Xiao-Dong Zhou a *, Qin-Fen Chen b *, Ming-Chun Zhang a, Sven Van Poucke c

Transcription

1 EXPERT REVIEW OF GASTROENTEROLOGY & HEPATOLOGY, ORIGINAL RESEARCH Scoring model to predict outcome in critically ill cirrhotic patients with acute respiratory failure: comparison with MELD scoring models and CLIF-SOFA score Xiao-Dong Zhou a *, Qin-Fen Chen b *, Ming-Chun Zhang a, Sven Van Poucke c, Wen-Yue Liu d, Yao Lu e, Ke-Qing Shi f,g, Wei-Jian Huang a and Ming-Hua Zheng f,g a Department of Cardiovascular Medicine, the Heart Center, the First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China; b Department of Gastroenterology, the First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China; c Department of Anesthesiology, Intensive Care, Emergency Medicine and Pain Therapy, Ziekenhuis Oost-Limburg, Genk, Belgium; d Department of Endocrinology, the First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China; e Department of Respiratory and Critical Care Medicine, the First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China; f Department of Hepatology, Liver Research Center, the First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China; g Institute of Hepatology, Wenzhou Medical University, Wenzhou, China ABSTRACT Background: Critically ill cirrhotic patients have a high mortality, particularly with concomitant respiratory failure on admission. There are no specific models in use for mortality risk assessment in critically ill cirrhotic patients with acute respiratory failure (CICRF). The aim is to develop a risk prediction model specific to CICRF in order to quantify the severity of illness. Methods: We analyzed 949 CICRF patients extracted from the MIMIC-III database. The novel model (ARF-CLIF-SOFA) was developed from the CLIF-SOFA score. Cox regression analysis and AUROC were implemented to test the predictive accuracy, compared with existing scores including the CLIF-SOFA score and MELD-related scores. Results: ARF-CLIF-SOFA contains PaO 2 /FiO 2 ratio, lactate, MAP, vasopressor therapy, bilirubin and creatinine (1 point each; score range: 0 6). Based on our patient cohort, the ARF-CLIF-SOFA score had good predictive accuracy for predicting the 30-, 90-day and 1-year mortality (AUROC = at 30-day, at 90-day, at 1-year, respectively). Additionally, the performance of the ARF-CLIF-SOFA is superior to existing scores (all P < 0.001). Conclusion: The ARF-CLIF-SOFA score can be considered a CICRF specific score with a better predictive accuracy compared to the existing scores. ARTICLE HISTORY Received 1 February 2017 Accepted 2 June 2017 KEYWORDS Critically ill cirrhotic; acute respiratory failure; CLIF- SOFA; MELD; prognosis 1. Introduction Critically ill cirrhotic patients, at risk for acute metabolic decompensation and multi-organ system failure, are associated with high short-term and long-term mortality [1 4]. Acuterespiratoryfailure(ARF)canbepartofmultipleorgan system failure syndrome in critically ill cirrhotic patients with rapid deterioration of disease [5 7]. Although critically ill cirrhotic patients with acute respiratory failure (CICRF) have a better prognosis related to more advanced care, CICRF patients still have a high mortality [2,3,8]. The pathophysiological factors of ARF may attribute to the other multi-organ system failure associated with increased mortality [8,9]. Based on reduced hemoglobin transit time, diffusion-perfusion failure has been described in patients with cirrhosis [10,11].Cirrhosis is associated with hepatopulmonary syndrome and portopulmonary hypertension [11]. The existing scores have not been specifically developed for CICRF patients. Chronic liver failure-sequential organ failure assessment (CLIF-SOFA) is the most widely used score [12,13]. In this score, the respiratory component is evaluated by the SpO 2 /FiO 2 ratio and PaO 2 /FiO 2 ratio. The Glasgow coma scale (GCS) included in the CLIF-SOFA score is subjective and potentially biased by the