If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Departamento de Gastroenterología, Escuela de Medicina, Pontificia Universidad Católica de Chile, Santiago, ChileDepartamento de Ciencias Médicas, Facultad de Medicina y Odontología, Universidad de Antofagasta, Antofagasta, Chile
Division of Gastroenterology and Hepatology, University of Florida, Gainesville, FL, USAGastroenterology unit, Massachusetts General Hospital, Boston, MA, USA
Gastroenterology and Hepatology Department. University Hospital Marqués de Valdecilla, Santander, SpainResearch Institute Valdecilla (IDIVAL), Santander, Spain
Gastroenterology and Hepatology Department. University Hospital Marqués de Valdecilla, Santander, SpainResearch Institute Valdecilla (IDIVAL), Santander, Spain
Department of Gastroenterology, Hospital de la Santa Creu i Sant Pau, Institut de Recerca Hospital de Sant Pau-IIB Sant Pau, Universitat Autònoma de Barcelona, CIBERehd, Barcelona, Spain
Department of Gastroenterology, Hospital de la Santa Creu i Sant Pau, Institut de Recerca Hospital de Sant Pau-IIB Sant Pau, Universitat Autònoma de Barcelona, CIBERehd, Barcelona, Spain
Department of Gastroenterology, Hospital de la Santa Creu i Sant Pau, Institut de Recerca Hospital de Sant Pau-IIB Sant Pau, Universitat Autònoma de Barcelona, CIBERehd, Barcelona, Spain
Division of Hepatology, Gastroenterology and Liver Transplantation, Department of Internal Medicine II, Slovak Medical University, F. D. Roosevelt University Hospital, Banska Bystrica, Slovak Republic
Division of Hepatology, Gastroenterology and Liver Transplantation, Department of Internal Medicine II, Slovak Medical University, F. D. Roosevelt University Hospital, Banska Bystrica, Slovak Republic
Liver Unit, Department of Digestive Diseases Hospital General Universitario Gregorio Marañón Madrid, SpainCIBERehd Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas Madrid, Spain
Liver Unit, Department of Digestive Diseases Hospital General Universitario Gregorio Marañón Madrid, SpainCIBERehd Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas Madrid, Spain
Department of Medicine, University of South Dakota Sanford School of Medicine and Transplant Hepatology, Avera Transplant Institute, Sioux Falls, SD, USA
Corresponding author. Division of Gastroenterology Department of Medicine, Schulich School of Medicine Western University & London Health Sciences Centre London, Ontario, Canada
Departamento de Gastroenterología, Escuela de Medicina, Pontificia Universidad Católica de Chile, Santiago, ChileDivision of Gastroenterology, Department of Medicine, Schulich School of Medicine, Western University & London Health Sciences Centre, London, Ontario, CanadaDepartment of Epidemiology and Biostatistics, Schulich School of Medicine, Western University, London, Ontario, Canada
Alcohol-associated (AH) hepatitis is associated with multi-organ failure and high short-term mortality.
-
MELD 3.0 predicts 30- and 90-day mortality better than the MELD-Na score and Maddrey’s discriminant function.
-
MELD 3.0 was not superior in predicting mortality to MELD and Age-Bilirubin-International Normalized Ratio-Creatinine (ABIC) scores, but its classification accuracy was similar between countries.
-
The new MELD 3.0 was the best predictor of renal replacement therapy requirements compared to other models.
Abstract
Background & Aims
MELD score better predicts mortality in alcohol-associated hepatitis (AH) but could underestimate severity in women and malnourished patients. We aimed to assess the performance of the MELD 3.0 score to predict short-term mortality in AH using a global cohort.
Methods
Retrospective cohort study including patients admitted with AH (2009-2019). The main outcome was all-cause 30-day mortality. We compared the area under the Receiver Operating Curve (AUC) using DeLong's method. We also performed a time-dependent AUC with competing risks analysis.
