Published online May 28, 2024. doi: 10.3748/wjg.v30.i20.2657
Revised: April 22, 2024
Accepted: May 9, 2024
Published online: May 28, 2024
Processing time: 133 Days and 2.6 Hours
Cirrhotic patients with acute-on-chronic liver failure (ACLF) in the intensive care unit (ICU) have a poor but variable prognoses. Accurate prognosis evaluation can guide the rational management of patients with ACLF. However, existing prognostic scores for ACLF in the ICU environment lack sufficient accuracy.
To develop a new prognostic model for patients with ACLF in ICU.
Data from 938 ACLF patients in the Medical Information Mart for Intensive Care (MIMIC) database were used to develop a new prognostic model (MIMIC ACLF) for ACLF. Discrimination, calibration and clinical utility of MIMIC ACLF were assessed by area under receiver operating characteristic curve (AUROC), calibration curve and decision curve analysis (DCA), respectively. MIMIC ACLF was then externally validated in a multiple-center cohort, the Electronic Intensive Care Collaborative Research Database and a single-center cohort from the Second Hospital of Hebei Medical University in China.
The MIMIC ACLF score was determined using nine variables: ln (age) × 2.2 + ln (white blood cell count) × 0.22 - ln (mean arterial pressure) × 2.7 + respiratory failure × 0.6 + renal failure × 0.51 + cerebral failure × 0.31 + ln (total bilirubin) × 0.44 + ln (internationalized normal ratio) × 0.59 + ln (serum potassium) × 0.59. In MIMIC cohort, the AUROC (0.81/0.79) for MIMIC ACLF for 28/90-day ACLF mortality were significantly greater than those of Chronic Liver Failure Consortium ACLF (0.76/0.74), Model for End-stage Liver Disease (MELD; 0.73/0.71) and MELD-Na (0.72/0.70) (all P < 0.001). The consistency between actual and predicted 28/90-day survival rates of patients according to MIMIC ACLF score was excellent and superior to that of existing scores. The net benefit of MIMIC ACLF was greater than that achieved using existing scores within the 50% threshold probability. The superior predictive accuracy and clinical utility of MIMIC ACLF were validated in the external cohorts.
We developed and validated a new prognostic model with satisfactory accuracy for cirrhotic patients with ACLF hospitalized in the ICU. The model-based risk stratification and online calculator might facilitate the rational management of patients with ACLF.
Core Tip: Patients with acute-on-chronic liver failure (ACLF) usually need to be hospitalized in intensive care unit for condition monitoring and organ support therapy where existing prognostic scores for ACLF have a limited predictive accuracy. In this study, we developed a new prognostic model [Medical Information Mart for Intensive Care (MIMIC) ACLF] using data in MIMIC database for patients with ACLF. MIMIC ACLF not only exhibited a satisfactory predictive accuracy and clinical utility in MIMIC cohort, but also performed well in external single and multiple-center validation cohorts. In addition, we performed a risk stratification and developed an online calculator based on MIMIC ACLF to facilitate its application in clinical practice.
- Citation: Zhu ZY, Huang XH, Jiang HQ, Liu L. Development and validation of a new prognostic model for patients with acute-on-chronic liver failure in intensive care unit. World J Gastroenterol 2024; 30(20): 2657-2676
- URL: https://www.wjgnet.com/1007-9327/full/v30/i20/2657.htm
- DOI: https://dx.doi.org/10.3748/wjg.v30.i20.2657
Liver cirrhosis is the common end stage of all chronic liver diseases. Acute decompensation (AD) and its final and most fatal stage, acute-on-chronic liver failure (ACLF), are major causes of death in patients with liver cirrhosis[1]. ACLF was first defined by European Association for the Study of Liver-Chronic Liver Failure (EASL-CLIF) Consortium as a distinct syndrome characterized by AD of cirrhosis, precipitating events, systemic inflammatory response, multiple organ failures and high short-term mortality[2]. Despite the progress in understanding of the pathophysiological mechanisms of ACLF, including intense systemic inflammatory response, immune dysfunction, metabolic changes, oxidative stress and mitochondrial dysfunction[3,4], no specific or definitive treatment is available for ACLF except for liver transplantation (LT)[5]. The main principle in management of ACLF is to treat precipitating events and associated complications and provide organ support therapy[6,7].
To monitor the disease condition and receive organ support therapy, patients with ACLF usually need to be transferred to the intensive care unit (ICU). In the ICU, reasonable management of ACLF should be consistent with the patient's prognosis. For patients with a relatively optimistic prognosis, ongoing aggressive therapy should be provided without hesitation as a bridge to LT; while for those with poor prognosis, considering their economic burden and limited medical resources, palliative therapy might be an optimal choice. Therefore, an excellent prognostic score that can accurately predict the prognosis of patients with ACLF is very important for decision-making by clinicians in the ICU. At present, existing prognostic scores for patients with ACLF in the ICU mainly include Child-Turcotte-Pugh (CTP), Model for End-stage Liver Disease (MELD), MELD-Na and Chronic Liver Failure Consortium (CLIF-C) ACLF (CLIF-C ACLF). Developed for cirrhotic patients, CTP, MELD and MELD-Na are traditional prognostic scores that do not consider the importance of multiple organ failure[8], resulting in a limited accuracy[9-11]. Compared with traditional scores, CLIF-C ACLF is a recently (2014) developed prognostic score by the EASL-CLIF Consortium[12], showing significantly superior prediction performance for 28/90 d mortality in ACLF patients to the traditional scores[12-14]. However, some later studies found that the prediction performance of CLIF-C ACLF was not so ideal[15-17] (concordance index < 0.8), especially in the ICU. In addition, CLIF-C ACLF was not fully developed for ACLF patients hospitalized in ICU and has several limitations, such as ceiling effect of serum bilirubin, international normalized ratio (INR) and creatinine, difficulty in differentiating hepatic encephalopathy (HE) grade 1 from grade 0; and subjective variability in determining use of vasopressors[18], which might reduce its predictive accuracy.
In this study, we aimed to develop a new prognostic model for cirrhotic patients complicated with ACLF in ICU using data from Medical Information Mart for Intensive Care (MIMIC) database and externally validate it in the multiple-center intensive care cohort, the electronic ICU (eICU) collaborative research database and a Chinese single-center general ward cohort from the Second Hospital of Hebei Medical University (SHHMU) in China. We hope that this new prognostic model has satisfactory accuracy and is convenient for predicting short-term mortality in cirrhotic patients with ACLF.
