Overall survival | ||||||
| BCLC | 471,590.960 | 353.832 | 395,798.441 | 409.405 | 54,985.640 | 0 |
| TNM | 471,237.128 | 0 | 395,389.036 | 0 | 55,041.219 | 55.579 |
| Cancer‐specific survival | ||||||
| BCLC | 354,806.385 | 989.868 | 300,661.481 | 1,081.639 | 38,718.478 | 0 |
| TNM | 353,816.517 | 0 | 299,579.842 | 0 | 38,796.275 | 77.797 |
| Disease‐free survival | ||||||
| BCLC | 516,029.339 | 402.357 | 433,492.857 | 229.534 | 57,521.868 | 0 |
| TNM | 515,626.982 | 0 | 433,263.323 | 0 | 57,523.222 | 1.354 |
| Local recurrence‐free | ||||||
| BCLC | 468,831.155 | 373.540 | 407,361.166 | 254.339 | 41,236.101 | 0 |
| TNM | 468,457.615 | 0 | 407,106.827 | 0 | 41,258.720 | 22.619 |
Abbreviations: −2LLL, −2 log‐likelihood; AIC, Akaike's information criterion; BCLC, Barcelona Clinic Liver Cancer; TNM, Tumor, Node, Metastasis.
Discussion
The present study demonstrated external prognostic validation of the 8th edition of the TNM system and the BCLC classification in an Asian cohort of patients with HCC in which only one fourth underwent surgery as an initial treatment. Using a prospective longitudinal database, both the AJCC 8th edition of the TNM staging system for HCC and the BCLC classification have shown excellent capability of predicting overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate in patients with HCC. The goodness‐of‐fit test indicated that the 8th edition of the TNM staging system outperformed the BCLC classification for the cohort. However, there existed disagreement of prognostic superiority between initial treatments, which was either early surgery or initial nonsurgery treatment. In the surgery group, the BCLC classification was more powerful than the AJCC 8th edition to correctly predict patients’ long‐term outcomes. Yet for all patients with HCC, the AJCC 8th staging system demonstrated the best overall performance in survival outcomes, as most patients with HCC could not undergo surgery initially.
Because the first edition of the AJCC Cancer Staging Manual was published in 1977, several revisions of the TNM staging system have improved prediction of outcomes compared with previous editions. A study queried the Surveillance, Epidemiology, and End Results (SEER) database (2010–2013), which comprised 20,281 patients with HCC, and revealed a small improvement of the 8th edition of the TNM system in predicting overall and cancer‐specific survivals compared with the 7th edition [22]. The authors reported that the 3‐year overall survival rates for stages I, II, III, and IV were 45%, 38%, 13%, and 5%, which were lower than the present study (75%, 52%, 16%, and 9%). The difference might be partially attributed to the smaller percent (9%) of surgery reported in their study. A similar study using SEER database (1998–2013), which comprised 8,918 patients (63% resected or 37% transplanted), reported that the c‐index for the AJCC 8th edition staging system was 0.60, which was parallel to that for the 7th edition (0.59) [23]. Another study using 792 Korean patients with HCC (underwent liver resection or liver transplantation) reported comparable results showing small improvement of the 8th edition of the TNM system in predicting overall and cancer‐specific survivals compared with the 7th edition [21].
The result of the surgery group of the present study was similar to a research by Liu et al., who investigated 774 Chinese patients with HCC (underwent surgical resection) between 2007 and 2009 and reported that the BCLC classification was the better staging system in predicting survival than the 7th edition of the TNM system [24]. In a study that analyzed 743 HBV‐related Chinese patients with HCC who underwent surgery initially, Zhao et al. have shown that the BCLC classification was a better staging model for HBV infected patients with HCC in Chinese population than the 7th and the 6th TNM staging systems [11]. They reported that the AIC values for overall survival regarding the BCLC and the 7th and the 6th TNM staging systems were 2,896.58, 2,899.98, and 2,902.17, respectively. Therefore, the ΔAIC value (the 7th TNM vs. the BCLC) was 3.4 (= 2,899.98–2,896.58), which corresponded to the ΔAIC value of 55.579 (the 8th TNM vs. the BCLC) in the present study (Table 4, right column), both indicating the BCLC classification was the better in the surgery group. But when this study considered all patients (surgery or nonsurgery) as a whole, the TNM system seemed to perform better. Consistent with the present cohort study, Su et al. investigated 307 northern Chinese patients with HCC (77.8% HBV infection, 56.1% surgery) and found that the BCLC had poorer prognosis prediction of overall survival than the 7th TNM staging system (c‐index: 0.69 vs. 0.70) [25].
