Alcoholic hepatitis (AH) is a severe clinical

Transcription

1 HEPATOLOGY, VOL. 66, NO. 6, 2017 AMERICAN ASSOCIATION FOR THE STUDY OFLIVERD I S E ASES STEATOHEPATITIS/METABOLIC LIVER DISEASE Alcohol Abstinence in Patients Surviving an Episode of Alcoholic Hepatitis: Prediction and Impact on Long-Term Survival Jose Altamirano, 1* Hugo Lopez-Pelayo, 2* Javier Michelena, 1 Patricia D. Jones, 5 Lluisa Ortega, 2 Pere Ginès, 1,3 Juan Caballerıa, 1,3 Antoni Gual, 2 Ramon Bataller, 4,6** and Anna Lligo~na 2** Alcoholic hepatitis (AH) is the most severe form of alcoholic liver disease. Most studies have focused on short-term prognosis, whereas factors associated with long-term survival are largely unknown. The aims of our study were to (1) determine the impact of complete abstinence from alcohol on long-term survival and (2) identify prognostic factors at admission capable of predicting abstinence during long-term follow-up in patients with AH. One hundred forty-two patients with biopsy-proven AH that survived the first episode were included. Demographic, psychiatric, and biochemical variables at admission and drinking status during follow-up were obtained. Cox regression, logistic regression, and classification and regression trees (CART) analyses were used for statistical analysis. Overall mortality was 38% with a median follow-up of 55 months. During follow-up, complete abstinence was reported in 39% and was associated with better long-term survival (hazard ratio, 0.53; P ). After adjustment for baseline prognostic scoring systems (Model for End-Stage Liver Disease and age, bilirubin, international normalized ratio, creatinine scores), complete abstinence was independently associated with survival (P < 0.05). Age and lack of past alcoholism treatments were independently associated with complete abstinence (P < and P , respectively) during follow-up. CART analysis generated a simple and practical algorithm based on the combination of past alcoholism treatments and age. Using CART analysis, we stratified 2 subgroups of patients with high (65%) and low (26%-29%) rates of complete abstinence after an episode of AH. Conclusion: Complete abstinence after an episode of AH positively impacts long-term survival. The combination of 2 variables easily obtained at admission might be useful to predict long-term abstinence after an episode of AH. Strategies aimed at promoting alcohol abstinence in these patients are necessary. (HEPATOLOGY 2017;66: ). SEE EDITORIAL ON PAGE 1722 Alcoholic hepatitis (AH) is a severe clinical entity characterized by acute onset of jaundice- and liver-related complications in patients with severe alcohol use disorders (AUD). It typically occurs in the setting of preexisting liver disease, and most patients have underlying cirrhosis. (1) The short-term mortality of AH remains very high globally, (1-3) and 15%-40% of patients die within the first 30 days. (4) Short-term prognosis can be predicted by prognostic models, including Maddrey s Abbreviations: ABIC, age, bilirubin, international normalized ratio, creatinine; AH, alcoholic hepatitis; ALD, alcoholic liver disease; ALT, alanine aminotransferase; AST, aspartate aminotransferase; AU, addiction unit; AUD, alcohol use disorders; AUROC, area under the receiving operator characteristics curve; CART, classification and regression trees; CBT, cognitive-behavior therapy; CMC, comprehensive medical care; DF, discriminant function; DSM-IV-TR, Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision; GGT, gamma-glutamyl transpeptidase; HRs, hazard ratios; HRAR, High-Risk Alcoholism Relapse; INR, international normalized ratio; IQR, interquartile range; MELD, Model for End- Stage Liver Disease; MET, motivational enhancement therapy; SD, standard drink; LT, liver transplantation; ORs, odds ratios; WBC, white blood cell. Received November 8, 2016; accepted June 21, Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/ /hep.29338/suppinfo. Supported by the National Institute on Alcohol Abuse and Alcoholism (1U01AA ). J.A. wishes to express his gratitude to the Mexican National Council of Science and Technology (CONACyT, Mexico City, Mexico) for partially supporting his predoctoral stay at IDIBAPS. *These authors share first co-authorship. **These authors share senior co-authorship. 1842

2 HEPATOLOGY, Vol. 66, No. 6, 2017 ALTAMIRANO, L OPEZ-PELAYO, ET AL. discriminant function (DF), (5) Model for End-Stage Liver Disease (MELD), (6) Glasgow Alcoholic Hepatitis Score, (7) and the age, bilirubin, international normalized ratio (INR), and creatinine (ABIC) score. (8) Moreover, the Lille model (9) incorporates response to prednisolone in determining risk of death. Medical management of AH has not evolved substantially in the last two decades. First-line therapy consists of corticosteroids. (5,9-11) For patients that do not respond to first-line therapy, there are no effective alternative drugs. (12-16) The only approach that improves survival in these patients is salvage liver transplantation (LT). (17,18) However, many patients with a severe episode are unlikely to survive the 6 months of abstinence typically required before consideration for transplant. The recent STOPAH (19) trial only showed a beneficial effect of prednisolone at day 28, whereas neither prednisolone nor pentoxifylline was effective after that period. Importantly, half of the patients died in 1 year and only one third remained abstinent. This high rate of recidivism has been confirmed in other studies. (20) This seminal study suggests that alcohol abstinence is only achieved in a minority of patients and that persistent alcohol intake could influence long-term survival. There are no studies assessing the predictors of alcohol abstinence in this patient population, and there are limited data on the influence of alcohol intake on patient outcome. The current study was undertaken to fill this gap. Although abstinence from alcohol is routinely recommended after an episode of AH, there are no studies assessing motivational or pharmacological therapies to promote abstinence. Similarly, no studies have identified parameters at admission that predict long-term abstinence. The risks of resuming alcohol use should not be understated. In one study, a significant