A simplified prognostic model to predict mortality in patients with acute variceal bleeding

G Model YDLD-3590; No. of Pages 7

ARTICLE IN PRESS Digestive and Liver Disease xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld

Alimentary Tract

A simplified prognostic model to predict mortality in patients with acute variceal bleeding Han Hee Lee a , Jae Myung Park a,∗ , Seunghoon Han b , Sung Min Park b , Hee Yeon Kim a , Jung Hwan Oh a , Chang Wook Kim a , Seung Kew Yoon a , Myung-Gyu Choi a a b

Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Republic of Korea Department of Pharmacology, College of Medicine, The Catholic University of Korea, Republic of Korea

a r t i c l e

i n f o

Article history: Received 19 June 2017 Received in revised form 22 October 2017 Accepted 13 November 2017 Available online xxx Keywords: Cirrhosis Logistic regression Prognostic model Survival Variceal bleeding

a b s t r a c t

Background: Acute variceal bleeding (AVB) is a major cause of death in patients with liver cirrhosis. The aim of this study was to investigate mortality predictors and develop a new simple prognostic model using easily verified factors at admission in AVB patients. Methods: Between January 2009 and May 2015, 333 consecutive patients with AVB were included. A simplified prognostic model was developed using multiple logistic regression after identifying significant predictors of 6-week mortality. Mortality prediction accuracy was assessed with area under the receiver operating characteristic (AUROC) curve. We compared the new model to existing models of model for end-stage liver disease (MELD) and Child–Pugh scores. Results: The 6-week overall mortality rate was 12.9%. Multivariate analysis showed that C-reactive protein (CRP), total bilirubin, and the international normalized ratio were independent predictors of mortality. A new logistic model using these variables was developed. This model’s AUROC was 0.834, which was significantly higher than that of MELD (0.764) or Child–Pugh scores (0.699). Two external validation studies showed that the AUROC of our model was consistently higher than 0.8. Conclusions: Our new simplified model accurately and consistently predicted 6-week mortality in patients with AVB using objective variables measured at admission. Our system can be used to identify high risk AVB patients. © 2017 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.

1. Introduction Patients with cirrhosis are at risk for developing one or more critical complications, such as ascites, spontaneous bacterial peritonitis, encephalopathy, or hepatorenal syndrome [1–4]. When these are accompanied by multiple organ failure and culminate in high short-term mortality, this becomes a distinct disease entity known as acute-on-chronic liver failure (ACLF) [5,6]. ACLF is an increasingly recognized disease entity. It has been postulated that ACLF could allow for early identification of patients at high risk for cirrhosis-related death. Acute variceal bleeding (AVB) is a major complication of portal hypertension in patients with cirrhosis. Endoscopic ligation ther-

∗ Corresponding author at: Division of Gastroenterology, Department of Internal Medicine, Seoul St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, 222 Banpo-daero, Seocho-u, Seoul 137-701, Republic of Korea. E-mail address: [email protected] (J.M. Park).

apy combined with vasoactive drugs and prophylactic antibiotics is the current standard of care for AVB patients [7,8]. Despite advances in diagnosis and management of esophageal and gastric varices, AVB is a main cause of upper gastrointestinal bleeding (UGIB). Mortality remains high in this situation (16%–24%) [7,9]. Development of sensitive and specific risk prediction models for AVB patients is important for reducing mortality in high-risk patients. Early transjugular intrahepatic portosystemic shunt (TIPS) use in selected high-risk patients reduces mortality [10,11]. Furthermore, rigorous application of treatments such as restricted blood transfusion or nonselective beta-blockers can improve AVB patient survival [12]. Although there are several AVB prognostic models, they have limited ability to predict patient outcomes. Rockall and Glasgow Blatchford scores are widely used UGIB risk predictors. However, these scores are poor at predicting clinical outcomes of patients with AVB [13]. The Child–Pugh score has subjective components that are inconsistently predictive, such as ascites or encephalopathy [14]. Model for end-stage liver disease (MELD) score is composed of objective variables [15]. However, this model

https://doi.org/10.1016/j.dld.2017.11.006 1590-8658/© 2017 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006

G Model YDLD-3590; No. of Pages 7

ARTICLE IN PRESS H.H. Lee et al. / Digestive and Liver Disease xxx (2017) xxx–xxx

