The Albumin-Bilirubin Score as a Predictor of Liver-Related Mortality in Primary Biliary Cholangitis with Compensated Cirrhosis

Background: Although several prognostic scores have been reported to correlate with the prognosis of primary biliary cholangitis (PBC) patients, there are limited tools to predict the prognosis of PBC with compensated cirrhosis. This study aimed to evaluate the prognostic performance of the albumin-bilirubin (ALBI) score in PBC patients with compensated cirrhosis. Methods: We conducted a retrospective longitudinal study of 219 patients with compensated PBC cirrhosis to evaluate the prognostic performance of the ALBI using Cox regression model, receiver operating characteristic (ROC) curve, and Kaplan-Meier method. Results: During follow-up, a total of 19 subjects (8.7%) met the primary endpoint of liver-related death or liver transplantation (LT). Patients who died/underwent LT have higher ALBI score (−1.06 vs. −2.06, p < 0.001) at baseline than those who survived. ALBI score (hazard ratio: 15.011, 95% confidence interval [CI]: 5.045–44.665, p < 0.001) was associated with an increase in liver-related mortality or LT. ALBI score had the best discriminative capacity to predict the 5-year liver-related mortality (area under the ROC curve: 0.871, 95% CI [0.820, 0.913]) compared with other prognostic scores. The ROC curve showed that the best cut-off value of ALBI score was −1.47, with 90.0% sensitivity and 76.6% specificity. Also, the probability of transplant-free survival decreased with increasing ALBI grade (log-rank p = 0.003). The 5-year transplant-free survival rates of patients in grade 1, grade 2, and grade 3 were 100.0%, 96.4%, and 89.4%, respectively. Conclusion: ALBI score is a simple and effective predictive factor estimating the clinical outcome of patients with compensated PBC cirrhosis and provides better prognostic performance compared with other prognostic scores.


Introduction
Primary biliary cholangitis (PBC) is an autoimmune, cholestatic, and chronic liver disease, which is characterized by the destruction of small intrahepatic bile ducts [1,2].A considerable proportion of patients suffer from cirrhosis-related complications or liver malignancies, though the overall 5-year survival rate of PBC patients is about 80-90% [3,4].Many studies have attempted to develop a classification system that can not only characterize the degree of liver injury but also predict the prognosis of patients according to clinical and laboratory variables.Thus, to distinguish patients at high risk of developing complications and death, several prognostic scores have been applied in patients with PBC, such as albumin-bilirubin (ALBI) score, Mayo risk score, modified model for end-stage liver disease (MELD) score for cholestatic disease, aspartate aminotransferase-to-platelet ratio index (APRI), and fibrosis-4 score (FIB-4) index [5][6][7][8][9].
Among these prognostic scores, the Mayo risk score is the most popular and has been externally validated by different studies [10,11].However, the Mayo risk score includes subjective component, such as edema.In addition, serum albumin (ALB) and edema are interrelated factors with the potential collinearity between them.There is no doubt that we always need to use simple, objective, and effective predictors to predict the prognosis of patients.The ALBI score, calculated using only serum total bilirubin (TBIL) and ALB values was recently reported by Johnson et al. [5] to assess hepatic function in patients with hepatocellular carcinoma (HCC).Subsequently, ALBI grade has also been evaluated as a prognostic factor estimating the clinical outcome of patients with PBC and provides better/similar prognostic significance compared with other well-established prognostic scores like Mayo risk score and MELD score [12].
Although several prognostic scores have been reported to correlate with the prognosis of PBC patients, there are limited prognostic tools to predict the prognosis of PBC with compensated cirrhosis.It remains unclear whether the ALBI score/grade can serve as a novel tool capable of predicting prognosis in patients with PBC and compensated cirrhosis.Therefore, in this retrospective longitudinal study, we evaluated the prognostic performance of the ALBI in comparison to Mayo risk score, Child-Pugh score (CPS), MELD score, APRI, and FIB-4 index in PBC patients with compensated cirrhosis at a large tertiary medical center in Northeast China.

