Gastrointestinal bleeding is one of the most important

Prospective Validation of Baveno V Definitions and Criteria for Failure to Control Bleeding in Portal Hypertension Sun Young Ahn, 1 Soo Young Park, 1 Won Young Tak, 1 Yu Rim Lee, 1 Eun Jeong Kang, 1 Jung Gil Park, 1 Won Kee Lee, 2 Kwan Lee, 3 and Young Oh Kweon 1 New definitions and criteria were released at the Baveno V consensus meeting. The purposes of this study were to verify Baveno V definitions and criteria for failure to control bleeding and to determine the usefulness of the combined use of the Adjusted Blood Requirement Index [ABRI: (number of blood units)/(final hematocrit-initial hematocrit)10.01] with Baveno V criteria. In all, 246 consecutive liver cirrhosis patients with acute bleeding associated with portal hypertension were enrolled prospectively between January 2010 and October 2012. The treatment outcome on day 5 was assessed by endoscopy. For the ABRI calculation, two hematocrit levels were used as the initial hematocrit: the first level measured upon patient arrival (ABRI-A) and the lowest level measured before transfusion (ABRI-B). Treatment failures were identified in 53 patients, of whom 24 died. Based on repeated endoscopic findings, 29 patients were identified as treatment failures, while according to Baveno V criteria, 47 patients were regarded as treatment failures. The area under the receiver operating characteristic curve (AUROC) of Baveno V criteria was 0.906, and the sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio were 83.0%, 98.4%, 93.6%, 95.5%, 53.41, and 0.17, respectively. The AUROC of Baveno V criteria was significantly greater than those of Baveno IV (P 5 0.0001) and Baveno II/ III (P < 0.0001) criteria. Adding ABRI-A or -B to Baveno V criteria resulted in a significant reduction of the AUROC (P < 0.05). Conclusion: The Baveno V criteria are good predictors of treatment failure of early-stage acute gastrointestinal bleeding in patients with portal hypertension, while the addition of ARBI does not improve the prediction accuracy of the outcome of bleeding. (HEPATOLOGY 2015;61:1033-1040) See Editorial on Page 762 Gastrointestinal bleeding is one of the most important causes of mortality in patients with liver cirrhosis. 1 Several factors have reduced the mortality rate for acute gastrointestinal bleeding in portal hypertension, including the introduction of new pharmacological agents, endoscopic technology, radiologic interventions, surgical interventions, and antibiotic prophylaxis. 1-3 However, despite improvements in management, the 6-week mortality rate of a single bleeding episode remains high, at 15-20%. 1,4 The mortality rate for acute gastrointestinal bleeding in portal hypertension is highest during the first 5 days and declines gradually over the following 6 weeks. 5 If treatments fail to control bleeding, additional treatment strategies need to be considered. 3 Thus, predicting and evaluating the adequacy of hemostasis by nonendoscopic methods is important in the management of acute gastrointestinal bleeding in portal hypertension. The definitions and criteria for failure to control gastrointestinal bleeding in portal hypertension were Abbreviations: ABRI, Adjusted Blood Requirement Index; AUROC, area under the receiver operating characteristic curve; EVL, endoscopic variceal ligation; HCC, hepatocellular carcinoma; NPV, negative predictive value; PPV, positive predictive value. From the 1 Division of Gastroenterology and Hepatology, Department of Internal Medicine, School of Medicine, Kyungpook National University, Daegu, Republic of Korea; 2 Center of Biostatistics, School of Medicine, Kyungpook National University, Daegu, Republic of Korea; 3 Department of Occupational and Environmental Medicine, Dongguk University Gyeongju Hospital, Gyeongju, Republic of Korea. Received November 26, 2013; accepted August 31, 2014. Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.27441/suppinfo. See Editorial on Page 762 1033

