Research Article Cirrhosis and Liver Failure PREDICT identifies precipitating events associated with the clinical course of acutely decompensated cirrhosis Jonel Trebicka1,2,*,†, Javier Fernandez1,4,†, Maria Papp5, Paolo Caraceni6, Wim Laleman13, Carmine Gambino7, Ilaria Giovo8, Frank Erhard Uschner2, Christian Jansen3, Cesar Jimenez9, Rajeshwar Mookerjee10, Thierry Gustot11, Agustin Albillos12, Rafael Bañares14, Peter Jarcuska15, Christian Steib16, Thomas Reiberger17, Juan Acevedo18, Pietro Gatti19, Debbie L. Shawcross20, Stefan Zeuzem2, Alexander Zipprich21, Salvatore Piano7, Thomas Berg22, Tony Bruns23,34, Karen Vagner Danielsen24, Minneke Coenraad25, Manuela Merli26, Rudolf Stauber27, Heinz Zoller28, José Presa Ramos29, Cristina Solé4, Germán Soriano30, Andrea de Gottardi31, Henning Gronbaek32, Faouzi Saliba33, Christian Trautwein34, Haluk Tarik Kani35, Sven Francque36, Stephen Ryder37, Pierre Nahon38,39,40, Manuel Romero-Gomez41, Hans Van Vlierberghe42, Claire Francoz43,44, Michael Manns45, Elisabet Garcia-Lopez1, Manuel Tufoni6, Alex Amoros1, Marco Pavesi1, Cristina Sanchez1, Michael Praktiknjo3, Anna Curto1, Carla Pitarch1, Antonella Putignano11, Esau Moreno1, William Bernal20, Ferran Aguilar1, Joan Clària1,4, Paola Ponzo8, Zsuzsanna Vitalis5, Giacomo Zaccherini6, Boglarka Balogh5, Alexander Gerbes16, Victor Vargas9, Carlo Alessandria8, Mauro Bernardi6, Pere Ginès4, Richard Moreau1,43,44, Paolo Angeli1,7, Rajiv Jalan1,10,‡, Vicente Arroyo 1,‡, for the PREDICT STUDY group of the EASL-CLIF CONSORTIUM# 1 European Foundation for Study of Chronic Liver Failure, EF-Clif, Barcelona, Spain; 2Department of Internal Medicine I, Goethe University Frankfurt, Frankfurt, Germany; 3Department of Internal Medicine I, University Hospital Bonn, Bonn, Germany; 4Hospital Clinic of Barcelona, University of Barcelona, CIBEReHD, IDIBAPS, Barcelona, Spain; 5University of Debrecen, Faculty of Medicine, Institute of Medicine, Department of Gastroenterology, Debrecen, Hungary; 6University of Bologna, Bologna, Italy; 7University of Padova, Padova, Italy; 8A.O.U. Città della Salute e della Scienza Torino, Torino, Italy; 9Liver Unit, Hospital Vall d’Hebron, Universitat Autònoma de Barcelona, CIBEREHD, Barcelona, Spain; 10 UCL Medical School,Royal Free Hospital, London, UK; 11C.U.B. Erasme, Bruxelles, Belgium; 12Department of Gastroenterology, Hospital Universitario Ramón y Cajal, IRYCIS, University of Alcalá, CIBEREHD, Madrid, Spain; 13Department of Gastroenterology and Hepatology, Section of Liver and Biliopancreatic disorders, University of Leuven, Leuven, Belgium; 14Hospital General Universitario Gregorio Marañón. Facultad de Medicina (Universidad Complutense of Madrid), CIBERehd, Madrid, Spain; 15Pavol Jozef Safarik University in Kosice, Kosice, Slovakia; 16Department of Medicine II, Liver Centre Munich, University Hospital, LMU, Munich, Germany; 17Medical University of Vienna, Vienna, Austria; 18University Hospitals Plymouth NHS Trust, Plymouth, UK; 19Internal Medicine PO Ostuni, ASL Brindisi, Italy; 20King’s College Hospital, London, UK; 21University Hospital Halle-Wittenberg, Halle(Saale), Germany; 22Division of Hepatology, Department of Medicine II, Leipzig University Medical Center, Leipzig, Germany; 23Jena University Hospital, Jena, Germany; 24Gastrounit, Medical Section, Hvidovre Hospital and Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark; 25Leiden University Medical Center, Leiden, The Netherlands; 26Universitá Sapienza Roma, Roma, Italy; 27Medical University of Graz, Graz, Austria; 28Medical University of Innsbruck, Innsbruck, Austria; 29CHTMAD Vila Real-Blueclinical, Vila Real, Portugal; 30Hospital de la Santa Creu i Sant Pau and CIBERehd, Barcelona, Spain; 31University Clinic of Visceral Surgery and Medicine-Inselspital, Bern and Ente Ospedaliero Cantonale, Universita della Svizzera Italiana, Lugano, Switzerland; 32Aarhus University Hospital, Aarhus, Denmark; 33AP-HP Hôpital Paul Brousse, Centre Hépato-Biliaire, Universite Paris Saclay, INSERM Unit 1193, Villejuif, France; 34Aachen University Hospital, Aachen, Germany; 35Marmara University, Kadiköy, Turkey; 36University Hospital Antwerp, Antwerp, Belgium; 37NIHR Biomedical Research Centre at Nottingham University Hospitals NHS Trust and the University of Nottingham, Nottingham, UK; 38AP-HP, Hôpital Jean Verdier, Service d’Hépatologie, Bondy; 39Université Paris 13, Sorbonne Paris Cité, “Equipe labellisée Ligue Contre le Cancer”, Saint-Denis; 40Inserm, UMR-1162, “Génomique fonctionnelle des tumeurs solides”, Paris, France; 41Virgen del Rocío University Hospital, Sevilla, Spain; 42Ghent University Hospital, Ghent, Belgium; 43APHP, Hôpital Beaujon, Service d’Hépatologie, Clichy, France; 44Inserm, Université de Paris, Centre de Recherche sur L’Inflammation, Paris, France; 45 Hannover Medical School, Hannover, Germany Keywords: Chronic liver disease; Non-elective admission; Acute complications; Outcome; Risk factors. Background & Aims: Acute decompensation (AD) of cirrhosis Received 14 July 2020; received in revised form 10 November 2020; accepted 10 may present without acute-on-chronic liver failure (ACLF) (AD- November 2020; available online 20 November 2020 * Corresponding author. Address: European Foundation for Study of Chronic Liver No ACLF), or with ACLF (AD-ACLF), defined by organ failure(s). Failure, EF-Clif, Travesera de Gracia 11, 7th Floor, 08021 Barcelona, Spain. Herein, we aimed to analyze and characterize the precipitants E-mail address: jonel.trebicka@kgu.de (J. Trebicka). leading to both of these AD phenotypes. † Shared first authorship. ‡ Shared senior authorship. Methods: The multicenter, prospective, observational PREDICT # List of collaborators are listed in Acknowledgements section. study (NCT03056612) included 1,273 non-electively hospitalized https://doi.org/10.1016/j.jhep.2020.11.019 Journal of Hepatology 2021 vol. 74 j 1097–1108 Research Article Cirrhosis and Liver Failure patients with AD (No ACLF = 1,071; ACLF = 202). Medical history, decompensated cirrhosis (SDC) patients presented an uncom- clinical data and laboratory data were collected at enrolment and plicated course during the 3-month follow-up period and during 90-day follow-up, with particular attention given to the showed lower 1-year mortality (9%). following characteristics of precipitants: induction of organ In the traditional view, the development of AD is initiated by dysfunction or failure, systemic inflammation, chronology, in- an acute worsening of stable cirrhosis through different patho- tensity, and relationship to outcome. physiological mechanisms considered as precipitants. Evidence Results: Among various clinical events, 4 distinct events were from the CANONIC and the PREDICT studies challenges this precipitants consistently related to AD: proven bacterial in- view,1,2 and suggests that AD manifests mainly as a result of fections, severe alcoholic hepatitis, gastrointestinal bleeding systemic inflammation, inducing multiple organ dysfunction and with shock and toxic encephalopathy. Among patients with presents with different clinical phenotypes.3,4 Indeed, systemic precipitants in the AD-No ACLF cohort and the AD-ACLF cohort inflammation increases across the sub-phenotypes of AD-no (38% and 71%, respectively), almost all (96% and 97%, respec- ACLF (SDC, UDC and pre-ACLF), and reaches its peak in patients tively) showed proven bacterial infection and severe alcoholic with AD-ACLF.5,6 Moreover, in AD-ACLF phenotype, the grade of hepatitis, either alone or in combination with other events. systemic inflammation correlated with the number of organ Survival was similar in patients with proven bacterial infections failures, clinical course severity and prognosis.3,4 Hence, for a or severe alcoholic hepatitis in both AD phenotypes. The number precipitant to be of importance, it must have the ability