Study Population
This study was a retrospective cohort analysis of all patients with compensated PBC cirrhosis diagnosed between July 2005 and December 2017 at the Department of Hepatology, The First Hospital of Jilin University, China.The diagnosis of PBC was confirmed if two out of the following criteria were fulfilled: (1) elevated alkaline phosphatase levels; (2) high-titer antimitochondrial antibody (AMA) or low-titer AMA confirmed by AMA-M 2 antibodies; and (3) histological characteristics consistent with PBC [13].The diagnosis of cirrhosis was based on the typical imaging features, histological features, and/or presence of varices on endoscopy.Patients were included if PBC was diagnosed before or at the same time as cirrhosis.Patients with previous liver decompensation (ascites, variceal bleed, and hepatic encephalopathy), acute-on-chronic liver failure, or liver transplantation (LT), patients with a known diagnosis of HCC, or patients with severe extrahepatic diseases with poor short-term prognosis were excluded from this study.Patients with lacking data, overlap syndromes, such as autoimmune hepatitis, primary sclerosing cholangitis, or other autoimmune diseases; patients with other liver diseases such as viral hepatitis, alcoholic, drug-induced, or metabolic liver disease; or follow-up <12 months were also excluded from this study.Finally, 219 adult cirrhotic patients with PBC were included in the present analysis.This study protocol was reviewed and approved by the Ethics Committee of The First Hospital of Jilin University, approval number (2021-226).Written informed consent from participants was not required in accordance with local/national guidelines.
For all patients, blood samples were collected on the same day and were assessed in the Clinical Laboratory of The First Hospital of Jilin University.Antibodies were tested by immunofluorescence or immunoblotting assays (AMA, anti-sp100, and anti-gp210) using a EUROLineMaster Plus-A fully automatic immunoblotting instrument.Immunoglobulin M was measured by turbidimetric inhibition immunoassay.The blood biochemistry indices were detected by an automatic biochemical analyzer (7600-210, Hitachi, Japan).Specifically, serum ALB level was measured by bromopotassium phenol green method (normal values: 40.0-55.0g/L); serum TBIL level was measured by enzymatic method involving bilirubin oxidase (normal values: 6.8-30.0μmol/L).Complete blood count was measured using a Sysmex XN-9000 hematology analyzer (Sysmex Corporation, Kobe, Japan).The coagulation tests were performed by the clotting method on the automatic coagulometer "Sysmex CS-5100" (Sysmex Corporation, Kobe, Japan).All assays were measured according to the manufacturer's instructions.

Clinical Outcomes
Usually, patients underwent regular follow-ups every 3-6 months in our hospital with liver biochemistry and ultrasonography regularly monitored.The follow-up started from the inclusion of the study and follow-up data were collected to December 20, 2019.The clinical outcomes were carefully recorded.The duration of follow-up was defined from the date of inclusion in the The Albumin-Bilirubin Score Predicts Liver-Related Mortality Dig Dis 2023;41:946-956 DOI: 10.1159/000531557 study to the last follow-up, the date of LT, or the date of death.For the primary outcome of liver-related death/LT, we identified the cause of death, which was considered liver related if it was attributable to hepatic decompensation, liver failure, HCC, or LT.Two authors independently reviewed the cause of death, and resolved disagreements through discussion.Transplant-free survival was defined from the date of the diagnosis of cirrhosis to the date of liver-associated death without LT.

Statistical Analysis
Continuous variables are presented as mean ± standard deviation (normally distributed) or median with quartile 25-quartile 75 (nonnormally distributed).Categorical variables are expressed as frequencies and percentages.Comparisons of parameters between groups were performed using independent Student's t or Mann-Whitney U tests for continuous variables and χ 2 test for categorical variables.Spearman's coefficient of correlation (r) analysis was conducted between ALBI score and other prognostic scores.Multivariable Cox regression model was carried out to identify independent factors associated with the outcome (liver-related death or LT) and cumulative overall survival was calculated to estimate the hazard ratios (HRs).Factors with p < 0.05 in the univariate Cox regression analysis were entered into the multivariable model.The timedependent receiver operating characteristic (ROC) curve was used to compare the prediction accuracy of several liver scores.The discriminative ability of prognostic scores (including ALBI) to predict the 5-year liver-related mortality was evaluated by using the area under the ROC curve (AUC).Survival rates were estimated by the Kaplan-Meier method with log-rank tests.Two-sided p < 0.05 was considered statistically significant.Statistical analyses were performed by SPSS Statistics 26 (IBM, New York, NY, USA), GraphPad Prism 6 (GraphPad Software, La Jolla, CA, USA), MedCalc Statistical Software version 15.8 (MedCalc Software bvba, Ostend, Belgium; https://www.medcalc.org;2015), and the timeROC package in R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria).
Due to the multicollinearity of different scores in multivariate analysis, only univariate Cox regression analysis was conducted.The results showed that all of prognostic scores used in this study were associated with liver-related death or LT in PBC with compensated cirrhosis.The HR (95% CI) was 15   power of the Mayo risk score for liver-related mortality or the incidence of LT was superior to the other scores within 5 years after the start of observation.However, over the course of the follow-up period, the ALBI score gradually became superior to the other scores, with the highest AUCs for liver-related mortality or LT.The AUCs of the APRI were the lowest of all the scores within 5 years after the start of observation.However, The AUCs of the CPS were the lowest of all the scores after 5 years of follow-up.