There are weakness and strengths of the BCLC classification. The weakness is that the BCLC classification was not developed from a cohort of patients with HCC based on multivariate analysis [5]. Because survival outcomes are usually confounded by treatment strategies that may be quite different from one center to another, it is noteworthy that the external validation that uses the natural history of untreated HCC cohorts might be the most useful way to compare the prognostic value of each staging system [5, 26, 27]. The strength is that the BCLC is the only staging system that highlights treatment algorithms based on specific prognostic subclasses stratified according to the patient's tumor characteristics, baseline liver function, and overall health status [28]. These algorithms might explain why the results of the present study showed more predictability for the BCLC classification in patients who underwent surgery initially. If treatment was more aggressive than what was prescribed by the staging system, then the outcomes might be better (if more aggressive treatment is beneficial) or worse (if more aggressive treatment has deleterious effects) [29]. Therefore, treatment might introduce bias if treatment itself was different in essential between what was prescribed by the TNM system or the BCLC classification. The treatment algorithm suggested by the BCLC classification homogenized the surgery group and in turn increased performance of outcome predictability of the BCLC classification. Based on these considerations and results of the current study, the 8th TNM staging system for HCC has better performance on prediction of long‐term outcomes in patients with HCC except for the surgery group. Unlike other tumors, 70.4% of patients with HCC (as shown in the present study) occur on a background of a liver disease, making the management complex. As such, only a few patients (25.6% in the present study) have chances to undergo liver surgery as initial treatment for HCC. Choosing a staging system should consider its applicability and generalizability.
We acknowledge that the study had several limitations. First, the inherent caveat is that the 8th TNM staging system for HCC is not treatment‐oriented, whereas the BCLC classification is. Therefore, HCC in early stage of the BCLC classification probably had been surgically removed. Second, comparison between the 8th TNM staging system and the BCLC classification is somewhat difficult because stages are incongruent (Ia, Ib, II, IIIa, IIIb, IVa, IVb vs. 0, A1, A2, A3, A4, B, C, D). For easy comparison, only the main stages (such as I vs. 0/A) were investigated. Third, data regarding alcoholic drinking habits, body mass index, and molecular information such as alpha‐fetoprotein (AFP) were not collected by the cancer registry until 8 years ago; nearly 50% of the studied patients had no related data recorded. Thus, we could not control for preoperative AFP level and other potential confounders in survival analysis. Similarly, liver functional reserve, such as Child‐Pugh grade or albumin‐bilirubin grade, was not controlled in the multivariate analysis as a prognostic value because of data unavailability. Finally, based on a nationwide database, the initial data set of this study covered only around 70% of patients with newly diagnosed liver cancer in Taiwan annually. The end results and interpretations might not be fully generalizable to entire population. However, we think these results could represent this population, as the health care insurance system in Taiwan is a single‐payer system and less than 0.4% of people are uninsured.
Conclusion
The present study of nationwide data demonstrated effectiveness of the prognostic significance of staging defined by the 8th edition of TNM staging system and the BCLC classification. Both staging systems offer excellent prediction regarding overall survival, cancer‐specific survival, disease‐free survival, and local recurrence‐free rate. For patients treated initially with surgery, the BCLC classification predicts better than the TNM system, whereas for patients treated initially without surgery, the TNM system predicts better than the BCLC classification. As over 70% of patients with HCC could not undergo surgery initially, the overall performance of the 8th AJCC staging system for prediction of survival outcomes is better.