to impair of precipitants was associated with significantly increased 90- end-organ function. day mortality and was paralleled by increasing levels of surro- Despite the fact that AD-ACLF phenotypes frequently develop gates for systemic inflammation. Importantly, adequate first-line in close chronological relationship with the precipitant(s), the antibiotic treatment of proven bacterial infections was associated critical time period prior to AD-ACLF has not yet been explored with a lower ACLF development rate and lower 90-day mortality. in detail. Moreover, no specific criteria for the diagnosis of pre- Conclusions: This study identified precipitants that are signifi- cipitants have been identified to date. Consequently, many clin- cantly associated with a distinct clinical course and prognosis in ically relevant aspects of precipitants remain ill-defined. patients with AD. Specific preventive and therapeutic strategies The current study is the second investigation derived from targeting these events may improve outcomes in patients with the PREDICT study. Its aim was to provide the rationale for the decompensated cirrhosis. diagnosis of precipitants and to investigate the association of Lay summary: Acute decompensation (AD) of cirrhosis is char- type and number of precipitants with early clinical course and acterized by a rapid deterioration in patient health. Herein, we prognosis in patients hospitalized with AD-No ACLF and AD-ACLF aimed to analyze the precipitating events that cause AD in pa- phenotypes. tients with cirrhosis. Proven bacterial infections and severe alcoholic hepatitis, either alone or in combination, accounted for Patients and methods almost all (96-97%) cases of AD and acute-on-chronic liver fail- Patients ure. Whilst the type of precipitant was not associated with The PREDICT study (ClinicalTrials.gov number, NCT03056612) is mortality, the number of precipitant(s) was. This study identified a European, investigator-initiated, multicenter, prospective, precipitants that are significantly associated with a distinct observational study performed in 48 university hospitals clinical course and prognosis of patients with AD. Specific pre- (approved by the respective ethics committees) from 15 coun- ventive and therapeutic strategies targeting these events may tries and promoted by the European Foundation for the Study of improve patient outcomes. Chronic Liver Failure. The design of the study has been reported © 2020 European Association for the Study of the Liver. Published by in detail elsewhere.2 Briefly, 1,071 cirrhosis patients with AD-No Elsevier B.V. This is an open access article under the CC BY-NC-ND ACLF phenotype and 202 with AD-ACLF phenotype non- license (http://creativecommons.org/licenses/by-nc-nd/4.0/). electively hospitalized for treatment were enrolled from March 2017 to July 2018 after providing their informed consent. AD was Introduction diagnosed as previously described2 and ACLF according to the Acute decompensation of cirrhosis (hereafter called AD) defines EASL-CLIF criteria.1,7 Stratification of patients with the AD-No the acute development of ascites, hepatic encephalopathy, ACLF phenotype into the AD-pre-ACLF, AD-UDC and AD-SDC gastrointestinal hemorrhage or bacterial infections, or any sub-phenotypes was performed using previously described combination of these. In 2013, the CANONIC study identified the criteria2 and outlined inFig. 1 (for detailed description please see syndrome of acute-on-chronic liver failure (ACLF), the most se- supplementary information). vere phenotype of AD, in 20% of 1,343 consecutive patients non- electively hospitalized for the treatment of an episode of AD.1 Study design ACLF was characterized by single or multiple organ failure and The PREDICTstudy2 was designed to explore the last 90 days prior to high 28-day mortality rate (30%). hospital admission (especially the last 2 weeks), and the first 3 In 2020, the PREDICT study, a prospective observational months after admission (follow-up period), in which the early investigation of 1,273 hospitalized patients with AD, showed that clinical course of patients was assessed. Pre-specified clinical and patients without ACLF (AD-No ACLF phenotype) comprised 3 standard laboratory data were obtained at enrolment and during distinct sub-phenotypes defined according to ACLF development follow-up visits. The design of the PREDICT study is described in and readmission within 3 months after AD.2 In brief, pre-ACLF detail in the supplementary information and elsewhere.2 patients developed ACLF and showed high short-term (90-day) mortality (67%); unstable decompensated cirrhosis (UDC) pa- Identification of precipitants of AD-No ACLF and AD-ACLF tients did not develop ACLF, but required readmission(s) and In order to identify the precipitants an adjudication committee of showed significant short-term mortality (35%); stable the PREDICT study, which included JT, JF, RM and VA, was 1098 Journal of Hepatology 2021 vol. 74 j 1097–1108 PREDICT STUDY COHORT (n = 1,273) AD-ACLF PHENOTYPE AD-No ACLF PHENOTYPE (n = 202) (n = 1,071) AD Pre-ACLF AD-UDC AD-SDC SUB-PHENOTYPE SUB-PHENOTYPE SUB-PHENOTYPE (n = 218) (n = 233 ) (n = 620) AD-NO ACLF COHORT AD Pre-ACLF AD-ACLF Progression to PHENOTYPE ACLF (n = 202) (n = 218) INTEGRATED AD-ACLF COHORT (n = 420) Fig. 1. Schematic outline of the study. AD phenotype groups and subgroups included in each of the AD cohorts used for the study analysis. For more explanation see the text. ACLF, acute-on-chronic liver failure; AD, acute decompensation; SDC, stable decompensated cirrhosis; UDC, unstable decompensated cirrhosis. nominated to elaborate a list of clinical events with the potential other investigations: bacterial infections, alcoholic hepatitis, to precipitate AD or ACLF, and also the general principles and gastrointestinal (GI) bleeding, drug-induced organ injury, ther- specific criteria for diagnosis. This committee identified pre- apeutic interventions. cipitants according to the criteria defined below. Bacterial infections (details in supplementary information). Infections were considered to be potential precipitants if they General principles for precipitant identification were diagnosed at the time of or solved within the 48-hour  Precipitants should consist of events that have the potential to period that preceded the onset of AD. Proven bacterial in- induce impairment in the function of the liver and/or other fections were defined as previously described9 and in accordance organs, either by direct organ injury (e.g. tissue hypo- with the EASL guidelines7 (detailed definition in the perfusion) or, indirectly, through significant dysregulation of supplementary information). important pathophysiological mechanisms (e.g. immune re- Alcohol-related liver injury (details in the supplementary sponses to microbial or endogenous factors). information). Alcoholic hepatitis was diagnosed according to  When assessing the potential of hepatotoxic, nephrotoxic or the clinical criteria of the NIAAA.10 These criteria are in line with neurotoxic drugs as precipitants, the lack of liver, kidney or the clinical diagnosis of alcoholic hepatitis according to the brain dysfunction or failure, respectively, as defined by the existing EASL guidelines.11 Alcoholic hepatitis was considered CLIF-C organ failure score8 rule out drug-induced organ severe if patients had CLIF-C acute decompensation scores of > −50 toxicity as a precipitant. points,12 or ACLF (Table 1).  As suggested by the results of the CANONIC study,1,7 clinically GI bleeding (details in the supplementary information). GI identifiable, relevant and true precipitants should have a bleeding was considered a precipitant if occurring within 7 days higher prevalence in patients with AD-ACLF than in those with prior to the onset of AD-ACLF. Moreover, hemorrhagic shock was AD-no ACLF. indicative of severe bleeding (Table 1).  