proportion of patients with recidivism developed a subsequent episode of AH that was more severe than the index episode with 60% mortality. (21) Identification of patients with high risk of recidivism is relevant because it can be used to select patients for salvage transplantation and identify those patients needing more-intense alcohol therapy. (22,23) However, there are no studies evaluating factors associated with long-term abstinence after an episode of AH. We hypothesize that persistent alcohol use is associated with poorer long-term survival. The aims of our study were to (1) determine impact of complete abstinence from alcohol on long-term survival and (2) investigate the predicting factors associated with long-term abstinence in patients with AH. Copyright VC 2017 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI /hep Potential conflict of interest: Dr. Ortega received grants from Lundbeck. Dr. Lligona received grants from Lundbeck. Dr. Bataller is on the speakers bureau for Echosens. Dr. Pelayo is on the speakers bureau for Teva, Janssen, and Lundbeck. He received grants from Pfizer, Otsuka, Rovi, Esteve, and Lilly. Dr. Gual is on the speakers bureau for Lundbeck and Teva. ARTICLE INFORMATION: From the 1 Institut d Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain; 2 Grup Recerca Addiccions Clinic (GRAC-GRE) Psychiatry Department, Neurosciences Institute, Hospital Clınic of Barcelona, Red de Trastornos Adictivos (RTA), Barcelona, Spain; 3 Liver Unit, CIBER de Enfermedades Hepaticas y Digestivas (CIBERehd), Institucio Catalana de Recerca i Estudis Avançats (ICREA), Universitat de Barcelona, Barcelona, Spain; 4 Division of Gastroenterology and Hepatology, Departments of Medicine and Nutrition, University of North Carolina at Chapel Hill, Chapel Hill, NC; 5 Division of Gastroenterology and Hepatology, University of Miami Miller School of Medicine, Miami, FL; and 6 Pittsburgh Liver Research Center, Division of Gastroenterology, Hepatology and Nutrition, University of Pittsburgh Liver Center, Pittsburgh, PA. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO: Anna Lligo~na, M.D. Grup de Recerca Addiccions Clinic (GRAC-GRE), Psychiatry Department, Neurosciences Institute, Hospital Clınic Mallorca Barcelona, Spain alligona@clinic.ub.es Tel: ; or Ramon Bataller, M.D., Ph.D. Division of Gastroenterology, Hepatology and Nutrition, University of Pittsburgh Medical Center (UPMC) Scaife Hall, 8th floor, S844A 3550 Terrace Street Pittsburgh, PA Bataller@pitt.edu Tel:

3 ALTAMIRANO, L OPEZ-PELAYO, ET AL. HEPATOLOGY, December 2017 Patients and Methods STUDY COHORT We included patients admitted to the liver unit (Hospital Clınic, Barcelona, Spain) from 1999 to 2012 with biopsy-proven AH included in several prospective studies. (2,4,8,24) Inclusion criteria were presence of heavy drinking (alcohol consumption >60 g/ day for more than 1 year) preceding admission, moderately elevated aminotransferases, aspartate aminotransferase (AST) levels higher than alanine aminotransferase (ALT), high gamma-glutamyl transpeptidase (GGT), and abrupt arise in bilirubin serum levels. In all cases, AH was confirmed histologically by the presence of hepatocellular damage (hepatocellular ballooning and presence of Mallory bodies), inflammatory infiltrate (predominantly polymorphonuclear cells), and pericellular fibrosis. Patients with hepatocellular carcinoma, any other potential cause of liver disease, and those who died during index hospitalization were excluded from the study. In our hospital, the vast majority of physicians use the ABIC score to stratify patients with severe AH, given that it was generated in our center and accurately identifies patients with high short-term mortality. Accordingly, severe AH was defined as Maddrey s DF >32 and/or ABIC score >6.71 at admission. All patients with AH received general support measures, and severe cases were treated with 40 mg of prednisone orally every 24 hours for 4 weeks followed by a taper period of 2 weeks. Treatment was discontinued in nonresponding patients as assessed by Lille score >0.45, according to the center s guidelines during patients admission. All patients underwent bacterial infection screening at admission. During hospitalization, patients with clinical complications, such as ascites, spontaneous bacterial peritonitis, renal dysfunction, overt hepatic encephalopathy, or gastrointestinal bleeding associated with portal hypertension, as well as bacterial infections nonrelated to portal hypertension, were diagnosed and treated according to international guidelines and the clinical protocol of the liver unit in effect when patients were admitted. As a part of the protocol for patients with alcoholic liver disease (ALD) admitted to our liver unit, a team composed of a psychiatrist, psychologist, and social worker carefully evaluated patients. After hospital discharge, patients were referred to the addiction unit (AU) and the liver unit of our center for subsequent evaluation and follow-up. The ethics committee of the Hospital Clinic approved the study and all patients gave written informed consent (CEIC 2011/707). CLINICAL AND PSYCHOSOCIAL EVALUATIONS AT ADMISSION Demographic and analytical parameters, assessed within 48 hours from hospital admission, included age, sex, serum glucose, creatinine, sodium, bilirubin, AST, ALT, GGT, albumin, INR, leukocyte count, platelet count, hematocrit, and hemoglobin levels. The different AH scoring systems were calculated based on the laboratory values at admission and or within 48 hours of admission. For purposes of this study, a psychiatrist from our AU performed a full interview of every included patient and, when needed, to close relatives. This interview included: psychiatric and social parameters, such as alcohol dependency (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision; DSM-IV-TR), cognitive deficits (Mini Mental State Examination), presence/absence of an underlying psychiatric disease (DSM-IV-TR), tobacco consumption, amount of alcohol consumption, years of heavy alcohol consumption, and previous treatments for alcoholism. When reports of alcohol consumption variables conflicted, the highest estimate was used. Diagnosis of alcohol abuse and dependence was made according