physicians judgment. The Model for End-stage Liver Disease (MELD) score, which was created based on objective laboratory parameters, has been widely implemented to quantify the mortality risk of critically ill cirrhotic patients [14]. In the past decades, the MELD-Na score, the MELD to SOdium ratio (MESO) score and updated MELD score gradually gained popularity. However, these MELD scores only contain creatinine, bilirubin, and international normalized ratio (INR), with sodium and/or age variables [15 18]. As such, the severity of critically ill cirrhotic with ARF might be underestimated using MELD scoring systems. Therefore, it is imperative to develop a more specific score able to stratify CICRF patients with a higher prognostic accu- CONTACT Ming-Hua Zheng zhengmh@wmu.edu.cn Department of Hepatology, Liver Research Center, the First Affiliated Hospital of Wenzhou Medical University, No. 2 Fuxue Lane, Wenzhou , China; Wei-Jian Huang weijianhuang69@126.com Department of Cardiovascular Medicine, the Heart Center, the First Affiliated Hospital of Wenzhou Medical University, No. 2 Fuxue Lane, Wenzhou , China *Co-first authors: Xiao-Dong Zhou and Qin-Fen Chen Informa UK Limited, trading as Taylor & Francis Group

2 2 X.-D. ZHOU ET AL. racy than the existing CLIF-SOFA, MELD, MELD-Na, MESO, and updated MELD scores. 2. Methods 2.1. Study design A retrospective analysis of the Multi-parameter Intelligent Monitoring in Intensive Care (MIMIC-III) database, version 1.3 was performed. The MIMIC-III database is a freely accessible large-scale critical care database with over 50,000 medical records from patients admitted to the Intensive Care Unit (ICU) at Beth Israel Deaconess Medical Center (Boston, MA) during the period Data from the first day of ICU admission of all study patients contain clinical parameters and laboratory parameters recorded from the hospital s online information systems. Clinical parameters included heart rate, respiration, temperature, systolic blood pressure (SBP), and diastolic blood pressure (DBP), while laboratory parameters included glucose, white blood cell count, platelet count, sodium, potassium, BUN, PaO 2, PCO 2, FIO 2, bicarbonate, creatinine, lactate, INR, and bilirubin. The following relevant variables were included: age, sex, height, weight, ethnicity, vasopressor therapy, and the diagnosis of cirrhosis. CLIF-SOFA, MELD, MELD-Na, MESO, and updated MELD scores were calculated to quantify the severity of illness at the first day of ICU admission. The CLIF-SOFA score was calculated according to the published formulas [13]. Additionally, the following formulas were used to calculate MELD, MELD-Na MESO, and updated MELD scores based on the laboratory parameters recorded at the first day of admission. MELD : R ¼ 9:57 log e ðcreatinine ðmg=dlþþ þ3:78 log e ðbilirubinðmg=dlþþþ11:2 log e ðinrþþ6:43 ½14Š; (1) MELD Na : R¼ MELD þ 1:59ð135 Na ðmmol=lþþ ½16Š; (2) MESO : R¼ðMELD=Na ðmmol=lþþ100 ½17Š; (3) Updated MELD : R¼ 1:27 log e ð1 þ creatinine ðmg=dlþþ þ 0:94log e ð1 þ bilirubinðmg=dlþþþ1:66 log e ð1 þ INRÞ½18Š: (4) The observation period was initiated from ICU admission to 1- year follow-up. The primary end points were defined as 30-, 90-day, and 1-year all-cause mortality. Our permission to access the database was approved after completion of the NIH web-based training course named Protecting Human Research Participants (Our certification number: ) Diagnosis of critically ill cirrhotic patients and ARF Liver cirrhosis was defined when at least two of the following criteria were satisfied: 1. Ultrasonographic evidence of a smallsized liver with and without splenomegaly/ascites; 2. Hypoalbuminemia (serum albumin <35 g/l); 3. An aminotransferase to platelet ratio ( 10 9 /L) 100 of more than 2. Alcoholic cirrhosis of liver was considered with a daily alcohol consumption of more than 80 g/day for at least five years. Critically ill cirrhotic patients were defined as patients with cirrhosis