Precipitants should precede or coincide with the onset of AD- Drug-induced liver injury was considered a potential precip- ACLF. The time period between the precipitants and the onset itant when the hepatotoxic drug was administered within 1 of AD-ACLF, however, is heterogeneous, depending on the month prior to the onset of AD-ACLF and the patient presented precipitants. with liver injury as defined by Hys law and FDA guidance as  Any event developing after the onset of AD-ACLF is a described in the recent EASL guidelines;13 as well as liver complication or a coincidental event but not a precipitant. dysfunction (in patients with AD-No ACLF, bilirubin >6 mg/dl) or liver failure (in patients with AD-ACLF, bilirubin >12 mg/dl). Only drugs from groups A and B of potential hepatotoxic drugs, Specific criteria for the identification of precipitants from the list described elsewhere,14 were considered potential candidates for proposed by the adjudication committee (for detailed description liver toxicity. see supplementary appendix) Drug-induced kidney injury was considered a potential pre- The adjudication committee evaluated the following events as cipitant when the nephrotoxic drug was administered within 7 potential precipitants as proposed by the CANONIC study and days prior to the onset of AD-ACLF and patients presented with Journal of Hepatology 2021 vol. 74 j 1097–1108 1099 Research Article Cirrhosis and Liver Failure Table 1. Clinical events, precipitants and combinations of precipitants in patients with AD-No ACLF and with AD-ACLF. AD-No ACLF (n = 1,071) AD-ACLF (n = 202) p valuea Clinical events, precipitants, n (%) Bacterial infections Any infection 314 (29.32) 101 (50.00) <0.0001 Suspected bacterial infection 74 (6.91) 12 (5.94) 0.61 Proven bacterial infectionsb 239 (22.32) 89 (44.06) <0.0001 Alcohol-related liver injury Alcoholic hepatitis 275 (25.68) 88 (43.56) <0.0001 Severe alcoholic hepatitisb 200 (18.67) 88 (43.56) <0.0001 GI bleeding Any GI bleeding 176 (16.43) 40 (19.80) 0.24 GI bleeding with hypovolemic shockb 13 (1.21) 12 (5.94) <0.0001 Drug-induced brain injury Patients treated with neurotoxic drugs 84 (7.84) 17 (8.42) 0.78 Toxic encephalopathyb 13 (1.21) 12 (5.94) <0.0001 Other candidates, n (%) Paracentesis without albumin 110 (10.28) 21 (10.40) 0.96 TIPS 49 (4.58) 8 (3.96) 0.69 Drug-induced liver injury 16 (1.49) 4 (1.98) 0.54 Viral hepatitis or other viral Infections 13 (1.21) 3 (1.49) 0.72 Drug-induced kidney injury 3 (0.28) 1 (0.50) - Surgery 3 (0.28) 0 (0.00) - Decompensated cardiopulmonary disease 4 (0.37) 3 (1.49) - Dehydration 3 (0.28) 1 (0.50) - Large hematomas 3 (0.28) 0 (0.00) - Acute pancreatitis 1 (0.09) 1 (0.50) - Portomesenteric vein thrombosis 2 (0.19) 1 (0.50) - Extrahepatic autoimmune disease 2 (0.19) 0 (0.00) - Cerebrovascular accident 0 (0.00) 1 (0.50) - Bowel occlusion 1 (0.09) 0 (0.00) - Number of precipitants Indeterminate 662 (61.81) 59 (29.21) <0.0001 1 354 (33.05) 93 (46.04) > −2 55 (5.14) 50 (24.75) Chi-square or Fisher’s tests performed in percentages comparisons. ACLF, acute-on-chronic liver failure; AD, acute decompensation; GI, gastrointestinal; TIPS, transjugular intrahepatic portosystemic shunt. a Certain p value were not determined because of the low number of patients. b Underlined precipitants are those considered as precipitants of AD-ACLF. either renal dysfunction or renal failure according to the CLIF-C for categorical variables, whereas the Student’s t test or analysis organ failure score. Diuretic-induced renal dysfunction or renal of variance were used for normally distributed continuous vari- failure was not considered a nephrotoxic condition. ables and the Wilcoxon rank-sum test or the Kruskal-Wallis test Toxic encephalopathy was considered a potential precipitant for continuous variables not normally distributed. For compari- when the neurotoxic drug was administered within 48 hours sons at different time-points in the same patients, paired tests prior to the onset of AD-ACLF and the patient presented with were used: McNemar test was applied for dichotomic variables encephalopathy in severity similar to brain dysfunction or brain and a test of symmetry was performed for variables with 3 failure according to the CLIF-C organ failure score. categories. In all statistical analyses, significance was set at Therapeutic interventions including transjugular intrahepatic p<0.05. portosystemic shunt (TIPS), major surgical procedures and large Overall, the proportion of missing values in the main reported volume paracentesis without albumin administration were characteristics (demographics, clinical variables, laboratory considered as potential precipitants if performed within 7 days values, precipitants and clinical outcomes) rounded 1% at most. prior to the onset of AD-ACLF. Only complete clinical blood counts and total cholesterol showed higher proportions of missing data, which were mainly due to Other potential precipitants identified by the investigators in the common problems with sample availability or with technical individual patients eCRF laboratory processes that occurred in several site laboratories The adjudication committee assessed 9 additional, infrequent and can be considered to be completely random. A simple clinical events (details in the supplementary information). imputation approach was used to impute the missing values for each of the 4 variables mentioned above (neutrophil, lymphocyte Statistical analysis and monocyte counts and total cholesterol). SAS PROC MI was Discrete variables are shown as counts (percentage) and used assuming an arbitrary pattern for missing values and continuous variables as mean ± SD. Non-normally distributed adopting a fully conditional specification (FCS) regression variables are summarized by the median (IQR). In univariate method. Model covariates included age, sex, CLIF-C organ failure statistical comparisons, the chi-square test or Fisher’s exact test, score and number of precipitants or presence of bacterial in- when at least 25% of expected counts were below 5, were used fections or alcoholism or ACLF at inclusion, depending on the 1100 Journal of Hepatology 2021 vol. 74 j 1097–1108 A Proven bacterial infection B 0.5 Cumulative incidence of Proven bacterial infections 9 Severe alcoholic hepatitis Severe alcoholic hepatitis GI bleeding with shock 0.4 2 2 mortality Toxic encephalopathy 0 0.3 0.2 0.1 p = 0.7666 0.0 0 30 60 90 188 44 155 Days from inclusion C 0.5 Cumulative incidence of Indeterminate PE 1 PE 0.4 ≥ 2 PEs mortality 0.3 1 4 2 0.2 6 0.1 p ≤0.0001 0.0 0 30 60 90 Days from inclusion D Leukocytes (x109/L) E Neutrophils (x109/L) F Monocytes (x109/L) G CRP (mg/L) 25 20 2.0 150 20 15 1.5 100 15 10 1.0 10 50 5 0.5 5 0 0 0.0 0 Indeterminate 1 PE ≥ 2 PEs Indeterminate 1 PE ≥2 PEs Indeterminate 1 PE ≥2 PEs Indeterminate 1 PE ≥2 PEs PE PE PE PE Fig. 2. Precipitants in AD-No ACLF. Combinations of PEs in the AD-No ACLF cohort shown in a 4-set circle Venn diagram (panel A). Cumulative incidence of mortality in patients with AD-No ACLF according to the type of precipitant (proven infections alone vs. severe alcoholic hepatitis alone; panel B) and the number >2 PEs; panel C) p values were obtained from Gray’s test. Blood levels of leukocytes (panel D), neutrophils (panel E), of precipitants (indeterminate PE, 1 PE, and − monocytes (panel F) and the serum concentration of CRP (panel G) in patients with AD-No ACLF and indeterminate PE, 1 PE and − >2 PEs. Boxes show median and IQR and whiskers show 10-90 percentiles. Kruskal-Wallis test was performed with all values in each comparison. Differences were statistically significant (p<0.0001) for all biomarkers. ACLF, acute-on-chronic liver failure; AD, acute decompensation; CRP, C-reactive protein; PE(s), precipitant(s). Table 