1. INTRODUCTION

Liver cirrhosis is a significant global health burden, ranking as the 11th leading cause of mortality and accounting for 2.2% of deaths worldwide.1 The age-standardized incidence rate of liver cirrhosis was 20.7 per 100,000 in 2015, representing a 13% increase since 2000, according to the Global Burden of Disease study.2,3 Viral hepatitis has been the primary cause of cirrhosis over the last three decades; however, due to advances in hepatitis B virus vaccination and hepatitis C virus treatment, the incidence and mortality rates have declined.1 Instead, non-alcoholic fatty liver disease (NAFLD) and alcoholic liver disease are on the rise.1–3

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer globally and ranks as the second leading cause of cancer-related deaths.4 The primary cause of HCC development remains viral hepatitis, including hepatitis B and hepatitis C.5,6 In recent years, NAFLD has emerged as a significant risk factor for the development of HCC.4,7

It is well-established that liver cirrhosis poses a significant risk for developing HCC, regardless of the underlying etiology.8 Liver functional reserve is a crucial determinant of the prognosis for patients with HCC. The Child-Turcotte-Pugh (CTP) score is a widely used tool for evaluating the severity of liver cirrhosis. According to the Barcelona Clinic Liver Cancer criteria,9 HCC patients with preserved liver function (CTP class A) are considered suitable for tumor-related treatment.9 However, in real-world clinical practice, many patients receive some form of treatment despite moderately severe liver disease (CTP class B), leading to ambiguity in the definition of preserved liver function and questioning whether current treatment strategies should be challenged.10

There are several models for assessing liver function for chronic liver disease. CTP score contains five variables, including total bilirubin, prothrombin time international normalized ratio, albumin, ascites, and hepatic encephalopathy.11 CTP score was categorized into three classes. Preserved liver function was defined as Child-Pugh class A, while CTP class C represents decompensated liver disease.11,12 However, there are several limitations in the CTP score system, which ascites and hepatic encephalopathy were subjective variables, and all the parameters had equal weights.13 Therefore, several novel models without subjective parameters, such as albumin-bilirubin (ALBI) grade14 and modified ALBI grade,15 were utilized to predict the prognosis of patients with HCC. Albumin and bilirubin were independently correlated with overall survival (OS) and have been validated by various studies to prove their values to provide an accurate prognosis.13–17 Furthermore, EZ(Easy)-ALBI grade was proposed by Kariyama et al. with a simplified formula and has demonstrated similar prognostic prediction ability compared to ALBI grade in several studies.18–20 Model for end-stage liver disease (MELD) score has been utilized to predict short-term survival in patients with end-stage liver disease.21 Several versions have been developed since its original version, including MELD-Na and MELD 3.0 and MELD 3.0 seemed to provide more accurate mortality prediction than traditional MELD score.22,23 However, there was no consensus regarding which staging system provides the most precise prognostic accuracy.

CTP class B includes subgroups CTP-B7, -B8, and -B9, but they are regarded as one subgroup based on established criteria. However, detailed patient demographics, tumor characteristics, treatment modalities, and long-term prognosis of HCC patients in each CTP-B7, -B8, and -B9 subgroup are not fully elucidated. Therefore, in this study, we aimed to investigate HCC patients with CTP class B and clarify which staging models provided the most accurate prognosis in these patients.

2. METHODS 2.1. Patients and follow-up

This retrospective study enrolled a total of 6141 patients who were newly diagnosed with HCC between October 2007 and May 2022 at Taipei General Veterans Hospital (Fig. 1). The diagnosis of HCC was certified based on the AASLD criteria.24 The CTP score was examined at the time of HCC diagnosis. The treatment plan was established based on recommendations from the BCLC staging system or specialists from a weekly multidisciplinary meeting. The treatment plan was well communicated to the patient and their family to ensure they received the most appropriate benefits.

The HCC registration organization at Taipei Veterans General Hospital is a prospective database that records patients’ demographic characteristics, laboratory data, tumor factors, treatment plans, and outcomes.25 Orthotopic liver transplantation, surgical resection, radiofrequency ablation, transarterial chemoembolization, and systemic therapy, including chemotherapy, immune checkpoint inhibitors, and tyrosine kinase inhibitors, were classified as tumor-related treatments, while supportive care was defined as not receiving any tumor-related treatment.