Author Contributions
Conception/design: Yun‐Jau Chang, Yao‐Jen Chang
Provision of study material or patients: Yun‐Jau Chang
Collection and/or assembly of data: Yun‐Jau Chang, Li‐Ju Chen
Data analysis and interpretation: Yun‐Jau Chang, Li‐Ju Chen
Manuscript writing: Yun‐Jau Chang, Li‐Ju Chen
Final approval of manuscript: Yun‐Jau Chang, Yao‐Jen Chang, Li‐Ju Chen
Disclosures
The authors indicated no financial relationships.
Supporting information
See http://www.TheOncologist.com for supplemental material available online.
Table S1 Definitions of the TNM system (8th edition) and the BCLC classification
Table S2 Descriptive statistics for overall survival, cancer‐specific survival, disease‐free survival and local recurrence regarding the TNM (8th edition), the BCLC classification in patients with hepatocellular carcinoma (N = 37,062)
Table S3 Multivariate Comparison of delta AIC (ΔAIC: BCLC vs. TNM) values from the TNM (8th edition) system and the BCLC classification for the best‐fit model of survival in patients with hepatocellular carcinoma (N = 37,062)
Acknowledgments
We gratefully thank to the Collaboration Center of Health Information Application, Ministry of Health and Welfare Bureau, Taiwan, for providing data for analysis.
Disclosures of potential conflicts of interest may be found at the end of this article.
No part of this article may be reproduced, stored, or transmitted in any form or for any means without the prior permission in writing from the copyright holder. For information on purchasing reprints contact Commercialreprints@wiley.com. For permission information contact permissions@wiley.com.
References
- 1. Bray F, Ferlay J, Soerjomataram I et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018;68:394–424. [DOI] [PubMed] [Google Scholar]
- 2. Chang YJ, Chung KP, Chang YJ et al. Long‐term survival of patients undergoing liver resection for very large hepatocellular carcinomas. Br J Surg 2016;103:1513–1520. [DOI] [PubMed] [Google Scholar]
- 3. Forner A, Reig M, Bruix J. Hepatocellular carcinoma. Lancet 2018;391:1301–1314. [DOI] [PubMed] [Google Scholar]
- 4. Pons F, Varela M, Llovet JM. Staging systems in hepatocellular carcinoma. HPB (Oxford) 2005;7:35–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Maida M, Orlando E, Camma C et al. Staging systems of hepatocellular carcinoma: A review of literature. World J Gastroenterol 2014;20:4141–4150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Burkhart RA, Pawlik TM. Staging and prognostic models for hepatocellular carcinoma and intrahepatic cholangiocarcinoma. Cancer Control 2017;24:1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Amin MB, Greene FL, Edge SB et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population‐based to a more “personalized” approach to cancer staging. CA Cancer J Clin 2017;67:93–99. [DOI] [PubMed] [Google Scholar]
- 8. Chun YS, Pawlik TM, Vauthey JN. 8th Edition of the AJCC Cancer Staging Manual: Pancreas and hepatobiliary cancers. Ann Surg Oncol 2018;25:845–847. [DOI] [PubMed] [Google Scholar]
- 9. Llovet JM, Bru C, Bruix J. Prognosis of hepatocellular carcinoma: The BCLC staging classification. Semin Liver Dis 1999;19:329–338. [DOI] [PubMed] [Google Scholar]
- 10. Marrero JA, Fontana RJ, Barrat A et al. Prognosis of hepatocellular carcinoma: Comparison of 7 staging systems in an American cohort. Hepatology 2005;41:707–716. [DOI] [PubMed] [Google Scholar]
- 11. Zhao JJ, Yan T, Zhao H et al. Evaluation of eight different clinical staging systems associated with overall survival of Chinese patients with hepatocellular carcinoma. Chin Med J (Engl) 2015;128:316–321. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Liu PH, Hsu CY, Hsia CY et al. Prognosis of hepatocellular carcinoma: Assessment of eleven staging systems. J Hepatol 2015;64:601–608. [DOI] [PubMed] [Google Scholar]
- 13. Wildi S, Pestalozzi BC, McCormack L et al. Critical evaluation of the different staging systems for hepatocellular carcinoma. Br J Surg 2004;91:400–408. [DOI] [PubMed] [Google Scholar]