2. Type and number of precipitants in patients with pre-ACLF, unstable decompensated cirrhosis and stable decompensated cirrhosis. Pre-ACLF (n = 218) At enrolment At ACLF development UDC (n = 233) SDC (n = 620) Type of precipitant, n (%) Proven bacterial infections 64 (29.4) 97 (44.5)** 49 (21.0) *,## 126 (20.3) **,## Severe alcoholic hepatitis 58 (26.6) 57 (26.1) 45 (19.3) + 97 (15.6) **,# GI bleeding with shock $ 2 (0.9) 8 (3.7) 2 (0.9) 9 (1.5) Toxic encephalopathy $ 3 (1.4) 4 (1.8) 3 (1.3) 7 (1.1) Number of precipitants, n (%) Indeterminate 111 (50.9) 88 (40.4)** 142 (60.9) *,## 409 (66.0) **,## One 88 (40.4) 98 (45.0)** 83 (35.6) ## 183 (29.5) ## Two or more 19 (8.7) 32 (14.7)** 8 (3.4) ## 28 (4.5) ## Comparison between all groups to the pre-ACLF group at enrolment is displayed by the following symbols: + p<0.07, *p<0.05 and **p<0.01 vs. the pre-ACLF group at enrolment. Comparison between all groups to the pre-ACLF group at ACLF development is displayed by the following symbols: # p<0.001 and ##p<0.0001 vs. pre-ACLF group at ACLF development. Chi-square or Fisher’s tests performed in percentages comparisons among groups. McNemar test used in paired comparisons for the types of precipitant between the 2 time-points in pre-ACLF group. Symmetry test used in paired comparisons for the number of precipitant between the 2 time-points in pre-ACLF group. $ p value not determined due to the low number of patients. ACLF, acute-on-chronic liver failure; AD, acute decompensation; SDC, stable decompensated cirrhosis; UDC, unstable decompensated cirrhosis. Journal of Hepatology 2021 vol. 74 j 1097–1108 1101 Research Article Cirrhosis and Liver Failure analysis that was to be performed. For each variable, missing infections, alcohol-related liver injury, GI bleeding and toxic values were imputed by computing the median of the values encephalopathy (Table 1). obtained by fitting the model on 100 repetitions generated from The prevalence of patients with proven bacterial infections the original dataset. was significantly higher in AD-ACLF than in AD-No ACLF cases, Cumulative incidence functions (CIF’s) were used to estimate while prevalence of suspected bacterial infections was very low survival curves accounting for liver transplantation as an event and similar in both groups. Therefore, only proven bacterial in- ’competing’ with mortality, as well as to estimate ACLF devel- fections were considered as precipitants of AD-ACLF, and this opment accounting for both mortality and liver transplantation was the most common precipitant (44% in AD-ACLF and in 22.3% as events “competing” with ACLF development, using common in AD-No ACLF [p<0.0001]). non-parametric methods. The equality of CIFs across groups was Prevalence of severe alcoholic hepatitis (alcoholic hepatitis evaluated by means of the Gray’s test.15 Statistical analysis was associated with CLIF-C AD score > −50 or ACLF) was significantly performed using SAS v9.4 and plots were performed with R higher in patients with AD-ACLF (43.6% vs. 18.7% in AD-No v1.2.5042 and GraphPad Prism v5 software. ACLF). Overall, alcoholic hepatitis was not always associated with organ dysfunction. Therefore, only severe alcoholic hepa- titis was identified as a precipitant, and was the second most Results frequent. Identification of precipitants for AD at enrolment in the Severe GI bleeding associated with hypovolemic shock was PREDICT study cohort the third most frequent precipitant, although its prevalence in The PREDICT study cohort includes 1,273 patients, of whom 202 the AD-ACLF and the AD-No ACLF group (5.9% and 1.2%, respec- patients presented with AD-ACLF and 1,071 patients with AD-No tively, p<0.0001) was low. ACLF (Fig. 1). There were 4 main precipitants: bacterial A Proven bacterial infection B 0.8 Proven bacterial infections Cumulative incidence of 6 Severe alcoholic hepatitis Severe alcoholic hepatitis 4 GI bleeding with shock 0.6 4 mortality Toxic encephalopathy 2 4 0.4 0 4 0.2 6 p = 0.5916 0.0 0 30 60 90 Days from ACLF diagnosis 111 55 73 C 0.8 Cumulative incidence of Indeterminate PE 1 PE ≥2 PEs 0.6 mortality 0.4 0.2 p = 0.0042 0.0 0 30 60 90 Days from ACLF diagnosis D Leukocytes (x109/L) E Neutrophils (x109/L) F Monocytes (x109/L) G CRP (mg/L) 30 30 3 150 20 20 2 100 10 10 1 50 0 0 0 0 Indeterminate 1 PE ≥2 PEs Indeterminate 1 PE ≥2 PEs Indeterminate 1 PE ≥2 PEs Indeterminate 1 PE ≥2 PEs PE PE PE PE Fig. 3. Precipitants in AD-ACLF. Combinations of PEs in the integrated AD-ACLF cohort shown in a 4-set circle Venn diagram (panel A). Cumulative incidence of mortality in patients with AD-ACLF according to the type of PE (proven infections alone vs. severe alcoholic hepatitis alone; panel B) and the number of PEs (indeterminate PE, 1 PE, and > −2 PEs; panel C); p-values were obtained from Gray’s test. Blood levels of leukocytes (panel D), neutrophils (panel E), monocytes (panel F) and the serum concentration of CRP (panel G) in patients with AD-ACLF and indeterminate PE, 1 PE and > −2 PEs. Boxes show median and IQR and whiskers show 10-90 percentiles. Kruskal-Wallis test was performed with all values in each comparison. Differences were statistically significant (p<0.0001) for all biomarkers. ACLF, acute-on-chronic liver failure; AD, acute decompensation; CRP, C-reactive protein; PE(s), precipitant(s). 1102 Journal of Hepatology 2021 vol. 74 j 1097–1108 Finally, of the 3 examined types of drug-induced organ with severe toxic encephalopathy were opioids or injury, only the prevalence of toxic encephalopathy benzodiazepines. was significantly higher in the AD-ACLF group than in the Neither therapeutic paracentesis without intravenous albu- AD-No ACLF group (5.9% and 1.2%, respectively, p<0.0001) and min nor TIPS qualified as precipitants, since their prevalence was it thus qualified as a precipitant. All drugs associated not significantly higher in patients with AD-ACLF than in patients with AC-no ACLF. Table 3. Demographic data and etiology, clinical and laboratory data at diagnosis, specific treatments during follow-up and mortality in patients included in the integrated AD-ACLF cohort according to the number of precipitants. Indeterminate precipitant (n = 147) 1 precipitant (n = 191) > −2 precipitants (n = 82) p value Demographic data and etiology of cirrhosis Age, year, mean ± SD 61.2 ± 11.38 60.5 ± 11.06 52.1 ± 11.41a <0.0001 Male sex, n (%) 99 (67.3) 137 (71.7) 52 (63.4) 0.36 Alcohol-related cirrhosis, n (%) 81 (55.1) 144 (75.4)b 77 (93.9)a <0.0001 Data at ACLF diagnosis Systemic hemodynamics, mean ± SD Mean arterial pressure (mmHg) 80.8 ± 12.51 79.0 ± 13.05 76.1 ± 13.65b 0.0419 Heart rate (bpm) 79.4 ± 15.80 82.0 ± 17.26 92.9 ± 19.93a <0.0001 Complications, n (%) Ascites 90 (73.2) 134 (74.9) 71 (88.8)a 0.0206 Hepatic encephalopathy 61 (49.6) 112 (62.6)b 62 (77.5)a 0.0003 GI bleeding 16 (13.1) 16 (8.9) 19 (23.8)c 0.0053 Organ failures, n (%) Liver failure 29 (23.6) 60 (33.5) 49 (61.3)a <0.0001 Renal failure 84 (68.3) 98 (54.7)b 33 (41.3)a 0.0006 Brain failure 13 (10.6) 31 (17.3) 27 (33.8)a 0.0002 Coagulation failure 25 (20.3) 41 (23.0) 28 (35.0)a 0.0474 Cardiovascular failure 6 (4.9) 25 (14.0)b 27 (33.8)a <0.0001 Respiratory failure 3 (2.4) 21 (11.9)b 13 (16.3)b 0.0022 Biomarkers of systemic inflammation, median (IQR) White blood cell count, x109/L 7.19 (5.03–9.40) 9.72 (6.39–13.50)b 12.14 (8.57–18.10)a <0.0001 Neutrophil count, x109/L 4.23 (2.25–6.85) 7.32 (4.60–10.45)b 9.56 (6.44–15.50)a <0.0001 Lymphocyte count, x109/L 0.85 (0.65–1.40) 0.94 (0.56–1.56) 1.20 (0.73–1.97)b 0.0794 Monocyte count, x109/L 0.60 (0.40–0.92) 0.92 (0.65–1.22)b 1.32 (0.95–1.77)a <0.0001 Serum C-reactive protein, mg/L 17.60 (8.80–32.00) 32.30 (15.00–58.90)b 36.15 (18.00–75.00)b <0.0001 Measurements estimating organ function Serum bilirubin, mg/dl, median (IQR) 2.29 (1.12–11.04) 5.70 (2.12–14.80)b 14.53 (6.55–23.08)a <0.0001 