2

was developed based on patients treated with TIPS more than 10 years ago and so might not apply to AVB patients. The chronic liver failure-sequential organ failure assessment (CLIF-SOFA) score has received the most attention recently and adequately predicts ACLF in chronic liver disease patients [6]. However, this score is complex and includes subjective components. The aim of the present study was to develop a new and simple prognostic model based on initial objective components in AVB patients. To achieve this aim, we first investigated mortality predictors in AVB patients. Based on this analysis, a new and simple prognostic model was produced using exclusively objective and easily verified factors. Finally, this new model was validated externally to ensure generalizability and implementation in clinical practice. 2. Patients and methods 2.1. Study population A consecutive database that included all patients admitted to Seoul St. Mary’s Hospital, a tertiary care center, for acute gastrointestinal (GI) bleeding from January 2009 to May 2015 was created. This study was approved by Seoul St. Mary’s Hospital Institutional Review Board (KC14RISI0606). From the total GI bleeding patient cohort, those who were found to have cirrhosis and variceal bleeding were identified and selected for this study. Patients with cirrhosis and acute bleeding from both esophageal and gastric varices were considered eligible for this study. Cirrhosis diagnosis was based on definite clinical data combined with imaging findings such as abdominal sonography or computed tomography. Only patients presenting with variceal bleeding confirmed by esophagogastroduodenoscopy (EGD) were included. According to Baveno guideline, antibiotic prophylaxis was instituted as early as possible on presentation of AVB and continued for 5–7 days in all patients with AVB in our institution. Vasoconstrictors were also administered as soon as possible when there is a clinical suspicion of AVB. We excluded the following cases: those under 18 years of age, those who did not undergo EGD, and follow-up loss within 6 weeks from initial endoscopic examination. 2.2. Data collection Baseline demographic characteristics, Child–Pugh and MELD scores, cirrhotic complications, previous episodes of variceal bleeding, and major comorbidities were recorded. Initial systolic blood pressure (SBP), diastolic blood pressure (DBP) and laboratory tests including measures of hemoglobin, serum blood urea nitrogen, creatinine, prothrombin time or international normalized ratio (INR), total bilirubin, albumin, serum CRP, and white blood cell count were also recorded. Bleeding focus and endoscopic findings were described by the endoscopist who performed the EGD. Endoscopic therapy was performed as soon as safely possible. Sclerotherapy or ligation choice was left to the endoscopist’s discretion according to recommended standards. A Sengstaken–Blakemore tube was placed when necessary. Transfusion requirements were defined as number of packed red blood cells (pRBC) products transfused on the day of bleeding or transfused continuously during the following hospital stay with initiation at the day of bleeding. Outcome data, including hospital stay, rebleeding, readmission, and 6-week mortality were also recorded. The primary outcome of the present study was 6-week mortality because it is a thoroughly validated end point during which most AVB deaths occur [7,11]. Rebleeding was defined as a new hematemesis or melena after 24 h of stable vital signs and hemoglobin level [12].

2.3. Definitions This study’s definitions were based on the criteria of the Baveno II consensus workshops [16]. Time zero of a variceal bleeding episode was defined as the first time a patient was admitted to the hospital presenting with AVB symptoms. The endoscopic findings of variceal bleeding were classified as active bleeding, stigmata, and no definite stigmata. Active bleeding was defined as a spurting or oozing lesion. Stigmata of recent hemorrhage was defined as presence of a white nipple on varices, adherent clots on varices or ulcers, or a visible vessel in an ulcer’s base. If varices were seen in the context of a recent UGIB with neither active bleeding nor stigmata and there were not any causative lesion for bleeding such as peptic ulcer or angioectasia, no definite stigmata was recorded. Bacterial infection was defined by one of the following criteria during the first 5 days after hemorrhage: spontaneous bacterial peritonitis, pneumonia, urinary tract infection (UTI), bacteremia, or other infection. Other infections were diagnosed according to clinical, radiologic, and bacteriologic data. Mean arterial pressure (MAP) was calculated as follows: MAP = DBP + [0.333 (SBP − DBP)]. 2.4. External validation External validation studies were performed in 2 series of cirrhotic patients presenting with variceal bleeding. The first validation was performed in 555 consecutive AVB patients in Uijeongbu St. Mary’s Hospital from June 2009 to October 2015. This facility is located in Gyeonggi-do, and is a tertiary care center reporting high variceal bleeding occurrence. The second sample included 105 consecutive AVB patients admitted to St. Paul’s Hospital from January 2009 to October 2013. This facility is located in Seoul, is a secondary care center, and reports a low variceal bleeding volume. Data from which to calculate the model was available for all patients in both series. 2.5. Statistical analysis and prognostic model generation Descriptive statistics were used to characterize the demographic features of the study population. For the univariate analysis, continuous variables were expressed as mean (±standard deviation) or median (interquartile range) and were compared using the Mann–Whitney U test. Categorical variables were expressed as number (percentage) and were compared between groups using Chi-square or Fisher’s exact test as appropriate. A logistic regression model was used to assess predictive factors of 6-week mortality. This model was considered a suitable alternative to the Cox model because the follow-up time was relatively short. Variables showing P-values <0.05 after univariate analysis and those that were considered clinically relevant were included in a multivariate logistic regression model to identify independent factors associated with 6-week mortality. Backward step-wise multivariate analyses were performed. For significant variables, coefficients and odds ratios with 95% confidence intervals were reported. After multivariate analysis, significant variables were selected for the final prognostic model of the outcome of interest. We decided to exclude variables that were subjective and hard to verify at initial presentation. To test discriminatory ability of the newly developed model, a prognostic index (PI) was computed for the model with the equation, PI = bxi + byi + . . . bzi + constant term. b was the regression coefficient for each variable in the final model, and xi, yi, zi, and so forth represented each variable’s value for patient i [9]. PI was subsequently used to compute individual probabilities for death and to produce receiver operating characteristic (ROC) curves. This new prognostic model then was validated by comparing the area under the receiver operating characteristic curve with other known prognostic models. Statistical analysis was