The study was conducted following the Declaration of Helsinki and current ethical guidelines and has been approved by the Institutional Review Board of Taipei Veterans General Hospital (VGHIRB No. 2023-05-007AC), Taiwan. As this was a retrospective cohort study, the Institutional Review Board waived the requirement for informed consent. Before the initiation of this study, patient information was de-identified to protect their privacy.

Fig. 1:

(A) Study flowchart.

2.2. Prognostic prediction models

The formula for the ALBI score was as follows: (log10 bilirubin [µmol/L] × 0.66) − (albumin [g/L] × 0.085).14 The cutoff values for ALBI grades 1/2/3 were ≤−2.60, −2.6 to −1.39, and >1.39, respectively. The modified ALBI grade further divided ALBI grade 2 into 2a and 2b, with 2a defined as a score of −2.60 to −2.270 and 2b as a score of −2.270 to −1.39.15 The EZ-ALBI score was calculated as total bilirubin (mg/dL) − (9 × albumin [g/dL]), with EZ-ALBI grades 1, 2, and 3 representing ≤−34.4, −34.4 to −22.2, and >−22.2, respectively.18 The MELD score was calculated using the formula 9.57 × loge (creatinine) + 3.78 × loge (total bilirubin) + 11.2 × loge (INR) + 6.43,21 while the MELD 3.0 score was calculated as 1.33 (if female) + (4.56 × loge [bilirubin]) + (0.82 × [137 – Na]) – (0.24 × [137 − Na] × loge [bilirubin]) + (9.09 × loge [INR]) + (11.14 × loge [creatinine]) + (1.85 × [3.5 − albumin]) – (1.83 × [3.5 − albumin] × loge [creatinine]) + 6.23.23 Both the MELD score and MELD 3.0 score were stratified into low risk, intermediate risk, and high risk with scores of ≤ 20, 20 to 30, and > 30, respectively.

2.3. Statistical analysis

The primary endpoint of this study was OS, which was calculated from the date of the patient’s HCC diagnosis to death, the last outpatient clinic visit, discharge, or loss of follow-up. Survival was censored on August 22, 2022. The Kaplan-Meier method was used to measure survival, and a Cox proportional hazards model was used to compare survival between groups. Categorical variables were compared using Pearson’s chi-squared analysis or Fisher’s exact test, while continuous variables were examined using the Mann-Whitney U test.

The survival analysis of each prediction model was measured using the Kaplan-Meier method with log-rank tests. Corrected Akaike information criteria (AICc) were used to illustrate how the prediction models correlated with patients’ survival. AICc was chosen over the Akaike information criterion (AIC) to compensate for different parameters in each prediction model. Homogeneity was measured using the Wald test to assess the disparities in survival among patients in the same stage within each model.

Statistical analyses were performed using IBM SPSS Statistics for Windows, version 25.0 (IBM Corp., Armonk, NY, USA). Variables that showed statistical significance in the univariate analysis were included in the multivariate analysis using a forward stepwise Cox regression model. A two-tailed p-value of less than 0.05 was considered statistically significant.

3. RESULTS 3.1. Patients’ demographic and characteristics

A total of 6141 patients were enrolled in the study, of whom 1143 (18.6%) were classified as CTP-B based on blood chemistry and image studies conducted before treatment for HCC. This group had 576 patients with CTP-B7, 363 with CTP-B8, and 204 with CTP-B9. Compared to patients in the CTP-B7 and CTP-B8 subgroups, those in the CTP-B9 subgroup were more likely to be in the terminal stage of the BCLC system (BCLC D), with poor performance status (PS) and less likely to receive tumor-related treatments. With regard to tumor features, CTP-B9 patients had more multiple tumors and more vascular invasion. Further demographic details are presented in Table 1.