admitted to ICU. ARF was defined with the presence of at least one of the following findings: (1) PaO 2 < 60 mmhg; (2) PaO 2 /FiO 2 ratio < 300; (3) respiratory acidosis: PaCO 2 45 mmhg with ph < 7.35; (4) ventilator support. Patients who meet the following criteria were excluded: past or current hepatocellular carcinoma; liver transplantation; other causes that might lead to respiratory failure (such as congenital vascular malformation, congenital chest wall deformity, empyrosis, etc.); pregnancy; intoxication; autoimmune diseases; or other identifiable potential causes of serious diseases in other organ systems (such as multiple myeloma, severe epilepsy, heart transplantation, etc.) Statistical analysis Statistical analysis was performed using SPSS version 23.0 software (IBM, Armonk, NY) and MedCalc version 12.7 (MedCalc Software, Ostend, Belgium). Descriptive statistics were represented at baseline using mean with standard derivations for continuous variables and frequencies (percentage) for categorical variables. Patient characteristics were compared between survivors and non-survivors. Student s -test was performed for comparisons of continuous baseline characteristics, while Chi-square test was performed for categorical values. Cox s proportional hazards regression were used to calculate hazard ratios (HRs) for relative risk of mortality calculation. Univariate logistic analyses were performed to determine the unadjusted association of clinical and laboratory parameters with the prognosis. All variables that found to be associated with mortality (p < 0.10) were included as candidate variables in a forward conditional stepwise logistic regression analysis to identify independent predictors for the prognosis of CICRF. Then we calculated area under the receiver operating characteristic (AUROC) curves to determine the optimal cut-off values that maximized the discriminative accuracy. Finally, based on the multiple logistic regression models and AUROC analysis, the ARF-CLIF-SOFA was created. Survivals rates were calculated using the Kaplan Meier method and different survival curves were compared using the log-rank test. To identify the performance of the novel scoring model, the AUROC curves were compared. Sensitivity and specificity were determined using the cut-off point with highest Youden index (sensitivity + specificity 1). p-values of less than 0.05 were considered statistically significant. 3. Results 3.1. Baseline characteristics of study population A total of 949 patients were enrolled with all-cause mortality of 53.8% (n = 511). During the 1-year follow-up period, 471 survivors (55.9 ± 10.2 years, 68.4% male) were identified in the study population along with 478 non-survivors (58.4 ± 12.0 years, 68.6% male) included. The Caucasian race was the most common ethnicity of patients (72.4%), followed by African American and

3 EXPERT REVIEW OF GASTROENTEROLOGY & HEPATOLOGY 3 others (6.7% and 20.2%). The majority of non-survivors had been graded higher scoring results, as reflected by scores of CLIF-SOFA (9.2 ± 3.1 vs ± 3.4, p < 0.001); MELD (15.3 ± 9.6 vs ± 11.5, p < 0.001); MELD-Na (16.8 ± 10.2 vs ± 11.2, p < 0.001); MESO (11.3 ± 7.4 vs ± 8.9, p <0.001),and updated MELD (3.8 ± 1.1 vs. 4.7 ± 1.4, p <0.001).Creatinine, lactate, and bilirubin were higher in non-survivors (creatinine: 1.4 ± 1.3 mg/dl vs. 2.0 ± 1.7 mg/dl, p < 0.001; lactate 2.4 ± 1.9 mg/dl vs. 3.7 ± 3.2 mg/dl, p < 0.001; bilirubin: 4.4 ± 6.9 mg/dl vs. 8.1 ± 10.0 mg/dl, p <0.001),whilePaO 2 / FiO 2 ratio and mean arterial pressure (MAP) were higher in survivors (MAP: 85.4 ± 16.9 mmhg vs ± 16.2 mmhg, p < 0.001); PaO 2 /FiO 2 ratio: ± mmhg vs ± mmhg, p < 0.001). Other details of baseline characteristics were shown in Table Predictors for 1-year mortality All variables were initially evaluated in a univariate model to assess the functional relation with survival. Parameters associated with 1-year mortality in the univariate analysis were