Patients included in this retrospective cohort study were from the MIMIC-Ⅳ database, the eICU database and the SHHMU cohort. MIMIC-Ⅳ is a large and freely available database that comprises anonymous medical records of patients admitted to the ICUs of Beth Israel Deaconess Medical Center between 2008 and 2019[19,20]. The eICU is a large, multi-center intensive care database jointly developed by the Philips Medical and Computational Physiology Laboratory of Massachusetts Institute of Technology on the basis of success of MIMIC database[21]. This database contains medical data of more than 200000 patients who were admitted to the ICUs at 208 hospitals in the mainland United States from 2014 to 2015. Because the source hospital of MIMIC database, Beth Israel Deacon Medical Center, is not involved in the eICU program, the eICU database is completely independent of the MIMIC database. The SHHMU cohort included cirrhotic patients with ACLF hospitalized in the Gastroenterology department at SHHMU. The MIMIC cohort was used for development and internal validation of a new prognostic model (MIMIC ACLF). The eICU and SHHMU cohorts were used for external validation of MIMIC ACLF in the environment of ICU and general ward, respectively. The reason we selected the SHHMU cohort is that although most ACLF patients require hospitalization in the ICU, there are still a few patients receiving treatment in the general ward. The inclusion criterion was cirrhotic patients with ACLF hospitalized for various ADs (ascites, bacterial infections, acute variceal bleeding and HE). The exclusion criteria were as follows: (1) < 18 years of age; (2) pregnancy; (3) incomplete records; (4) hepatocellular carcinoma (HCC) or malignant tumors from other organs; (5) human immunodeficiency virus (HIV) infection or undergoing immunosuppressive therapy; (6) death within 24 h of ICU admission; and (7) loss of follow-up.
Data in MIMIC-Ⅳ (version: 2.0) and EICU (Version: 2.0.1) databases, including age, gender, race, category of AD, first day vital signs, first day laboratory parameters, vasopressors therapy, mechanical ventilation therapy, renal replacement treatment therapy, severity scores, and survival information were extracted through Postgres Structured Query Language (PostgreSQL) programming in Navicat Premium software. For those admitted multiple times to the ICU, data from their first admission were used. For laboratory tests that were measured more than one time, maximum or minimum values were selected according to the clinical implications, such as the maximum of total bilirubin (TB) or minimum of albumin. Patient’s data in the SHHMU cohort was obtained through electronic medical record system.
For the MIMIC and eICU cohorts, MELD and APACHE II scores were obtained directly from the MIMIC and eICU databases, respectively. For the SHHMU cohort, MELD was calculated as: 3.8 × ln [bilirubin (mg/dL)] + 11.2 × ln (INR) + 9.6 × ln [creatinine (mg/dL)] + 6.4 × (cirrhosis etiology: Biliary or alcoholic 0, others 1)[22]. The MELD-Na score was calculated as: MELD + 1.59 (135-Na) with maximum and minimum Na of 135 and 120 mEq/L, respectively[23]. The CLIF-C ACLF score were calculated as: 10 × {0.33 × CLIF- organ failures (OFs) + 0.04 × age + 0.63 × ln [white blood cell (WBC) count] – 2}[12].
ACLF was defined by EASL-CLIF consortium[2] and diagnosed/graded according to CLIF-OF score[12]. Specifically, liver failure was defined as TB ≥ 12 mg/dL; coagulation failure was defined as INR ≥ 2.5; renal failure was defined as creatinine > 2 mg/dL or requirement of renal replacement therapy; circulatory failure was defined as requirement of vasopressor therapy to maintain blood pressure; cerebral failure was defined as HE grade III or IV (West Haven criterion); respiratory failure was defined as PaO2/FiO2 ≤ 200 or SpO2/FiO2 ≤ 214 or mechanical ventilation for reasons other than airway protection in cerebral failure[12]. ACLF grade 1 was defined as: (1) Single renal failure; (2) single failure of liver, coagulation, circulation or respiration along with a serum creatinine ranging from 1.5 to 1.9 mg/dL and/or mild to moderate HE; and (3) single cerebral failure along with serum creatinine ranging from 1.5 and 1.9 mg/dL. ACLF grade 2 and grade3 were defined as: 2 and ≥ 3 OFs, respectively[12].
The outcomes of this study were 28-day and 90-day all-cause mortality. MIMIC-IV database contains patient survival information, both within or outside the hospital. The eICU database contains patient in-hospital survival information. For the SHHMU cohort, survival information was obtained from electronic medical records system or by contacting patients’ relatives through reserved phone numbers. Follow-up started on the date of patient admission and ended at 28 days and 90 days after admission or upon patient death.
Continuous variables with a normal or skewed distribution are described as the means ± SD and medians (interquartile range), respectively. Categorical variables are described as numbers (percentages). Univariate and multivariate proportional hazards regression analyses considering LT as the competing risk were performed to identify the independent variables for 28-day mortality[12,24]. Variables with P values < 0.05 in the univariate analysis were transformed to their natural logarithms and considered for multivariable analysis. To evaluate the performance of prognostic scores in predicting mortality, discrimination, calibration and overall performance of each score were evaluated. Discrimination were evaluated by plotting receiver operating characteristic (ROC) curve and calculating the area under curve (AUC). Bootstraps with 1000 resamples were applied to correct the AUC and explain the variance due to overoptimism. Calibration was evaluated according to the P value in Hosmer-Lemeshow “goodness-of-fit” test and the visually observed concordance between predicted and actual survival proportions in calibration curves. Overall performance was assessed using the Brier score and R2 value. A lower Brier score or higher R2 value indicates a superior performance[25]. Clinical utility was determined by determining the net benefit in decision curve analysis (DCA)[26]. Cumulative survival curves were plotted using the Kaplan-Meier method and overall survival proportion were compared by log-rank test. Statistical analysis was performed using MedCalc statistical software version 19.0.4, STATA 15.0 and R version 4.2.0. A two-tailed P value < 0 .05 was considered to be statistically significant.
In the MIMIC cohort, 3256 consecutive cirrhotic patients with various forms of ADs were screened and 2005 patients were excluded for the following reasons: Lack of ACLF diagnosis (n = 900), incomplete records (n = 903), HCC or malignant tumors from other organs (n = 40), HIV infection or undergoing immunosuppressive therapy (n = 68) and death within 24 h of ICU admission (n = 94). Finally, 1251 cirrhotic patients complicated with ACLF were enrolled in this study. These patients were then randomly divided into a derivation cohort (n = 938) and an internal validation cohort (n = 313) at a ration of 3:1. In the eICU cohort, 1042 cirrhosis patients admitted due to various ADs were screened and 491 patients were excluded, resulting in a total of 551 cirrhotic patients with ACLF participating in the study. In the SHHMU cohort, 1729 consecutive cirrhotic patients with various forms of ADs admitted to the Department of Gastroenterology at SHHMU between August 1, 2021 and August 31, 2023 were screened, and 1608 patients were excluded, resulting in a total of 121 cirrhotic patients with ACLF participating in the study. The patients’ baseline characteristics in the three cohorts are shown in Table 1.