Results
A total of 2,124 patients were included from 28 centers from 10 countries on three continents (median age 47.2±11.2 years, 29.9% women, 71.3% had underlying cirrhosis). The median MELD 3.0 score at admission was 25,
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
, with an estimated survival of 73.7% at 30 days. The MELD 3.0 score had a better performance in predicting 30–day mortality (AUC:0.761, 95%CI:0.732–0.791) compared to MELD-Na (AUC:0.744, 95%CI:0.713–0.775; p=0.042) and Maddrey’s discriminant function (mDF) (AUC:0.724, 95%CI:0.691–0.757; p=0.013). However, MELD 3.0 did not perform better than traditional MELD (AUC:0.753, 95%CI:0.723–0.783; p=0.300) and Age-Bilirubin-International Normalized Ratio-Creatinine (ABIC) (AUC:0.757, 95%CI:0.727–0.788; p=0.765). These results were consistent in competing-risk analysis, where MELD 3.0 (AUC:0.757, 95%CI:0.724–0.790) predicts better 30-day mortality than MELD-Na (AUC:0.739, 95%CI:0.708–0.770; p=0.028) and mDF (AUC:0.717, 95%CI:0.687–0.748; p=0.042). MELD 3.0 score was significantly better in predicting renal replacement therapy (RRT) requirements during admission than other scores (AUC:0.844, 95%CI:0.805–0.883).
Conclusions
MELD 3.0 demonstrated better performance than MELD-Na and mDF in predicting 30-day and 90-day mortality, and was the best predictor of RRT requirements during admission for AH. Further prospective studies are needed to validate its extensive use.
Lay Summary
Severe alcohol-associated hepatitis (AH) has a high short-term mortality. The establishment of treatments and liver transplantation depends on mortality prediction. We evaluated in a large global cohort the performance of the new MELD 3.0 score to predict short-term mortality in AH. MELD 3.0 has a better performance in predicting 30- and 90-day mortality than MELD-Na and Maddrey’s discriminant function, but was similar to MELD and Age-Bilirubin-International Normalized Ratio-Creatinine (ABIC) scores. MELD 3.0 was the best predictor of renal replacement therapy requirements. Further prospective studies are needed to support the wide use of MELD 3.0 in AH.
National Institute on Alcohol Abuse and Alcoholism
RRT
renal replacement therapy
Conflict of interest statement
The authors declare that there is no conflict of interest.
Financial support/Acknowledgments
Juan Pablo Arab and Marco Arrese receive support from the Chilean government through the Fondo Nacional de Desarrollo Científico y Tecnológico (FONDECYT 1200227 to JPA and 1191145 to MA). Ramón Bataller is recipient of NIAAA U01AA021908 and U01AA020821. Graphical abstract was partially made with Biorender.
Data availability statement
The datasets generated and analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.
Authors' contributions
LAD and JPA conceived and designed the study; all authors collected the data, contributed to data analysis and interpretation; JPA, LAD, EF, WD, AKS, and RB performed final analysis and drafted the manuscript; all the authors participated in drafting the article and revising it critically for important intellectual content; and all the authors gave final approval of the version submitted.
Introduction
Alcohol-associated liver disease (ALD) constitutes a leading cause of alcohol-related deaths worldwide
European Association for the Study of the Liver Electronic address: [email protected], European Association for the Study of the Liver. EASL Clinical Practice Guidelines: Management of alcohol-related liver disease.
The Model of End-Stage Liver Disease (MELD) and modified Maddrey’s discriminant function (mDF) scores have been shown to accurately predict mortality, where a score of >20 or ≥32, respectively, is associated with lower 90-day survival
. MELD sodium (MELD-Na) did not further improve the accuracy of MELD to predict mortality, while the mDF demonstrated the poorest performance as a static score to predict 90-day mortality in AH
Mortality Risk Factors Among Patients With Cirrhosis and a Low Model for End-Stage Liver Disease Sodium Score (≤15): An Analysis of Liver Transplant Allocation Policy Using Aggregated Electronic Health Record Data.
. This version added female gender and albumin in the score, demonstrating a slightly more accurate mortality prediction than MELD-Na in cirrhotic patients
. It seems critical to have a score with high accuracy in predicting mortality in different scenarios and renal replacement therapy (RRT) to facilitate decision-making. Therefore, we aimed to assess the new MELD 3.0 in predicting 30- and 90-day mortality in AH in a worldwide cohort study. We also explored the performance in predicting mortality, assessing different subgroups according to corticosteroid use, sex, and prediction of RRT requirements during hospitalization.