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006

G Model YDLD-3590; No. of Pages 7

ARTICLE IN PRESS H.H. Lee et al. / Digestive and Liver Disease xxx (2017) xxx–xxx

3

Table 1 Demographic characteristics of the study population at admission. Overall (N = 333)

Age (years, mean), range Male sex Etiology of liver disease Alcohol Viral Others Child–Pugh class A B C MELDa , range HCC Hepatic encephalopathy Ascites Infection Previous episodes of variceal bleeding Comorbidities Diabetes Hypertension Heart failure ESRD Systolic pressure (mmHg)a Diastolic pressure (mmHg)a Mean arterial pressure (mmHg)a Laboratory resultsa Hemoglobin (g/dl) BUN (mg/dl) Creatinine (mg/dl) INR Total bilirubin (mg/dl) Albumin (g/dl) CRP (mg/dl)b WBC count (×109 /L) Platelet count (×109 /L) Time from admission to perform endoscopy (hours, median), range Endoscopic finding Active bleeding Stigmata No definite stigmata Treatment EVL EVS Combination Medical Sengstaken–Blakemore tube PRC transfusion (units)a

Groups according to bleeding site

P

Esophageal varix (n = 253)

Gastric varix (n = 80)

57.5 (22–98) 247 (74.2%)

58.0 (25–98) 189 (74.7%)

55.8 (22–89) 58 (72.5%)

119 (35.7%) 174 (52.3%) 40 (12.0%)

88 (34.8%) 135 (53.4%) 30 (11.9%)

31 (38.8%) 39 (48.8%) 10 (12.5%)

77 (23.1%) 216 (64.9%) 40 (12.0%) 14.0 ± 5.2 (6.7–36.4) 105 (31.5%) 16 (4.8%) 207 (62.2%) 24 (7.3%) 161 (48.4%)

47 (18.6%) 137 (54.2%) 69 (27.3%) 14.2 ± 5.2 82 (32.4%) 14 (5.5%) 173 (68.4%) 18 (7.1%) 125 (49.4%)

30 (37.5%) 39 (48.8%) 11 (13.8%) 13.2 ± 5.2 23 (28.8%) 2 (2.5%) 34 (42.5%) 6 (7.5%) 36 (45.0%)

0.144 0.634 0.420 <0.001 >0.999 0.576

88 (26.4%) 67 (20.1%) 3 (0.9%) 4 (1.2%) 108.5 ± 22.7 63.8 ± 14.8 78.7 ± 16.5

65 (25.7%) 48 (19.0%) 2 (0.8%) 4 (1.6%) 108.7 ± 22.3 63.9 ± 14.3 78.8 ± 16.1

23 (28.8%) 19 (23.8%) 1 (1.3%) 0 (0%) 108.1 ± 23.9 63.6 ± 16.4 78.4 ± 17.9

0.693 0.442 >0.999 0.587 0.840 0.885 0.858

8.7 ± 2.2 34.0 ± 22.1 1.0 ± 0.8 1.6 ± 0.4 2.9 ± 4.1 2.8 ± 0.6 1.3 ± 2.7 8.3 ± 5.5 115.8 ± 94.6 5 (1–81)

8.6 ± 2.3 34.3 ± 22.5 1.1 ± 0.7 1.5 ± 0.4 2.9 ± 4.5 2.8 ± 0.6 1.3 ± 2.4 8.2 ± 5.7 117.4 ± 102.3 5 (1–81)

8.8 ± 2.0 32.8 ± 21.1 1.0 ± 0.9 1.4 ± 0.4 3.1 ± 5.1 2.9 ± 0.6 1.5 ± 3.4 8.6 ± 5.1 110.9 ± 64.9 6 (1–72)

0.605 0.586 0.475 0.362 0.676 0.209 0.220 0.594 0.594 0.463 0.377

95 (28.5%) 209 (62.8%) 29 (8.7%)

70 (27.7%) 158 (62.5%) 25 (9.9%)

25 (31.3%) 51 (63.8%) 4 (5.0%)