age, temperature, SBP, DBP, MAP, vasopressor therapy, glucose, PLT, sodium, potassium, BUN, PaO 2, FiO 2, PaO 2 /FiO 2 ratio, lactate, bilirubin, creatinine, and INR. In the final multivariate model, PaO 2 /FiO 2 ratio (HR = 0.823, 95% CI: , p < 0.001), lactate (HR = 1.126, 95% CI: , p < 0.001), MAP (HR = 0.977, 95% CI: , p < 0.001), vasopressor therapy (HR = 2.217, 95% CI: , p < 0.001), bilirubin (HR = 1.035, 95% CI: , p < 0.001), and creatinine (HR = 1.155, 95% CI: , p < 0.001) were considered independent predictors of 1-year mortality (Table 2). Table 1. Characteristics of critically ill cirrhotic patients with acute respiratory failure on the first day of admission, stratified by survival. Variable Total (n = 949) Survivors (n = 471) Non-survivors (n = 478) p-value Demographic parameters Age, year 57.1 ± ± ± Sex, male no. (%) 650 (68.5%) 322 (68.4%) 328 (68.6%) Height, cm ± ± ± Weight, kg 82.5 ± ± ± Vital signs Heart rate 91.4 ± ± ± Respiratory rate 39.7 ± ± ± Temperature, C 36.6 ± ± ± 1.1 <0.001 SBP, mmhg ± ± ± 22.8 <0.001 DBP, mmhg 62.2 ± ± ± 15.7 <0.001 MAP, mmhg 81.0 ± ± ± 16.2 <0.001 Vasopressors therapy, no. (%) 510 (53.7%) 191 (40.6%) 319 (66.7%) <0.001 Renal replacement therapy 128 (13.5%) 90 (19.1%) 38 (8.0%) <0.001 Ethnicity Caucasian no. (%) 688 (72.5%) 364 (77.3%) 324 (67.8%) Africa American no. (%) 64 (6.7%) 27 (5.7%) 37(7.7%) Other no. (%) 197 (20.8%) 80 (17.0%) 117 (24.5%) Etiologies Alcoholic, n. (%) 468 (49.3%) 220 (46.7%) 248 (51.9%) Biliary, n. (%) 17 (1.8%) 10 (2.1%) 7 (1.5%) Non-alcoholic, n. (%) 464 (48.9%) 241 (51.2%) 223(46.7%) Laboratory parameters Glucose, mg/dl ± ± ± WBC, 10 9 /L 10.8 ± ± ± PLT, 10 9 /L ± ± ± 92.4 <0.001 Sodium, meq/l ± ± ± 7.1 <0.001 Potassium, meq/l 4.2 ± ± ± 1.0 <0.001 BUN, mg/dl 32.0 ± ± ± 26.8 <0.001 PH 7.3 ± ± ± 0.1 <0.001 PaO 2, mmhg ± ± ± <0.001 PCO 2, mmhg 38.6 ± ± ± FIO ± ± ± PaO 2 /FiO 2 ratio, mmhg ± ± ± <0.001 Bicarbonate, meq/l 22.6 ± ± ± 5.9 <0.001 Creatinine, mg/dl 1.7 ± ± ± 1.7 <0.001 INR 1.9 ± ± ± 3.1 <0.001 Lactate, mg/dl 3.1 ± ± ± 3.2 <0.001 Bilirubin, mg/dl 6.3 ± ± ± 10.0 <0.001 Respiratory acidosis, n. (%)* 146 (15.4%) 63(13.4%) 83(17.4%) ARF-CLIF-SOFA: acute respiratory failure-chronic liver failure-sequential organ failure assessment; CICRF: critically ill cirrhotic patients with acute respiratory failure; CLIF-SOFA: chronic liver failure-sequential organ failure assessment; DBP: diastolic blood pressure; FiO 2 : fraction of inspired oxygen; INR: international normalized ratio; MAP: mean arterial pressure; MELD: model for end-stage liver disease; MESO: model for end-stage liver disease to sodium ratio; PaO 2 : partial pressure of arterial oxygen; PCO 2 : partial pressure of arterial carbon dioxide; SBP: systolic blood pressure. *Respiratory acidosis: PaCO 2 45 mmhg with ph < 7.35.

4 4 X.-D. ZHOU ET AL. Table 2. Univariate and multivariate analysis of association between clinical/laboratory characteristics and 1-year mortality in critically ill cirrhotic patients with acute respiratory failure. Univariate analysis Multivariate analysis Variables HR 95% CI p HR 95% CI p Demographic parameters Age ( ) <0.001 Sex, male* ( ) Vital signs Heart rate ( ) Respiratory rate ( ) Temperature ( ) <0.001 SBP ( ) <0.001 DBP ( ) <0.001 MAP ( ) < ( ) <0.001 Vasopressors therapy* ( ) < ( ) <0.001 Laboratory parameters Glucose ( ) WBC ( ) PLT ( ) <0.001 Sodium ( ) <0.001 Potassium ( ) <0.001 BUN ( ) <0.001 PaO ( ) <0.001 PCO ( ) FiO ( ) <0.001 FiO 2 /PaO 2 ratio ( ) < ( ) <0.001 Lactate ( ) < ( ) <0.001 Bilirubin ( ) < ( ) <0.001 Creatinine ( ) < ( ) <0.001 INR ( ) <0.001 BUN: blood urea nitrogen; CICRF: critically ill cirrhotic patients with acute respiratory failure; DBP: diastolic blood pressure; INR: international normalized ratio; MAP: mean arterial pressure; PaO 2 : partial pressure of arterial oxygen; PCO 2 : partial pressure of arterial carbon dioxide; PLT: platelet; SBP: systolic blood pressure; WBC: white blood cells. *Dichotomous values