Material And Methods
Study design and participants
We conducted a retrospective registry-based study of patients admitted with severe AH. We included patients who met the National Institute on Alcohol Abuse and Alcoholism (NIAAA) clinical criteria of severe AH
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
. The specific inclusion criteria were 1. A history of alcohol use of >60 g/day in men and >40 g/day in women; 2. an AST <400 U/l with AST/ALT ratio >1.5; 3. Serum γ-glutamyl transpeptidase (GGT) >80 mg/dl; 3. Abnormal coagulation tests (prolonged prothrombin time and/or International Normalized Ratio [INR] values); 4. Serum bilirubin levels >3 mg/dl. According to clinical criteria, those with uncertain AH diagnosis had undergone a liver biopsy, which must confirm the diagnosis of AH. Only patients meeting all the clinical criteria (probable AH) or biopsy-proven AH (definite AH) independent of corticosteroid use were included.
The diagnosis of cirrhosis was based on prior medical history and imaging (ultrasound, transient elastography, computed tomography, or magnetic resonance imaging). Liver biopsy was done in uncertain cases by the attending physician’s criteria. Treatment with corticosteroids and their continuation or discontinuation in the light of an assessment of response was based on discretion of the treating physician. We excluded patients with 1. Age under 18 years old; 2. Pregnancy; 3. AST and/or ALT levels above 400 IU/ml; 4. Prolonged alcohol abstinence (>60 days) before the presentation; 5. Presence of drug-induced liver injury (DILI), ischemic hepatitis, biliary duct obstruction, viral hepatitis, autoimmune hepatitis, or Wilson disease; 6. Hepatocellular carcinoma beyond Milan criteria; 7. Extrahepatic neoplasia with a life expectancy of fewer than six months; or 8. History of severe extrahepatic disease that confers a survival of fewer than six months. For patients with more than one admission, information was registered only for the first episode of AH.
Data collection
We performed a retrospective review of the records of patients hospitalized with the diagnosis of severe AH (from January 2009 to January 2019). We recorded laboratory results performed during admission, as well as the type of steroids and length of use. We also recorded the MELD, MELD-Na, mDF, and Age-Bilirubin-International Normalized Ratio-Creatinine (ABIC) scores at admission, infections, mortality, and causes of death at 90 days
. The data collected was recorded in a confidential electronic case report form, which was used at all the centers collecting the data. The electronic database was only managed by the main researchers of the study. We requested an informed consent waiver at each participating center, and de-identified data were analyzed.
Statistical analysis
The primary outcome was 30-day mortality in patients with severe AH. The secondary outcomes were 90-day mortality and renal replacement therapy (RRT) requirements during admission. As an exploratory analysis, we assessed the performance of scores to predict 30-day and 90-day mortality in patients stratified by sex and those who underwent corticosteroid treatment. We also evaluated response to corticosteroid treatment at day 7 using the Lille score. Categorical variables were summarized using frequencies and percentages. We assessed normal distribution in continuous data using the Kolmogorov-Smirnov test and histograms. Continuous variables with normal distribution were described with mean and standard deviation. Variables without a normal distribution were summarized using the median and interquartile ranges. Analyses were completed using the Chi-square test for categorical variables, the Student's T for normally distributed continuous variables, and non-parametric tests for continuous variables that are not normally distributed.
We constructed receiver operating characteristics (ROC) curves to assess the accuracy of prognosis scores, and we calculated the area under the ROC curves (AUC). We used DeLong's method to test for statistically significant differences between ROC curves
. We specified a tie-corrected nonparametric estimate (trapezoidal approximation) since the assumption that the true ROC curve was smooth. This means that the classifiers we measured (i.e., MELD 3.0, MELD-Na, MELD, mDF, and ABIC) were a discretized approximation of a true latent and a continuous classifier. The standard errors and confidence intervals were estimated through bootstrapping. We also estimated mortality in time-dependent AUC with competing risk (liver transplantation) using the inverse probability of censoring weighted method
. We determined the optimal cutpoint to predict RRT requirements using the Youden index. We also assessed the sensitivity and specificity of each cut-off value. Those patients who were lost to follow-up were censored in the analyses. Heterogeneity among countries was assessed by comparing the AUC adjusted by country using the STATA "comproc" command. With "comproc" the Wald test results for marker comparisons are based on the bootstrap standard errors for the difference between markers
. We evaluated the country's incremental value by comparing the AUC for logistic models predicting 30-day mortality when including or not the country of each patient
. A second logit regression model was formulated to assess the relationship of predictions to the true probabilities of the event, based upon a polynomial transformation of the predictions, the degree of the polynomial (beginning with second order) being forwardly selected based on a sequence of likelihood ratio tests. The calibration belt's deviation from the identity line was reported with a p-value. For all analyses, a p<0.05 was considered significant. This manuscript adheres to the TRIPOD statement for reporting prediction models
. The analyses were performed with STATA software version 14 (StataCorp, College Station, Texas) and R software (R Foundation for Statistical Computing, Vienna, Austria). All authors had access to the study data, reviewed, and approved the final manuscript.