to the criteria detailed in the DSM-IV. For each patient, alcohol consumption was classified using the High-Risk Alcoholism Relapse (HRAR) scale. The HRAR scale was originally developed from a cohort of male U.S. veterans and includes three parameters (duration of heavy drinking, usual number of daily drinks, and number of past alcoholism treatment experiences) empirically evaluated to estimate the risk of alcoholism recidivism. (25) The HRAR scale has recently shown good accuracy to predict harmful drinking in patients undergoing LT for alcoholic cirrhosis. (25) We adapted the number of drinks to the amount of alcohol of one standard drink (SD) in Europe (e.g., 1 SD 5 10 g of pure ethanol). Each item can be scored 0, 1, or 2 for a total possible score ranging from 0 to 6 (Supporting Table S1). FOLLOW-UP AND DEFINITIONS After hospital discharge from the index AH episode, patients were referred to the AU for regular follow-up. Our AU has extensive experience in the psychological and psychiatric treatments of patients with ALD. (26-28) Data about past alcohol consumption, consumption of 1844

4 HEPATOLOGY, Vol. 66, No. 6, 2017 ALTAMIRANO, L OPEZ-PELAYO, ET AL. psychotropic drugs, alcohol abstinence, time since last alcohol consumption, and the number of previous alcohol detoxifying treatments were carefully collected. All collected data from patients were always compared with reports from relatives or accompanying persons. In the event of recidivism, information about the number of SDs consumed was collected. For the purpose of this study, data about long-term abstinence and alcohol recidivism were retrospectively collected by (1) telephone survey to the patient and/or patient s relatives (see the Supporting Information for Survey; n 5 98), (2) in-person interviews (n 5 15), and/or (3) review of clinical records from hospital and primary care medicine units (n 5 29). All data recorded from the telephonic interviews were also contrasted with data from electronic charts of the primary care centers (shared electronic charts). Data about clinical decompensations after index admission for AH and LT during follow-up were also recorded retrospectively. Alcohol abstinence was evaluated on the basis of the patient s self-report, family member interviews, and/or presence of alcohol in urine in those attending the hospital AU. Alcohol recidivism was defined as any amount of regular alcohol use (consumption at least once a week) after index hospitalization for AH. Overall survival was calculated from the date of hospitalization for the initial episode of AH until the date of death or the date the patient was last known to be alive. Those undergoing LT were censored alive at the date of transplantation. Current hospital protocols for LT consider only those patients having >6 months of alcohol abstinence to candidates. None of the patients in this cohort underwent salvage LT. STATISTICAL ANALYSIS Continuous variables were described as median (interquartile range [IQR], 25-75). Categorical variables were described by means of counts and percentages. Comparisons between groups were performed using the Student t test or Mann-Whitney U test, when appropriate. Differences between categorical variables were assessed by the chi-square test or Fisher s exact test, when necessary. To investigate variables with prognostic information for long-term survival and alcohol recidivism, those that were statistically significant (P < 0.05), and those considered clinically relevant at univariate analyses were entered into multivariate analysis. P values for the univariate tests were not corrected for multiple testing, because those tests were exploratory. Results of the multivariate logistic regression and Cox regression analyses (odds ratios [ORs] and hazard ratios [HRs], respectively) determined those variables independently associated with the main outcomes survival and recidivism (after adjusting for the contributions of other variables). In order to avoid colinearity, those variables included in a prognostic score (e.g., ABIC score and MELD score) were not included separately in the multivariate models. To avoid overfitting, a predefined ratio of candidate prognostic variables to the number of observed events (e.g., deaths or alcohol recidivism [y/n]) was set at 1:10. In order to evaluate the influence of abstinence in long-term survival, comparative risk analysis using the Kaplan-Meier method compared by the log-rank test was performed. The SPSS statistical package (version 15.0; SPSS, Inc., Chicago, IL.) was used for all analyses. A P value <0.05 was required for significance. Finally, to evaluate the interaction of variables independently associated with alcohol recidivism after an index episode of AH, a classification and regression trees (CART) analysis was performed. We used CART Pro software (v6.0; Salford systems, San Diego, CA), based on the original Breinman s code. See the Supporting Data for a detailed explanation on CART analysis. Results CLINICAL CHARACTERISTICS OF THE STUDY COHORT One hundred sixty-two patients comprised the original cohort admitted for AH, with 142 surviving index hospitalization of AH and finally included in this study. Study participants were predominantly male (69%), and median age at admission was 50 years. Almost one third of patients (30%) had a psychiatric comorbidity, predominantly depression; however, some patients had anxiety and/or personality disorders. Median reported daily alcohol intake at time of admission was 100 g/day (25-75; IQR, ). Most patients were long-time drinkers and 28 of 142 (20%) reported over 20 daily alcoholic drinks. From the whole cohort, 46% of patients had previously been diagnosed with liver disease and 81% were with cirrhosis admission. Patients had severe liver disease on presentation, with a median MELD of 16, Maddrey s DF of 35, and ABIC score of 7.2. As expected, total median bilirubin was elevated at admission (7 mg/dl), AST was elevated, and median AST/ALT ratio at admission was 2.7. Median platelet count was /L and median INR The majority of patients 1845