Construction of ARF-CLIF-SOFA The following scoring model was generated by selecting the optimal cut-off value revealing the best discriminative accuracy. ARF-CLIF-SOFA contained PaO 2 /FiO 2 ratio of 195 mmhg or less, lactate of 2.7 mg/dl or more, MAP of 77.3 mmhg or less, vasopressor therapy, bilirubin of 5.7 mg/dl or more, and creatinine of 1.3 mg/dl or more (1 point each; score range, 0 6) (Table 3) PaO 2 /FiO 2 ratio, lactate, MAP, vasopressor therapy, bilirubin, and creatinine related to mortality The Kaplan Meier analysis demonstrated that survival of patients with PaO 2 /FiO 2 ratio 195 or lactate 2.7 mg/dl or vasopressor therapy or MAP 77.3 mmhg, bilirubin 5.7 mg/dl or Creatinine 5.7 mg/dl was significantly lower than that of patients without these characteristics (all p < 0.001) (Figure 1). Table 3. Acute respiratory failure-chronic liver failure-sequential organ failure assessment (ARF-CLIF-SOFA) score. Score. Variables 0 1 FiO 2 /PaO 2 ratio, mmhg > Lactate, mg/dl < MAP, mmhg > Vasopressors therapy No Yes Bilirubin, mg/dl < Creatinine, mg/dl < ARF-CLIF-SOFA: acute respiratory failure-chronic liver failure-sequential organ failure assessment; FiO 2 : fraction of inspired oxygen; MAP: mean arterial pressure; PaO 2 : partial pressure of arterial oxygen Performance analysis of ARF-CLIF-SOFA Figure 2 showed the performance analysis of discriminative accuracy of the ARF-CLIF-SOFA score for each period mortality, with AUROC of (95% CI: ) in predicting 30-day mortality and (95% CI: ) at 90-day mortality. Compared to CLIF-SOFA, MELD, MELD-Na, MESO, and updated MELD scores, the presented ARF-CLIF-SOFA score indicated a 4 6% improvement in discriminative accuracy (all p <0.001). In the same dataset, an analysis of AUROC at 1-year period showed that ARF-CLIF-SOFA was also superior to CLIF-SOFA and MELD-related scores. Furthermore, all existing MELDrelated scores failed to predict 1-year mortality (AUROC < 0.7). The goodness-of-fit χ 2 and AUROC were: ARF-CLIF-SOFA: and (p = 0.652); CLIF-SOFA: and (p = 0.157); MELD: and (p = 0.144); MELD-Na: and (p = 0.012); MESO: and (p = 0.466), and updated MELD and (p = 0.517) (Table 4). Figure 3 showed the cumulative survival curve for the study population which could be divided into four groups (group 1: score = 0; group 2: score = 1 2; group 3: score 3 4; group 4: score = 5 6) related to patients with higher mortality at 1-year follow-up period with a p < As a result of multivariate analysis, the ARF-CLIF-SOFA score was the only independent predictor for 1-year mortality (HR = 1.509, 95% CI: , p < 0.001) (Table 5). For ARF-CLIF-SOFA, the optimal cut-off value of 2.50 was related with a sensitivity of 0.71, specificity of 0.71, respectively, positive likelihood ratio (PLR) of 2.43, negative likelihood ratio (NLR) of 0.41, positive predictive value (PPV) of 0.55, and negative predictive values

5 EXPERT REVIEW OF GASTROENTEROLOGY & HEPATOLOGY 5 Figure 1. Kaplan-Meier curves stratified by PaO2/FiO2 ratio (A), lactate (B), mean arterial pressure (C), vasopressors (D), bilirubin (E) and creatinine (F), respectively (all P < 0.05). Figure 2. Receiver operating characteristic curves analysis of the prognostic efficiency of ARF-CLIF-SOFA and other models to predict 30- (A), 90-day (B), 1-year (C) mortality of critically ill cirrhotic patients with acute respiratory failure (all P < 0.05). (NPV) of 0.71, respectively (Table 6). The mortality of CICRF for ARF-CLIF-SOFA, MELD, CLIF-SOFA scores and the combination of these scores were shown in the Table Discussion To our knowledge, this is the first study developing a prognostic model specific for CICRF patients in predicting mortality. Our study established a clinical risk prediction score (ARF- CLIF-SOFA) relevant to CICRF from the MIMIC database. In addition, the prediction accuracy revealed that the ARF-CLIF- SOFA score demonstrated a better discrimination than the original CLIF-SOFA, MELD, MELD-Na, MESO, and updated MELD scores at 30-, 90-day, 1-year follow-up periods. The ARF-CLIF-SOFA score included PaO 2 /FiO 2 ratio, lactate, MAP, vasopressor therapy, bilirubin, and creatinine. It remains uncertain whether PaO 2 /FiO 2 ratio is superior to SpO 2 /FiO 2 ratio for predicting mortality in respiratory disease patients [19]. Our final model selected the PaO 2 /FiO 2 ratio instead of the combination of PaO 2 /FiO 2 and SpO 2 /FiO 2 to simplify the calculation of the CLIF-SOFA score. Increased bilirubin