Results
Baseline characteristics of the cohort
We included a total of 2,124 patients from 28 centers (10 countries on three continents). The median number of patients included per center was 26 [10–72] (Supplementary Table 1). The mean age was 47.2±11.2 years old, 29.9% were women, and 71.3% had a prior history of cirrhosis. The median MELD, MELD-Na, and MELD 3.0 scores at admission were 25
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
, respectively. Also, the median mDF and ABIC scores were 58 [40–83] and 8.0 [6.9–9.0], respectively. Patients presented with median albumin of 2.6 [2.0–3.0] g/dL, bilirubin of 13.3 [7.0–23.7] mg/dL, and INR of 1.9 [1.6–2.3]. At admission, the median creatinine was 0.8 [0.6–1.5] mg/dL, BUN of 13
European Association for the Study of the Liver Electronic address: [email protected], European Association for the Study of the Liver. EASL Clinical Practice Guidelines: Management of alcohol-related liver disease.
Mortality Risk Factors Among Patients With Cirrhosis and a Low Model for End-Stage Liver Disease Sodium Score (≤15): An Analysis of Liver Transplant Allocation Policy Using Aggregated Electronic Health Record Data.
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
mg/dL, plasma sodium of 132 [128–136] mEq/lt, and 12.1% of patients required RRT during the hospitalization. The main characteristics of the cohort and its differences according to the prior history of cirrhosis are summarized in Table 1.
Table 1- Baseline characteristics of all patients, and according to the presence of cirrhosis. A total of 1,652 (77.8%) had data about the presence of cirrhosis. Comparisons were performed using the Chi-square test for categorical variables, the Student's T for normally distributed continuous variables, and non-parametric tests for continuous variables that are not normally distributed.
The median follow-up was 183 [27-799] days. A total of 167 (7.9%) and 254 (12%) were lost to follow-up at 30 and 90 days, respectively. The overall estimated survival since admission was 73.7% (95%CI:71.5–75.8%) at 30 days, 62.5% (95%CI:59.9–65.0%) at 90 days, and 57.2% (95%CI:54.5–59.7%) at 180 days. Thirteen percent of patients underwent liver transplantation and the median time between admission and liver transplantation was 204 [65–437] days. The main attributed causes of death were multi-organ failure (37.2%), infections (17.6%), gastrointestinal bleeding (12.1%), acute kidney injury (AKI)(6.3%), and several patients had more than one cause of death (5%). A total of 46.3% of patients developed an infection during hospitalization. The most common infections were from the urinary tract (37.6%), respiratory tract infections (18.1%), and spontaneous bacterial peritonitis (6.0%). A total of 44.3% of infected patients had bacteremia with positive blood cultures.
Performance MELD 3.0 and other models in alcohol-associated hepatitis
Mortality Risk Factors Among Patients With Cirrhosis and a Low Model for End-Stage Liver Disease Sodium Score (≤15): An Analysis of Liver Transplant Allocation Policy Using Aggregated Electronic Health Record Data.