| Baseline characteristics | MIMIC cohort | External validation cohorts | ||
| Derivation set (n = 938) | Validation set (n = 313) | eICU (n = 551) | SHHMU (n = 121) | |
| Age (yr) | 59 (52-68) | 56 (53-66) | 57 (50-64) | 58 ± 13 |
| Male, n (%) | 617 (65.8) | 202 (64.5) | 334 (60.6) | 81 (66.9) |
| Race, n (%) | ||||
| White | 613 (65.4)a | 209 (66.8) | 448 (81.3) | Eastern Asian |
| Non white | 325 (34.6)a | 104 (33.2) | 103 (18.7) | Eastern Asian |
| Etiology of cirrhosis, n (%) | ||||
| Alcohol | 524 (55.9)b | 180 (57.5) | 376 (68.2) | 43 (35.5) |
| Non-alcohol | 414 (44.1)b | 133 (42.5) | 175 (31.8) | 78 (64.5) |
| AD at admission, n (%) | ||||
| Ascites | 520 (55.4)a,b | 173 (55.3) | 169 (30.7) | 32 (26.4) |
| Hepatic encephalopathy | 390 (41.6)b | 136 (43.5) | 156 (28.3) | 31 (25.6) |
| Bacterial infection | 422 (45.0)b | 152 (48.6) | 216 (39.2) | 16 (13.2) |
| Acute variceal bleeding | 75 (8.0)a,b | 34 (10.9) | 127 (23.0) | 42 (34.7) |
| Laboratory indicators | ||||
| WBC (109/L) | 12 (8-18)b | 13 (9-18) | 10 (6-14) | 6 (5-11) |
| Hemoglobin | 85 (73-100)b | 83 (72-98) | 88 (75-103) | 92 (77-107) |
| Platelet (109/L) | 81 (50-135) | 72 (49-120) | 78 (49-126) | 75 (48-116) |
| Total bilirubin (mg/dL) | 4.1 (1.6-11.7) | 4.9 (1.8-12.3) | 4.0 (1.5-12.9) | 6.2 (1.6-17.5) |
| Albumin (g/dL) | 2.9 (2.4-3.3) | 2.9 (2.4-3.4) | 2.8 (2.2-3.3) | 2.9 ± 0.6 |
| ALT (U/L) | 39 (22-98)b | 39 (23-101) | 36 (21-97) | 34 (16-64) |
| AST (U/L) | 83 (44-216)b | 94 (47-266) | 91 (43-253) | 62 (35-138) |
| INR | 2.1 (1.6-2.8)a,b | 2.0 (1.6-2.7) | 1.7 (1.4-2.4) | 1.7 (1.2-2.6) |
| Prothrombin time (s) | 22 (17-30)b | 22 (17-29) | 21 (16-28) | 19 (14-29) |
| Serum creatinine (mg/dL) | 2.1 (1.3-3.4) | 2.0 (1.2-3.2) | 2.0 (1.1-3.2) | 2.1 (1.1-3.0) |
| Serum sodium (mEq/L) | 135 (131-139) | 135 (131-139) | 137 (133-142) | 134 ± 7 |
| Serum potassium (mEq/L) | 4.6 (4.1-5.3)a,b | 4.6 (4.2-5.4) | 4.0 (3.6-4.6) | 4.2 (3.7-4.7) |
| Organ failure, n (%) | ||||
| Circulatory failure | 382 (40.7)a,b | 148 (47.3) | 305 (55.4) | 18 (14.9) |
| Respiratory failure | 427 (45.5)a,b | 156 (49.8) | 162 (29.4) | 22 (18.2) |
| Cerebral failure | 129 (13.8)a,b | 46 (14.7) | 223 (40.5) | 38 (31.4) |
| Renal failure | 658 (70.1)a,b | 201 (64.2) | 280 (50.8) | 75 (62.0) |
| Coagulation failure | 332 (35.4)b | 106 (33.9) | 131 (23.8) | 37 (30.6) |
| Liver failure | 233 (24.8)a,b | 82 (26.2) | 273 (49.5) | 54 (44.6) |
| Prognostic scores at admission | ||||
| MELD | 28 (22-34)a,b | 27 (21-34) | 25 ± 10 | 25 (18-31) |
| MELD-Na | 31 (23-40)b | 31 (22-40) | 27 (19-35) | 30 (21-38) |
| CLIF-C ACLF | 58 (51-64)b | 58 (52-64) | 50 (45-57) | 47 (42-54) |
| Liver transplantation, n (%) | 44 (4.7) | 11 (3.5) | 24 (4.4) | 2 (1.7) |
| ACLF grade, n (%) | ||||
| Grade 1 | 266 (28.4)a | 86 (27.5) | 263 (46.3) | 30 (33.1) |
| Grade 2 | 317 (33.8) | 105 (33.5) | 177 (31.1) | 50 (41.3) |
| Grade 3 | 355 (37.8)a | 122 (39.0) | 128 (22.6) | 41 (33.9) |
| 28-day mortality, n (%) | 376 (40.1) | 122 (39.0) | -- | 46 (38.0) |
| 90-day mortality, n (%) | 465 (49.6) | 144 (46.0) | -- | 59 (48.7) |
| In-ICU mortality, n (%) | -- | -- | 112 (20.3) | -- |
| In-hospital mortality, n (%) | -- | -- | 179 (32.4) | -- |
In the derivation set of MIMIC cohort, significant variables in univariate analysis included each OF defined by CLIF-OF score, ln (age), ln [mean arterial pressure (MAP)], ln (WBC count), ln (hemoglobin), ln (platelet count), ln (TB), ln [aspartate aminotransferase (AST)], ln (INR), ln (sodium level) and ln (potassium level), see Table 2. As the OFs criteria in CLIF-OF score determining CLIF-C ACLF with the maximum weight were proposed to have some limitations in clinical application, such as a ceiling effect of serum bilirubin and INR, and subjective variability in determining use of vasopressors, we used ln (TB), ln (INR) and ln (MAP) to reflect the state of liver, coagulation system and circulatory system, respectively. As renal replacement therapy was also a significant criterion in defining renal failure, we remained the renal failure criterion in CLIF-OF score. Respiratory and cerebral failure were defined according to CLIF-OF criteria. In addition, significant variables not indicating OF in the univariate analysis were also included in the multivariate analysis.