6 6 X.-D. ZHOU ET AL. Table 4. Calibration and discrimination for different scoring models in predicting hospital mortality. Calibration Discrimination Goodness-of-fit (χ 2 ) df p Cut-off AUROC 95% CI ARF-CLIF-SOFA ( ) CLIF-SOFA * ( ) MELD * ( ) MELD-Na * ( ) MESO * ( ) Updated MELD * ( ) ARF-CLIF-SOFA: acute respiratory failure-chronic liver failure-sequential organ failure assessment; CLIF-SOFA: chronic liver failure-sequential organ failure assessment; MELD: model for end-stage liver disease; MESO: model for end-stage liver disease to sodium ratio. *p < vs. ARF-CLIF-SOFA. Figure 3. Kaplan-Meier curves stratified by different risk levels of ARF-CLIF-SOFA (P < 0.001). included in CLIF-SOFA and MELD has been previously shown to be associated with adverse outcome in critically ill cirrhotic patients [20 22]. The existence of hyperlactemia may be directly related to liver dysfunction but of critical illness and mortality. Hyperlactemia at the first day of admission to the ICU has been considered as a potential predictor for evaluating mortality risk with a high discriminative accuracy for adverse outcome [23]. Lactate dehydrogenase under hypoxic conditions converts pyruvate to lactate [23,24]. Lactic acid levels relate to the oxygen insufficiency corresponding to tissue hypoperfusion [25]. Serum creatinine is a common variable used to quantify renal function and define acute renal failure associated with progressive multiple organ failure and mortality in critically ill cirrhotic patients [26,27]. Acute circulatory failure is considered a complication related to mortality in critically ill cirrhotic patients [28]. Low arterial pressure or vasopressors requirement determined hypoperfusion, cellular dysfunction, and metabolic disorders related to the survival [28,29]. Vasopressor therapy from the original CLIF-SOFA indicated the status of circulatory function. Vasopressor therapy was included the final ARF-CLIF-SOFA score. Hepatic encephalopathy and INR were excluded in the final score based on their mulifactorial variability. Any specific score should improve the discriminative accuracy compared to existing scores [30]. The parameters in ARF-CLIF-SOFA can be easily obtained in the management of CICRF. In addition, it had advanced diagnostic accuracy as compared to other scores. However, our study still has certain limitations. The ARF-CLIF-SOFA score was developed based on a single institution cohort requiring external validated. 5. Conclusions In summary, the ARF-CLIF-SOFA score is specific to CICRF patients score able to stratify the mortality risk, but also has more excellent predictive accuracy as compared to CLIF-SOFA, MELD, MELD-Na, MESO, and updated MELD scores. Table 5. Univariate and multivariate analysis of association between different scoring models and 1-year mortality in critically ill cirrhotic patients with acute respiratory failure. Univariate analysis Multivariate analysis HR 95% CI p HR 95% CI p ARF-CLIF-SOFA: 0 Ref ( ) ( ) < ( ) < ( ) < ( ) < ( ) <0.001 ARF-CLIF-SOFA ( ) < ( ) <0.001 CLIF-SOFA ( ) < ( ) MELD ( ) < ( ) MELD-Na ( ) <0.001 MESO ( ) <0.001 Updated MELD ( ) <0.001 ARF-CLIF-SOFA: acute respiratory failure-chronic liver failure-sequential organ failure assessment; CLIF-SOFA: chronic liver failure-sequential organ failure assessment; CICRF: critically ill cirrhotic patients with acute respiratory failure; MELD: model for end-stage liver disease; MESO: model for end-stage liver disease to sodium ratio.