Serum albumin, g/dl, mean ± SD 3.0 ± 0.82 2.9 ± 0.68 2.9 ± 0.65 0.45 Total cholesterol, mg/dl, median (IQR) 86.70 (57.75–123.80) 70.50 (48.50–104.00)b 64.50 (42.00–83.50)b 0.0145 International normalized ratio, median (IQR) 1.53 (1.32–2.13) 1.75 (1.45–2.34)b 2.18 (1.80–2.78)a <0.0001 Serum creatinine, mg/dl, median (IQR) 2.15 (1.54–2.80) 2.00 (1.04–2.50)b 1.55 (0.82–2.81)b 0.0024 Serum sodium, mEq/L, mean ± SD 133.6 ± 6.77 133.6 ± 6.36 134.4 ± 8.71 0.70 Prognostic scores, mean ± SD Child-Pugh score 9.5 ± 2.41 10.5 ± 2.18b 11.8 ± 1.50a <0.0001 MELD score 24.3 ± 6.21 25.6 ± 6.41 29.8 ± 6.13a <0.0001 MELD-Na score 26.6 ± 6.11 27.9 ± 5.81 31.2 ± 5.83a <0.0001 CLIF-C organ failure score 8.9 ± 1.70 9.7 ± 1.97b 11.3 ± 2.20a <0.0001 CLIF-C ACLF score 45.7 ± 7.45 50.1 ± 8.05b 54.1 ± 10.86a <0.0001 ACLF grades, n (%) ACLF grade I 93 (76.2) 105 (59.7)b 24 (30.0)a <0.0001 ACLF grade II 23 (18.9) 53 (30.1)b 34 (42.5)a ACLF grade III 6 (4.9) 18 (10.2)b 22 (27.5)a Specific treatments and mortality Specific treatments from ACLF, n (%) Intensive care 15 (10.2) 41 (21.5)b 32 (39.0)a <0.0001 Renal replacement 8 (5.4) 13 (6.8) 14 (17.1)a 0.0055 Mechanical ventilation 3 (2.4) 22 (12.3)b 22 (27.5)a <0.0001 Vasopressors 35 (23.8) 72 (37.7)b 52 (63.4)a <0.0001 90-day liver transplantation 19 (13.1) 25 (13.4) 5 (6.3) 0.22 Mortality after ACLF diagnosis, n (%) 90-day mortality 62 (42.2) 95 (49.7) 52 (63.4)a 0.0087 a p<0.05 vs. no precipitant and 1 precipitant. b p<0.05 vs. indeterminate precipitant. c p<0.05 vs. 1 precipitant. Chi-square or Fisher’s tests performed in percentages comparisons. For continuous variables comparisons, analysis of variance for normally distributed variables or Kruskal-Wallis test for not normally distributed variables were used. ACLF, acute-on-chronic liver failure; AD, acute decompensation; GI, gastroin- testinal; MELD, model for end-stage liver disease. Journal of Hepatology 2021 vol. 74 j 1097–1108 1103 Research Article Cirrhosis and Liver Failure In total, 721 patients (56.6%) included in the PREDICT study Interestingly, patients with a single precipitant of the 2 major cohort did not present any identifiable precipitant (indetermi- groups of precipitants (proven bacterial infection and severe nate precipitant), 447 (35.1%) presented 1 precipitant, and 105 alcoholic hepatitis) showed a comparable 90-day mortality (8.2%) presented >−2 precipitants. (Fig. 2B). This is the case, despite the significant differences in The clinical characteristics, laboratory data, prognostic scores, clinical and laboratory parameters between patients with either and 90-day mortality rate of patients with AD-No ACLF and AD- proven bacterial infection or severe alcoholic hepatitis as sole ACLF are presented in Table S1. precipitant (Table S2), indicating that the type of precipitant is not crucial for outcome, if correctly defined. Prevalence and association of precipitants with As shown in Fig. 2C, 90-day mortality was highest in patients characteristics, clinical course and prognosis of patients with >−2 precipitants and lowest in patients without any identi- included in the AD-No ACLF cohort fiable precipitant (Fig. 2C). In parallel, levels of leukocytes, neu- Prevalence of precipitants and their combinations trophils, monocytes and C-reactive protein (CRP) (Fig. 2D-G), AD-No ACLF (n = 1,071) was associated with 1 precipitant in 354 organ dysfunction and failures and overall scores increased with patients (33.0%), and with > the number of precipitants (Table S3). −2 precipitants in 55 patients (5.1%), as illustrated in Fig. 2A. In the AD-No ACLF cohort, 662 patients (61.8%) presented with indeterminate precipitants (Table 1). Therefore, in 394 patients (96.3%), AD-No ACLF was related to proven bacterial infections or severe acute alcoholic hepatitis, Results derived from the integrated ACLF cohort either alone or in combination. AD-No ACLF was unrelated to This integrated cohort included 202 patients with AD-ACLF at the bacterial infections or alcoholic hepatitis in only 15 (3.7%) time of enrolment (AD-ACLF group) and 218 patients in AD-pre- patients. ACLF group who developed AD-ACLF during the study and who were included at the time of development of ACLF (Fig. 1). The Precipitants are associated with the clinical course and integrated AD-ACLF cohort was developed with 2 objectives: i) a survival of patients with AD-No ACLF. further characterization of the AD-ACLF phenotype in patients Prevalence of patients with proven bacterial infections and se- with community-acquired and hospital-acquired ACLF; and ii) an vere alcoholic hepatitis at enrolment was higher in AD-pre-ACLF analysis of precipitants in a sufficiently sized AD-ACLF cohort. (29.4% and 26.6%, respectively) than in AD-UDC (21.0% and 19.3%) or AD-SDC (20.3%. and 15.6%) phenotypes. Moreover, the number of patients with indeterminate precipitants was signif- icantly lower (50.9%) and the number of patients with 1 or > −2 precipitants was higher (40.4% and 8.7%) in patients with AD- Prevalence of precipitants and their combinations pre-ACLF than in those with AD-UDC (60.9%, 35.6% and 3.4%, Of the 420 patients included in the integrated AD-ACLF cohort, respectively) and AD-SDC (66.0%, 29.5% and 4.5%). Moreover, AD-ACLF was triggered by 1 precipitant in 191 patients (45.5%), these differences were even more pronounced, when UDC or and by or > −2 in 82 patients (19.5%), while precipitant was SDC groups at baseline were compared with the AD-pre-ACLF indeterminate in 147 patients (35.0%) (Table S4). Fig. 3A shows group at the time point of ACLF development. These observa- the different combinations of precipitants in the Integrated AD- tions suggest that the presence and the number of precipitants at ACLF cohort. Like the AD-No ACLF cohort, 266 (97.4%) of pa- enrolment are important determinants in the development of tients with identifiable precipitants had proven bacterial in- AD-pre-ACLF, the most severe sub-phenotype in patients with fections or severe acute alcoholic hepatitis as either a single or as AD-No ACLF (Table 2). combined precipitants. Table 4. Adequacy of initial antibiotic strategies. Type of empirical antibiotic strategies Total Classic* Piperacillin- MDR p value tazobactam coverage** Number of all proven bacterial infections 440 273 70 92 Resolution of infection without further escalation or bacterial susceptibility to initial antibiotics in 62.5 54.2 68.6 82.6 <0.0001 culture positive infections (%) Bacterial susceptibility to initial antibiotics in culture positive infections (%) 61.6 48.3 70.7 93.4 <0.0001 Number of proven bacterial infections precipitating AD 265 187 34 39 Resolution of infection without further escalation or bacterial susceptibility to initial antibiotics in 68.1 62.0 73.5 92.3 0.0008 culture positive infections (%) Bacterial susceptibility to initial antibiotics in culture positive infections (%) 63.9 54.4 75.0 100.0 <0.0001 Number of proven bacterial infections precipitating ACLF 175 86 36 53 Resolution of infection without further escalation or bacterial susceptibility to initial antibiotics in 54.3 37.2 63.9 75.5 <0.0001 culture positive infections (%) Bacterial susceptibility to initial antibiotics in culture positive infections (%) 58.5 36.7 66.7 89.2 <0.0001 Adequacy based on clinical criteria (resolution of infection without further escalation or bacterial susceptibility to initial antibiotics in culture positive infections) and the microbiological criterion (bacterial susceptibility to initial antibiotics in culture positive infections) in the whole series of proven infections and in infections precipitating AD and ACLF, according to empirical antibiotic strategies. *One to third generation cephalosporins, amoxicillin-clavulanic acid, quinolones; **carbapenem±glycopeptide/linezolid/daptomycin or tigecycline; Chi-square or Fisher’s tests used to compare percentages. ACLF, acute-on-chronic liver failure; AD, acute decompensation; MDR, multidrug resistant. 