Table 1 - Demographic of study cohort Characteristics All patients CPS:7 CPS:8 CPS:9 p N = 1143 N = 576 N = 363 N = 204 Age 66 (56-77) 66 (56-78) 67 (57-77) 64 (55-74) 0.868 Sex (M/F) 849/294 (74.3%/25.7%) 430/146 (74.7%/25.3%) 263/100 (72.5%/27.5%) 156/48 (76.5%/23.5%) 0.052 HBsAg (+/−) 577/557 (50.9%/49.1%) 289/285 (50.3%/49.7%) 176/183 (49.0%/51.0%) 112/89 (55.7%/44.3%) 0.294 HCV (+/−) 300/833 (26.5%/73.5%) 163/410 (28.4%/71.6%) 93/266 (25.9%/74.1%) 44/157 (21.9%/78.1%) 0.178 ALT (U/L) 48 (30-86) 45 (28-80) 49 (30-87) 62 (33-102) 0.110 Creatinine (mg/dL) 0.93 (0.73-1.24) 0.91 (0.73-1.20) 0.95 (0.74-1.34) 0.93 (0.74-1.23) 0.263 Bilirubin (mg/dL) 1.79 (1.01-2.79) 1.31 (0.86-2.05) 2.1 (1.13-3.09) 3.2 (2.3-5.9) <0.001 ALK-p (U/L) 150 (103-252) 134 (96-228) 141 (106-254) 187 (134-299) <0.001 Albumin (g/dL) 3.0 (2.6-3.3) 3.1 (2.8-3.4) 2.9 (2.6-3.2) 2.7 (2.5-3.1) <0.001 PT-INR 1.20 (1.11-1.31) 1.17 (1.09-1.26) 1.23 (1.14-1.33) 1.28 (1.18-1.43) <0.001 Platelets (×1000/mm3) 140 (86-232) 144 (87-242) 138 (76-230) 134 (93-217) 0.574 Modified ALBI grade (1/2A/2B/3) (%) 5/18/423/697 (0.4%/1.6%/37%/61.0%) 5/17/332/222 (0.9%/3.0%/57.6%/38.5%) 0/1/79/283 (0%/0.3%/21.8%/78.0%) 0/0/12/192 (0%/0%/5.9%/94.1%) <0.001 MELD score 12.15 (9.90-15.01) 10.76 (8.92-12.68) 12.94 (10.78-15.98) 15.18 (13.11-18.35) <0.001 MELD 3.0 14.27 (11.38-18.28) 12.22 (10.01-15.02) 15.38 (12.88-19.31) 18.42 (15.05-21.29) <0.001 AFP (ng/mL) 166 (10.9-6164.2) 120.7 (9.5-4724.5) 107.5 (11.0-5658.5) 407.2 (17.9-1211.5) 0.424 BCLC (0/A/B/C/D) (%) 6/256/146/313/422 (0.5%/22.4%/12.8%/27.4%/36.9%) 3/158/83/190/14 (0.5%/27.4%/14.4%/33.0%/24.7%) 3/77/47/86/150 (0.8%/21.2%/12.9%/23.7%/41.3%) 0/21/16/37/130 (0%/10.3%/7.8%/18.1%/63.7%) <0.001 Performance status (0/≧1) (%) 634/509 (55.5%/44.5%) 355/221 (61.6%/38.4%) 198/165 (54.5%/45.5%) 81/123 (39.7%/60.3%) <0.001 Tumor size (cm) 6.8 (3.1-10.5) 6.5 (3.1-10.9) 6.3 (3.0-10.3) 8.5 (3.5-10.2) 0.461 Number of tumors (1/>1) (%) 505/637 (44.2%/55.8%) 267/309 (46.4%/53.6%) 167/195 (46.1%/53.9%) 71/133 (34.8%/65.2%) 0.011 Vascular invasion (+/−) (%) 492/651 (43.0%/57.0%) 229/347 (39.8%/60.2%) 144/219 (39.7/60.3%) 119/85 (58.3%/41.7%) <0.001 Metastasis (+/−) (%) 234/907 (20.5%/79.5%) 112/462 (19.5%/80.5%) 71/292 (19.6%/80.4%) 51/153 (25.0%/75.0%) 0.228 Ascites (+/−) (%) 689/454 (60.3%/39.7%) 301/275 (52.3%/47.7%) 231/132 (63.6%/36.4%) 157/47 (77.0%/23.0%) <0.001 Hepatic encephalopathy (+/−) (%) 36/1107 (3.1%/96.9%) 6/570 (1.0%/99.0%) 15/348 (4.1%/95.9%) 15/189 (7.4%/92.6%) <0.001 Treatment modality  SR + OLT 85 (7.6%) 52 (9.4%) 21 (5.9%) 12 (6.0%) 0.116  RFA 134 (12.1%) 83 (14.9%) 38 (10.7%) 13 (6.5%) 0.004  TACE 241 (21.7%) 143 (25.7%) 82 (23.2%) 16 (8.0%) <0.001  Systemic therapy 207 (18.6%) 126 (22.7%) 54 (15.3%) 27 (13.4%) 0.003  Supportive care 413 (37.2%) 139 (25.0%) 148 (41.8%) 126 (62.7%) <0.001  Tumor-related treatment 698 (62.8%) 417 (59.7%) 206 (29.5%) 75 (10.7%) <0.001

Continuous variables are expressed as the median with the 25th and 75th percentiles.