5 ALTAMIRANO, L OPEZ-PELAYO, ET AL. HEPATOLOGY, December 2017 Variable TABLE 1. Baseline Characteristics of Patients with AH, Stratified by Outcome After Index Hospitalization (n 5 142) Alive (n 5 88) Deceased (n 5 54) P Value Clinical and epidemiological variables Age (years) 48 (40-53) 52 (47-57) Sex (male) 62.5 (55) 79.6 (43) Alcohol intake at admission (g/day) 100 (80-130) 100 (80-130) Duration of heavy drinking (years) (17) 3.7 (2) (57) 70.4 (38) (14) 25.9 (14) Active smoking 67.0 (59) 74.1 (40) Psychiatric comorbidity 27.3 (24) 33.3 (18) Cirrhosis 78.4 (69) 85.2 (46) Follow-up (months) 66 (46-99) 14 (6-50) <0.001 Biochemical variables at admission Hemoglobin (g/dl) 11.2 ( ) 11.1 ( ) WBC (310 9 /L) 8.7 ( ) 7.8 ( ) Platelets (310 9 /L) 136 (86-213) 103 (56-167) Creatinine (mg/dl) 0.8 ( ) 0.8 ( ) Bilirubin (mg/dl) 6.6 ( ) 8.8 ( ) AST (IU/L) 119 (78-172) 138 (92-202) ALT (IU/L) 41 (31-57) 52 (36-73) GGT (IU/L) 379 ( ) 394 ( ) Albumin (g/l) 27 (23-30) 26 ( ) INR 1.56 ( ) 1.63 ( ) Sodium (meq/l) 134 ( ) 134 ( ) Scoring systems Baseline ABIC 7.02 ( ) 7.48 ( ) Baseline MELD 14.4 ( ) 17.5 ( ) Data shown as median (interquartile range) or % (n). Psychiatric comorbidity defined as the presence of at least one of the following: depressive syndrome, anxiety disorder, bipolar disorder, or type A or B cluster personality disorder. presented with jaundice, hepatomegaly, ascites, and edema. Infection at admission was present in 18% of patients, and 30% of patients developed an infection during the index hospitalization. Fifteen percent of patients developed acute kidney injury (according to Acute Kidney Injury Network criteria 2 ) during admission, 10.6%, and 17% of patients required admission to the intensive care unit. Clinical, demographic, and biochemical data at admission are shown in Supporting Table S2. FACTORS ASSOCIATED WITH LONG-TERM SURVIVAL AFTER INDEX ADMISSION Overall, median follow-up was 55 months (IQR, 17-84). At the end of follow-up, 62% of the patients were alive (n 5 88) and 38% were deceased (n 5 54). Specific cause of death during follow-up was obtained in 95% of the 54 deceased patients. Main causes of death were liver-related causes in 47 patients (80%), non-liver-related in 11 (15%), and unknown in 3 (4%). From liver-unrelated death, 3 patients had an upper aerodigestive tract cancer, 4 lethal cardiovascular events, and 1 perforated appendicitis. Patients deceased at follow-up were older at baseline, with median age of 52 years compared to 48 years in those alive at follow-up (P < 0.05). Sex was associated with survival, and of those deceased at follow-up, 80% were men compared to 62% of those alive at follow-up (P ). Importantly, a higher proportion of those deceased at follow-up had a long duration of heavy drinking (>11 years; Table 1). There were no significant differences in the number of patients with psychiatric comorbidities or current smokers or cirrhosis when comparing those alive versus deceased at followup. Importantly, complete abstinence after the index admission was associated with better survival. Of those alive at follow-up, 55% had resumed active drinking, compared to 70% of those deceased at follow-up. In order to explore those factors associated with long-term survival, we conducted univariate and multivariate Cox survival analyses. In univariate analysis, age (P ), sex (P ), duration of heavy drinking 25 years (P ), INR (P ), baseline ABIC score (P ), MELD score (P ), 1846

6 HEPATOLOGY, Vol. 66, No. 6, 2017 ALTAMIRANO, L OPEZ-PELAYO, ET AL. TABLE 2. Univariate Analysis of Variables Associated With Survival After Index Hospitalization in Patients with AH Variable HR 95% CI P Value Clinical and epidemiological variables Age (years) Sex (male) Alcohol intake at admission (g/day) Duration of heavy drinking (years) 11 Ref Active smoking Psychiatric comorbidity Cirrhosis Biochemical variables at admission Hemoglobin (g/dl) WBC (310 9 /L) Platelets (310 9 /L) Creatinine (mg/dl) Bilirubin (mg/dl) AST (IU/L) ALT (IU/L) GGT (IU/L) Albumin (g/l) INR Sodium (meq/l) Scoring systems at admission Baseline ABIC Baseline MELD Follow-up variables Decompensations during follow-up <0.001 Complete abstinence during follow-up Psychiatric comorbidity defined as the presence of at least one of the following: depressive syndrome, anxiety disorder, bipolar disorder, or type A or B cluster personality disorder. Abbreviation: 95% CI, 95% confidence interval. and presence of liver decompensations during follow-up (P < 0.001) were associated with poor survival. Conversely, white blood cells (WBC; P ) and complete abstinence (P ) were associated with improved long-term survival (Table 2). After adjusting for all variables by univariate analysis and stratifying by the presence of decompensations during follow-up, two multivariate Cox survival models were fitted. On separate multivariate analyses, baseline ABIC score with complete abstinence and age, along with baseline MELD and complete abstinence, were independently associated with long-term survival (Table 3). Finally, we analyzed the subset of patients with a serum bilirubin >4.7 mg/dl at admission (n 5 88), confirming that complete abstinence was the main factor influencing long-term survival (Supporting Tables S3, S4, and S5). IMPACT OF ALCOHOL ABSTINENCE ON LONG-TERM SURVIVAL We next compared the impact of alcohol recidivism on long-term survival. Overall mortality of patients with alcohol recidivism versus those that remain completely abstinent was 44% versus 29% (P ), respectively. Kaplan-Meier analysis showed that presence of alcohol recidivism after an episode of AH negatively influenced long-term survival (Fig. 1). As expected, median follow-up of patients with alcohol recidivism was significantly shorter when compared to those completely abstinent (46 vs. 63 months; P ). When analyzing time frames, we did not find significant differences on survival between those abstinent and patients who recidivated at 3, 6, and 12 months of follow-up (P > 0.05, for all cases). However, differences in survival became significant after 18 months of follow-up (P ; Supporting Fig. S1). Finally, 36% of patients with complete abstinence presented at least 1 liver decompensation requiring admission during follow-up, compared with 58% of patients with alcohol recidivism (P , Supporting Fig. S2). FACTORS INFLUENCING LONG- TERM ABSTINENCE AFTER THE INDEX ADMISSION Based on our criteria to