| Variable | Univariate analysis | Multivariate analysis | ||
| HR (95%CI) | P value | HR (95%CI) | P value | |
| Female | 1.08 (0.88-1.33) | 0.47 | ||
| Ln (age) | 2.18 (1.32-3.61) | 0.002 | 9.14 (5.16-16.19) | < 0.001 |
| Race (Non-white) | 1.15 (0.93-1.41) | 0.20 | ||
| Cirrhosis etiology (non-alcohol) | 1.00 (0.82-1.23) | 0.998 | ||
| Ln (MAP) | 0.03 (0.01-0.06) | < 0.001 | 0.06 (0.03-0.14) | < 0.001 |
| Organ failures | ||||
| Circulatory failure | 1.87 (1.53-2.29) | < 0.001 | ||
| Respiratory failure | 1.56 (1.27-1.93) | < 0.001 | 1.84 (1.48-2.29) | < 0.001 |
| Cerebral failure | 1.63 (1.26-2.11) | < 0.001 | 1.36 (1.05-1.77) | 0.02 |
| Renal failure | 1.89 (1.53-2.34) | < 0.001 | 1.68 (1.36-2.09) | < 0.001 |
| Liver failure | 2.31 (1.67-2.50) | < 0.001 | ||
| Coagulation failure | 2.04 (1.87-2.84) | < 0.001 | ||
| Laboratory indicators | ||||
| Ln (WBC) | 1.58 (1.33-1.88) | < 0.001 | 1.25 (1.05-2.10) | 0.01 |
| Ln (hemoglobin) | 0.65 (0.44-0.97) | 0.04 | ||
| Ln (platelet) | 0.84 (0.73-0.97) | 0.02 | ||
| Ln (TB) | 1.50 (1.38-1.63) | < 0.001 | 1.57 (1.42-1.73) | < 0.001 |
| Ln (Albumin) | 0.68 (0.45-1.04) | 0.07 | ||
| Ln (ALT) | 1.04 (0.97-1.12) | 0.29 | ||
| Ln (AST) | 1.09 (1.02-1.17) | 0.01 | ||
| Ln (INR) | 1.27 (1.20-1.33) | < 0.001 | 1.82 (1.43-2.32) | < 0.001 |
| Ln (sodium) | 0.04 (0.01-0.23) | < 0.001 | ||
| Ln (potassium) | 2.05 (1.23-3.41) | 0.006 | 1.84 (1.10-3.07) | 0.02 |
Results of the multivariate analysis were as follows: ln (age) (HR: 9.14, 95%CI: 5.16-16.19, P < 0.001), ln (WBC) (HR: 1.25, 95%CI: 1.05-2.10, P = 0.01), ln (TB) (HR: 1.57, 95%CI: 1.42-1.73, P < 0.001), ln (INR) (HR: 1.82, 95%CI: 1.43-2.32, P < 0.001), ln (potassium) (HR: 1.84, 95%CI: 1.10-3.07, P = 0.02), ln (MAP) (HR: 0.06, 95%CI: 0.03-0.14, P < 0.001), respiratory failure (HR: 2.74, 95%CI: 1.54-4.88, P < 0.001), cerebral failure (HR: 1.36, 95%CI: 1.05-1.77, P = 0.02) and renal failure (HR: 1.68, 95%CI: 1.36-2.09, P < 0.001) remained in the final regression model (Figure 1).
Based on the coefficients of the above mentioned independent variables in the final regression model, the following new prognostic model for ACLF patients hospitalized in the ICU, MIMIC ACLF, was developed: MIMIC ACLF = ln (age) × 2.2 + presence of renal failure × 0.51 – ln (MAP) × 2.7 + presence of respiratory failure × 0.6 + presence of cerebral failure × 0.31 + ln (WBC) × 0.22 + ln (TB) × 0.44 + ln (INR) × 0.59 + ln (potassium) × 0.59. The probability of death at time “t” can be estimated as P = 1 - e-CI(t) × exp [β(t) × MIMIC ACLF score]. CI(t) and β(t) are the cumulative baseline hazard and score coefficient estimated by the model fitted for time “t”. At the main time, they are as follows: CI (28) = 0.281, β (28) = 0.9987; CI (90) = 0.443, β (90) = 0.90.
Discrimination: To determine the discrimination ability of MIMIC ACLF, we plotted the ROC curve and calculated the AUC. The AUC of MIMIC ACLF for the 28/90-day mortality in patients with ACLF were 0.81/0.79, which were significantly greater than those of CLIF-C ACLF (0.76/0.74), MELD (0.73/0.71) and MELD-Na (0.72/0.70; all P < 0.001; Figure 2).
Calibration: To determine the calibration ability of MIMIC ACLF, we performed Hosmer-Lemeshow test and plotted the calibration curves. The actual and predicted 28/90-day survival probabilities were highly consistent across the deciles of MIMIC ACLF score (Hosmer-Lemeshow test X2 = 5.26/8.46, P = 0.73/0.39; Brier score = 0.172/0.188; R2 = 0.364/0.315; Figure 3; Table 3). Calibration performance of CLIF-C ACLF were slightly inferior to that of MIMIC ACLF (Figure 3). MELD and MELD-Na, especially the latter obviously underestimated the 28-day and 90-day mortality in ACLF patients (Figure 3).
| Scores | Youden index | Cutoff value | SEN (%) | SPE (%) | PPV (%) | NPV (%) | Brier score | R2 value | P in H-L test |
| Derivation set in MIMIC cohort | Development of MIMIC ACLF score | ||||||||
| MIMIC ACLF | 0.48 | 0.57 | 71.81 | 75.98 | 66.7 | 80.1 | 0.17 | 0.36 | 0.73 |
| CLIF-C ACLF | 0.40 | 59 | 65.43 | 74.56 | 63.2 | 76.3 | 0.19 | 0.26 | 0.70 |
| MELD | 0.38 | 27 | 73.14 | 65.12 | 58.4 | 78.4 | 0.20 | 0.20 | 0.04 |
| MELD-Na | 0.36 | 29 | 75.53 | 59.96 | 55.8 | 78.6 | 0.21 | 0.18 | 0.11 |
| Validation set in MIMIC cohort | Internal validation | ||||||||
| MIMIC ACLF | 0.46 | 0.32 | 81.97 | 64.40 | 69.5 | 84.8 | 0.17 | 0.36 | 0.26 |
| CLIF-C ACLF | 0.40 | 61 | 55.74 | 84.29 | 69.4 | 74.9 | 0.19 | 0.26 | 0.92 |
| MELD | 0.37 | 27 | 72.13 | 64.40 | 56.4 | 78.3 | 0.20 | 0.21 | 0.67 |
| MELD-Na | 0.35 | 30 | 72.95 | 62.30 | 55.3 | 78.3 | 0.20 | 0.23 | 0.09 |
| SHHMU cohort | External validation | ||||||||
| MIMIC ACLF | 0.49 | 1.51 | 65.22 | 84.00 | 71.4 | 79.7 | 0.17 | 0.34 | 0.84 |
| CLIF-C ACLF | 0.54 | 49 | 71.74 | 82.67 | 71.7 | 82.7 | 0.17 | 0.35 | 0.14 |
| MELD | 0.30 | 33 | 36.96 | 93.33 | 77.3 | 70.7 | 0.21 | 0.13 | 0.08 |
| MELD-Na | 0.44 | 34 | 60.87 | 82.67 | 68.3 | 77.5 | 0.20 | 0.22 | 0.52 |
To determine the clinical utility of MIMIC ACLF, we performed the decision curves analysis. The net benefit with MIMIC ACLF was greater than that indicated by the MELD, MELD-Na and CLIF-C ACLF in the entire range of the threshold probability (Figure 4), which indicated that MIMIC ACLF had superior clinical utility to the three existing prognostic scores.
To stratify patients with ACLF according to MIMIC ACLF score, we performed survival analysis on enrolled patients based on the two optimal cutoff values (0.44 and 1.55) of MIMIC ACLF score obtained through X-tile software (version 3.6.1; Yale University, United States). The results showed that a total of 90 (18.7%)/137 (28.4%), 132 (53.9%)/208 (65.2%) and 114 (83.2%)/120 (87.6%) patients among those with new score of ≤ 0.44, 0.45-1.54 and ≥ 1.55 died during a 28-day/90-day follow-up period, respectively (P value in log-rank test < 0.0001; Figure 5). Patients with MIMIC ACLF score of ≤ 0.44, 0.45-1.54 and ≥ 1.55 were thus stratified into low, moderate and high risk of death groups, respectively.