AFP = alpha-fetoprotein; ALBI = albumin-bilirubin; ALK-P = alkaline phosphatase; ALT = alanine aminotransferase; BCLC = the Barcelona Clinic Liver Cancer; HBsAg = hepatitis B surface antigen; HBV = hepatitis B virus; HCV = hepatitis C virus; MELD = Model For End-Stage Liver Disease; OLT = orthotopic liver transplantation; PT-INR = prothrombin time international normalized ratio; RFA = radiofrequency ablation; SR = surgical resection; TACE = transarterial chemoembolization.

3.2. Long-term OS and prognostic factors between CTP-B7, CTP-B8 and CTP-B9 subgroups

After a median follow-up of 4.6 months (interquartile range IQR 1.8–17.2 months), 963 patients had died, and the 5-year OS rate was 11.4%. The 5-year OS rates for the CTP-B7, CTP-B8, and CTP-B9 groups were 11.6%, 13.6%, and 8.3%, respectively. Patients in the CTP-B7 and CTP-B8 groups had comparable OS (p = 0.089, Fig. 2A), both of which were better than that of the CTP-B9 group (p < 0.001, Fig. 2B, C). Multivariate analysis showed that serum creatinine >1.0 mg/dL, platelet count >150,000/mm3, alpha-fetoprotein (AFP) levels >400 ng/mL, tumor size ≥5 cm, multiple tumors, vascular invasion, extra-hepatic metastasis, presence of ascites, poor PS and receipt of supportive treatment were independent predictors of poor OS. The detailed prognostic factors are presented in Table 2. We also conducted multivariate analysis for each individual CTP score group, and detailed information is provided in Supplementary Table S1, https://links.lww.com/JCMA/A200.

Table 2 - Analysis of factors associated with poor prognosis in patients with Child-Pugh Score B Parameter Univariate Multivariate HR 95% CI p HR 95% CI p       Age >65 vs ≦ 65 1.150 1.011-1.308 0.033       Sex Male vs Female 1.155 0.999-1.335 0.052     HBsAg (+) No vs Yes 1.080 0.951-1.226 0.237    Anti-HCV (+) No vs Yes 0.828 0.718-0.956 0.009   Albumin (g/dL) >3.5 vs ≦ 3.5 1.069 0.862-1.327 0.543  Creatinine (mg/dL) >1.0 vs ≦1.0 1.377 1.211-1.567 <0.001 1.261 1.085-1.467 0.003     ALT (U/L) >40 vs ≦ 40 1.178 1.035-1.340 0.013  Bilirubin (mg/dL) >3.0 vs ≦ 3.0 1.016 0.871-1.184 0.844    ALK-P (U/L) >100 vs≦ 100 1.616 1.372-1.905 <0.001  PLT (x1000/mm3) ≦150 vs > 150 0.488 0.429-0.555 <0.001 0.784 0.656-0.937 0.007      PT-INR >1.2 vs ≦1.2 0.957 0.844-1.087 0.501     AIBI grade 1(reference)    ALBI grade 2A 0.723 0.238-2.197 0.567    ALBI grade 2B 1.034 0.386-2.722 0.947    ALBI grade 3 1.125 0.421-3.010 0.814    AFP (ng/mL) ≦20 (reference) 20-400 1.326 1.112-1.583 0.002 >400 3.213 2.737-3.773 <0.001 1.735 1.425-2.113 <0.001    MELD score >15 vs ≦15 1.448 1.251-1.676 <0.001     MELD 3.0 >20 vs ≦20 1.697 1.435-2.007 <0.001   Tumor size (cm) ≧5 vs <5 3.540 3.059-4.097 <0.001 1.894 1.552-2.310 <0.001  Numbers of Tumor >1 vs ≦1 1.685 1.479-1.919 <0.001 1.180 1.014-1.370 0.032  Vascular invasion Yes vs No 3.460 3.008-3.979 <0.001 1.576 1.325-1.874 <0.001     Metastasis Yes vs No 2.626 2.246-3.070 <0.001 1.325 1.105-1.590 0.002  Performance status ≧1 vs 0 3.120 2.724-3.573 <0.001 1.196 1.003-1.426 0.047  Treatment modality Supportive vs Tumor-related 4.046 3.514-4.660 <0.001 2.734 2.313-3.232 <0.001      Ascites No vs Yes 1.572 1.378-1.793 <0.001 1.249 1.069-1.459 0.005      Hepatic encephalopathy No vs Yes 1.190 0.836-1.693 0.346