1104 Journal of Hepatology 2021 vol. 74 j 1097–1108 A of ACLF development 0.8 Inadequate B 0.8 Inadequate C 0.8 Inadequate Cumulative incidence Cumulative incidence Cumulative incidence Adequate Adequate Adequate 0.6 0.6 0.6 of mortality of mortality 0.4 0.4 0.4 0.2 0.2 0.2 p = 0.0005 p = 0.0004 p = 0.0017 0.0 0.0 0.0 0 30 60 90 0 30 60 90 0 30 60 90 Days from AD diagnosis Days from AD diagnosis Days from ACLF diagnosis Fig. 4. Treatment of bacterial infections. Prognostic impact of inappropriate empirical antibiotic therapy in patients with AD and ACLF. (A) Probability of ACLF at day 90 in infected patients with AD receiving adequate or inadequate empirical antibiotic strategies (B-C) Probability of death at day 90 in patients with AD (B) and ACLF (C). Inadequacy of empirical strategies significantly increased the probability of ACLF and death in the different populations. p values were obtained from Gray’s test. ACLF, acute-on-chronic liver failure; AD, acute decompensation. The type of precipitant is significantly associated with clinical (MDR) strains were involved in 18.9% of all infections and 29.4% characteristics, but not clinical course and mortality of patients with of culture positive episodes. Also, prevalence of infections caused AD-ACLF in the integrated cohort by MDR strains was significantly higher in infections precipi- As in AD-No ACLF patients (Table S2), patients with AD-ACLF had tating ACLF during follow-up and in those causing severe sepsis/ different clinical characteristics (among others higher bilirubin shock (Table S6). Overall, resolution of infection was significantly but lower CRP values in severe alcoholic hepatitis) depending on lower in episodes caused by MDR bacteria (57.8% vs. 82.1%, the type of single precipitant: proven bacterial infections or se- p<0.0001). The lower resolution rate of MDR-infections was vere alcoholic hepatitis (Table S5). Similar to AD-No ACLF pa- associated with higher 28-day and 90-day mortality in patients tients, these differences did not impact the clinical course and with AD-ACLF, but not in infections precipitating AD (Table S7). prognosis, as shown in Fig. 3B. Classic antibiotic strategies were used frequently as first-line therapy in community-acquired and healthcare-associated in- fections (Table 4). In contrast, nosocomial episodes were more Number of precipitants is significantly associated with the clinical frequently treated with piperacillin-tazobactam (20.4%) or with course and mortality of patients with AD-ACLF. The number of precipitants in patients included in the integrated broader MDR-covering strategies (38.8%). Remarkably, a signifi- cant percentage of patients with severe sepsis/shock still AD-ACLF cohort (indeterminate, 1 precipitant, and 2 or 3 pre- received classic schemes not covering MDR strains (40.5%). cipitants) correlated positively with the prevalence of liver, brain, coagulation and cardio-circulatory failure and inversely with the Empirical MDR-covering strategies were more effective in infection resolution (with regard to clinical response and prevalence of renal failure. These findings were due to differ- microbiological susceptibility) than classic schemes (Table 4, ences in the predominance of specific organ failures among pa- tients with a distinct number of precipitants. The predominant Table S8). Adequacy of empirical antibiotic therapy was defined as resolution of infection without further escalation or bacterial organ failure in patients with an indeterminate precipitant or susceptibility to initial antibiotics in culture positive infections. with only 1 precipitant was renal failure. By contrast, liver failure Importantly, adequacy of first-line antibiotic strategies decreased was the predominant organ failure in patients with 2 or 3 pre- cipitants. Moreover, the prevalence of other organ failures was the cumulative incidence of developing ACLF in patients with AD (21.3% vs. 39.2%, Fig. 4A) and 90-day mortality in both AD (16.9% also higher in patients with 2 or 3 precipitants. Consistent with vs. 36.5%, Fig. 4B) and ACLF patients (44.2% vs. 66.2%, Fig. 4C). these results, the number of precipitants at diagnosis also correlated directly with the grade of severity of ACLF (I, II or III), the severity of prognostic scores, the need for intensive care, the Severe alcoholic hepatitis frequency of treatment with mechanical ventilation or renal Steroids were administered in 49 patients with severe alcoholic replacement therapy, and the 90-day cumulative incidence of hepatitis (18.9%), 30 patients with AD and 19 patients with ACLF mortality (Table 3, Fig. 3C). Systemic inflammation, as estimated at inclusion. The 28-day and 90-day mortality rates were not by the white blood cell count and blood levels of neutrophils and significantly different between patients receiving or not monocytes, increased in parallel with the number of precipitants receiving steroids, neither in the whole population nor in pa- (Table 3, Fig. 3D-G). Serum levels of CRP were also significantly tients with AD or ACLF at inclusion (Table S9). higher in patients with 1 or −>2 precipitants than in patients with indeterminate precipitants. Discussion The PREDICT study offers a comprehensive investigation char- Role of treatment of precipitant in prevention of AD-ACLF and acterizing the precipitants of AD and demonstrating their impact improvement of survival on the development of AD-ACLF and prognosis. Proven bacterial infections (details in the supplementary The CANONIC study characterized the AD-ACLF phenotype information) and attributed an important role to precipitants in its develop- A total of 376 patients (29.5%) developed 440 bacterial infections, ment. The PREDICT study, designed to assess the period prior to of which 66.2% were culture positive. Nosocomial episodes and ACLF,2 identified 3 different clinical courses in AD-No ACLF: pre- severe sepsis or shock predominated in infections diagnosed as a ACLF, AD-UDC and AD-SDC. Moreover, the PREDICT study precipitant of ACLF during follow-up, while multidrug resistant assessed how type and number of precipitants influence the Journal of Hepatology 2021 vol. 74 j 1097–1108 1105 Research Article Cirrhosis and Liver Failure clinical course and the prognosis in patients with both AD-No course severity and the 90-day cumulative incidence of mortal- ACLF and AD-ACLF. This prospective and detailed characteriza- ity. Not only that multiple (2 or more) precipitants trigger AD- tion offers diagnostic criteria for precipitants and rationalizes the ACLF (one in 5 patients) and is exceptional (one in 20 patients) identification of precipitants in patients with cirrhosis and AD. in AD-No ACLF, but also the intensity of systemic inflammation, The criteria used for the diagnosis of precipitants considered the the prevalence of organ failures, the need for organ support, and severity of the precipitant, the time interval between onset/ the prognostic scores increased progressively from patients with resolution of the precipitant and onset of the AD episode, and indeterminate precipitants to patients with 1 and multiple pre- their higher prevalence in patients with AD-ACLF than in pa- cipitants. Therefore, when precipitants are defined according to tients with AD-No ACLF, which are more objective than the these criteria, they are synergistic and additive in the worsening traditional principles of chronology and potential of organ injury. of outcome, despite different clinical characteristics. Among the events recorded and evaluated in the PREDICT Almost all (>96%) patients with precipitants showed proven study, only 4 fulfilled the properties of precipitants (chronology, bacterial infection and/or severe alcoholic hepatitis, either