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
in patients who died at 30 days (p<0.001). The MELD 3.0 score had a better performance in predicting 30-day mortality (AUC 0.761, 95%CI:0.732–0.791) compared to MELD-Na (AUC 0.744, 95%CI:0.713–0.775; p=0.042) and mDF (AUC 0.724, 95%CI:0.691–0.757; p=0.013)(Figure 1A). MELD 3.0 was also superior (AUC of 0.744, 95%CI:0.718–0.771) than MELD-Na (AUC 0.721, 95%CI:0.694–0.749; p=0.003) and mDF (AUC 0.706, 95%CI:0.677–0.735; p=0.004) in predicting 90-day mortality (Figure 1B). However, MELD 3.0 was similar to traditional MELD and ABIC in predicting 30- and 90-day mortality (Table 2). A MELD 3.0 higher than 20 had a sensitivity of 91.5% and a specificity of 32.6% in predicting 30-day mortality and was slightly better than the prior versions of MELD (Supplementary Tables 2–4). The sensitivity and specificity of mDF and ABIC are also described in Supplementary Tables 5–6. In deceased patients at 90 days, the new MELD 3.0 increased the score in 279 (54.2%) patients, decreased the score in 193 (37.5%), and 43 (8.3%) patients did not have any change compared to MELD-Na (Supplementary Table 7). In addition, we performed a time-dependent AUC with competing risk analysis to better understand the performance of models in predicting mortality, weighting the benefit in survival of liver transplant. In this analysis, assessing transplant as a competing risk, the MELD 3.0 was also superior to MELD-Na and mDF in predicting 30- and 90-day mortality (Supplementary Table 8).
Figure 1- Comparison of MELD-Na and MELD 3.0 in predicting mortality in AH. Receiver operating characteristic curves and AUC were generated, and MELD 3.0 score was superior to MELD-Na and mDF predicting (A) 30-day mortality and (B) 90-day mortality. The 95% confidence intervals are described in parentheses.
Table 2- Performance of MELD 3.0, MELD-Na, MELD, modified Maddrey’s discriminant function (mDF), and Age-Bilirubin-International Normalized Ratio-Creatinine (ABIC) scores in predicting 30 or 90-day mortality in AH. We performed comparisons between ROC AUC using DeLong's method.
We assessed the calibration of models in predicting 30- and 90-day mortality. Although most models have an acceptable calibration, the MELD 3.0 demonstrated the best calibration to predict 30- and 90-day mortality, while mDF showed the poorest calibration. Of note, ABIC model showed a regular calibration in patients with higher scores (Supplementary Figure). Heterogeneity between countries was also assessed by comparing AUC adjusted per country. We observed that the discriminatory accuracy of MELD 3.0 did not significantly differ by adjusting per country (Supplementary Table 9). However, MELD and ABIC scores significantly improve their discriminatory accuracy adjusting by country (p=0.047 and p=0.012 for testing the incremental predicting value of the country, respectively)(Supplementary Table 9).
A total of 45.7% of patients underwent corticosteroid therapy. The mean age of them was 47.2±10.9 years old, 29.8% were women, 69.7% had a prior history of cirrhosis, and they scored a median MELD 3.0 of 26
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
. In these patients, the overall performance of MELD 3.0 score (AUC 0.728, 95%CI:0.681–0.776) was only superior to mDF in predicting 30-day mortality (AUC 0.681, 95%CI:0.628–0.733; p=0.048)(Supplementary Table 10). However, the MELD 3.0 demonstrated a better performance in predicting 90-day mortality (AUC 0.720, 95%CI:0.679–0.760) compared to MELD-Na (AUC 0.687, 95%CI:0.645–0.729; p=0.013) and traditional MELD (AUC 0.693, 95%CI:0.651–0.735; p=0.049)(Supplementary Table 10). Only 51.7% of patients who underwent corticosteroid treatment achieved the criteria of responders at day 7. ABIC demonstrated a superior performance (AUC 0.746, 95%CI:0.708–0.784) in predicting response to corticosteroid treatment versus MELD 3.0 (AUC 0.685, 95%CI:0.644–0.727; p=0.002), while MELD 3.0 was superior to the other models (Supplementary Table 11).
Impact of sex on survival prediction
The median MELD 3.0 score at admission according to sex was 26
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
Mortality Risk Factors Among Patients With Cirrhosis and a Low Model for End-Stage Liver Disease Sodium Score (≤15): An Analysis of Liver Transplant Allocation Policy Using Aggregated Electronic Health Record Data.
