Comparison of Four Albumin-Based Liver Reserve Models (ALBI/EZ-ALBI/PALBI/PAL) against MELD for Patients with Hepatocellular Carcinoma Undergoing Transarterial Chemoembolization

Simple Summary Liver functional reserve plays a critical role in the management of hepatocellular carcinoma (HCC) patients. The model for end-stage liver disease (MELD) is used to assess liver functional reserve in HCC patients. Alternatively, the albumin-bilirubin (ALBI) score and easy (EZ)-ALBI score, platelet-albumin-bilirubin (PALBI) score and platelet-albumin (PAL) score are also proposed to indicate liver reserve in HCC. We aimed to compare these four albumin-based liver reserve models (ALBI/EZ-ALBI/PALBI/PAL) against MELD for HCC patients undergoing transarterial chemoembolization (TACE). Among these models, the PALBI grade had the highest homogeneity and lowest corrected Akaike information criteria value, followed by EZ-ALBI, PAL, ALBI and lastly, MELD. All four albumin-based liver reserve models are better prognostic tools than MELD scores in patients undergoing TACE. Of these, the PALBI score is the best model to evaluate the liver reserve and should be considered a surrogate marker in these patients. Abstract (1) Background: The severity of liver functional reserve plays an important role in the management of hepatocellular carcinoma (HCC). Noninvasive models such as the model for end-stage liver disease (MELD), albumin-bilirubin (ALBI) grade and easy (EZ)-ALBI grade, platelet-albumin-bilirubin (PALBI) and platelet-albumin (PAL) are used to evaluate liver dysfunction. We aimed to compare the prognostic performance of these four albumin-based models against MELD in HCC patients undergoing transarterial chemoembolization (TACE). (2) Methods: A total of 1038 treatment naïve HCC patients who received TACE as the primary treatment were enrolled. A multivariate Cox model was used to determine independent survival predictors. (3) Results: Multivariate analysis revealed that higher serum creatinine and α-fetoprotein level, vascular invasion, large tumor size, ALBI grades 2–3, EZ-ALBI grades 2–3, PALBI grades 2–3, PAL grades 2–3, but not the MELD score, were independent predictors associated with decreased survival in different Cox models. Among these models, the PALBI grade had the highest homogeneity and lowest corrected Akaike information criteria value, followed by EZ-ALBI, PAL, ALBI and, lastly, MELD. (4) Conclusions: All four albumin-based liver reserve models are better prognostic tools than MELD score in HCC patients undergoing TACE. Of these, the PALBI score is the best model to evaluate the liver reserve and should be considered a surrogate marker in these patients.


Introduction
Hepatocellular carcinoma (HCC) is the fourth leading cause of cancer-related death globally [1]. Its main risk factors are chronic hepatitis B and C virus (HBV, HCV) infections, alcoholism and non-alcoholic fatty liver disease (NAFLD) [2,3]. The prognosis of HCC is usually poor because many patients are diagnosed at an advanced stage. According to current HCC practice guidelines [4,5], surgical resection, local ablation and transplantation are recommended for very early and early-stage HCC. Patients at intermediate or advanced stages are suggested to receive transarterial chemoembolization (TACE) or systemic therapy, including targeted-and immuno-therapy [6,7].
TACE is a safe, effective and usually the first-line treatment to prolong survival in patients with multinodular HCC unsuitable for curative treatments [8][9][10][11]. For these patients, the severity of liver dysfunction is tightly linked with their long-term outcome. The Child-Turcotte-Pugh (CTP) score is a commonly used tool to evaluate liver injury [12]. However, the CTP score has drawbacks due to inter-related objective variables and subjective variables such as ascites and hepatic encephalopathy. The model for end-stage liver disease (MELD) was shown to be a better alternative method initially used in selecting patients for timely liver transplantation [12,13]; however, it also harbors the disadvantage of very complex calculation.
The albumin-bilirubin (ALBI) score, based solely on the serum albumin and bilirubin levels, is an objective marker of the liver functional reserve [14][15][16][17]. The easy (EZ)-ALBI score is an updated version of the ALBI with adequate predictive accuracy [18][19][20]. A potential shortcoming of the ALBI and EZ-ALBI grades is that they do not include any parameters to indicate portal hypertension in liver cirrhosis. Therefore, the platelet-albumin-bilirubin (PALBI) score, which includes a platelet count, a surrogate marker in portal hypertension, was proposed to evaluate liver reserve in HCC [21][22][23][24][25][26]. More recently, a simplified version, the platelet-albumin (PAL) score, was developed to assess short-and long-term survival in HCC patients receiving surgical resection [27,28]. In this study, we specifically compared the prognostic performance of these four albumin-based liver reserve models (ALBI, EZ-ALBI, PALBI and PAL) against MELD in HCC patients undergoing TACE.