218 (65.5%) 32 (9.6%) 4 (1.2%) 79 (23.7%) 21 (6.3%) 4.7 ± 4.7

178 (70.4%) 2 (0.8%) 2 (0.8%) 71 (28.1%) 13 (5.1%) 4.3 ± 4.3

40 (50.0%) 30 (37.5%) 2 (2.5%) 8 (10.0%) 8 (10.0%) 5.7 ± 5.8

0.172 0.806 0.765

0.001

<0.001

0.119 0.060

MELD, model for end-stage liver disease; HCC, hepatocellular carcinoma; ESRD, end-stage renal disease; BUN, blood urea nitrogen; INR, international normalized ratio; CRP, C-reactive protein; WBC, white blood cell; EVL, endoscopic variceal ligation; EVS, endoscopic variceal sclerotherapy; PRC, packed red blood cells. a Mean ± SD. b CRP was evaluated in 224 and 72 cases in the esophageal and gastric varix groups, respectively.

performed using Statistical Analysis System software (version 8.02, SAS Institute, Cary, NC, USA). P-values <0.05 were considered to be significant. 3. Results 3.1. Patients features and outcomes From January 2009 to May 2015, 343 consecutive patients with AVB who were admitted to Seoul St. Mary’s Hospital were identified. After excluding those who were under 18 years of age (n = 1), who did not undergo EGD (n = 2), and those lost to follow-up within 6 weeks from the initial endoscopic exam (n = 7), 333 patients remained and were included. The demographic characteristics of the studied population are summarized in Table 1. The mean age was 57.5 years (range: 22–98) and there were 247 men. Viral hepatitis was the leading cause of liver cirrhosis in more than half of the patients (n = 174, 52.3%). The

mean MELD was 14.0 (range: 6.7–36.4). Infection was documented in 24 patients (7.3%), including 3 spontaneous bacterial peritonitis, 9 pneumonia, 2 UTI, 6 bacteremia, and 4 other infections. On endoscopy, active bleeding was found in 95 of 333 patients (28.5%). Overall, esophageal and gastric varices were the primary cause of bleeding in 253 (76.0%) and 80 (24.0%) patients, respectively. There were no significant demographic characteristic differences between patients except Child–Pugh class, ascites, and initial treatment. The mean hospitalization time for patients was 13 days. During the 6-week period after index bleeding, 70 (21.2%) patients experienced rebleeding. The overall 6-week mortality was 12.9% (43/333). 3.2. Predictive factors of 6-week mortality Results of the univariate and multivariate analyses for 6-week mortality are shown in Table 2. All significant univariate variables

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006

G Model

ARTICLE IN PRESS

YDLD-3590; No. of Pages 7

H.H. Lee et al. / Digestive and Liver Disease xxx (2017) xxx–xxx

4

Table 2 Univariate and multivariate analyses of predictive factors associated with 6-week mortality in patients with AVB. Measures

Survivors (n = 290)

Age ≥ 60 years, n (%) Male sex, n (%) Child–Pugh class, n (%) A B C MELD score, mean ± SD HCC Hepatic encephalopathy Ascites Infection Previous episodes of variceal bleeding Blood in stomach Bleeding site Esophageal varices Gastric varices Endoscopic finding Active bleeding Stigmata No definite stigmata Initial treatment EVL EVS Combination Medical Systolic pressure (mmHg) Mean arterial pressure (mmHg) Hemoglobin (g/dl) BUN (mg/dl) Creatinine (mg/dl) INR Albumin (g/dl) Total bilirubin CRPb Platelet count (×109 /L) PRC transfusion 6-week rebleeding

Deaths (n = 43)

119 (41.0%) 213 (73.4%)

13 (30.2%) 34 (79.1%)

75 (25.9%) 189 (65.2%) 26 (9.0%) 13.1 ± 4.3 87 (30.0%) 11 (3.8%) 172 (59.3%) 17 (5.9%) 149 (51.4%) 233 (80.3%)

2 (4.7%) 27 (62.8%) 14 (32.6%) 19.5 ± 7.1 18 (41.9%) 5 (11.6%) 35 (81.4%) 7 (16.3%) 12 (27.9%) 36 (83.7%)

223 (76.9%) 67 (23.1%)

30 (69.8%) 13 (30.2%)

82 (28.3%) 182 (62.8%) 26 (9.0%)

13 (30.2%) 27 (62.8%) 3 (7.0%)

192 (66.2%) 28 (9.7%) 4 (1.4%) 66 (22.8%) 108.9 ± 22.5 78.7 ± 16.3 8.6 ± 2.2 32.1 ± 19.0 0.9 ± 0.5 1.5 ± 0.4 2.9 ± 0.6 2.3 ± 3.1 1.0 ± 2.0 117.1 ± 98.4 4.2 ± 4.2 51 (17.6%)

26 (60.5%) 4 (9.3%) 0 (0.0%) 13 (30.2%) 105.79 ± 23.6 78.7 ± 18.2 8.7 ± 2.5 46.5 ± 34.9 1.7 ± 1.7 1.7 ± 0.5 2.6 ± 0.6 6.9 ± 9.2 4.0 ± 4.9 107.2 ± 64.0 7.7 ± 6.8 19 (44.2%)