define recidivism (see Patients and Methods), 60% were considered recidivists during follow-up. Recidivist patients were younger at admission (47 vs. 51 years; P ), reported a higher number of daily drinks (P ), and more past alcoholism treatments before the index hospitalization (P ; Table 4). Importantly, patients that achieved complete abstinence during follow-up had a lower median HRAR scale (2 vs. 3 points; P ). However, when comparing the proportion of patients with HRAR scale >3 points (e.g., high risk for alcohol recidivism) at baseline between the two groups, no significant differences were found (P ). In addition, there were no significant differences in the proportion of smokers, presence and type of psychiatric comorbidity, and baseline MELD score when comparing abstinent patients versus those that reported recidivism. We found a significantly higher baseline ABIC score among abstinent patients (P ). However, we found that this difference was related to inclusion of age as a variable of the ABIC scoring system. Time to recidivism was available for 120 patients; 62 (52%) 1847

7 ALTAMIRANO, L OPEZ-PELAYO, ET AL. HEPATOLOGY, December 2017 TABLE 3. Multivariate Cox Regression Analyses of Variables Associated With Survival after Index Hospitalization in Patients With AH Multivariate Cox Regression Analysis Model #1 HR (95% CI) P Value Baseline ABIC 1.35 ( ) <0.001 Complete abstinence during 0.51 ( ) follow-up Model #2 Multivariate Cox Regression Analysis HR (95% CI) P Value Age (years) 1.06 ( ) Baseline MELD 1.04 ( ) Complete abstinence during follow-up 0.43 ( ) Variables included in Model #1 were HRAR scale, ABIC at admission, abstinence during follow-up, sex, and WBC count. Variables included in Model #2 were HRAR scale, MELD at admission, age, abstinence during follow-up, sex, and WBC count. Analysis was stratified by the presence of decompensations during follow-up for both models. Abbreviation: 95% CI, 95% confidence interval. recidivated 1 year after the index episode and 85 (71%) after 24 months. In order to identify factors associated with longterm abstinence, we next performed univariate and multivariate logistic regression analyses including variables obtained at the index hospitalization. In univariate analysis, age (P ), number of daily drinks 11 (P ), no past alcoholism treatments (P ), HRAR scale (P ), and ABIC score (P ) were associated with complete abstinence during follow-up. Importantly, when categorizing HRAR scale, we did not find any significant association with long-term abstinence. This last finding likely reflects the lack of associative effect of one of the parameters included in this scale (e.g., duration of heavy drinking) with long-term abstinence (Table 5). After including variables significant after univariate analysis and adjusting for baseline liver function (e.g., baseline ABIC and MELD scores), four multivariate logistic regression models for predicting abstinence were fitted. On multivariate models, only age and lack of past alcoholism treatments were independently associated with long-term abstinence (Table 6). We also analyzed the subset of patients with serum bilirubin >4.7 mg/dl at admission (n 5 88), confirming the association of age and lack of previous alcoholism treatments with complete abstinence (Supporting Tables S6, S7, and S8). Finally, in order to evaluate the interactions between the two significant variables and aiming to generate a more intuitive and practical model for accurate prediction of long-term abstinence, we fitted a CART model. The generated decision tree discriminated two subpopulations with different rates of long-term abstinence: a high-rate abstinence group (65% complete abstinence rate at follow-up) composed by those patients without history of past alcoholism treatments and >48 years at baseline and a low-rate abstinence group (26%-29% complete abstinence rate at complete follow-up) composed by those with past alcoholism treatments and those without past alcoholism treatments but 48 years (Figure 2). This CART model showed good usefulness estimated by an area under the receiving operator characteristics curve (AUROC) of Discussion Most clinical and translational research on AH focuses on short-term prognosis, whereas studies assessing the main determinants of long-term prognosis are scarce. Based on our clinical experience, we hypothesized that alcohol drinking status heavily determines long-term outcome of patients surviving an episode of AH. We followed up for 4.6 years a cohort of well-characterized, biopsy-proven cohort of patients with AH. We confirmed that the main parameter influencing mortality in the long term was resuming alcohol consumption. Interestingly, we found that alcohol recidivism after the index episode of AH strongly predicted survival. Whether the pattern of FIG. 1. Impact of complete alcohol abstinence on long-term survival of patients that survive an episode of AH. The Kaplan- Meier analysis shows that the presence of alcohol recidivism after an episode of AH negatively influenced long-term survival. 1848

8 HEPATOLOGY, Vol. 66, No. 6, 2017 ALTAMIRANO, L OPEZ-PELAYO, ET AL. Variable TABLE 4. Baseline Characteristics of Patients With AH, Stratified by Abstinence During Follow-up (n 5 142) Complete Abstinence (n 5 56) Alcohol Recidivism (n 5 86) P Value Clinical, epidemiological, and psychiatric variables Age (years) 51 (47-56) 47 (41-54) Male sex 66.1 (37) 70.9 (61) Follow-up (months) 63 (37-104) 46 (14-71) Duration of heavy drinking (years) (9) 11.6 (10) (37) 67.4 (58) (19) 20.9 (18) No. of daily drinks (22) 20.9 (18) (25) 57 (49) (9) 22.1 (19) Past alcoholism treatments (41) 50 (43) (11) 31.4 (27) >1 7.1 (4) 18.6 (16) HRAR 2 (1-3) 3 (2-4) HRAR (35) 46.5 (40) (18) 41.9 (36) (3) 11.6 (10) HRAR (>3 points) 9 (16) 26 (30) 0.08 Active smoking 66.1 (37) 72.1 (62) Psychiatric comorbidity 30.4 (17) 30.2 (26) Type of psychiatric comorbidity (n 5 43) Anxiety and depression 82.4 (14) 73.1 (19) Personality disorders 17.6 (3) 26.9 (7) Biochemical variables at admission Haemoglobin (g/dl) 11.2 ( ) 11.2 ( ) WBC (310 9 /L) 8.7 ( ) 8.4 ( ) Platelets (310 9 /L) 134 (84-204) 115 (69-181) Creatinine (mg/dl) 0.80 ( ) 0.80 ( ) Bilirubin (mg/dl) 8.0 ( ) 6.1 ( ) AST (IU/L) 117 (82-175) 132 (85-195) ALT (IU/L) 45 (32-57) 46 (33-67) GGT (IU/L) 376 ( ) 432 ( ) Albumin (g/l) 26 (24-30) 27 (23-31) INR 1.59 ( ) 1.57 ( ) Sodium (meq/l) 134 ( ) 135 ( ) Scoring systems at admission Baseline ABIC 7.69 ( ) 7.03 ( ) Baseline MELD ( ) ( ) Data shown as median (IQR) or % (n). One daily drink 5 10 grams of alcohol. Alcoholism treatments refer to any outpatient or inpatient treatment for alcoholism. Psychiatric comorbidity defined as the presence of at least one of the following: depressive syndrome, anxiety disorder, bipolar disorder, or type A or B cluster personality disorder. drinking or occasional drinking (i.e., slips ) also influence survival was not assessed in our study and deserves further investigation. Likewise, the prognostic value of previous admissions to hospitalization and parameters at discharge has not been assessed in this study. Other scoring systems, such as the alcoholic hepatitis histological score or the presence of systemic inflammatory response syndrome, predicted short-term survival in our series of patients with AH, (4,24) but not alcohol relapse or long-term survival, indicating that parameters from the index hospitalization other than age do not play a role in the long-term outcome of patients surviving an episode of AH. Larger studies should investigate whether genetic factors or ethnicity (given that 95% of our cohort was white) influence the deleterious effects of alcohol intake on long-term outcome. Our results strongly indicate that maneuvers aimed at promoting alcohol abstinence should be initiated as early as possible in patients hospitalized with AH. 1849

9 ALTAMIRANO, L OPEZ-PELAYO, ET AL. HEPATOLOGY, December 2017 Ideally, multidisciplinary teams, including addiction therapists, should contribute in the integrated treatment of these patients. According to previous research, alcohol recidivism prevention in patients with cirrhosis caused by ALD depends on multidisciplinary collaboration, including psychological treatment based on cognitive-behavior therapy (CBT), motivational enhancement therapy (MET), comprehensive medical care (CMC), and pharmacological treatments. (29-32) In addition, a recent open-label study using baclofen (a gamma-aminobutyric acid receptor agonist) seems promising in the prevention of alcohol recidivism of patients with AH. (33) Other psychological treatments (e.g., CBT, MET, CMC, or their combination) have not been prospectively evaluated for abstinence maintenance in the treatment of patients with AH. Besides determining the key role of alcohol consumption in the outcome of these patients, our study attempted to identify factors that predict alcohol recidivism. In our center, addiction specialists assess all patients during the index hospitalization as part of the clinical protocol. We routinely collect data on the type and amount of alcohol consumption and previous detoxification attempts. These parameters are included in the HRAR score, which was initially designed to predict the recidivism in patients undergoing an LT. We have also data on smoking status and use of illicit drugs; however, employment status and other psychosocial variables were not collected. Interestingly, we found that neither the amount of daily consumption nor the duration of alcohol intake, current smoking status, and use of illicit drugs predicted recidivism in patients that survive an episode of AH. This can be partially explained because all these patients were heavy drinkers (defined as >60 g/day), constituting a homogeneous sample with only slight differences in daily amount of alcohol intake (IQR, and <20% taking >20 standard drink units) that do not allow to predict any change in recidivism risk in this population. In contrast, age and previous detoxification attempts heavily predicted alcohol recidivism. By combining these two parameters, we were able to build a simple algorithm to identify patients with high and low risk of recidivism. In this study, we found that age >48 years was associated with higher rate of abstinence, which goes in line with previous literature. In previous studies, age has been inversely associated with lower response to alcohol treatment because of lack of adherence and was also inversely associated with more relapse in AUD. (34) The reason for this might be the presence of higher craving in younger people. In fact, TABLE 5. Univariate Logistic Regression Analysis of Variables Associated With Alcohol Recidivism in Patients with AH Variable OR 95% CI P Value Clinical, epidemiological, and psychiatric variables Age (years) Male sex Duration of heavy drinking (years) 11 Ref No. of daily drinks 11 Ref No. of daily drinks (11 daily drinks) Past alcoholism treatments 0 Ref > Past alcoholism treatments (1 treatment) HRAR (points) HRAR (points) 0-2 Ref Modified HRAR (>3 points) Active smoking Psychiatric comorbidity Scoring systems at admission Baseline ABIC Baseline MELD One daily drink 5 10 grams of alcohol. Alcoholism treatments refer to any outpatient or inpatient treatment for alcoholism. Psychiatric comorbidity defined as the presence of at least one of the following: depressive syndrome, anxiety disorder, bipolar disorder, or type A or B cluster personality disorder. Abbreviation: CI, confidence interval. there are studies demonstrating that some anticraving drugs, such as ondansertron, were successfully tested in adolescents and adults with early onset of AUD, but being ineffective in adults with late-onset AUD. (35,36) In this context, it seems that two types of alcoholism exist according to age, and craving plays a role as mediator. Finally, age at onset is an essential element to differentiate two types of alcoholism according to the classical classification of Gilligan et al. (37) However, a physiological explanation of increased craving in young alcohol dependents is still lacking; further prospective studies in patients with AH, taking into account craving, tools for craving assessment, and other psychological phenomena, are required. With regard to the importance of previous AUD treatments, our algorithm shows that those patients 1850