AFP = alpha-fetoprotein; ALBI = Albumin-Bilirubin; ALK-P = alkaline phosphate; ALT = alanine aminotransferase; CI = confidence interval; HBsAg = hepatitis B surface marker; HCV = hepatitis C virus; HR = hazard ratio; MELD = Model for End-Stage Liver Disease; PLT = platelet; PT-INR = prothrombin time international normalized ratio.

Fig. 2:

Comparison of OS between (A) CTP-B7 and CTP-B8, (B) CTP-B8 and CTP-B9, (C) CTP-B7 and CTP-B9.

3.3. Comparison of the prognostic performance among MELD score, MELD 3.0 score, ALBI grade, modified ALBI grade, EZ-ALBI grade, and Child-Turcotte Pugh score

We evaluated six prediction models separately using Kaplan-Meier OS analysis. Among them, CTP score, EZ-ALBI grade, MELD score, and MELD 3.0 score showed clinical significance in survival across all subgroups, whereas both ALBI grade and mALBI grade were not associated with significant differences in survival (p = 0.311 and p = 0.263, respectively). The OS of each prediction model is shown in Fig. 4A–F.

Fig. 3:

Comparison of the OS between tumor-related treatment subgroup and the supportive care subgroup before (A) CTP-B7, (B) CTP-B8, (C) CTP-B9; and after propensity score matching (D) CTP-B7, (E) CTP-B8, (F) CTP-B9.

Fig. 4:

Comparison of the OS between each prognostic models (A) CTP score, (B) ALBI grade, (C) mALBI grade, (D) EZ-ALBI grade, (E) MELD score, and (F) MELD 3.0 score.

We presented the prognostic performance of the six prediction models in Table 3. Overall, the MELD 3.0 score had the highest homogeneity and lowest AICc value among all patients. After stratifying by each CTP score in class B, the MELD 3.0 score consistently showed the lowest AICc levels and highest homogeneity in each subgroup (Supplementary Table S5, https://links.lww.com/JCMA/A200).

Table 3 - Comparison of prognostic performance among six prediction models Model Homogeneity (Wald test) Corrected Akaike information criteria (AICc) All patients (N = 1143) CTP score 31.175 11,990.632 ALBI grade 3.184 12,018.623 mALBI grade 2.320 12,019.487 EZ-ALBI grade 8.418 12,013.389 MELD score 19.676 11,974.009 MELD 3.0 score 51.319 11,837.725

AIBI = Albumin-Bilirubin; CTP = Child-Turcotte-Pugh; MELD = Model for End-stage Liver Disease;

Long-term OS of individual CTP class B between tumor-related treatment and supportive care subgroups

Overall, we compared the OS rates between the tumor-related treatment subgroup and the supportive care subgroup. Specifically, in the CTP-B7, CTP-B8, and CTP-B9 subgroups, the 1-year OS rates were 50.7% vs 5.4%, 50.3% vs 7.5%, and 41.9% vs 6.2%, respectively (p < 0.001, Fig. 3A–C). Patients who received any form of tumor-related treatment had better OS rates than those who received supportive care in all three CTP score subgroups (i.e., CTP-B7, CTP-B8, and CTP-B9).

We conducted propensity score matching to reduce selection bias between patients in the tumor-related treatment and supportive care subgroups. As a result, 190, 170, and 108 patients with CTP scores of -B7, -B8, and -B9, respectively, were equally distributed into the treatment and supportive care subgroups (Supplementary Table S2–4, https://links.lww.com/JCMA/A200). After propensity score matching, no clinical significance was observed between the tumor-related treatment and supportive care subgroups for each parameter.

After conducting propensity score matching, we analyzed the OS rates between the tumor-related treatment subgroup and the supportive care subgroup for each individual CTP score group. Specifically, in the CTP-B7, CTP-B8, and CTP-B9 subgroups, the 1-year OS rates were 28.3% vs 6.8%, 42.3% vs 9.8%, and 37.9% vs 10.0%, respectively (