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
in women (p=0.012). When we assessed the predicting performance of 30-day survival according to sex, the MELD 3.0 score did not perform better over the other models in predicting 30-day mortality for men or women, exclusively (Supplementary Table 12). However, MELD 3.0 score demonstrated a better performance in predicting 90-day mortality than MELD-Na (AUC 0.724, 95%CI:0.691–0.758; p=0.013) in men, and mDF (AUC 0.686, 95%CI:0.636–0.737; p=0.031) in women (Supplementary Table 12).
Renal replacement therapy requirements in AH
We also explored the performance of the scores in predicting RRT requirements during hospitalization. The estimated survival was lower in patients who required RRT compared to those who did not at 30 days (47.7% [95%CI:40.6–54.5%] versus 78.2% [95%CI:75.7–80.4%]) and at 90 days (30.5% [95%CI:24.1–37.1%] versus 68% [95%CI:65.1–70.6%]), p<0.001 (Figure 2A). The MELD 3.0 demonstrated the best performance predicting RRT requirement during hospitalization (AUC 0.844, 95%CI:0.805–0.883) compared to all the other models (Figure 2B and Supplementary Table 13). A MELD 3.0 score of 35 or higher had a sensitivity of 74.1% and a specificity of 81.6% to predict RRT requirements during admission (Supplementary Table 14).
Figure 2- Short-term survival of patients per renal replacement therapy (RRT) requirement (A) and comparison of models in predicting RRT requirements (B). Survival was estimated using Kaplan-Meier curves, and comparisons were performed using log-rank test. Receiver operating characteristic curves and AUC were generated to compare performance between models in predicting RRT requirement. The 95% confidence intervals are described in parentheses.
. However, there are several concerns about the use of MELD and MELD-Na scores due to potential disparities in women and malnourished patients. In this large cohort study, we evaluated the performance of the third iteration of the MELD score (MELD 3.0) predicting mortality in severe AH. We identified a slight but significantly better performance of the MELD 3.0 score over MELD-Na and mDF in predicting 30-day mortality, which was similar at 90 days. Also, the MELD 3.0 score up-categorized 54.2% of deceased patients at 90 days over the MELD-Na score. Interestingly, we observed that MELD 3.0 had the highest performance in predicting RRT during admission compared to other models (AUC 0.84, 95%CI:0.81-0.88). Although MELD 3.0 did not demonstrate a better performance than traditional MELD and ABIC scores, the calibration of MELD 3.0 was slightly better, especially in patients with higher scores, and its classification accuracy was similar among different countries.
Historically, the mDF was used as a predictor of mortality risk in patients with severe AH (with a score over 32) based on a retrospective analysis
. However, the mDF requires the use of the prothrombin time (PT), which has been largely replaced by the INR, and its reference value is not always reported by many clinical laboratories. In addition, mDF had the poorest performance in this study. Thus, a MELD score of 21 or higher has demonstrated a better accuracy yield predicting mortality in severe AH
. Since 2016, the MELD-Na score has been used for liver allocation instead of MELD, since hyponatremia was recognized as a prominent independent risk factor for mortality in end-stage liver disease
. As a novel feature, MELD 3.0 includes the addition of two variables (female gender and serum albumin), demonstrating a more accurate mortality prediction than MELD in cirrhotic patients and improving the prior disparities in allocation for liver transplantation among patients with decompensated cirrhosis
The first MELD score included three variables: serum bilirubin, serum creatinine, and the INR, which could imply several shortcomings. For example, inter-laboratory variability in the INR and creatinine measurement might contribute to an overall mean variation in calculated MELD of 4.8 points (range 2–11)
. Likewise, hyperbilirubinemia can substantially affect the result of a colorimetric assay used to measure creatinine. Also, key factors that contribute to rising mortality are not assessed (e.g. hypoalbuminemia). Some patients with AH have chronic liver disease, alcohol use disorder, or other causes that lead to malnutrition, which impacts mortality and transplant requirements
Mortality Risk Factors Among Patients With Cirrhosis and a Low Model for End-Stage Liver Disease Sodium Score (≤15): An Analysis of Liver Transplant Allocation Policy Using Aggregated Electronic Health Record Data.