P

Multivariate analysis

P

Odd ratio

95% confidence interval

3.03 1.27 2.08 0.52

0.20–50.00 0.46–3.52 0.42–10.00 0.20–1.36

0.421 0.649 0.370 0.180

1.74 4.22 0.87 1.16 1.26

0.84–3.62 1.60–11.15 0.37–2.06 1.04–1.30 1.13–1.42

0.136 0.004 0.752 0.009 <0.001

0.177 0.432 <0.001

<0.001 0.118 0.042a 0.005 0.024a 0.004 0.600 0.307

0.896

0.644

0.399 0.990 0.840 <0.001 <0.001 <0.001 0.002 <0.001 <0.001 0.525 0.002 <0.001

MELD, model for end-stage liver disease; HCC, hepatocellular carcinoma; EVL, endoscopic variceal ligation; EVS, endoscopic variceal sclerotherapy; BUN, blood urea nitrogen; INR, international normalized ratio; CRP, C-reactive protein; PRC, packed red blood cells. a Fisher’s exact test. b CRP was evaluated in 264 and 32 cases in the survivor and death groups, respectively. Table 3 Prognostic models of 6-week mortality in patients with AVB using multivariate logistic regression analysis.

Constant term CRP Total bilirubin INR

B

SE

P

OR

95% CI

−5.3076 0.2333 0.1480 1.4388

0.8340 0.0589 0.0568 0.4960

<0.001 <0.001 0.009 0.004

1.26 1.16 4.22

1.13–1.42 1.04–1.30 1.60–11.15

CRP, C-reactive protein; INR, international normalized ratio.

were chosen for multivariate analysis. BUN was excluded from the multivariate analysis because there was multicollinearity between BUN and creatinine. PRC transfusion and 6-week rebleeding were significantly higher in patients who subsequently died compared to survivors. However, these were not included in the multivariate analysis because they were outcome variables and could not be measured at initial presentation.

calculated as follows: (0.2333 × CRP) + (0.1480 × total bilirubin) + (1.4388 × INR) − 5.3076. Discriminatory performance was evaluated by analyzing the ROC curves. Our model was compared with various known prognostic models including Child and MELD scores (Fig. 1). As shown, our model demonstrated the best predictive accuracy among all prognostic models, with an AUROC significantly better than the Child–Pugh score (P = 0.014). This model’s performance was compared to MELD score, and its superiority was close to significant (P = 0.076). UGIB scores not specific for variceal bleeding including AIMS65, Glasgow Blatchford, and Rockall scores were poor predictors of patient mortality, as expected. An additional sensitivity analysis that included all 95 active variceal bleeding patients on initial endoscopy was conducted (Fig. 2). This sensitivity analysis result did not differ from the analysis based on all included patients. In addition, Child–Pugh and MELD score AUROCs were not statistically significant in this setting (0.599, 95% CI 0.424–0.773 and 0.593, 95% CI 0.391–0.796, respectively).

3.3. Model generation and performance of 6-week mortality 3.4. External validation The newly developed prognostic model obtained with logistic regression analysis is presented in Table 3. This new model was developed only from independent prognostic factors using multivariate analysis and was

Two external validation datasets were used to evaluate the external validity of our model. In the first dataset from Uijeongbu St. Mary’s Hospital, the model was consistently effective, with an

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006

G Model YDLD-3590; No. of Pages 7

ARTICLE IN PRESS H.H. Lee et al. / Digestive and Liver Disease xxx (2017) xxx–xxx

Fig. 1. Receiver operating characteristic (ROC) curves of various prognostic models used to predict 6-week mortality in patients with acute variceal bleeding. MELD, model for end-stage liver disease score; CTP, Child–Turcotte–Pugh score; GBS, Glasgow–Blatchford score.

AUROC of 0.840 (95% CI, 0.780–0.899) (Fig. 3A). In the second dataset from St. Paul’s Hospital, the AUROC of our model was 0.803 (95% CI, 0.670–0.936) (Fig. 3B). 4. Discussion Although several prognostic models have been developed to predict outcomes of patients with AVB, these are not widely used because some indicators in the models are subjective and difficult to assess at AVB onset. In the present study, independent predictive factors of 6-week mortality were identified, and a simple yet novel prognostic model calculated from a cohort of patients with AVB was proposed. This model was validated externally in two additional patient series. All model components could be measured objectively and verified easily at variceal hemorrhage onset. In addition, this model was shown to be superior to existing models currently used to predict AVB prognosis. This study was conducted on a good cohort with a considerable patient sample size over a consecutive five-year period. Using this

5

Fig. 2. Receiver operating characteristic (ROC) curve analysis of patients with active variceal bleeding used to predict 6-week mortality. MELD, model for end-stage liver disease score; CTP, Child–Turcotte–Pugh score.