Patients
Over a 16-year period between 2002 and 2018, a total of 1038 treatment-naïve HCC patients receiving TACE as their primary treatment at Taipei Veterans General Hospital were enrolled. Their clinical information, including baseline demographics, the extent of tumor involvement (number and size of the tumor, vascular invasion and metastasis), laboratory data, the severity of liver functional reserve (CTP class, ALBI grade, EZ-ALBI grade, PALBI grade, PAL grade and MELD score) and cancer staging, were recorded at the time of diagnosis. This study was approved by the Institutional Review Board of the Taipei Veterans General Hospital and complies with current ethical guidelines.

Diagnosis and Definition
The diagnosis of HCC was based on typical radiological characteristics (early arterial enhancement in the arterial phase and delayed wash-out in the venous phase) by contrastenhanced computed tomography (CT) scan or magnetic resonance imaging (MRI) [4,5]. The performance status (PS) was assessed by the Eastern Cooperative Oncology Group (ECOG) criteria [29]. The calculation and grading of the liver reserve models (MELD, ALBI, EZ-ALBI, PALBI and PAL) are summarized in Table 1 [14,18,25,27,30].

Treatments
The inclusion criteria of TACE were: (1) unresectable HCC or patients not eligible or unwilling to receive other therapies; (2) CTP class A or B; (3) no main portal vein thrombosis; (4) serum creatinine level < 1.5 mg/dL and (5) no gross ascites by sonography or CT scan [31]. Treatment selection was discussed at the multidisciplinary HCC Board of the Taipei Veterans General Hospital. The information, such as benefits and risks, was provided to the individual patient based on shared decision making. Written informed consent was obtained prior to initiation of the treatment. The details of the TACE procedure were described in our previous studies [32,33]. After TACE, patients were followed up every 2-3 months for serum α-fetoprotein (AFP) level, ultrasonography, CT or MRI to confirm therapeutic response and further treatment planning.

Statistics
Statistical analyses were performed using IBM SPSS Statistics for Windows, version 25.0 (IBM Corp., Armonk, NY, USA). The Mann-Whitney rank-sum test was used to compare continuous variables; the comparison of categorical data was evaluated by the Pearson chi-squared test or the two-tailed Fisher's exact test. Overall survival (OS) was estimated using the Kaplan-Meier method and compared with the log-rank test. Independent prognostic predictors associated with survival were assessed by the multivariate Cox proportional hazards model to generate a hazard ratio (HR) and 95% confidence interval (CI). For all tests, a p < 0.05 was considered statistically significant. The prognostic performance among different liver reserve models was compared using the corrected Akaike information criteria (AICc) and homogeneity. The lower the AICc, the more explanatory and informative the model is [34,35].

Patient Characteristics
The baseline characteristics of patients are summarized in Table 1. Their mean age was 67 years and the majority were male. HBV and HCV were the most common etiologies of HCC. About 53% of patients had a single tumor nodule, and the mean tumor diameter was 6.3 cm. Approximately 79% of patients were CTP classification A and 18% of patients had ascites at the time of diagnosis. The grading and mean values of ALBI, EZ-ALBI, PALBI, PAL and MELD scores are given in Table 2. The mean scores of these models according to CTP class A vs. B are shown in Figure 1. All the scores in CTP B display a higher (or less negative) value in comparison with the scores in CTP class A (all p < 0.001). Table 2. Formula and grading of noninvasive liver reserve models. Taipei Veterans General Hospital. The information, such as benefits and risks, was provided to the individual patient based on shared decision making. Written informed consent was obtained prior to initiation of the treatment. The details of the TACE procedure were described in our previous studies [32,33]. After TACE, patients were followed up every 2-3 months for serum α-fetoprotein (AFP) level, ultrasonography, CT or MRI to confirm therapeutic response and further treatment planning.

Statistics
Statistical analyses were performed using IBM SPSS Statistics for Windows, version 25.0 (IBM Corp., Armonk, NY, USA). The Mann-Whitney rank-sum test was used to compare continuous variables; the comparison of categorical data was evaluated by the Pearson chi-squared test or the two-tailed Fisher's exact test. Overall survival (OS) was estimated using the Kaplan-Meier method and compared with the log-rank test. Independent prognostic predictors associated with survival were assessed by the multivariate Cox proportional hazards model to generate a hazard ratio (HR) and 95% confidence interval (CI). For all tests, a p < 0.05 was considered statistically significant. The prognostic performance among different liver reserve models was compared using the corrected Akaike information criteria (AICc) and homogeneity. The lower the AICc, the more explanatory and informative the model is [34,35].