alone severe organ injury or higher prevalence in the AD-ACLF or in combination with other precipitants. This overwhelming phenotype): proven bacterial infections, severe alcoholic hepa- prevalence of proven bacterial infections and/or severe alcoholic titis, GI bleeding with shock and toxic encephalopathy. While hepatitis as precipitants suggests that diagnosing, preventing paracentesis without intravenous albumin administration and and treating these precipitants is paramount to improve the TIPS did not induce organ impairment (TIPS even improves prognosis of patients with decompensated cirrhosis. survival in GI bleeding and ACLF16,17), the prevalence of drug- PREDICT demonstrates that proven bacterial infections induced liver or renal injury and of other potential precipitants require specific and adequate treatment for prevention of AD- proposed by the investigators was extremely low, frequently ACLF. This is of particular importance since MDR may challenge below 1%, suggesting that they could be coincidental rather than empirical treatments. The overall prevalence of MDR bacterial precipitants. infections in PREDICT was in line with that reported in recently Proven bacterial infections and severe alcoholic hepatitis published multicenter investigations19,20 and MDR bacterial in- were by far the most prevalent precipitants observed in the fections were more severe (higher rate of severe sepsis/shock PREDICT study. Prevalence of GI bleeding associated with shock and of ACLF), associated with a lower resolution rate and higher and toxic encephalopathy was considerably lower in both 28-day and 90-day mortality,19–21 underlining the importance of groups. In patients with AD-No ACLF, the prevalence of proven treatment of precipitants, thus confirming that definition and bacterial infections or severe alcoholic hepatitis and the number selection of precipitants has been chosen adequately. Impor- of precipitants present at enrolment were higher in patients with tantly, classic antibiotics (1st-3rd generation cephalosporins, AD-Pre-ACLF than in patients with AD-UDC and AD-SDC. In quinolones) have an unacceptable efficacy (<40%) in nosocomial contrast, no differences were found in the prevalence of these infections, or in those with severe sepsis or shock. These findings precipitants between patients with UDC or SDC. These findings support the current recommendations on the use of empirical suggest that precipitants are determinants of the development of broad schemes, according to specific epidemiological pattern of the AD-Pre-ACLF sub-phenotype, which is associated with a antibiotic resistance,22–24 in the nosocomial setting and in severe worse clinical course and prognosis in patients with AD-No ACLF. sepsis/shock with rapid de-escalation strategies.13 Importantly, in >60% of the patients in the AD-No ACLF cohort, In summary, of the clinical events explored as potential pre- precipitating events were indetermined at enrolment, while this cipitants in the PREDICT study, only 4 (proven bacterial in- was the case in only 35% of the patients with AD-ACLF. These fections, severe acute alcoholic hepatitis, GI bleeding associated data suggest that AD-No ACLF develops more frequently in the with shock and toxic encephalopathy) fulfilled the diagnostic context of endogenous mechanisms (e.g. progressive liver dis- criteria of precipitants. Proven bacterial infections and severe ease, bacterial translocation), confirming the CANONIC study and alcoholic hepatitis were present in the absolute majority (>96%) underlining the solidity of the PREDICT study. of patients. However, no precipitating event could be identified The type of precipitant was associated with different clinical in 2/3 of AD-No ACLF patients and in 1/3 AD-ACLF patients. The characteristics, but a similar clinical course and mortality. This prevalence and number of precipitants increased with severity of finding is not surprising, since reactivation or superimposed the AD-sub-phenotype form SDC/UDC to pre-ACLF and ACLF, hepatitis, also showed different prevalences of specific organ which were also directly related with clinical course severity and failures in AD-ACLF, but similar outcome as AD-ACLF triggered by short-term mortality in patients with AD. Our data, therefore, extrahepatic precipitants (e.g GI bleeding).18 The explanation strongly suggest that precipitants are significantly associated may be due to the sequence of mechanisms. Bacterial infections with the clinical course and prognosis of patients with AD and would induce systemic inflammation as the primary mechanism, specific preventive and therapeutic strategies for these pre- leading to predominantly circulatory and renal dysfunction or cipitants are required to improve outcomes in decompensated failure. In contrast, the direct insult of alcohol toxicity induces cirrhosis. hepatic inflammation and cell death as primary mechanisms culminating in liver and coagulation dysfunction or failure. Abbreviations Importantly, systemic inflammation aggravates and leads to an ACLF, acute-on-chronic liver failure; ALT, alanine aminotrans- identical syndrome through distinct pathophysiological path- ferase; AST, aspartate aminotransferase; CLIF, chronic liver fail- ways. For this reason, the criterion of severity (either systemic ure; CIF, cumulative incidence of function; CRP, C-reactive inflammation or organ injury) of the event is crucial to identify protein; GI, gastrointestinal; OF, organ failure; MDR, multidrug the precipitant. resistant; MELD, model of end-stage liver disease; PE, precipi- Finally, our results show that the number of precipitants was tating event; SDC, stable decompensated cirrhosis; UDC, unsta- an important determinant for the characteristics, the clinical ble decompensated cirrhosis. 1106 Journal of Hepatology 2021 vol. 74 j 1097–1108 Financial support Florian Rainer27, Jörg Tobiasch Moritz28, Mónica Mesquita29, The study was supported by the European Foundation for the Edilmar Alvarado-Tapias30, Osagie Akpata31, Luise Aamann32, Study of Chronic Liver Failure (EF-Clif). EF-Clif is a non-profit Didier Samuel33, Sylvie Tresson33, Pavel Strnad34, Roland Ama- private organization. EF-Clif receives unrestricted donations thieu35, Macarena Simón-Talero36, Francois Smits11, Natalie van from Cellex Foundation and Grifols. EF-Clif is partner, contributor den Ende13, Javier Martinez12, Rita Garcia14, Harald Rup- and coordinator in several EU Horizon 2020 program projects. JT prechter17, Cornelius Engelmann22, Osman Cavit Özdogan37 was appointed visiting Professor in EF-Clif for the execution of the study by a grant from Cellex Foundation. The funders had no Affiliations: influence on study design, data collection and analysis, decision 1 Munster University Hospital, Münster, Germany to publish or preparation of the manuscript. 