cohort, marked improvement in AVB patient prognosis was identified compared to previous studies executed over the past few decades. Although a 1981 study reported a 60% mortality rate in AVB [17], recent studies have described mortality as low as 15% [7,18,19], which is similar to our result of 12.9%. Reduced AVB mortality in the context of cirrhotic portal hypertension is considered to result from advancement in varices and variceal bleeding management. In addition, several clinical guidelines recommend managing these patients based on animal research models and many randomized clinical trials [20–22]. Emergent EGD is available at our institution on a 24/7 basis with an on-call GI endoscopist proficient in endoscopic hemostasis and on-call support staff with technical expertise in endoscopic device usage, which corresponds with the recommendation from the 6th Baveno Consensus Workshop [23]. Within the context of this policy, only two patients did not undergo EGD during this study period. Active bleeding was not associated with prognosis in our study, which seemed to be related with as appropriate and immediate endoscopic treatment for variceal bleeding.

Fig. 3. Discrimination efficacy results for our model and model for end-stage liver disease (MELD) score after external validation analyses. (A) Uijeongbu St. Mary’s Hospital dataset. (B) St. Paul’s Hospital dataset.

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006

G Model YDLD-3590; No. of Pages 7 6

ARTICLE IN PRESS H.H. Lee et al. / Digestive and Liver Disease xxx (2017) xxx–xxx

Various predictive factors associated with 6-week mortality were analyzed. As a consequence, only 3 factors of INR, total bilirubin, and CRP were ultimately found to be independent prognostic factors. Whereas INR and total bilirubin are well known prognostic factors in existing models such as MELD and Child–Pugh score, CRP has never been included in a prognostic model for patients with liver cirrhosis or GI bleeding. However, several studies have reported an association between CRP and cirrhotic patient prognosis. Serum CRP levels were significantly higher in patients evaluated with Child–Pugh score C than the other models [24]. Patients with ACLF, which has become an increasingly recognized and universally accepted diagnosis, also had higher CRP plasma levels [6]. In addition, CRP has been reported to be a prognostic factor of cirrhotic patient mortality [24–27]. Why does CRP have such prognostic value for cirrhotic patients with AVB? First, CRP is strongly associated with overt bacterial infection and systemic inflammatory response syndrome, both of which account for significant mortality in cirrhotic patients [28]. During infection, pathogens are present in the systemic circulation and cause hepatic endothelial dysfunction resulting in increased vasoconstrictor production, which leads to increased portal hypertension and variceal bleeding risk [29]. This is supported by a recent national cohort study that showed that timely antibiotic administration can reduce the mortality of cirrhotic patients with UGIB [30]. In our study, patients with overt infection at admission presented with higher CRP levels compared to those without infection. Second, whether or not bacterial infection was documented, CRP could be related to low grade systemic inflammation caused by bacterial translocation. In advanced cirrhosis, viable bacteria passage and bacterial components of gut origin are elevated due to increased intestinal permeability, impaired reticuloendothelial function, and decreased complement factor synthesis [31]. This process, termed bacterial translocation, may cause inflammatory response activation and aggravate circulatory failure. It is well known that CRP actually indicates interleukin-6 synthesis, which involves activation of a proinflammatory response in cirrhotic patients [32,33]. Third, CRP levels are closely linked to hepatic venous pressure gradient (HVPG). HVPG directly correlated with portal hypertension, and HVPG >20 mmHg is independently associated with mortality in variceal rupture [34]. However, HVPG measurement is an invasive procedure. Proinflammatory markers including CRP have been correlated with HVPG and systemic vascular resistance [27]. In addition, intestinal decontamination with rifaximin significantly lowered HVPG values in patients with alcohol-related decompensated cirrhosis [35]. The new model developed in this study was based on independent and objective prognostic factors in AVB patients. Our model showed the highest discrimination. Rockall score and Glasgow Blatchford score, which are well known scoring models for UGIB, as well as variceal bleeding, were poor predictors of prognosis in these patients. These results correspond well with those from an earlier study that reported GBS, admission Rockall, or full Rockall scores to not be useful for predicting clinical outcomes in variceal bleeding patients [13]. The performance of our model is obviously superior to the Child–Pugh score, whereas the difference of performance between our model and the MELD score was not statistically significant in our cohort and two external validation sets. Our model’s discriminatory performance was especially superior for predicting active spurting variceal bleeding. It is unclear why these models show such discrepancy. However, patient CRP levels in the presence of active bleeding were somewhat higher than in other patients (data not shown), which might explain the new model’s better performance in this context. A strength of the new model is simplicity. Only three factors are required to produce better predictive power. The CLIF-SOFA scoring system reported by the European Association for Study of