1034 AHN ET AL. HEPATOLOGY, March 2015 proposed at the Baveno Consensus Workshops II 6 and were reviewed at the Baveno Consensus Workshop III. 7 However, some of the proposed methods did not adequately reflect clinical practice, 8 which prompted new proposed definitions in the Baveno IV consensus meeting in 2005, and the consensus statements were modified in 2010 based on an observational study assessing the diagnostic performance of Baveno IV definitions and criteria. 5,9,10 The current Baveno V definitions and criteria for failure to control bleeding require that the time frame for the acute bleeding episode should be 120 hours (i.e., 5 days), and define failure as death or the need to change therapy according to one of the following criteria 5 : 1. Fresh hematemesis or nasogastric aspiration of 100 ml of fresh blood 2 hours after the start of a specific drug treatment or therapeutic endoscopy. 2. Development of hypovolemic shock. 3. A 3-g drop in hemoglobin (equivalent to a 9% drop in hematocrit) within any 24-hour period if no transfusion is administered. 4. The potential value of an index of blood transfusion requires prospective validation (e.g., the Adjusted Blood Requirement Index [ABRI] from Baveno IV). While the criteria for failure to control bleeding have been modified between Baveno II to Baveno V, no prospective study has validated the Baveno V criteria. In addition, the usefulness of ABRI has yet to be confirmed. The purpose of this study was to validate the Baveno V criteria for failure to control bleeding based on repeated endoscopic findings on day 5 of an acute bleeding episode, and to determine the clinical usefulness of combined use of Baveno V and ABRI for better prediction of treatment failure. Patients and Methods Study Population and Design. This was a prospective cohort study performed in accordance with the ethical guidelines of the 1975 Declaration of Helsinki. Written consent was obtained from each participant or responsible family member after the possible complications of the diagnostic procedures had been fully explained. This study was approved by the Institutional Review Board of Kyungpook National University Hospital in Daegu, Korea. In total, 246 patients with acute gastrointestinal bleeding from portal hypertension were prospectively recruited between January 2010 and October 2012 at Kyungpook National University Hospital. Inclusion criteria were the presence of liver cirrhosis, being older than 20 years, experiencing onset of melena or hematemesis within 24 hours before arrival, and surviving the first hematocrit measurement. Patients who were pregnant, had a history of variceal bleeding within 6 weeks of admission, had bleeding from lesions unrelated to portal hypertension confirmed by endoscopy, or were diagnosed with cancer other than hepatocellular carcinoma (HCC) were excluded. We defined hypovolemic shock as systolic blood pressure falls to 90 mmhg or less and central venous pressure falls to 3cmH 2 O or less, and evidence of organ hypoperfusion such as oliguria or alteration in mental status or cold sweating with pale skin. The treatment outcome was assessed for 5 days after intensive treatment. When Baveno criteria were met or rebleeding was suspected after intensive treatment began, repeat endoscopy was conducted immediately following the event. We suspected rebleeding in cases of ongoing melena over 48 hours. When there was no event or evidence of bleeding on that repeat endoscopy, endoscopic evaluation was repeated again on day 5 after treatment to determine treatment failure. Definition of Treatment Failure as a Reference Standard. Death within 5 days and active bleeding such as spurting or oozing blood or fresh blood contents in the stomach on repeated endoscopy were regarded as treatment failure. Management of Acute Bleeding From Portal Hypertension. Patients were managed according to the Baveno V therapeutic options. 5 After taking blood samples for laboratory tests, the patients were resuscitated with normal saline fluid and packed red blood cells. Terlipressin acetate was used as a vasoactive agent at an initial intravenous dose of 2 mg, followed by a 1 mg intravenous dose every 4 hours for 3 days. Address reprint requests to: Won Young Tak, M.D., Ph.D., Division of Gastroenterology and Hepatology, Department of Internal Medicine, Kyungpook National University Hospital, 130 Dongdeok-ro, Jung-gu, Daegu, 700-721, Republic of Korea. E-mail: wytak@knu.ac.kr; fax: 182 53 426 2046. Copyright VC 2014 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.27441 Potential conflict of interest: Dr. Tak received grants from Ferring.