10 HEPATOLOGY, Vol. 66, No. 6, 2017 ALTAMIRANO, L OPEZ-PELAYO, ET AL. TABLE 6. Multivariate Logistic Regression Analyses of Variables Associated With Alcohol Recidivism in Patients With AH Multivariate Logistic Regression Analysis Model #1 OR (95% CI) P Value Baseline ABIC 0.71 ( ) No. of daily drinks ( ) ( ) No. of past alcohol treatments ( ) > ( ) Model #2 OR (95% CI) Multivariate Logistic Regression Analysis P Value Baseline ABIC 0.72 ( ) No. of daily drinks (11 daily drinks) 1.89 ( ) Number of prior alcohol treatments ( 1 treatment) 2.59 ( ) Model #3 OR (95% CI) Multivariate Logistic Regression Analysis P Value Baseline MELD 0.98 ( ) Age (years) 0.94 ( ) No. of daily drinks ( ) ( ) No. of past alcohol treatments ( ) > ( ) Model #4 OR (95% CI) Multivariate Logistic Regression Analysis P Value Baseline MELD 0.97 ( ) Age (years) 0.94 ( ) No. of daily drinks (11 daily drinks) 1.95 ( ) No. of past alcohol treatments (1 treatment) 2.51 ( ) Model #1 AUROC: 0.71 (95% CI, ); Model #2 AUROC: 0.71 (95% CI, ); Model #3 AUROC: 0.71 (95% CI, ); Model #4 AUROC: 0.72 (95% CI, ). 95% CI, 95% confidence interval. with previous AUD treatments had a higher risk for long-term alcohol recidivism. Previous literature has shown that the probability of success of any given treatment decreases with each previous failure. (38) This means that the patient s expectancy to achieve treatment goals decreases as the number of treatments increases, suggesting an altered role in self-efficacy. (39,40) Of note, our study was done in the Barcelona area, and it is well known that alcohol consumption varies among different geographical areas and is influenced by socioeconomic and cultural factors. Traditionally, two cultural patterns have been proposed: (1) that based on binge drinking (e.g., UK or Ireland with >30% prevalence of heavy episodic drinkers among drinkers vs. <20% in sourthern countries such as Spain or Italy) and (2) the regular drinking pattern (mainly seen in Southern Europe). Although these patterns are quickly evolving because of globalization, our predicting algorithm should be externally validated in these two populations. This study has important clinical and research implications. First, it highlights the crucial role of alcohol consumption in the long-term prognosis of this disease. Specialized liver centers should prioritize early interventions and close follow-up of these patients to assist in the AUD. Second, it suggests that any clinical trial assessing the impact of medications on long-term survival after an episode of AH should take into 1851

11 ALTAMIRANO, L OPEZ-PELAYO, ET AL. HEPATOLOGY, December 2017 system and, most important, test pharmacological and psychological interventions to promote alcohol abstinence among these patients. REFERENCES FIG. 2. Prognostic algorithm for complete abstinence prediction on long-term follow-up of patients that survived an episode of AH. The CART analysis discriminated two subpopulations: (1) a high-rate abstinence group (65% complete abstinence rate at complete followup) including those patients without past alcoholism treatments and older than 48 years and (2) a low-rate abstinence group (26%-29% complete abstinence rate at complete follow-up) composed by those with past alcoholism treatments and those without past alcoholism treatments but younger than 48 years. account the presence/absence of alcohol consumption. This is important because medications aimed at improving liver function could also affect alcohol behavior. This is a topic of recent interest in the field of alcohol addiction given that experimental and human data suggest that systemic inflammation could play a role in AUD (both in the etiology and prognosis). (41,42) Whether this link is present in patients with AH deserves future translational studies. Finally, the identification of those with higher risk of recidivism might allow for testing and implementation of specific pharmacological strategies to promote abstinence during the follow-up of these patients. In conclusion, the current study investigates the main determinants of long-term prognosis in patients with biopsy-proven AH. We found that continued alcohol consumption is associated with a poorer longterm survival after an episode of AH, which has important implications in the management of these patients. Moreover, we developed a simple algorithm capable of identifying patients with low and high risk of alcohol recidivism. Future prospective studies including patients with more-severe AH should validate this 1) Gao B, Bataller R. Alcoholic liver disease: pathogenesis and new therapeutic targets. Gastroenterology 2011;141: ) Altamirano J, Fagundes C, Dominguez M, Garcıa E, Michelena J, Cardenas A, et al. Acute kidney injury is an early predictor of mortality for patients with alcoholic hepatitis. Clin Gastroenterol Hepatol 2012;10:65-71.e3. 3) Serste T, Njimi H, Degre D, Deltenre P, Schreiber J, Lepida A, et al. The use of beta-blockers is associated with the occurrence of acute kidney injury in severe alcoholic hepatitis. Liver Int 2015;35: ) Altamirano J, Miquel R, Katoonizadeh A, Abraldes JG, Duarte- Rojo A, Louvet A, et al. A histologic scoring system for prognosis of patients with alcoholic hepatitis. Gastroenterology 2014; 146: ) Carithers RL, Jr., Herlong HF, Diehl AM, Shaw EW, Combes B, Fallon HJ, et al. Methylprednisolone therapy in patients with severe alcoholic hepatitis. A randomized multicenter trial. Ann Intern Med 1989;110: ) Dunn W, Jamil LH, Brown LS, Wiesner RH, Kim WR, Menon KVN, et al. MELD accurately predicts mortality in patients with alcoholic hepatitis. HEPATOLOGY 2005;41: ) Forrest EH, Evans CD, Stewart S, Phillips M, Oo YH, McAvoy NC, et al. Analysis of factors predictive of mortality in alcoholic hepatitis and derivation and validation of the Glasgow alcoholic hepatitis score. Gut 2005;54: ) Dominguez M, Rincon D, Abraldes JG, Miquel R, Colmenero J, Bellot P, et al. A new scoring system for prognostic stratification of patients with alcoholic hepatitis. Am J Gastroenterol 2008;103: ) Louvet A, Naveau S, Abdelnour M, Ramond MJ, Diaz E, Fartoux L, et al. The Lille model: a new tool for therapeutic strategy in patients with severe alcoholic hepatitis treated with steroids. HEPATOLOGY 2007;45: ) Mathurin P, Duchatelle V, Ramond MJ, Degott C, Bedossa P, Erlinger S, et al. Survival and prognostic factors in patients with severe alcoholic hepatitis treated with prednisolone. Gastroenterology 1996;110: ) Mathurin P, O Grady J, Carithers RL, Phillips M, Louvet A, Mendenhall CL, et al. Corticosteroids improve short-term survival in patients with severe alcoholic hepatitis: meta-analysis of individual patient data. Gut 2011;60: ) Mathurin P, Louvet A, Duhamel A, Nahon P, Carbonell N, Boursier J, et al. Prednisolone with vs without pentoxifylline and survival of patients with severe alcoholic hepatitis: a randomized clinical trial. JAMA 2013;310: ) Louvet A, Diaz E, Dharancy S, Coevoet H, Texier F, Thevenot T, et al. Early switch to pentoxifylline in patients with severe alcoholic hepatitis is inefficient in non-responders to corticosteroids. J Hepatol 2008;48: ) Naveau S, Chollet-Martin S, Dharancy S, Mathurin P, Jouet P, Piquet MA, et al. A double-blind randomized controlled trial of 1852