Liver/Chronic Liver Failure consortium could better discriminate between survivors and nonsurvivors with ACLF [6]. However, this scoring is more complicated and requires evaluation of six organs associated with hepatic, renal, cerebral, coagulatory, circulatory and respiratory function. We suppose that CRP may reflect the presence of overall organ failure and an intense systemic inflammatory response. Indeed, ACLF patients showed higher plasma CRP levels [6]. Another strength is that our model only uses objective and easily verified laboratory factors collected from patients presenting with variceal bleeding. Including many different risk scores limits the utility of other models in clinical practice [36]. Lastly, the predictive ability of our model did not decline, but maintained consistency when applied to external datasets with different characteristics in terms of hospital region, patient care volume, and health care type (secondary or tertiary health center). On the other hand, the nature of this study based on a retrospective analysis of prospectively collected data of GI bleeding cohort is a major limitation. In conclusion, this new prognostic model showed that 6-week mortality could be simply predicted by variables collected at admission and its discriminatory ability was as accurate as MELD score and far superior to other well-known UGIB scoring systems utilized with AVB patients. Prospective studies to further validate the efficacy of our risk model in predicting AVB patient mortality are needed to confirm its performance described retrospectively here. Conflict of interests None declared. Funding This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (Ministry of Science, ICT & Future Planning) (NRF-2016R1C1B2015185) and by Leading Foreign Research Institute Recruitment Program through the NRF funded by the Ministry of Science, ICT and Future Planning (NRF-2011-0031644). Acknowledgement The authors thank Soung Won Jeong, MD, Associate Professor at Soonchunhyang University College of Medicine, for discussions regarding the data analysis. References [1] Moore KP, Wong F, Gines P, Bernardi M, Ochs A, Salerno F, et al. The management of ascites in cirrhosis: report on the consensus conference of the International Ascites Club. Hepatology 2003;38(1):258–66. [2] Wiest R, Krag A, Gerbes A. Spontaneous bacterial peritonitis: recent guidelines and beyond. Gut 2012;61(2):297–310. [3] Blei AT, Cordoba J. Hepatic encephalopathy. Am J Gastroenterol 2001;96(7):1968–76. [4] Salerno F, Gerbes A, Gines P, Wong F, Arroyo V. Diagnosis, prevention and treatment of hepatorenal syndrome in cirrhosis. Gut 2007;56(9):1310–8. [5] Jalan R, Gines P, Olson JC, Mookerjee RP, Moreau R, Garcia-Tsao G, et al. Acute-on chronic liver failure. J Hepatol 2012;57(6):1336–48. [6] Moreau R, Jalan R, Gines P, Pavesi M, Angeli P, Cordoba J, et al. Acute-onchronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology 2013;144(7):1426–37, 37 e1-9. [7] Augustin S, Altamirano J, Gonzalez A, Dot J, Abu-Suboh M, Armengol JR, et al. Effectiveness of combined pharmacologic and ligation therapy in highrisk patients with acute esophageal variceal bleeding. Am J Gastroenterol 2011;106(10):1787–95. [8] de Franchis R. Evolving consensus in portal hypertension. Report of the Baveno IV consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol 2005;43:167–76. [9] Augustin S, Muntaner L, Altamirano JT, Gonzalez A, Saperas E, Dot J, et al. Predicting early mortality after acute variceal hemorrhage based on classification and regression tree analysis. Clin Gastroenterol Hepatol 2009;7(12):1347–54. [10] Monescillo A, Martinez-Lagares F, Ruiz-del-Arbol L, Sierra A, Guevara C, Jimenez E, et al. Influence of portal hypertension and its early decompres-

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006

G Model YDLD-3590; No. of Pages 7

ARTICLE IN PRESS H.H. Lee et al. / Digestive and Liver Disease xxx (2017) xxx–xxx

[11]

[12]

[13]

[14]

[15]

[16]

[17] [18]

[19]

[20]

[21]

[22] [23]

[24]