HEPATOLOGY, Vol. 61, No. 3, 2015 AHN ET AL. 1035 Antibiotic prophylaxis was instituted. Endoscopies were performed within 12 hours. Endoscopic therapy included either endoscopic variceal ligation (EVL) or endoscopic variceal obturation with N-butyl-2-cyanoacrylate performed as soon as possible by an expert endoscopist. Initial blood samples were taken for complete blood counts, prothrombin time, and liver function tests. The hemoglobin level was monitored every 6 hours for the first 2 days, and every 12 hours for the following 3 days. The target hemoglobin level for transfusion was 8 g/dl. Vital signs were checked every 2 hours for 48 hours and then every 6 hours for the following 3 days. Calculation of ABRI. ABRI values were calculated using the formula defined by Baveno IV as follows: ABRI 5 (blood units transfused)/(final hematocrit initial hematocrit)10.01, where the total amount of packed cells transfused during the first 5 days of treatment was used to quantify the blood units transfused. 9 To determine the appropriate timing for hematocrit measurement, two hematocrit levels measured at different timepoints were used for the initial hematocrit of the ABRI formula: the level measured upon arrival at the emergency room (ABRI-A), and the lowest level measured before transfusion (ABRI-B). The hematocrit on day 5 was used as the final hematocrit measurement. Statistical Analysis. Statistical analysis was performed using the Statistical Package for the Social Science software (v. 20.0 for Windows, IBM SPSS Statistics, Chicago, IL). Data are presented as the mean 6 standard deviation values for continuous variables, and as frequencies and percentages for categorical variables. Logistic regression analyses were performed to assess factors influencing the day-5 treatment outcome. To assess the diagnostic performance of Baveno V criteria, receiver operating characteristic (ROC) curves were constructed and the areas under the ROC curves (AUROCs) were calculated. The sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), and positive and negative likelihood ratio (1LR and LR, respectively) of Baveno V criteria were determined from the ROC curves. To assess the usefulness of ABRI, ROC curves with AUROCs were calculated and optimal ABRI cutoff values were selected using the Youden test. Comparisons of the AUROC values of Baveno V and Baveno II/III or IV, or Baveno V combined with ABRI were determined using the method suggested by DeLong s test using MedCalc software (v. 12.7 for Windows, Mariakerke, Belgium). 11,12 In all analyses, P < 0.05 was considered statistically significant. Table 1. Baseline Characteristics of Study Population Characteristic Results Study Population n 5 246 Age (years) 56.5 6 10.9 (31-88) Male/female 199/47 (80.9/19.1) Initial vital signs Systolic blood pressure (mmhg) 110.0 6 23.6 (50-180) Diastolic blood pressure (mmhg) 65.9 6 14.9 (24-134) Heart rate (bpm) 94.4 6 20.9 (42-160) Child-Turcotte-Pugh classification Class A/B/C 62/120/64 (25.2/48.8/26.0) Underlying liver disease HBV/HCV/alcohol/others 95/18/120/13 (38.6/7.3/48.8/0.4/5.3) Presence of HCC 52 (21.1) Portal vein thrombosis 45 (19.1) First episode of bleeding/ 136/110 (55.3/44.7) recurrent bleeding Source of bleeding Esophageal varix/gastric varix/portal 189/42/15 (76.8/17.1/6.1) hypertensive gastropathy Preventative use of b-blocker 66 (26.8) Laboratory (initial) Hemoglobin (g/dl) 8.7 6 2.3 Hematocrit (mg/dl) 25.7 6 6.4 Platelet count (1000/mm 3 ) 104.6 6 63.6 Aspartate aminotransferase (IU/L) 124.1 6 421.3 Alanine aminotransferase (IU/L) 43.5 6 91.6 Total bilirubin (mg/dl) 2.9 6 4.2 Albumin (g/dl) 2.8 6 0.6 Prothrombin time (sec) 16.3 6 4.9 Data are mean 6 standard deviation, mean 6 standard deviation (range), or n (%) values. HCC, hepatocellular carcinoma; HBV, hepatitis B virus; HCV, hepatitis C virus. Study Population. In total, 280 consecutive patients were screened for possible inclusion in the study. Thirty-four patients were excluded: 21 patients who had no bleeding from portal hypertension, 12 patients with a history of bleeding within 6 weeks of admission, and one patient with esophageal cancer. The baseline characteristics of the 246 included patients are presented in Table 1. There were 199 male patients (80.9%) and 47 female patients (19.1%) with a mean age of 56.5 years (range, 31-88 years). According to the Child-Turcotte-Pugh classification performed on admission day, 62 (25.2%) of the population were class A, 120 (48.8%) were class B, and 64 (26.0%) were class C. The causes of cirrhosis were hepatitis B virus (n 5 95, 38.6%), hepatitis C virus (n 5 18, 7.3%), alcohol (n 5 120, 48.8%), and others (n 5 13, 5.3%). Sources of bleeding were esophageal varices (n 5 189, 76.8%), gastric varices (n 5 42, 17.1%), and acute diffuse bleeding with portal hypertensive gastropathy (n 5 15, 6.1%). There were 52 (21.1%) cases with HCC and 45 (19.1%) cases with portal vein thrombosis including benign portal vein