Patient Characteristics
The baseline characteristics of patients are summarized in Table 1. Their mean age was 67 years and the majority were male. HBV and HCV were the most common etiologies of HCC. About 53% of patients had a single tumor nodule, and the mean tumor diameter was 6.3 cm. Approximately 79% of patients were CTP classification A and 18% of patients had ascites at the time of diagnosis. The grading and mean values of ALBI, EZ-ALBI, PALBI, PAL and MELD scores are given in Table 2. The mean scores of these models according to CTP class A vs. B are shown in Figure 1. All the scores in CTP B display a higher (or less negative) value in comparison with the scores in CTP class A (all p < 0.001).

Prognostic Performance of the Five Non-Invasive Liver Reserve Models
The comparison of the prognostic performance of the five non-invasive liver reserve models is shown in Table 8. Of these, the PALBI grade had the highest homogeneity and lowest AICc value, followed by the EZ-ALBI, PAL, ALBI grades and lastly, the MELD score. Thus, all four albumin-based models had better prognostic performance than the MELD.

Discussion
The prognosis of HCC depends not only on the extent of the tumor but also on the severity of the liver reserve. So far, there are four albumin-based liver reserve models specifically for liver cirrhosis and HCC. Our data demonstrate that the PALBI grade is not only an independent survival predictor in HCC patients undergoing TACE but also shows the best prognostic ability in terms of discriminating long-term survival in comparison with the other three albumin-based models and MELD scores. In addition, all four albumin-based models (PALBI, ALBI, EZ-ALBI and PAL grade) were better prognostic tools compared to the MELD for outcome prediction, suggesting the indispensable role of albumin in this clinical setting.
HCC typically develops in the background of chronic liver injury or cirrhosis. The rationale for adopting the four albumin-based models and MELD in HCC can also be demonstrated by the fact that their scores are higher (or less negative) in CTP class B than those in CTP class A (Figure 1).
The ALBI grade, based solely on serum albumin and bilirubin levels, is an objective tool to indicate liver functional reserve in HCC [14]. Consistent with previous studies [36][37][38][39], this study shows that the ALBI grade can discriminate long-term survival in HCC patients undergoing TACE. However, the calculation of the ALBI score is quite complex. Its updated version, the EZ-ALBI grade, is a more feasible method to estimate liver reserve and has been validated in the previous study [18]. The EZ-ALBI grade was shown to highly correlate with the ALBI grade, and it was consistently identified as an independent prognostic predictor in patients undergoing TACE in this study [18,19,40].
Although the ALBI and EZ-ALBI grades are useful methods to evaluate liver injury in HCC, the factor of portal hypertension is not considered in the formula. Notably, thrombocytopenia occurs as a result of hypersplenism in cirrhosis and is a surrogate marker for portal hypertension. Therefore, platelet count has been identified as an independent survival predictor in HCC [21][22][23]27,28]. The PALBI grade, based on serum albumin, bilirubin and platelet count, can well discriminate survival in patients undergoing TACE in our study. This result is consistent with previous studies [24][25][26]. Alternatively, the PAL score, based on platelet count and albumin level, can also well discriminate long-term outcomes in patients undergoing TACE, indicating this newly proposed marker is also a robust prognostic marker in HCC.
However, this study has some limitations. Firstly, this is a single-center study from a predominantly HBV-infected patient cohort. This feature is quite different from Western countries, where HCV infection and alcoholism are more prevalent. Secondly, this study was performed in patients undergoing TACE and the prognostic role of these models in other patient cohorts needs further study to be validated. Thirdly, treatment decisions were primarily based on the multidisciplinary HCC Board in our hospital and thus, they may not strictly adhere to the current Barcelona Clinic Liver Cancer (BCLC) recommendations.

Conclusions
Our study shows that the PALBI score is the best prognostic model in the setting of HCC patients undergoing TACE. All four albumin-based liver reserve models (PALBI, EZ-ALBI, PAL and ALBI) are better prognostic tools as compared with the MELD score. These data indicate the crucial role of albumin in the outcome prediction for HCC. External studies are required to confirm our findings. Funding: This study was funded by grants from the Taipei Veterans General Hospital (V112C-206, VN112-13), Taipei, Taiwan.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of Taipei Veterans General Hospital (protocol code: 2020-02-004CC, date of approval: 28 February 2020).
Informed Consent Statement: Patient consent was waived by the IRB due to the retrospective nature of this study.

Data Availability Statement:
The data presented in this study are available in this article.

Acknowledgments:
The authors thank Chia-Lin Wang from the Department of Medical Research, Taipei Veterans General Hospital, for her help in the preparation of this manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.