7 EXPERT REVIEW OF GASTROENTEROLOGY & HEPATOLOGY 7 Table 6. Diagnostic accuracy of different prognostic models in predicting 1-year mortality using the optimal cut-off point. Prognostic models AUROC Cut-off point Sensitivity Specificity Youden index PLR NLR PPV NPV ARF-CLIF-SOFA ( ) CLIF-SOFA ( ) * MELD ( ) * MELD-Na ( ) * MESO ( ) * Updated MELD ( ) * ARF-CLIF-SOFA: acute respiratory failure-chronic liver failure-sequential organ failure assessment; AUROC: area under the receiver operating characteristic curve; CLIF- SOFA: chronic liver failure-sequential organ failure assessment; MELD: model for end-stage liver disease; MESO: model for end-stage liver disease to sodium ratio; NLR: negative likelihood ratio; NPV: negative predictive value; PLR: positive likelihood ratio; PPV: positive predictive value. *p < vs. ARF-CLIF-SOFA. Table 7. Mortality of critically ill cirrhotic patients with acute respiratory failure at 30-, 90-day as well as 1-year. N 30-day 90-day 1-year ARF-CLIF-SOFA (60.2%) 169 (70.1%) 182 (75.5%) MELD (53.5%) 148 (60.9%) 162 (66.7%) CLIF-SOFA (61.9%) 161 (69.7%) 174 (75.3%) ARF-CLIF-SOFA 4 & CLIF-SOFA (67.7%) 128 (76.6%) 135 (80.8%) ARF-CLIF-SOFA 4 & MELD (66.7%) 117 (75%) 125 (80.1%) CLIF-SOFA 13 and MELD (65.5%) 125 (73.1%) 135 (78.9%) ARF-CLIF-SOFA: acute respiratory failure-chronic liver failure-sequential organ failure assessment; CLIF-SOFA: chronic liver failure-sequential organ failure assessment; MELD: model for end-stage liver disease. Key issues Critically ill cirrhosis with acute respiratory failure (CICRF) is associated with high short-term mortality. CLIF-SOFA is a general scoring system for the mortality of critically ill cirrhosis patients. There are no specific models for CICRF to identify the severity the illness. We developed a specific to CICRF score based on CLIF- SOFA, named ARF-CLIF-SOFA. ARF-CLIF-SOFA has a more excellent discriminative accuracy to predict mortality than CLIF-SOFA, MELD, MELD-Na, MESO and updated MELD scores. Authors contributions Zhou XD, Chen QF, Huang WJ and Zheng MH designed the study. Zhou XD, Chen QF and Liu WY screened and extracted data. Zhang MC, Zhou XD, Lu Y did the statistical analyses. Liu WY, Zhang MC and Shi KQ prepared figures. Zhou XD, Chen QF, Poucke SV, Liu WY, Huang WJ and Zheng MH reviewed the results, interpreted data, and wrote the manuscript. All authors have made an intellectual contribution to the manuscript and approved the submission. Funding The authors were supported by grants from the National Natural Science Foundation of China ( ), Scientific Research Foundation of Wenzhou (Y ), High Level Creative Talents from Department of Public Health in Zhejiang Province, and Project of New Century 551 Talent Nurturing in Wenzhou. Declaration of interest The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. ORCID Sven Van Poucke Wen-Yue Liu Ming-Hua Zheng References Papers of special note have been highlighted as either of interest ( ) or of considerable interest ( ) to readers. 1. Berry PA, Thomson SJ, Rahman TM, et al. Review article: towards a considered and ethical approach to organ support in critically-ill patients with cirrhosis. Aliment Pharmacol Ther. 2013;37(2): Leon DA, McCambridge J. Liver cirrhosis mortality rates in Britain from 1950 to 2002: an analysis of routine data. Lancet (London, England). 2006;367(9504): A detailed review including the epidemiology and demographic characteristics of cirrhosis in Britain from 1950 to Thomson SJ, Moran C, Cowan ML, et al. Outcomes of critically ill patients with cirrhosis admitted to intensive care: an important perspective from the non-transplant setting. Aliment Pharmacol Ther. 2010;32(2): Olson JC, Wendon JA, Kramer DJ, et al. Intensive care of the patient with cirrhosis. Hepatology (Baltimore, Md). 2011;54(5): Ginès P, Fernández J, Durand F, et al. Management of critically-ill cirrhotic patients. J Hepatol. 2012;56 Suppl 1:S Tsai MH, Peng YS, Lien JM, et al. Multiple organ system failure in critically ill cirrhotic patients. A comparison of two multiple organ dysfunction/failure scoring systems. Digestion. 2004; 69(3): Chen YC, Tian YC, Liu NJ, et al. Prospective cohort study comparing sequential organ failure assessment and acute physiology, age, chronic health evaluation III scoring systems for hospital mortality prediction in critically ill cirrhotic patients. Int J Clin Pract. 2006;60 (2):