1036 AHN ET AL. HEPATOLOGY, March 2015 Table 2. Factors Affecting the Outcome of Acute Gastrointestinal Bleeding Associated With Portal Hypertension Univariate Analysis Multivariate Analysis Variable Odds ratio 95% CI p Odds ratio 95% CI p Age >55 years 0.933 0.520-1.765 0.824 Male 1.718 0.721-4.093 0.222 Initial vital sign Systolic blood pressure <80 mmhg, 0.612 0.435-4.115 0.612 Diastolic blood pressure <60 mmhg 1.313 0.720-2.459 0.361 Heart rate >100/min 0.608 0.313-1.180 0.141 Child-Turcotte-Pugh classification B (vs. A) 1.687 0.709-4.015 0.237 C (vs. A) 3.297 1.330-8.169 0.010 2.199 1.133-4.267 0.02 Underlying liver disease HBsAg (vs. anti-hcv) 0.775 0.410-1.465 0.433 Alcohol (vs. anti-hcv) 1.105 0.602-2.031 0.747 HCC 1.882 0.876-4.045 0.105 Portal vein thrombosis 1.679 0.787-3.582 0.208 More than one previous bleeding 1.969 1.036-3.743 0.039 2.104 1.091-4.061 0.027 Esophageal varix 1.161 0.552-2.440 0.694 Gastric varix 0.861 0.410-1.811 0.694 Preventative use of b-blocker 0.515 0.217-1.222 0.132 Laboratory (initial) Hematocrit <24 mg/dl 0.82 0.439-1.532 0.319 Platelet count <100,000/mL 0.626 0.340-1.152 0.312 Total bilirubin >2.0 mg/dl 1.585 0.856-2.935 0.143 Albumin <3.5 g/dl 1.295 0.467-3.590 0.62. Prothrombin time >14 sec 1.986 0.934-4.221 0.074 CI, confidence interval; HBsAg, hepatitis B surface antigen; anti-hcv, antibody to HCV. thrombosis. The volume of mean packed cells transfused was 2.6 units (range, 0-10 units) over the 5-day observation period. Band ligations were performed in 165 patients. Varices in 26 patients were obturated endoscopically. Treatment Results on Day 5. After initial treatment began, 47 patients experienced the following events: fresh hematemesis (n 5 33), hypovolemic shock (n 5 10), and a 3-g drop in hemoglobin (n 5 4). Of these 47 patients, 7 were unable to submit to repeated endoscopy due to unstable vital signs from recurrent hematemesis or a confused mental state. All seven of these patients died and so were regarded as treatment failures. The remaining 40 patients underwent immediate endoscopy, which revealed active bleeding in 32 patients; the other eight patients showed no evidence of bleeding on endoscopy. Of these eight patients, five died and so were regarded as treatment failures. Two patients experienced ongoing melena over 48 hours without Baveno V criteria and showed active bleeding at immediate endoscopy. Repeated endoscopy was performed on day 5 in 171 patients who had no event and three patients who had a 3-g drop in hemoglobin but clean endoscopic findings. The repeated endoscopy revealed active bleeding in 7 of these 171 patients who had no event (Supporting Figure). None of the three patients who had events but clean endoscopic findings showed evidence of bleeding on repeated endoscopy. Of the total 246 patients, 193 (78.5%) were treated successfully. Treatment failures were identified in 53 patients (21.5%) over the 5-day observation period, including 24 patients who died during this period. The causes of death were rebleeding (n 5 12, 50%), hepatic failure (n 5 7, 29.2%), and sepsis (n 5 5, 20.8%). The factors affecting the control of acute gastrointestinal bleeding in portal hypertension were Child-Turcotte-Pugh class C and history of a previous bleeding event on multivariate analysis (P < 0.05; Table 2). Assessment of Baveno V for Failure to Control Bleeding. According to Baveno V criteria, 47 patients were regarded as treatment failures, 24 of whom died during the 5-day period of the study. Fresh hematemesis or nasogastric aspiration of 100 ml of fresh blood was observed in 18 patients, and a 3-g drop in hemoglobin within the 24-hour period occurred in four patients without a blood transfusion. One patient went into hypovolemic shock (Table 3). The results of standard reference and Baveno V criteria were discordant in 12 cases (Table 4). Among these patients, nine were identified as treatment failures based on repeated endoscopic findings. In seven out of nine patients, oozing blood was found following band