Internal validation: In the internal validation cohort, the AUC of MIMIC ACLF for 28-day mortality (0.80) was significantly greater than those of CLIF-C ACLF (0.75, P = 0.01), MELD (0.74, P = 0.02) and MELD-Na (0.75, P = 0.05; Figure 6A). With regard to 90-day mortality, the AUC of MIMIC ACLF (0.76) was significantly or numerically greater than that of CLIF-C ACLF (0.71, P = 0.02), MELD (0.71, P = 0.07) and MELD-Na (0.72, P = 0.17; Figure 6B).
The actual and predicted 28-day survival probability were highly consistent across the sextile of MIMIC ACLF score (Hosmer-Lemeshow test X2 = 5.26 P = 0.73; Brier score = 0.172; R2 = 0.364; Figure 7A-D; Table 3). The calibration performances of MELD, MELD-Na and CLIF-C ACLF were inferior to that indicated by MIMIC ACLF (Figure 7). With regard to the 90-day mortality, the performances of all prognostic scores were similar and satisfactory (Figure 7E-H; Table 3).
DCA showed that the net benefit with MIMIC ACLF was greater than that indicated by MELD, MELD-Na and CLIF-C ACLF in almost the entire range of threshold probabilities (Figure 8).
Survival analysis revealed that a total of 33 (20.4%)/47 (29.0%), 56 (49.6%)/62 (54.9%) and 33 (86.8%)/35 (92.1%) patients in the low, moderate and high-risk groups died during the 28-day/90-day follow-up period, respectively (P value in log-rank test < 0.0001; Figure 9).
SHHMU cohort: In the SHHMU cohort, the AUC of MIMIC ACLF for 28/90 day mortality (0.80/0.78) was comparable with that of CLIF-C ACLF (0.82/0.80, P > 0.05) but significantly or numerically greater than that of MELD (0.65/0.67, P < 0.05) and MELD-Na (0.74/0.73, P > 0.05) (Figure 10A and B). The calibration curve showed that the calibration ability of MIMIC ACLF for 28-day mortality was superior to that indicated by MELD, MELD-Na and CLIF-C ACLF; the calibration ability of MIMIC ACLF for 90-day mortality was similar with that of MELD but superior to that indicated by MELD-Na and CLIF-C ACLF (Figure 10C and D; Table 3). DCA showed that the net benefit of MIMIC ACLF was comparable with that of CLIF-C ACLF but greater than indicated by MELD and MELD-Na (Figure 10E and F). Survival analysis showed that a total of 33 (20.4%)/47 (29.0%), 56 (49.6%)/62 (54.9%) and 33 (86.8%)/35 (92.1%) patients among those stratified into low, moderate and high risk groups died during a 28-day/90-day follow-up period, respectively (P value in log-rank test < 0.0001; Figure 10G and H).
eICU cohort: In the eICU cohort, the AUC (0.75) of MIMIC ACLF for in-ICU mortality was equal to CLIF-C ACLF (0.75, P > 0.05), but significantly or numerically greater than that indicated by MELD-Na (0.67, P = 0.006), MELD (0.70, P > 0.05) and APACHE (0.71, P > 0.05; Figure 11A). The AUC (0.76) of MIMIC ACLF for in-hospital mortality was comparable to CLIF-C ACLF (0.77, P > 0.05), significantly or numerically greater than that indicated by MELD (0.65/0.67, P < 0.05), MELD-Na, and APACHE (0.74/0.73, P > 0.05; Figure 11B). The calibration curve showed that the calibration ability of MIMIC ACLF for in-ICU mortality is similar to MELD and CLIF-C ACLF, but better than MELD-Na and APACHE (Figure 12A-E); The calibration ability of MIMIC ACLF for in-hospital mortality is similar to MELD, MELD-Na, APACHE, and CLIF-C ACLF (Figure 12F-J). The overall performance of MIMIC ACLF in predicting ICU mortality in ACLF patients (Brier score = 0.14, R2 value = 0.21) is superior to CLIF-C ACLF (Brier score = 0.14, R2 value = 0.20), MELD (Brier score = 0.15, R2 value = 0.13), MELD-Na (Brier score = 0.15, R2 value = 0.09), and APACHE (Brier score = 0.15, R2 value = 0.14; Figure 12A-E); The overall performance of MIMIC ACLF in predicting in-hospital mortality (Brier score = 0.18, R2 value = 0.26) was slightly lower than CLIF-C ACLF (Brier score = 0.17, R2 value = 0.28), but better than MELD (Brier score = 0.19, R2 value = 0.18), MELD-Na (Brier score = 0.20, R2 value = 0.14), and APACHE (Brier score = 0.20, R2 value = 0.15; Figure 12F-J). DCA showed that the net benefit of MIMIC ACLF for ICU mortality is comparable to CLIF-C ACLF, but greater than that indicated by MELD, MELD-Na, and APACHE (Figure 13A). The net benefit of MIMIC ACLF for in-hospital mortality is slightly lower than CLIF-C ACLF, but greater than that indicated by MELD, MELD-Na, and APACHE (Figure 13B).
To facilitate the clinical application of MIMIC ACLF prognostic model, we constructed an online calculator based on MIMIC ACLF model (https://www.scidb.cn/en/anonymous/anllWU52). After the easily available variables (i.e., age, WBC, serum potassium, TB, INR, MAP, respiratory failure, cerebral failure and renal failure) are input into the MIMIC ACLF online calculator, predicted individual 28/90-day mortality in ACLF patients can be easily obtained (Figure 14).
ACLF is the most severe stage of chronic liver disease with single/multiple OF (s) and high short-term mortality. Patients with ACLF usually need to be hospitalized in the ICU for monitoring of their condition and organ support therapy. Reasonable treatment for ALCF should be specified based on the corresponding prognosis. Therefore, an accurate and robust prognostic score is very important for decision-making by clinicians in ICU. At present, existing prognostic scores used in ICU mainly include CTP, MELD and MELD-Na and CLIF-C ACLF. However, these scores either do not reflect the impact of multiple OFs for ACLF or have some limitations, resulting in a limited predictive accuracy. In this study, we developed a new prognostic model (MIMIC ACLF) using medical data in MIMIC database for patients with ACLF. MIMIC ACLF not only exhibited a satisfactory predictive accuracy and clinical utility in MIMIC cohort, but also performed well in the external single and multiple-center validation cohorts. In addition, we constructed an online calculator based on MIMIC ACLF to facilitate the its application in clinical practice.
Compared with results of recent studies, the short-term mortality in ACLF patients included in this study was highly consistent. In the derivation cohort, 28/90-day mortality in all patients with ACLF, ACLF grade 1, ACLF grade 2 and ACLF grade 3 were 40.1%/49.6%, 20.7%/33.1%, 31.5%/40.1% and 62.3%/70.4%, respectively. This is highly consistent with the result in a recent large sample observational study including 867 European and Canadian ACLF patients hospitalized in ICU where the 28/90-day mortality in all patients with ACLF, ACLF grade 1, ACLF grade 2 and ACLF grade 3 were 43.9%/54.0%, 22.0%/33.0%, 30.0%/40.0% and 64.0%/74.0%, respectively[10]. Compared to the mortality in another recent large sample retrospective cohort study in United States[27] where the 28-day/90-day mortality in all ACLF, ACLF grade 1, grade 2 and ACLF grade 3 were 25.5%/40.0%, 16.9%/30.8%, 26.8%/41.6% and 53.3%/68.8%, respectively, our mortality seems somewhat higher. This might be due to the fact that they included hospitalized patients with ACLF in both ICU and general wards while we only included those hospitalized in ICU. In summary, the diagnosis/grade of ACLF which plays an important role in determining the mortality in ACLF patients was accurate enough to lay a foundation for the subsequent analysis in this study.