8 8 X.-D. ZHOU ET AL. 8. Huffmyer JL, Nemergut EC. Respiratory dysfunction and pulmonary disease in cirrhosis and other hepatic disorders. Respir Care. 2007;52(8): A detailed review including the mutual effects of cirrhosis on the respiratory system. 9. Ramalingam VS, Ansari S, Fisher M. Respiratory complication in liver disease. Crit Care Clin. 2016;32(3): A detailed review included the most important complications that cause significant morbidity and mortality: hepatopulmonary syndrome, hepatic hydrothorax, and portopulmonary hypertension. 10. Fritz JS, Fallon MB, Kawut SM. Pulmonary vascular complications of liver disease. Am J Respir Crit Care Med. 2013;187(2): Porres-Aguilar M, Gallegos-Orozco JF, Garcia H, et al. Pulmonary vascular complications in portal hypertension and liver disease: a concise review. Rev Gastroenterol Mex. 2013;78(1): Pan HC, Jenq CC, Tsai MH, et al. Scoring systems for 6-month mortality in critically ill cirrhotic patients: a prospective analysis of chronic liver failure - sequential organ failure assessment score (CLIF-SOFA). Aliment Pharmacol Ther. 2014;40(9): A study to validate that CLIF-SOFA scores are excellent prognosis evaluation tools for critically ill cirrhotic patients. 13. Moreau R, Jalan R, Gines P, et al. Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology. 2013;144(7): , 37.e1 37.e Kamath PS, Kim WR. The model for end-stage liver disease (MELD). Hepatology (Baltimore, Md). 2007;45(3): A study to validate that MELD scores are excellent prognosis evaluation tools for critically ill cirrhotic patients. 15. Luca A, Angermayr B, Bertolini G, et al. An integrated MELD model including serum sodium and age improves the prediction of early mortality in patients with cirrhosis. Liver Transpl. 2007;13(8): Biggins SW, Kim WR, Terrault NA, et al. Evidence-based incorporation of serum sodium concentration into MELD. Gastroenterology. 2006;130(6): Huo T-I, Wang Y-W, Yang -Y-Y, et al. Model for end-stage liver disease score to serum sodium ratio index as a prognostic predictor and its correlation with portal pressure in patients with liver cirrhosis. Liver Int. 2007;27(4): Sharma P, Schaubel DE, Sima CS, et al. Re-weighting the model for end-stage liver disease score components. Gastroenterology. 2008;135(5): Chen W, Janz DR, Shaver CM, et al. Clinical characteristics and outcomes are similar in ARDS diagnosed by oxygen saturation/ Fio2 ratio compared with Pao2/Fio2 ratio. Chest. 2015;148 (6): Temme EH, Zhang J, Schouten EG, et al. Serum bilirubin and 10- year mortality risk in a Belgian population. Cancer Causes Control: CCC. 2001;12(10): Zhai R, Sheu CC, Su L, et al. Serum bilirubin levels on ICU admission are associated with ARDS development and mortality in sepsis. Thorax. 2009;64(9): Campbell J, McPeake J, Shaw M, et al. Validation and analysis of prognostic scoring systems for critically ill patients with cirrhosis admitted to ICU. Crit Care. 2015;19: Tas A, Akbal E, Beyazit Y, et al. Serum lactate level predict mortality in elderly patients with cirrhosis. Wien Klin Wochenschr. 2012;124 (15 16): A study to investigate the association between serum lactate level and the mortality in elderly patients with cirrhosis. 24. Funk GC, Doberer D, Kneidinger N, et al. Acid-base disturbances in critically ill patients with cirrhosis. Liver Int. 2007;27(7): Ranucci M, De Toffol B, Isgro G, et al. Hyperlactatemia during cardiopulmonary bypass: determinants and impact on postoperative outcome. Crit Care. 2006;10(6):R Jenq CC, Tsai MH, Tian YC, et al. RIFLE classification can predict short-term prognosis in critically ill cirrhotic patients. Intensive Care Med. 2007;33(11): Dunn W, Jamil LH, Brown LS, et al. MELD accurately predicts mortality in patients with alcoholic hepatitis. Hepatology (Baltimore, Md). 2005;41(2): Fede G, Privitera G, Tomaselli T, et al. Cardiovascular dysfunction in patients with liver cirrhosis. Ann Gastroenterol. 2015;28(1): Hollenberg SM, Waldman B. The circulatory system in liver disease. Crit Care Clin. 2016;32(3): Jalan R, Saliba F, Pavesi M, et al. Development and validation of a prognostic score to predict mortality in patients with acute-onchronic liver failure. J Hepatol. 2014;61(5):