HEPATOLOGY, Vol. 61, No. 3, 2015 AHN ET AL. 1037 Table 3. Treatment Failure to Control Bleeding According to Each Set of Criteria Number of Cases Baveno V 47 Death or 24 Fresh hematemesis or nasogastric aspiration of 18 100 ml of fresh blood 2 h after the start of a specific drug treatment or therapeutic endoscopy 3-g drop in hemoglobin within any 24-h period 4 if no transfusion is administered Hypovolemic shock 1 Baveno II/III 124 Transfusion of 4 units of blood or more within 8 the first 6 h Inability to achieve an increase in systolic 13 blood pressure of 20 mmhg or to 70 mmhg or more within the first 6 h Inability to achieve a pulse reduction of 20/ 29 min from baseline or to <100/min within the first 6 h Occurrence of hematemesis after 6 h 18 Reduction in blood pressure of more than 20 10 mmhg from the 6-h point Increase in pulse rate of more than 20/min 7 from the 6-h point on two consecutive readings taken 1 h apart Transfusion of 2 units of blood or more (over 39 and above the previous transfusions) required to increase the hematocrit to above 27% or hemoglobin to above 90 g/l Baveno IV One criterion defines failure, whichever occurs 105 first Death 6 Fresh hematemesis or nasogastric aspiration of 36 100 ml of fresh blood 2 h after the start of a specific drug treatment or therapeutic endoscopy 3-g drop in hemoglobin within any 24-h period 4 if no transfusion is administered ABRI0.75 59 ABRI, Adjusted Blood Requirement Index. ligation or variceal obturation of an ulcer on day 5. Two of the nine patients had melena over 48 hours, and they exhibited active bleeding on repeated endoscopy. Three patients were considered treatment successes based on repeated endoscopic findings, even though they had a 3-g drop in hemoglobin. These patients initially underwent EVL, and the EVL sites were well preserved, with no active bleeding and no evidence of bleeding (e.g., fresh blood clots) on repeated endoscopy. They underwent a further followup endoscopy on day 5, with the same endoscopic findings. The AUROC of Baveno V was 0.906 (95% confidence interval [95% CI]: 0.844-0.967). Se, Sp, PPV, NPB, 1LR, and LR were 83.0% (95% CI: 70.2-91.9), 98.4% (95% CI: 95.5-99.7), 93.6% (95% CI: 82.4-98.6), 95.5% (95% CI: 91.6-97.9), 53.41 (95% CI: 17.26-165.23), and 0.17 (95% CI: 0.10-0.31), respectively (Table 5). Comparison of Baveno V Criteria With Baveno IV and Baveno II/III Criteria. Based on Baveno IV and Baveno II/III criteria, 105 (42.7%) and 124 (50.4%) cases were regarded as treatment failures. The AUROC was significantly larger for Baveno V than for Baveno IV (P 5 0.0001) and Baveno II/III (P < 0.0001; Fig. 1, Table 5). Definition of ABRI and Evaluation of the Clinical Usefulness of Combined Baveno V and ABRI. The AUROCs of ABRI-A and ABRI-B were 0.637 6 0.048 (P 5 0.002; 95% CI: 0.0536-0.732) and 0.605 6 0.050 (P 5 0.021; 95% CI: 0.508-0.702), respectively. The cutoff values of ABRI-A and ABRI-B were 0.99 (Se: 60.4%, Sp: 78.8%) and 0.66 (Se: 50.9%, Sp: 80.3%). For Baveno V with ABRI-A 0.99, the AUROC was 0.844 6 0.029 (95% CI: 0.786-0.902); Se, Sp, PPV, NPV, 1LR, and LR were 90.6% (95% CI: 79.3-96.8), 78.2% (95% CI: 71.7-83.8), 53.3% (95% CI: 42.5-63.9), 96.8% (95% CI: 92.7-98.9), 4.16 (95% CI: 3.14-5.51), and 0.12 (95% CI: 0.05-0.28), respectively (Table 6). For Baveno V with ABRI-B 0.66, the AUROC was 0.852 6 0.029 (95% CI: 0.795-0.909); Se, Sp, PPV, NPV, 1LR, and LR were 90.6% (95% CI: 79.3-96.8), 79.8% (95% CI: 73.4-85.2), 55.2% (95% CI: 44.1-65.8), 96.7% (95% CI: 92.8-98.9), 4.48 (95% CI: 3.34-6.01), 0.12 (95% CI: 0.05-0.27), respectively (Table 6). The AUROC was larger for Baveno V alone than for Baveno V plus ABRI-A (the difference between the two AUROCs was 0.062, P 5 0.0067) and Baveno V plus ABRI-B (the difference between the AUROCs was 0.054, P 5 0.0162), with discrepant findings in 52 patients when using ABRI-A and 31 patients when using ABRI-B. Discussion It is important to have standard definitions for key events because they are used as endpoints in clinical studies, as well as treatment endpoints in clinical practice. Although gastrointestinal bleeding is one of the most serious complications in portal hypertension, the definition of failure to control acute bleeding remains Table 4. Assessment of Baveno V Definitions and Criteria Compared to Repeated Endoscopic Results Standard Reference Success Failure Total Baveno V Success (n) 190 9 199 Failure (n) 3 44 47 Total (n) 193 53 246