sion by TIPS placement on the outcome of variceal bleeding. Hepatology 2004;40(4):793–801. Garcia-Pagan JC, Caca K, Bureau C, Laleman W, Appenrodt B, Luca A, et al. Early use of TIPS in patients with cirrhosis and variceal bleeding. N Engl J Med 2010;362(25):2370–9. de Franchis R. Revising consensus in portal hypertension: report of the Baveno V consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol 2010;53(4):762–8. Reed EA, Dalton H, Blatchford O, Ashley D, Mowat C, Gaya DR, et al. Is the Glasgow Blatchford score useful in the risk assessment of patients presenting with variceal haemorrhage? Eur J Gastroenterol & Hepatol 2014;26(4):432–7. Bosch J, Thabut D, Albillos A, Carbonell N, Spicak J, Massard J, et al. Recombinant factor VIIa for variceal bleeding in patients with advanced cirrhosis: a randomized, controlled trial. Hepatology 2008;47(5):1604–14. Malinchoc M, Kamath PS, Gordon FD, Peine CJ, Rank J, ter Borg PC. A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology 2000;31(4):864–71. de Franchis R, Pascal JP, Ancona E, Burroughs AK, Henderson M, Fleig W, et al. Definitions, methodology and therapeutic strategies in portal hypertension. A Consensus Development Workshop, Baveno, Lake Maggiore, Italy, April 5 and 6, 1990. J Hepatol 1992;15(1-2):256–61. Graham DY, Smith JL. The course of patients after variceal hemorrhage. Gastroenterology 1981;80(4):800–9. Carbonell N, Pauwels A, Serfaty L, Fourdan O, Levy VG, Poupon R. Improved survival after variceal bleeding in patients with cirrhosis over the past two decades. Hepatology 2004;40(3):652–9. Seo YS, Park SY, Kim MY, Kim JH, Park JY, Yim HJ, et al. Lack of difference among terlipressin, somatostatin, and octreotide in the control of acute gastroesophageal variceal hemorrhage. Hepatology 2014;60(3):954–63. Garcia-Tsao G, Sanyal AJ, Grace ND, Carey W. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology 2007;46(3):922–38. Garcia-Tsao G, Sanyal AJ, Grace ND, Carey WD. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Am J Gastroenterol 2007;102(9):2086–102. Garcia-Tsao G, Bosch J. Management of varices and variceal hemorrhage in cirrhosis. N Engl J Med 2010;362(9):823–32. de Franchis R, Baveno VIF. Expanding consensus in portal hypertension: report of the Baveno VI Consensus Workshop: stratifying risk and individualizing care for portal hypertension. J Hepatol 2015;63(3):743–52. Mortensen C, Andersen O, Krag A, Bendtsen F, Moller S. High-sensitivity Creactive protein levels predict survival and are related to haemodynamics in alcoholic cirrhosis. Eur J Gastroenterol Hepatol 2012;24(6):619–26.

7

[25] Cervoni JP, Thevenot T, Weil D, Muel E, Barbot O, Sheppard F, et al. C-reactive protein predicts short-term mortality in patients with cirrhosis. J Hepatol 2012;56(6):1299–304. [26] Waidmann O, Brunner F, Herrmann E, Zeuzem S, Piiper A, Kronenberger B. Macrophage activation is a prognostic parameter for variceal bleeding and overall survival in patients with liver cirrhosis. J Hepatol 2013;58(5):956–61. [27] Wiese S, Mortensen C, Gotze JP, Christensen E, Andersen O, Bendtsen F, et al. Cardiac and proinflammatory markers predict prognosis in cirrhosis. Liver Int 2014;34(6):e19–30. [28] Arvaniti V, D’Amico G, Fede G, Manousou P, Tsochatzis E, Pleguezuelo M, et al. Infections in patients with cirrhosis increase mortality four-fold and should be used in determining prognosis. Gastroenterology 2010;139(4):1246–56, 56.e15. [29] La Mura V, Pasarin M, Rodriguez-Vilarrupla A, Garcia-Pagan JC, Bosch J, Abraldes JG. Liver sinusoidal endothelial dysfunction after LPS administration: a role for inducible-nitric oxide synthase. J Hepatol 2014;61(6):1321–7. [30] Moon AM, Dominitz JA, Ioannou GN, Lowy E, Beste L. Use of antibiotics among patients with cirrhosis and upper gastrointestinal bleeding is associated with reduced mortality. Clin Gastroenterol Hepatol 2016;14(11):1629–37. [31] Cirera I, Bauer TM, Navasa M, Vila J, Grande L, Taura P, et al. Bacterial translocation of enteric organisms in patients with cirrhosis. J Hepatol 2001;34(1):32–7. [32] Heikkila K, Ebrahim S, Rumley A, Lowe G, Lawlor DA. Associations of circulating C-reactive protein and interleukin-6 with survival in women with and without cancer: findings from the British Women’s Heart and Health Study. Cancer Epidemiol Biomark Prev 2007;16(6):1155–9. [33] Rosenbloom AJ, Pinsky MR, Bryant JL, Shin A, Tran T, Whiteside T. Leukocyte activation in the peripheral blood of patients with cirrhosis of the liver and SIRS. Correlation with serum interleukin-6 levels and organ dysfunction. JAMA 1995;274(1):58–65. [34] Garcia-Tsao G, Bosch J, Groszmann RJ. Portal hypertension and variceal bleeding–unresolved issues. Summary of an American Association for the study of liver diseases and European Association for the study of the liver single-topic conference. Hepatology 2008;47(5):1764–72. [35] Vlachogiannakos J, Saveriadis AS, Viazis N. Theodoropoulos I, Foudoulis K, Manolakopoulos S, et al. Intestinal decontamination improves liver haemodynamics in patients with alcohol-related decompensated cirrhosis. Aliment Pharmacol Ther 2009;29(9):992–9. [36] Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology 2010;21(1):128–38.

Please cite this article in press as: Lee HH, et al. A simplified prognostic model to predict mortality in patients with acute variceal bleeding. Dig Liver Dis (2017), https://doi.org/10.1016/j.dld.2017.11.006