. As an acute phase reactant, albumin may also have other causes to be altered in these patients. Those differences in laboratory values can also be observed according to sex. Indeed, in a steady state, the main determinant of serum creatinine is its endogenous production, in which several factors, unrelated to renal function play a role, mainly muscle mass which, in turn, is influenced by sex
. A prior model described in the MELD-GRAIL-Na study showed that the addition of glomerular filtration rate (GFR) allows for improved discrimination among women and those with the highest risk of premature mortality due to cirrhosis
. However, this model was designed for chronic patients with stable GFR, and multiple factors make it difficult to estimate renal function during an acute decompensation, discouraging its use in AH
. Also, we found that MELD 3.0 better predicts the need for dialysis during hospitalization. This better performance could be explained by several factors. First, the GFR is overestimated in women compared to men with the same creatinine level
. Second, women are disadvantaged by MELD and are estimated to receive 1 to 2.4 fewer creatinine-derived MELD points than men with the same renal function
. Third, the higher creatinine values add more points compared to prior versions, and the ceiling of serum creatinine was lowered from 4.0 mg/dl to 3.0 mg/dl
. In the future, a MELD 3.0 score cut-off could be determined to prevent and treat AKI early in severe AH, with potential clinical benefit in this population.
The mDF and MELD scores have been previously used to define corticosteroid therapy in severe AH
. Although the STOPAH study suggested a narrow therapeutic window, a recent study demonstrated a short-term benefit of corticosteroids even with higher MELD scores, where the highest effect was observed in patients with MELD scores between 25 and 39
The Impact of Malnutrition on the Hospital and Infectious Outcomes of Patients Admitted With Alcoholic Hepatitis: 2011 to 2017 Analysis of US Hospitals.
. Therefore, an adequate selection of candidates for corticosteroids and the early detection of infections during treatment is the key to decreasing death due to infectious diseases
. MELD 3.0 could better represent the malnourished patients and potentially impact a better selection of patients for corticosteroid treatment. Although a new model called the Mortality Index for Alcohol-Associated Hepatitis (MIAAH) was developed to predict short-term mortality, its performance was lower than the traditional MELD score in the validation cohort
. Thus, further studies should compare the performance of MIAAH and MELD 3.0 scores.
Our retrospective cohort study includes a vast number of patients, ethnicities, and centers. However, our study suffers several limitations related to its retrospective nature, including potential losses or errors in the data records. Another limitation was the diagnosis of cirrhosis since it was performed by the attending physician using clinical data, laboratory results, and imaging (mostly without a liver biopsy). Furthermore, the albumin serum levels in the blood could be elevated due to the prior administration of albumin, which could modify the MELD 3.0 values. Also, there are important differences in nutrition and muscle mass among ethnicities, not specifically assessed in this study. On the other hand, we considered RRT requirements as an outcome, but some centers could not have RRT available for AH patients, and the criteria for establishing RRT could be different among centers. Future studies are needed to define its capacity to determine the best candidates for treatments such as corticosteroids or early establishment of RRT.
In conclusion, our large global cohort study demonstrated that MELD 3.0 predicts better 30-day and 90-day mortality in AH compared to MELD-Na and mDF. MELD 3.0 also predicted better the RRT requirements during admission compared to other models, and it was strongly associated with decreased survival. The MELD 3.0 has adequate calibration, and its classification accuracy did not differ between countries. Our results suggest that MELD 3.0 is promising and utility in determining mortality risk and RRT in AH. Further prospective studies are needed to validate our findings supporting MELD 3.0 score use.
Appendix A. Supplementary data
The following is/are the supplementary data to this article:
Electronic address: [email protected], European Association for the Study of the Liver. EASL Clinical Practice Guidelines: Management of alcohol-related liver disease.
Mortality Risk Factors Among Patients With Cirrhosis and a Low Model for End-Stage Liver Disease Sodium Score (≤15): An Analysis of Liver Transplant Allocation Policy Using Aggregated Electronic Health Record Data.
Standard Definitions and Common Data Elements for Clinical Trials in Patients With Alcoholic Hepatitis: Recommendation From the NIAAA Alcoholic Hepatitis Consortia.
The Impact of Malnutrition on the Hospital and Infectious Outcomes of Patients Admitted With Alcoholic Hepatitis: 2011 to 2017 Analysis of US Hospitals.
00058-7&id=ga1.jpg){kind=link}
00058-7&id=gr1.jpg){kind=link}
00058-7&id=gr2.jpg){kind=link}