1038 AHN ET AL. HEPATOLOGY, March 2015 Table 5. Comparison of Baveno V Definitions and Criteria and Baveno IV or Baveno II/III Criteria AUROC 6 SE Se (%) Sp (%) PPV (%) NPV (%) Positive LR Negative LR Differences Between AUROCs p Baveno V 0.906 6 0.031 (0.844-0.967) Baveno IV 0.779 6 0.035 (0.711-0.848) Baveno II/III 0.594 6 0.043 (0.505-0.680) 83.0 (70.2-91.9) 84.9 (72.4-93.2) 66.0 (51.7-78.5) 98.4 (95.5-99.7) 70.9 (64.0-77.3) 53.9 (46.6-61.1) 93.6 (82.4-98.6) 44.6 (34.7-54.8) 28.2 (20.5-37.0) 95.5 (91.6-97.9) 94.5 (89.4-97.6) 85.2 (77.7-91.0) 53.41 (17.26-165.23) 2.93 (2.29-3.75) 1.43 (1.12-1.83) 0.17 (0.10-0.31) 0.21 (0.11-0.41) 0.63 (0.42-0.94) 0.127 0.0001 0.312 <0.0001 SE, standard error; Se, sensitivity; Sp, specificity; PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio; AUROC, area under the receiver operating characteristic curve. controversial. The need for a reliable and noninvasive method for assessing treatment failure led to the introduction of the Baveno definitions and criteria for failure to control acute bleeding in portal hypertension, which were modified as the Baveno V criteria in 2010. 5 However, these definitions and criteria had yet to be validated. This is the first prospective study to assess the diagnostic performance of Baveno V criteria in controlling bleeding in cirrhosis patients. The main problem with verifying the definition of failure to control bleeding is that there are few standard references. Furthermore, few studies have investigated the Baveno IV criteria. One study used expert clinical consensus as a standard definition, while another found that ABRI 0.75 was correlated with day-7 mortality. Yet another study compared ABRI 0.75 with other Baveno criteria. 10,13,14 An objective reference is thus needed to validate the new criteria. Fig. 1. Receiver operating characteristic curves for Baveno V, Baveno IV, and Baveno II/III criteria. The present study confirmed the treatment results based on repeated endoscopy for the diagnosis of recurrent bleeding within 5 days. Decision-making based on endoscopic findings has the potential for interobserver variability, but many studies have demonstrated that endoscopy is the most objective method for confirming upper gastrointestinal bleeding. We defined the criteria of rebleeding as active bleeding such as spurting or oozing blood, or fresh blood contents in the stomach on repeated endoscopy. All patients regarded as exhibiting rebleeding on repeated endoscopy exhibited active bleeding, and none of the patients exhibited blood contents without active bleeding in this study. Bleeding control failure was confirmed by repeated endoscopy in 213 patients. Of the 33 patients who did not undergo repeated endoscopy, 7 died and 26 were maintained in a stable condition. The latter 26 patients were regarded as treatment successes because they had stable vital signs and laboratory findings without transfusion during the 5-day observation period, and maintained these findings for over 2-6 weeks after discharge from the hospital. These stable patients refused to submit to a repeat endoscopy, but were included in the study to avoid further selection bias, even though it is possible that determining a treatment result without endoscopy can lead to overestimating the value of the Baveno V criteria. Statistically, about one patient might have active bleeding in these patients because 4.0% (7 out of 174) patients who did not require an endoscopy within 5 days eventually displayed active bleeding at endoscopy on day 5 in our study. However, that risk was reduced by observing these patients for 2-6 weeks after discharge. The early mortality and treatment failure rates yielded by the present study (9.8% and 21.5%, respectively) are lower than those reported previously (15-20% and 28%, respectively), 1,10 demonstrating that strictly following the Baveno V therapeutic recommendation of acute bleeding in portal hypertension provides a higher success rate. Although vasoactive drugs

HEPATOLOGY, Vol. 61, No. 3, 2015 AHN ET AL. 1039 Table 6. Comparison of Baveno V Definitions and Criteria and Baveno V Plus ABRI Criteria AUROC 6 SE Se (%) Sp (%) PPV (%) NPV (%) Positive LR Negative LR Differences Between AUROCs p Baveno V 0.906 6 0.031 (0.844-0.967) Baveno V 1 0.844 6 0.029 ABRI-A (0.786-0.902) Baveno V 1 0.852 6 0.029 ABRI-B (0.795-0.909) 83.0 (70.2-91.9) 90.6 (79.3-96.8) 90.6 (79.3-96.8) 98.4 (95.5-99.7) 78.2 (71.7-83.8) 79.8 (73.4-85.2) 93.6 (82.4-98.6) 53.3 (42.5-63.9) 55.2 (44.1-65.8) 95.5 (91.6-97.9) 96.8 (92.7-98.9) 96.7 (92.8-98.9) 53.41 (17.26-165.23) 4.16 (3.14-5.51) 4.48 (3.34-6.01) 0.17 (0.10-0.31) 0.12 (0.05-0.28) 0.12 (0.05-0.27) 0.062 0.0067 0.054 0.0162 should be continued for 5 days based on Baveno V therapeutic options for acute bleeding in portal hypertension, we used only 20 vials per patient because of national insurance reimbursement guidelines. However, the optimal duration of pharmacological therapy is not well established and there is consensus that the most appropriate duration of terlipressin therapy would be anywhere between 2 and 5 days. 3,5 A recent randomized, double-blind, dummy-controlled trial showed that a 24-hour course of adjuvant terlipressin in variceal bleeding was as effective as a 72-hour course; thus, we believe that the administration of terlipressin was sufficient in the present study. 