- 14. Dhir M, Are C. Response: “We Still in an Update Process of the BCLC System.” Ann Surg 2018;267:e24–e25. [DOI] [PubMed] [Google Scholar]
- 15. Sohn JH, Duran R, Zhao Y et al. Validation of the Hong Kong Liver Cancer Staging System in determining prognosis of the North American patients following intra‐arterial therapy. Clin Gastroenterol Hepatol 2017;15:746–755.e744. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Chen LJ, Chung KP, Chang YJ. Ratio and log odds of positive lymph nodes in breast cancer patients with mastectomy. Surg Oncol 2015;24:239–247. [DOI] [PubMed] [Google Scholar]
- 17. Chung KP, Chen LJ, Chang YJ. Can composite performance measures predict survival of patients with colorectal cancer? World J Gastroenterol 2014;20:15805–15814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Wu CY, Chen YJ, Ho HJ et al. Association between nucleoside analogues and risk of hepatitis B virus‐related hepatocellular carcinoma recurrence following liver resection. JAMA 2012;308:1906–1914. [DOI] [PubMed] [Google Scholar]
- 19. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD‐9‐CM administrative databases. J Clin Epidemiol 1992;45:613–619. [DOI] [PubMed] [Google Scholar]
- 20. Kawaoka T, Aikata H, Takaki S et al. Transcatheter chemoembolization for unresectable hepatocellular carcinoma and comparison of five staging systems. Hepatol Res 2010;40:1082–1091. [DOI] [PubMed] [Google Scholar]
- 21. Kim IG, Hu XG, Wang HJ et al. The 7th/8th American Joint Committee on Cancer and the Modified Union for International Cancer Control Staging System for Hepatocellular Carcinoma. Yonsei Med J 2019;60:140–147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Abdel‐Rahman O. Assessment of the discriminating value of the 8th AJCC stage grouping for hepatocellular carcinoma. HPB (Oxford) 2018;20:41–48. [DOI] [PubMed] [Google Scholar]
- 23. Kamarajah SK, Frankel TL, Sonnenday C et al. Critical evaluation of the American Joint Commission on Cancer (AJCC) 8th edition staging system for patients with Hepatocellular Carcinoma (HCC): A Surveillance, Epidemiology, End Results (SEER) analysis. J Surg Oncol 2018;117:644–650. [DOI] [PubMed] [Google Scholar]
- 24. Liu C, Duan LG, Lu WS et al. Prognosis evaluation in patients with hepatocellular carcinoma after hepatectomy: Comparison of BCLC, TNM and Hangzhou criteria staging systems. PLoS One 2014;9:e103228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Su L, Zhou T, Zhang Z et al. Optimal staging system for predicting the prognosis of patients with hepatocellular carcinoma in China: A retrospective study. BMC Cancer 2016;16:424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Llovet JM, Bustamante J, Castells A et al. Natural history of untreated nonsurgical hepatocellular carcinoma: Rationale for the design and evaluation of therapeutic trials. Hepatology 1999;29:62–67. [DOI] [PubMed] [Google Scholar]
- 27. Ryder S. Predicting survival in early hepatocellular carcinoma. Gut 2005;54:328–329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Fong ZV, Tanabe KK. The clinical management of hepatocellular carcinoma in the United States, Europe, and Asia: A comprehensive and evidence‐based comparison and review. Cancer 2014;120:2824–2838. [DOI] [PubMed] [Google Scholar]
- 29. Sherman M. Staging for hepatocellular carcinoma: An embarrassment of riches. J Hepatol 2016;64:535–536. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
See http://www.TheOncologist.com for supplemental material available online.
Table S1 Definitions of the TNM system (8th edition) and the BCLC classification
Table S2 Descriptive statistics for overall survival, cancer‐specific survival, disease‐free survival and local recurrence regarding the TNM (8th edition), the BCLC classification in patients with hepatocellular carcinoma (N = 37,062)
Table S3 Multivariate Comparison of delta AIC (ΔAIC: BCLC vs. TNM) values from the TNM (8th edition) system and the BCLC classification for the best‐fit model of survival in patients with hepatocellular carcinoma (N = 37,062)