Although the MIMIC ACLF prognostic model was developed based on the CLIF-C ACLF model, it overcomes the limitations of the CLIF-C ACLF model. There are 9 variables that constitute MIMIC ACLF, namely age, WBC count, serum potassium, TB, INR, MAP, respiratory failure, cerebral failure and renal failure. The natural logarithms of TB, INR and MAP reflect the weight of liver failure, coagulation failure and circulatory failure, respectively, on the short-term mortality of patients with ACLF. In this way, the proposed shortcomings of CLIF-C ACLF, such as ceiling effect of TB and INR, subjective variability in determining use of vasopressors can be avoided. In addition, the presence/absence of cerebral failure, renal failure and respiration failure as defined by the CLIF-OF criterion in CLIF-C ACLF was also included in MIMIC ACLF. In this way, the limitation of difficulty in distinguishing HE grade 1 and grade 0 in CLIF-C ACLF can also be avoided because it is relatively simple to distinguish between presence and absence of cerebral failure (HE grade 3 or 4 vs HE grade 0). Similar to the CANONIC study, age and WBC were significantly and positively associated with the mortality in patients with ACLF, indicating that ACLF is a distinct syndrome highly associated with systemic inflammation[2]. A new finding is that the natural logarithm of serum potassium was identified as an independent risk factor for the 28-day mortality in ACLF patients (HR = 1.84, P = 0.02). This finding coincided with that of a recent study in which presence of hyperkalemia during admission was independently associated with the 90-day mortality in ACLF patients (HR = 2.4, P < 0.001)[28]. To date, data on the impact of serum potassium on the prognosis of ACLF patients are scarce. We assume that hyperkalemia might represent a more severe condition involving intensive systemic inflammation, renal failure, liver failure and sarcopenia[26]. However, the pathophysiological association between hyperkalemia and poor prognosis of ACLF patients remains to be elucidated.
An ideal prognostic score should perform well in the following aspects: discrimination, calibration and DCA. Discrimination refers to the ability to stratify patients according to their risk of developing the outcome. Calibration refers to the ability to predict absolute risks (how closely the predicted probabilities agree with the actual outcomes). The DCA is a suitable analytical method evaluating which prognostic score benefits patients most and has the best clinical utility. In the MIMIC cohort, whether in the derivation or validation set, MIIMC ACLF performed well and outperformed existing prognostic scores, including CLIF-C ACLF, in the assessment of discrimination, calibration and clinical utility for 28/90 d mortality in ACLF patients. MELD-Na markedly underestimated the 28/90-day mortality in ACLF patients, which was consistent with the results in recent large sample studies[9,29]. According to the multiple-center external validation using data from the eICU cohort, the predictive accuracy of MIMIC ACLF for the in-ICU and in-hospital mortality was comparable to that of CLIF-C ACLF but superior to that of traditional scores, such as MELD, MELD-Na and APACHE. In addition, we also validated the prediction performance of MIMIC ACLF in SHHMU, a Chinese ACLF cohort in which the hospitalization location, demographic characteristics and cirrhosis etiology differed from those in western cohorts. Results were that MIMIC ACLF combined with CLIF-C ACLF performed well, outperforming CTP, MELD and MELD-Na in the prediction of 28/90 mortality. In conclusion, MIMIC ACLF model showed a prediction accuracy comparable to that of CLIF-C ACLF score but outperformed other existing scores for short-term mortality in ACLF patients hospitalized both in the ICU and general ward environments.
Importantly, in this study, patients with ACLF were stratified according to the MIMIC ACLF score. Patients with MIMIC ACLF score of ≤ 0.44, 0.45-1.54 and ≥ 1.55 were respectively stratified into low, moderate and high risk of death groups. The 28/90-day mortality significantly increased from low (18.7%/28.4%) to moderate (53.9%/65.2%) to high risk groups (83.2%/87.6%; P <0.0001), which was confirmed in the internal and external validation cohorts. Compared with the risk stratification with CLIF-C ACLF score (28/90-day mortality in ACLF grade 1, grade 2 and grade 3: 20.7%/33.1%, 31.5%/40.1% and 62.3%/70.4%, respectively), a lower and higher mortality rate were respectively seen in our low and high risk groups than that in the equal risk grades of CLIF-C ACLF. In addition, we also developed an online calculator of MIMIC ACLF score to make the individual mortality prediction easier. This online calculator, combined with the risk grades, will be very useful in assisting clinicians in ICU to evaluate the prognosis of patients with ACLF. After risk stratification, patients in low and moderate risk groups can be considered for ongoing intensive treatment without hesitation in order to prevent the progression of ACLF, and meanwhile LT assessment should be performed as soon as possible[5,30]; for those in high risk group, if not the LT candidates, a careful consideration of transition to palliation should be discussed between patient families and ICU clinicians after a short treatment trial[31,32].
Our study has two limitations that warrant consideration. First, as this study was a retrospective cohort study, selection, information and confounding bias are inevitable. For example, diagnosis of covert HE (minimal HE and grade Ⅰ) might be partly affected by the subjective factors of observers; distinguishing between respiratory failure and cerebral failure became difficult when mechanical ventilation therapy was provided because it was hard to determine the exact indication for mechanical ventilation (respiratory failure, airway protection due to cerebral failure or both). Second, the sample size in the SHHMU cohort was relatively small. However, we are planning to conduct a multi-center and longer-term study to continue externally validating the MIMIC ACLF prognostic model.
In conclusion, we developed and validated a new prognostic model (MIMIC ACLF) for cirrhotic patients with ACLF hospitalized in ICU. MIMIC ACLF overcomes the shortcomings of existing prognostic scores and has demonstrated a satisfactory ability to predict the short-term mortality in patients with ACLF. The MIMIC ACLF based risk stratification and online calculator could be useful for assisting in risk evaluation and guiding rational management of patients with ACLF. More large-sample and prospective studies are needed to validate the predictive accuracy of this new model.