15 In the present study, the primary endpoint confirming the usefulness of Baveno V criteria was based on the endoscopy performed on day 5 and death. Baveno V criteria can be used to discriminate treatment failure of acute gastrointestinal bleeding in portal hypertension with good values for AUROC, Se, Sp, PPV, NPV, 1LR, and LR. Discrepancy was found for only 12 cases (4.9%), three of which had treatment success, as confirmed by repeated endoscopy, with initial hemoglobin levels of 12.3-13.8 g/dl. Although they did not initially need a transfusion, previous blood loss and initial hydration during resuscitation might cause a rapid 3-g drop in hemoglobin. The Asian Pacific Association of the Study of the Liver (APASL) has suggested that a 2-g drop in hemoglobin occurs if no transfusion is administered. 3 In the present study, 44 cases did not receive blood transfusions of red blood cells, and seven of them experienced a 2-g drop in hemoglobin over a 24-hour period. However, they did not all show evidence of treatment failure. Previous blood loss and initial hydration during resuscitation might cause a rapid 2-g drop in hemoglobin. If such a drop in hemoglobin is regarded as a treatment failure, there is a high probability of an unnecessary repeat endoscopy. It is therefore clear that further study of the decreased hemoglobin rate is needed. Among the 12 discordant cases, nine were treatment failures, of whom seven exhibited blood oozing at the endoscopic treatment sites with stable vital signs and a gradually decreased hemoglobin level, and two had sustained melena for over 48 hours, with active bleeding revealed on endoscopy. These results suggest that treatment failure should be considered when the patient s hemoglobin declines continuously without signs of rebleeding or when melena is sustained for over 48 hours. The Baveno V definitions and criteria for failure of treatment for acute bleeding are superior to previous Baveno criteria. Baveno II/III criteria exhibited a small AUROC and low Se, Sp, PPV, and NPV, which were difficult to apply in practice. Although Baveno IV criteria had a relatively high Se (84.9%), we identified 64 discrepant cases (26.0%) with lower Sp (70.9%) and PPV (44.6%). The main difference between Baveno IV and Baveno V criteria was the inclusion of ABRI 0.75; applying ABRI 0.75 as a criterion resulted in many more cases being considered treatment failures. Further study is therefore required to establish the timing of hemoglobin measurement and the ABRI cutoff. The usefulness of ABRI, which was proposed in the Baveno IV criteria, was also evaluated in the present study. We compared ABRI in combination with Baveno V criteria to the application of Baveno V criteria alone. For ABRI calculation, two timepoints were accessed to determine the optimum time for determining the initial hematocrit: the first level measured and the lowest level measured before the start of transfusion. These timepoints were chosen because packed red cells were not always available until the initial treatment had begun, and the fluid administered before transfusion would have diluted the blood. These timepoints were used to determine new cutoff values for ABRI-A and ABRI-B. Combining Baveno V criteria with ABRI-A 0.99 or ABRI-B 0.66 identified 52 and 31 discrepant cases in this study, respectively, and the AUROCs of Baveno V and Baveno V with ABRI differed significantly. Consistent with previous reports, this suggests that using the blood transfusion index provides no benefit. 10,13 A negative ABRI-A was seen in 21 of the 52 discrepant cases, which indicates that