Impact of ALBI on Survival
As shown in Figure 4, we compared transplant-free survival by the Kaplan-Meier method among the three groups based on ALBI grade.Not surprisingly, the probability of transplant-free survival decreased with increasing ALBI grade (log-rank p = 0.003).The 5-year transplant-free survival rates of patients in grade 1, grade 2, and grade 3 were 100.0%,96.4%, and 89.4%, respectively.Chen/Zhong/Wang/Kui/Xiaoyu Wen/ Huang/Jin

Discussion
Despite being well validated in PBC, the impact of the ALBI score on liver-related mortality/LT in PBC patients who have developed cirrhosis is unclear.Because most studies in PBC have relatively low numbers of patients with cirrhosis, we chose this cohort of subjects with compensated PBC cirrhosis to better understand the role of ALBI score.Our results showed that a simple evidence-based score incorporating only serum ALB and TBIL concentrations can stratify patients with compensated PBC cirrhosis into different risk categories.The ALBI score is higher in patients who died/underwent LT during follow-up.The higher ALBI score was independently associated with liver-related mortality or LT.Diagnostic accuracy of ALBI score for the prediction of 5-year liverrelated mortality in patients is at least as good as that achieved by the conventional Mayo risk score and MELD score.Moreover, the probability of transplant-free survival decreased with increasing ALBI grade.These findings reveal the prognostic role of ALBI score in PBC patients with compensated cirrhosis, which can identify high-risk patients early and will carry out alternative or additional therapies early for individuals with an unfavorable prognosis.
Liver cirrhosis and its complications are the main causes of mortality in patients with chronic liver disease [15,16].More and more scholars have attempted to develop a classification system that can predict the prognosis of patients with cirrhosis based on clinical and laboratory parameters [17].Currently available survival scoring systems, such as the CPS and the MELD score, have been widely validated as accurate predictors of short-or medium-term survival in patients with liver cirrhosis [18].However, their accuracy is limited in compensated cirrhosis because they    Chen/Zhong/Wang/Kui/Xiaoyu Wen/ Huang/Jin depend on variables reflecting pathophysiological changes associated with advanced disease [19].The CPS is further limited due to subjective assessment of ascites and hepatic encephalopathy [20].A recent meta-analysis reviewed data to assess the diagnostic accuracy of CPS and MELD scores in cirrhotic patients and outlined the need for developing new scores that more accurately evaluate the prognosis of patients with liver disease [21].
It is well known that serum TBIL and ALB are part of a group of tests widely known as liver function tests.We demonstrated that higher TBIL and lower ALB levels  were independently associated with an increase in liverrelated mortality or LT.Our findings share many similarities with previous findings [4,[22][23][24][25][26].In their study, multivariate regression analysis showed that serum ALB and TBIL were independent prognostic factors for adverse outcomes.In fact, the guidelines of the American Association for the Study of Liver Diseases (AASLD) stated that serum TBIL level is the most important factor for predicting prognosis in PBC patients [27].In a systematic review of prognostic indicators in cirrhotic patients, serum TBIL and ALB were the most significant individual prognostic variables in good studies [18].
Both TBIL and ALB were components of the ALBI score, a well-established predictive score [5].Regarding the analysis of clinical outcomes in PBC, some previous studies reported the utility of the ALBI score as a predictor [12,28,29], which was consistent with the findings of our study.As for our study, it is the first in which the ALBI grade was assessed in a cohort of patients with PBC with compensated cirrhosis.We confirmed the discriminative capacity of ALBI (AUC: 0.871), which was better than CPS (AUC: 0.750, p = 0.038), APRI (AUC: 0.693, p = 0.024), and the FIB-4 index (AUC: 0.716, p = 0.043) and was more comparable to the Mayo risk score (AUC: 0.861, p = 0.796) and the MELD score (AUC: 0.780, p = 0.291), with respect to 5-year liver-related mortality in the total cohort.Furthermore, ALBI was a significant risk factor for liver-related mortality/LT (HR: 15.011, 95% CI [5.045, 44.665], p < 0.001).Patients with ALBI grade 1 had the longest survival, followed by patients with ALBI grades 2 and 3 (log-rank p < 0.001).The data show that the ALBI score has the potential to classify the stage of compensated PBC cirrhosis solely on the basis of serum ALB and TBIL values.Moreover, serum ALB and TBIL for the ALBI score are easily obtained by an easily accessible blood test and objectively assessed.Ideally, this could allow us to classify our patients according to their grade of ALBI and manage them based on the expected clinical outcome.
The patients in our cohort have a low 5-year mortality (4.6%) compared to the literature [22,30,31].This may be due to the following reasons: First, the number of cases is relatively small.Second, as one of the largest tertiary medical centers for liver diseases in Northeast China, our data may not represent the general population due to referral bias and geographical constraints.Third, in the retrospective cohort study, 137 patients were lost to follow-up, which could increase the risk of bias.Thus, a prospective study with large samples is required in the future.
The Mayo risk score is one of the most widely accepted and frequently used noninvasive indicators for estimating the prognosis of patients with PBC [32,33].It was developed by Dickson et al. [6] and based on a cohort of 312 PBC patients using clinical and biochemical variables.It has been cross-validated with other independent cohorts and could be generally used for patients with PBC in different stages [10,11].The high AUC, sensitivity, and specificity of the Mayo risk score in our study indicated its efficacy in predicting 5-year liver-related mortality.Furthermore, the ALBI score strongly correlates with Mayo risk score (r = 0.901, p < 0.001).However, one of its drawbacks is its calculation, that requires the presence of edema, a subjective parameter.The CPS is one of the most commonly used methods worldwide for assessing hepatic function [14].However, the CPS also includes subjective components, such as ascites and encephalopathy.The MELD score, is based on serum creatinine, TBIL, and INR and is applicable to patients of diverse etiologies and extensive severity [7].MELD score predicts short mortality with good sensitivity and specificity characteristics [34,35].Nevertheless, their accuracy is limited in compensated cirrhosis [19].Recently, conventional noninvasive biomarkers for liver fibrosis, the APRI and FIB-4 index, were reported to predict prognosis in patients with PBC [36].However, our data indicated that the ALBI score showed larger AUC for 5-year liver-related mortality and more significant HR for transplant-free survival than CPS, APRI, and FIB-4 index did.In other words, CPS, APRI, and FIB-4 index have limited abilities to predict prognosis in the compensated PBC cirrhosis.Therefore, the ALBI score might provide another simple tool as an alternative to CPS, APRI, and FIB-4 index.
This study has several limitations.First, the major weakness of the study is the heterogeneous observation period of the patients.The performance of ALBI may change significantly if only subjects observed over an equal period are included.Second, in this hospital-based and retrospective cohort study, 137 patients were lost to follow-up, which could increase the risk of bias.Therefore, we will conduct prospective studies to confirm our findings in the future.Third, our sample population was relatively small and confined to Chinese patients.However, there is no significant difference in prognosis between Asian and Caucasian patients with PBC.Fourth, most of the subjects with liver cirrhosis were diagnosed by ultrasound, which is not specific.Although liver biopsy is considered to be the best way to evaluate fibrosis stage, it is an invasive procedure.According to the guidelines, liver biopsy is not essential for the diagnosis of PBC [13].Despite these limitations, we were able to identify the predictive utility of the ALBI score regarding PBC patients with compensated cirrhosis.
In summary, our results presented strong evidence that the ALBI score is a simple and effective predictive factor estimating the clinical outcome of patients with compensated PBC cirrhosis and provides better prognostic performance compared with other prognostic scores.Further prospective studies are needed to confirm our findings.