We would like to thank the researchers in Massachusetts Institute of Technology and medical staff in Beth Israel Deaconess Medical Center for their contribution to the MIMIC database. We would also like to thank the Philips eICU Research Institute and Philips Healthcare for their contribution to the eICU database and thank patients in the SHHMU cohort for their contribution to the SHHMU data.
| 1. | Ferstl P, Trebicka J. Acute Decompensation and Acute-on-Chronic Liver Failure. Clin Liver Dis. 2021;25:419-430. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 8] [Cited by in F6Publishing: 8] [Article Influence: 2.7] [Reference Citation Analysis (1)] |
| 2. | Moreau R, Jalan R, Gines P, Pavesi M, Angeli P, Cordoba J, Durand F, Gustot T, Saliba F, Domenicali M, Gerbes A, Wendon J, Alessandria C, Laleman W, Zeuzem S, Trebicka J, Bernardi M, Arroyo V; CANONIC Study Investigators of the EASL–CLIF Consortium. Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology. 2013;144:1426-1437, 1437.e1. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 1720] [Cited by in F6Publishing: 1894] [Article Influence: 172.2] [Reference Citation Analysis (5)] |
| 3. | Engelmann C, Clària J, Szabo G, Bosch J, Bernardi M. Pathophysiology of decompensated cirrhosis: Portal hypertension, circulatory dysfunction, inflammation, metabolism and mitochondrial dysfunction. J Hepatol. 2021;75 Suppl 1:S49-S66. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 49] [Cited by in F6Publishing: 132] [Article Influence: 44.0] [Reference Citation Analysis (1)] |
| 4. | Moreau R, Clària J, Aguilar F, Fenaille F, Lozano JJ, Junot C, Colsch B, Caraceni P, Trebicka J, Pavesi M, Alessandria C, Nevens F, Saliba F, Welzel TM, Albillos A, Gustot T, Fernández J, Moreno C, Baldassarre M, Zaccherini G, Piano S, Montagnese S, Vargas V, Genescà J, Solà E, Bernal W, Butin N, Hautbergue T, Cholet S, Castelli F, Jansen C, Steib C, Campion D, Mookerjee R, Rodríguez-Gandía M, Soriano G, Durand F, Benten D, Bañares R, Stauber RE, Gronbaek H, Coenraad MJ, Ginès P, Gerbes A, Jalan R, Bernardi M, Arroyo V, Angeli P; CANONIC Study Investigators of the EASL Clif Consortium; Grifols Chair; European Foundation for the Study of Chronic Liver Failure (EF Clif). Blood metabolomics uncovers inflammation-associated mitochondrial dysfunction as a potential mechanism underlying ACLF. J Hepatol. 2020;72:688-701. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 144] [Cited by in F6Publishing: 197] [Article Influence: 49.3] [Reference Citation Analysis (1)] |
| 5. | Schulz MS, Gu W, Schnitzbauer AA, Trebicka J. Liver Transplantation as a Cornerstone Treatment for Acute-On-Chronic Liver Failure. Transpl Int. 2022;35:10108. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 7] [Cited by in F6Publishing: 7] [Article Influence: 3.5] [Reference Citation Analysis (1)] |
| 6. | European Association for the Study of the Liver. EASL Clinical Practice Guidelines for the management of patients with decompensated cirrhosis. J Hepatol. 2018;69:406-460. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 1177] [Cited by in F6Publishing: 1498] [Article Influence: 249.7] [Reference Citation Analysis (2)] |
| 7. | Nanchal R, Subramanian R, Karvellas CJ, Hollenberg SM, Peppard WJ, Singbartl K, Truwit J, Al-Khafaji AH, Killian AJ, Alquraini M, Alshammari K, Alshamsi F, Belley-Cote E, Cartin-Ceba R, Dionne JC, Galusca DM, Huang DT, Hyzy RC, Junek M, Kandiah P, Kumar G, Morgan RL, Morris PE, Olson JC, Sieracki R, Steadman R, Taylor B, Alhazzani W. Guidelines for the Management of Adult Acute and Acute-on-Chronic Liver Failure in the ICU: Cardiovascular, Endocrine, Hematologic, Pulmonary, and Renal Considerations. Crit Care Med. 2020;48:e173-e191. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 33] [Cited by in F6Publishing: 36] [Article Influence: 9.0] [Reference Citation Analysis (1)] |
| 8. | Jalan R, Gustot T, Fernandez J, Bernal W. 'Equity' and 'Justice' for patients with acute-on chronic liver failure: A call to action. J Hepatol. 2021;75:1228-1235. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 11] [Cited by in F6Publishing: 11] [Article Influence: 3.7] [Reference Citation Analysis (1)] |
| 9. | Wu T, Li J, Shao L, Xin J, Jiang L, Zhou Q, Shi D, Jiang J, Sun S, Jin L, Ye P, Yang L, Lu Y, Li T, Huang J, Xu X, Chen J, Hao S, Chen Y, Xin S, Gao Z, Duan Z, Han T, Wang Y, Gan J, Feng T, Pan C, Li H, Huang Y, Xie Q, Lin S, Li L; Chinese Group on the Study of Severe Hepatitis B (COSSH). Development of diagnostic criteria and a prognostic score for hepatitis B virus-related acute-on-chronic liver failure. Gut. 2018;67:2181-2191. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 167] [Cited by in F6Publishing: 236] [Article Influence: 39.3] [Reference Citation Analysis (2)] |
| 10. | Sundaram V, Jalan R, Wu T, Volk ML, Asrani SK, Klein AS, Wong RJ. Factors Associated with Survival of Patients With Severe Acute-On-Chronic Liver Failure Before and After Liver Transplantation. Gastroenterology. 2019;156:1381-1391.e3. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 154] [Cited by in F6Publishing: 206] [Article Influence: 41.2] [Reference Citation Analysis (1)] |
| 11. | Karvellas CJ, Garcia-Lopez E, Fernandez J, Saliba F, Sy E, Jalan R, Pavesi M, Gustot T, Ronco JJ, Arroyo V; Chronic Liver Failure Consortium and European Foundation for the Study of Chronic Liver Failure. Dynamic Prognostication in Critically Ill Cirrhotic Patients With Multiorgan Failure in ICUs in Europe and North America: A Multicenter Analysis. Crit Care Med. 2018;46:1783-1791. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 26] [Cited by in F6Publishing: 39] [Article Influence: 7.8] [Reference Citation Analysis (1)] |
| 12. | Jalan R, Saliba F, Pavesi M, Amoros A, Moreau R, Ginès P, Levesque E, Durand F, Angeli P, Caraceni P, Hopf C, Alessandria C, Rodriguez E, Solis-Muñoz P, Laleman W, Trebicka J, Zeuzem S, Gustot T, Mookerjee R, Elkrief L, Soriano G, Cordoba J, Morando F, Gerbes A, Agarwal B, Samuel D, Bernardi M, Arroyo V; CANONIC study investigators of the EASL-CLIF Consortium. Development and validation of a prognostic score to predict mortality in patients with acute-on-chronic liver failure. J Hepatol. 2014;61:1038-1047. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 557] [Cited by in F6Publishing: 639] [Article Influence: 63.9] [Reference Citation Analysis (1)] |
| 13. | Engelmann C, Thomsen KL, Zakeri N, Sheikh M, Agarwal B, Jalan R, Mookerjee RP. Validation of CLIF-C ACLF score to define a threshold for futility of intensive care support for patients with acute-on-chronic liver failure. Crit Care. 2018;22:254. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 55] [Cited by in F6Publishing: 99] [Article Influence: 16.5] [Reference Citation Analysis (1)] |