1040 AHN ET AL. HEPATOLOGY, March 2015 the final hematocrit was lower than the initial hematocrit. Since we assumed that a negative ABRI was the result of an ineffective transfusion, we defined a negative ABRI as treatment failure, as reported elsewhere. 10 To avoid a negative ABRI, we recommend that the initial hematocrit values be checked more often than every 6 hours and that transfusion be performed earlier, although this may be difficult in practice. Thus, the ABRI value might be inconsistent under various conditions and might not provide any benefit. Study Limitations. This study was subject to several limitations. First, we chose endoscopy for a reference standard. There was a possibility that our methods, performing endoscopy within 5 days on every patient who met one of the Baveno criteria, or had a suspicion of bleeding, favor Baveno criteria. However, currently, endoscopy is the most objective method for confirming upper gastrointestinal bleeding. Thus, further study is needed if a new, more perfect monitoring method is available in the future. Second, we were unable to perform repeated endoscopy in all patients because they were either in critical condition or else stable but refused the follow-up endoscopy. The study design followed normal clinical practice and repeated endoscopy was performed in 17 of the 24 patients who died. However, all 24 of these patients were classified as treatment failures, regardless of the endoscopic findings. Rebleeding was confirmed on repeated endoscopy in all treatment failure patients, with the exception of the seven who died. About 10.6% (n 5 26) of patients who refused the repeat endoscopy were regarded as treatment successes because they were stable without transfusion during the 5-day observation period or for over 2-6 weeks during the follow-up period after discharge from the hospital. This is a small number, and represents a limitation of this study. Another limitation is that we did not assess bleeding control at 6 weeks, thus presuming that the 5-day result was related to the 6-week treatment result. Further study is needed to determine whether this 5-day criterion can actually predict the results at 6 weeks. Finally, our study population was of homogeneous ethnicity; the influence of race or ethnicity should be addressed in a prospective multicenter, multiethnicity study. In conclusion, this prospective study shows that Baveno V definitions and criteria, but not ABRI, can predict the outcome of acute gastrointestinal bleeding management in patients with portal hypertension during the initial stages. References 1. 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;40:652-659. 2. Majid S, Azam Z, Shah HA, Salih M, Hamid S, Abid S, et al. Factors determining the clinical outcome of acute variceal bleed in cirrhotic patients. Ind J Gastroenterol 2009;28:93-95. 3. Sarin SK, Kumar A, Angus PW, Baijal SS, Baik SK, Bayraktar Y, et al. Diagnosis and management of acute variceal bleeding: Asian Pacific Association for Study of the Liver recommendations. Hepatol Int 2011;5:607-624. 4. Jansen PL. Life expectancy in liver cirrhosis after the first variceal hemorrhage: how can survival be improved? Scand J Gastroenterol Suppl 1990;178:106-110. 5. 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;53:762-768. 6. de Franchis R. Developing consensus in portal hypertension. J Hepatol 1996;25:390-394. 7. de Franchis R. Updating consensus in portal hypertension: report of the Baveno III Consensus Workshop on definitions, methodology and therapeutic strategies in portal hypertension. J Hepatol 2000;33:846-852. 8. de Franchis R. Review article: definition and diagnosis in portal hypertension continued problems with the Baveno consensus? Aliment Pharmacol Ther 2004;20(Suppl 3):2-6; discussion 7. 9. 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;43:167-176. 10. Thabut D, D Amico G, Tan P, De Franchis R, Fabricius S, Lebrec D, et al. Diagnostic performance of Baveno IV criteria in cirrhotic patients with upper gastrointestinal bleeding: analysis of the F7 liver-1288 study population. J Hepatol 2010;53:1029-1034. 11. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-845. 12. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839-843. 13. Akhtar N, Zuberi BF, Hasan SR, Kumar R, Afsar S. Determination of correlation of Adjusted Blood Requirement Index with outcome in patients presenting with acute variceal bleeding. World J Gastroenterol WJG 2009;15:2372-2375. 14. Duvnjak M, Barsic N, Tomasic V, Jukic LV, Lerotic I, Pavic T. Adjusted blood requirement index as indicator of failure to control acute variceal bleeding. Croat Med J 2006;47:398-403. 15. Azam Z, Hamid S, Jafri W, Salih M, Abbas Z, Abid S, et al. Short course adjuvant terlipressin in acute variceal bleeding: a randomized double blind dummy controlled trial. J Hepatol 2012;56:819-824. Supporting Information Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.27441/suppinfo.