Statement of Ethics
This study protocol was reviewed and approved by the Ethics Committee of The First Hospital of Jilin University, approval number .Written informed consent from participants was not required in accordance with local/national guidelines.
Time-Dependent ROC Analysis for Liver-Related Mortality or LT Figure2shows the plots of AUCs for the ALBI score, Mayo risk score, CPS, MELD score, APRI, and FIB-4 index for liver-related mortality or LT during follow-up.Time-dependent ROC analysis showed that the predictive

Fig. 1 .
Fig. 1.Correlation of the baseline ALBI score with other prognostic scores.a Correlation of the baseline ALBI score with Mayo risk score.b Correlation of the baseline ALBI score with CPS.c Correlation of the baseline ALBI score with MELD score.d Correlation of the baseline ALBI score with APRI.e Correlation of the baseline ALBI score with FIB-4 index.ALBI, albumin-bilirubin; CPS, Child-Pugh score; MELD, model for end-stage liver disease; APRI, aspartate aminotransferase-to-platelet ratio index; FIB-4, fibrosis-4 score.

Fig. 4 .
Fig. 4. Survival analysis of patients with compensated PBC cirrhosis by the ALBI grade using the Kaplan-Meier method.Probability of survival free of LT calculated by the Kaplan-Meier method in PBC patients with compensated cirrhosis.The probability of transplant-free survival decreased with increasing ALBI grade (p = 0.003) by log-rank test.PBC, primary biliary cholangitis; ALBI, albumin-bilirubin.

Table 4 .
Univariate analyses of different prognostic scores in predicting liver-related death or transplantation

Table 5 .
Diagnostic accuracy of different prognostic scores for the prediction of 5-year liver-related mortality in patients ALBI, albumin-bilirubin; CPS, Child-Pugh score; MELD, model for end-stage liver disease; APRI, aspartate aminotransferase-toplatelet ratio index; FIB-4, fibrosis-4 score; AUC, area under the receiver operating characteristic curve.