| 14. | Barosa R, Roque Ramos L, Patita M, Nunes G, Fonseca J. CLIF-C ACLF score is a better mortality predictor than MELD, MELD-Na and CTP in patients with Acute on chronic liver failure admitted to the ward. Rev Esp Enferm Dig. 2017;109:399-405. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 23] [Cited by in F6Publishing: 26] [Article Influence: 3.7] [Reference Citation Analysis (0)] |
| 15. | Costa E Silva PP, Codes L, Rios FF, Esteve CP, Valverde Filho MT, Lima DOC, de Almeida Filho GF, Morais MCA, Lima BC, Chagas PBO, Boa-Sorte N, Bittencourt PL. Comparison of General and Liver-Specific Prognostic Scores in Their Ability to Predict Mortality in Cirrhotic Patients Admitted to the Intensive Care Unit. Can J Gastroenterol Hepatol. 2021;2021:9953106. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 2] [Cited by in F6Publishing: 8] [Article Influence: 2.7] [Reference Citation Analysis (1)] |
| 16. | Jahn M, Raschidi L, Özçürümez MK, Arzideh F, Korth J, Kribben A, Canbay A, Katsounas A. Comparison of Mortality Prediction Scores in Intermediate-Care Patients with Liver Cirrhosis at a German University Transplant Centre: A Prospective Study. Dig Dis. 2023;41:96-106. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 1] [Cited by in F6Publishing: 1] [Article Influence: 0.5] [Reference Citation Analysis (1)] |
| 17. | Gustot T, Fernandez J, Garcia E, Morando F, Caraceni P, Alessandria C, Laleman W, Trebicka J, Elkrief L, Hopf C, Solís-Munoz P, Saliba F, Zeuzem S, Albillos A, Benten D, Montero-Alvarez JL, Chivas MT, Concepción M, Córdoba J, McCormick A, Stauber R, Vogel W, de Gottardi A, Welzel TM, Domenicali M, Risso A, Wendon J, Deulofeu C, Angeli P, Durand F, Pavesi M, Gerbes A, Jalan R, Moreau R, Ginés P, Bernardi M, Arroyo V; CANONIC Study Investigators of the EASL-CLIF Consortium. Clinical Course of acute-on-chronic liver failure syndrome and effects on prognosis. Hepatology. 2015;62:243-252. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 372] [Cited by in F6Publishing: 425] [Article Influence: 47.2] [Reference Citation Analysis (1)] |
| 18. | Moreau R, Gao B, Papp M, Bañares R, Kamath PS. Acute-on-chronic liver failure: A distinct clinical syndrome. J Hepatol. 2021;75 Suppl 1:S27-S35. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 22] [Cited by in F6Publishing: 52] [Article Influence: 17.3] [Reference Citation Analysis (1)] |
| 19. | Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, Pollard TJ, Hao S, Moody B, Gow B, Lehman LH, Celi LA, Mark RG. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10:1. [PubMed] [DOI] [Cited in This Article: ] [Cited by in F6Publishing: 203] [Reference Citation Analysis (1)] |
| 20. | Goldberger AL, Amaral LA, Glass L, Hausdorff JM, Ivanov PC, Mark RG, Mietus JE, Moody GB, Peng CK, Stanley HE. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation. 2000;101:E215-E220. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 7935] [Cited by in F6Publishing: 4648] [Article Influence: 193.7] [Reference Citation Analysis (1)] |
| 21. | Pollard TJ, Johnson AEW, Raffa JD, Celi LA, Mark RG, Badawi O. The eICU Collaborative Research Database, a freely available multi-center database for critical care research. Sci Data. 2018;5:180178. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 657] [Cited by in F6Publishing: 532] [Article Influence: 88.7] [Reference Citation Analysis (1)] |
| 22. | Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, D'Amico G, Dickson ER, Kim WR. A model to predict survival in patients with end-stage liver disease. Hepatology. 2001;33:464-470. [PubMed] [DOI] [Cited in This Article: ] |
| 23. | Biggins SW, Kim WR, Terrault NA, Saab S, Balan V, Schiano T, Benson J, Therneau T, Kremers W, Wiesner R, Kamath P, Klintmalm G. Evidence-based incorporation of serum sodium concentration into MELD. Gastroenterology. 2006;130:1652-1660. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 495] [Cited by in F6Publishing: 509] [Article Influence: 28.3] [Reference Citation Analysis (1)] |
| 24. | Wolbers M, Koller MT, Witteman JC, Steyerberg EW. Prognostic models with competing risks: methods and application to coronary risk prediction. Epidemiology. 2009;20:555-561. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 366] [Cited by in F6Publishing: 440] [Article Influence: 29.3] [Reference Citation Analysis (1)] |
| 25. | Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, Pencina MJ, Kattan MW. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology. 2010;21:128-138. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 3195] [Cited by in F6Publishing: 3014] [Article Influence: 215.3] [Reference Citation Analysis (1)] |
| 26. | Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26:565-574. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 2176] [Cited by in F6Publishing: 2986] [Article Influence: 175.6] [Reference Citation Analysis (1)] |
| 27. | Hernaez R, Kramer JR, Liu Y, Tansel A, Natarajan Y, Hussain KB, Ginès P, Solà E, Moreau R, Gerbes A, El-Serag HB, Kanwal F. Prevalence and short-term mortality of acute-on-chronic liver failure: A national cohort study from the USA. J Hepatol. 2019;70:639-647. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 73] [Cited by in F6Publishing: 87] [Article Influence: 17.4] [Reference Citation Analysis (1)] |
| 28. | Mezzano G, Cárdenas A, Aguilar F, Pavesi M, Solé C, Napoleone L, Graupera I, Juanola A, Carol M, Pose E, Fabrellas N, Hernaez R, Martínez J, Saliba F, Arroyo V, Sola E, Gines P. Hyperkalemia influences the outcome of patients with cirrhosis with acute decompensation (AD) and acute-on-chronic liver failure (ACLF). Dig Liver Dis. 2021;53:738-745. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 3] [Cited by in F6Publishing: 3] [Article Influence: 1.0] [Reference Citation Analysis (1)] |
| 29. | Hernaez R, Liu Y, Kramer JR, Rana A, El-Serag HB, Kanwal F. Model for end-stage liver disease-sodium underestimates 90-day mortality risk in patients with acute-on-chronic liver failure. J Hepatol. 2020;73:1425-1433. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 62] [Cited by in F6Publishing: 74] [Article Influence: 18.5] [Reference Citation Analysis (1)] |
| 30. | Trebicka J, Sundaram V, Moreau R, Jalan R, Arroyo V. Liver Transplantation for Acute-on-Chronic Liver Failure: Science or Fiction? Liver Transpl. 2020;26:906-915. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 32] [Cited by in F6Publishing: 30] [Article Influence: 7.5] [Reference Citation Analysis (1)] |
| 31. | Arshad A, Irshad L, Nabavi T, Whitehouse T. Acute-on-chronic liver failure: to admit to intensive care or not? Br J Hosp Med (Lond). 2020;81:1-6. [PubMed] [DOI] [Cited in This Article: ] [Reference Citation Analysis (1)] |
| 32. | Dong V, Karvellas CJ. Acute-on-chronic liver failure: Objective admission and support criteria in the intensive care unit. JHEP Rep. 2019;1:44-52. [PubMed] [DOI] [Cited in This Article: ] [Cited by in Crossref: 9] [Cited by in F6Publishing: 21] [Article Influence: 4.2] [Reference Citation Analysis (1)] |