12 HEPATOLOGY, Vol. 66, No. 6, 2017 ALTAMIRANO, L OPEZ-PELAYO, ET AL. infliximab associated with prednisolone in acute alcoholic hepatitis. HEPATOLOGY 2004;39: ) Boetticher NC, Peine CJ, Kwo P, Abrams GA, Patel T, Aqel B, et al. A randomized, double-blinded, placebo-controlled multicenter trial of etanercept in the treatment of alcoholic hepatitis. Gastroenterology 2008;135: ) Halle P, Pare P, Kaptein E, Kanel G, Redeker AG, Reynolds TB. Double-blind, controlled trial of propylthiouracil in patients with severe acute alcoholic hepatitis. Gastroenterology 1982;82: ) Mathurin P, Moreno C, Samuel D, Dumortier J, Salleron J, Durand F, et al. Early liver transplantation for severe alcoholic hepatitis. N Engl J Med 2011;365: ) Im GY, Kim-Schluger L, Shenoy A, Schubert E, Goel A, Friedman SL, et al. Early liver transplantation for severe alcoholic hepatitis in the United States a single-center experience. Am J Transplant 2015;16: ) Thursz MR, Richardson P, Allison M, Austin A, Bowers M, Day CP, et al. Prednisolone or pentoxifylline for alcoholic hepatitis. N Engl J Med 2015;372: ) Potts JR, Goubet S, Heneghan MA, Verma S. Determinants of long-term outcome in severe alcoholic hepatitis. Aliment Pharmacol Ther 2013;38: ) Potts JR, Howard MR, Verma S. Recurrent severe alcoholic hepatitis: clinical characteristics and outcomes. Eur J Gastroenterol Hepatol 2013;25: ) Dumortier J, Dharancy S, Cannesson A, Lassailly G, Rolland B, Pruvot FR, et al. Recurrent alcoholic cirrhosis in severe alcoholic relapse after liver transplantation: a frequent and serious complication. Am J Gastroenterol 2015;110: ) Lucey MR. Liver transplantation for alcoholic liver disease. Nat Rev Gastroenterol Hepatol 2014;11: ) Michelena J, Altamirano J, Abraldes JG, Affo S, Morales-Ibanez O, Sancho-Bru P, et al. Systemic inflammatory response and serum lipopolysaccharide levels predict multiple organ failure and death in alcoholic hepatitis. HEPATOLOGY 2015;62: ) De Gottardi A, Spahr L, Gelez P, Morard I, Mentha G, Guillaud O, et al. A simple score for predicting alcohol relapse after liver transplantation: results from 387 patients over 15 years. Arch Intern Med 2007;167: ) Altamirano J, Bataller R, Cardenas A, Michelena J, Freixa N, Monras M, et al. Predictive factors of abstinence in patients undergoing liver transplantation for alcoholic liver disease. Ann Hepatol 2012;11: ) Lligo~na A, Freixa N, Bataller R, Monras M, Rimola A. [Clinical guideline for the evaluation of liver transplant candidates with addictions]. [Article in Spanish]. Gastroenterol Hepatol 2009;32: ) Bravo F, Gual A, Lligo~na A, Colom J. Gender differences in the long-term outcome of alcohol dependence treatments: an analysis of twenty-year prospective follow up. Drug Alcohol Rev 2013;32: ) Addolorato G, Mirijello A, Barrio P, Gual A. Treatment of alcohol use disorders in patients with alcoholic liver disease. J Hepatol 2016; 65: ) Addolorato G, Russell M, Albano E, Haber PS, Wands JR, Leggio L. Understanding and treating patients with alcoholic cirrhosis: an update. Alcohol Clin Exp Res 2009;33: ) Dom G, Wojnar M, Crunelle CL, Thon N, Bobes J, Preuss UW, et al. Assessing and treating alcohol relapse risk in liver transplantation candidates. Alcohol Alcohol 2015;50: ) Khan A, Tansel A, White DL, Kayani WT, Bano S, Lindsay J, et al. Efficacy of psychosocial interventions in inducing and maintaining alcohol abstinence in patients with chronic liver disease: a systematic review. Clin Gastroenterol Hepatol 2016;14: ) Yamini D, Lee SH, Avanesyan A, Walter M, Runyon B. Utilization of baclofen in maintenance of alcohol abstinence in patients with alcohol dependence and alcoholic hepatitis with or without cirrhosis. Alcohol Alcohol 2014;49: ) Oslin DW, Pettinati H, Volpicelli JR. Alcoholism treatment adherence: older age predicts better adherence and drinking outcomes. Am J Geriatr Psychiatry 2002;10: ) Johnson BA, Roache JD, Javors MA, DiClemente CC, Cloninger CR, Prihoda TJ, et al. Ondansetron for reduction of drinking among biologically predisposed alcoholic patients: a randomized controlled trial. JAMA 2000;284: ) Kranzler HR, Pierucci-Lagha A, Feinn R, Hernandez-Avila C. Effects of ondansetron in early- versus late-onset alcoholics: a prospective, open-label study. Alcohol Clin Exp Res 2003;27: ) Gilligan SB, Reich T, Cloninger CR. Etiologic heterogeneity in alcoholism. Genet Epidemiol 1987;4: ) Siegel C, Alexander MJ, Lin S. Severe alcoholism in the mental health sector: II. Effects of service utilization on readmission. J Stud Alcohol 1984;45: ) Greenfield SF, Hufford MR, Vagge LM, Muenz LR, Costello ME, Weiss RD. The relationship of self-efficacy expectancies to relapse among alcohol dependent men and women: a prospective study. J Stud Alcohol 2000;61: ) Kadden RM, Litt MD. The role of self-efficacy in the treatment of substance use disorders. Addict Behav 2011;36: ) Gonzalez-Reimers E, Santolaria-Fernandez F, Martın-Gonzalez MC, Fernandez-Rodrıguez CM, Quintero-Platt G. Alcoholism: a systemic proinflammatory condition. World J Gastroenterol 2014;20: ) Kebir O, Gorsane MA, Blecha L, Krebs MO, Reynaud M, Benyamina A. Association of inflammation genes with alcohol dependence/abuse: a systematic review and a meta-analysis. Eur Addict Res 2011;17: Author names in bold designate shared co-first authorship. Supporting Information Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/ /hep.29338/suppinfo. 1853