2 University of Basel-St Gall Cantonal Hospital, Switzerland 3 Hôpitaux Universitaires de Genève, Genève, Switzerland 4 Conflict of interest University of Birmingham, Birmingham, UK 5 None of the authors have conflicts of interest for the reported University of Debrecen, Faculty of Medicine, Institute of Medi- study. cine, Department of Gastroenterology, Debrecen, Hungary 6 Please refer to the accompanying ICMJE disclosure forms for University of Bologna, Bologna, Italy 7 further details. University of Padova, Padova, Italy 8 A.O.U. Città della Salute e della Scienza Torino, Torino, Italy 9 Authors’ contributions University Hospital Bonn, Bonn, Germany 10 JT, JF, WL, JC, RJ, RM, PG, PA, VA: study concept and design, Hospital Clinic of Barcelona, Barcelona, Spain 11 acquisition of data, analysis and interpretation of data, drafting of C.U.B. Erasme, Bruxelles, Belgium 12 the manuscript, funding recipient, administrative, technical and Department of Gastroenterology, Hospital Universitario Ramón material support, study supervision; EG, AA, AC, CP, MP, CS, AC, y Cajal, IRYCIS, University of Alcalá, CIBEREHD, Madrid, Spain 13 AM, FA: acquisition of data, analysis of data, technical and ma- Department of Gastroenterology and Hepatology, Section of terial support; TT, MB, PA, CA, FEU, CJ, MST, TG, DLS, AA, WL, ES, Liver and Biliopancreatic disorders, University of Leuven, Leuven, RB, MJ, CS, TR, JA, PG, WB, SZ, CR, TB, AS, KVD, MC, OR, RS, HZ, AC, Belgium 14 GSP, AdG, HG, FS, CT, OCÖ, FS, SR, RA, MRG, HVV, CF, MM, MP, PC, Hospital General Universitario Gregorio Marañón. Facultad de SP, IG, MP, VV, RM, ZV, MB, EB: acquisition of data, interpretation Medicina (Universidad Complutense of Madrid), CIBERehd, of data, critical revision of the manuscript regarding important Madrid, Spain 15 intellectual content Pavol Jozef Safarik University in Kosice, Kosice, Slovakia 16 Department of Medicine II, Liver Centre Munich, University Data availability statement Hospital, LMU, Munich, Germany 17 While some of the data of this paper will be available upon Medical University of Vienna, Vienna, Austria 18 request, the majority of the data are not suitable for posting as Department of Internal Medicine I, Goethe University Frank- they are confidential. furt, Frankfurt Germany 19 Internal Medicine PO Ostuni, ASL Brindisi, Italy 20 King’s College Hospital, London, UK Acknowledgements 21 University Hospital Halle-Wittenberg, Halle(Saale), Germany The authors are very grateful to the patients, their families and 22 Division of Hepatology, Department of Medicine II, Leipzig the personnel of the hospitals for making this possible. In University addition, a special thank you is dedicated to Mrs. Yolanda Godoy, Medical Center, Leipzig, Germany Dr. Anna Bosch, Dr. Josep-Maria Torner, Mrs. Cecilia Ducco and 23 Jena University Hospital, Jena, Germany Montserrat Carreras for excellent assistance in the accomplish- 24 Gastrounit, Medical Section, Hvidovre Hospital and Depart- ment of the study. We thank Marites Abans, Paul Sauerbruch, ment of Clinical Medicine, University of Copenhagen, Copenha- Gudrun Hack, Nadine Köstlmeier, Kristin Gehrmann for technical gen, Denmark and administrative assistance. 25 Leiden University Medical Center, Leiden, The Netherlands 26 Universitá Sapienza Roma, Roma, Italy 27 Appendix Medical University of Graz, Graz, Austria 28 Collaborators: Medical University of Innsbruck, Innsbruck, Austria Miriam Maschmeier1, David Semela2, Laure Elkrief3, Ahmed 29 CHTMAD Vila Real-Blueclinical, Vila Real, Portugal Elsharkawy4, Tamas Tornai5, Istvan Tornai5, Istvan Altorjay5, 30 Hospital de la Santa Creu i Sant Pau and CIBERehd, Barcelona, Agnese Antognoli6, Maurizio Baldassarre6, Martina Gagliardi6, Spain Eleonora Bertoli7, Sara Mareso7, Alessandra Brocca7, Daniela 31 UCL Medical School,Royal Free Hospital, London, UK Campion8, Giorgio Maria Saracco8, Martina Rizzo8, Jennifer 32 Aarhus University Hospital, Aarhus, Denmark Lehmann9, Alessandra Pohlmann9, Maximilian J. Brol9, Johannes 33 AP-HP Hôpital Paul Brousse, Centre Hépato-Biliaire, Universite Chang9, Robert Schierwagen18, Elsa Solà10, Nesrine Amari11, Paris Saclay, INSERM Unit 1193, Villejuif, France Miguel Rodriguez12, Frederik Nevens13, Ana Clemente14, Martin 34 Aachen University Hospital, Aachen, Germany Janicko15, Daniel Markwardt16, Mattias Mandorfer17, Christoph 35 AP-HP, Hôpital Jean Verdier, Service d’Hépatologie, Bondy; Welsch18, Tanja M. Welzel18, Emanuela Ciraci19, Vish Patel20, Université Paris 13, Sorbonne Paris Cité, “Equipe labellisée Ligue Cristina Ripoll21, Adam Herber22, Paul Horn23, Flemming Bend- Contre le Cancer”, Saint-Denis; Inserm, UMR-1162, “Génomique tsen24, Lise Lotte Gluud24, Jelte Schaapman25, Oliviero Riggio26, fonctionnelle des tumeurs solides”, Paris, France Journal of Hepatology 2021 vol. 74 j 1097–1108 1107 Research Article Cirrhosis and Liver Failure 36 Liver Unit, Hospital Vall d’Hebron, Universitat Autònoma de [11] European Association for the Study of the Liver. Electronic address eee, Barcelona, CIBEREHD, Barcelona, Spain European association for the study of the L. EASL clinical practice guide- 37 lines: management of alcohol-related liver disease. J Hepatol Marmara University, Kadiköy, Turkey 2018;69:154–181. [12] Jalan R, Pavesi M, Saliba F, Amoros A, Fernandez J, Holland-Fischer P, et al. Supplementary data The CLIF Consortium Acute Decompensation score (CLIF-C ADs) for Supplementary data to this article can be found online at https:// prognosis of hospitalised cirrhotic patients without acute-on-chronic liver failure. J Hepatol 2015;62:831–840. doi.org/10.1016/j.jhep.2020.11.019. [13] European Association for the Study of the Liver. Electronic address eee, clinical practice guideline panel C, panel m, representative EGB. EASL References clinical practice guidelines: drug-induced liver injury. J Hepatol Author names in bold designate shared co-first authorship. 2019;70:1222–1261. [14] Bjornsson ES, Hoofnagle JH. Categorization of drugs implicated in causing [1] Moreau R, Jalan R, Gines P, Pavesi M, Angeli P, Cordoba J, et al. Acute-on- liver injury: critical assessment based on published case reports. Hep- chronic liver failure is a distinct syndrome that develops in patients with atology 2016;63:590–603. acute decompensation of cirrhosis. Gastroenterology 2013;144. 1426- [15] Gray R. A class of K-sample tests for comparing the cumulative incidence 1437, 1437 e1421-1429. of a competing risk. Ann Stat 1988;16:1141–1154. [2] Trebicka J, Fernandez J, Papp M, Caraceni P, Laleman W, Gambino C, et al. [16] Trebicka J, Gu W, Ibanez-Samaniego L, Hernandez-Gea V, Pitarch C, The PREDICT study uncovers three clinical courses in acutely decom- Garcia E, et al. Rebleeding and mortality risk are increased by ACLF but pensated cirrhosis with distinct pathophysiology. J Hepatol reduced by pre-emptive TIPS. J Hepatol 2020;73(5):1082–1091. 2020;73(4):842–854. [17] Kumar R, Kerbert AJC, Sheikh MF, Roth N, Calvao JAF, Mesquita MD, et al. [3] Arroyo V, Moreau R, Kamath PS, Jalan R, Gines P, Nevens F, et al. Acute-on- Determinants of mortality in patients with cirrhosis and uncontrolled chronic liver failure in cirrhosis. Nat Rev Dis Prim 2016;2:16041. variceal bleeding. J Hepatol 2020;74(1):66–79. [4] Arroyo V, Moreau R, Jalan R. Acute-on-Chronic liver failure. New Engl J [18] Shi Y, Yang Y, Hu Y, Wu W, Yang Q, Zheng M, et al. Acute-on-chronic liver Med 2020;382:2137–2145. failure precipitated by hepatic injury is distinct from that precipitated by [5] Claria J, Stauber RE, Coenraad MJ, Moreau R, Jalan R, Pavesi M, et al. extrahepatic insults. Hepatology 2015;62:232–242. Systemic inflammation in decompensated cirrhosis: characterization and [19] Piano S, Singh V, Caraceni P, Maiwall R, Alessandria C, Fernandez J, et al. role in acute-on-chronic liver failure. Hepatology 2016;64:1249–1264. Epidemiology and effects of bacterial infections in patients with cirrhosis [6] Trebicka J, Amoros A, Pitarch C, Titos E, Alcaraz-Quiles J, Schierwagen R, worldwide. Gastroenterology 2019;156. 1368-1380 e1310. et al. Addressing profiles of systemic inflammation across the different [20] Fernandez J, Prado V, Trebicka J, Amoros A, Gustot T, Wiest R, et al. clinical phenotypes of acutely decompensated cirrhosis. Front Immunol Multidrug-resistant bacterial infections in patients with decompensated 2019;10:476. cirrhosis and with acute-on-chronic liver failure in Europe. J Hepatol 2019 [7] European Association for the Study of the Liver, Collaborators, Angeli P, Mar;70(3):398–411. Bernardi M, Villanueva C, Francoz C, et al. EASL Clinical Practice Guide- [21] Fernández J, Acevedo J, Castro M, Garcia O, Rodríguez de Lope C, Roca D, lines for the management of patients with decompensated cirrhosis. et al. Prevalence and risk factors of infections by multiresistant bacteria in J Hepatol 2018;69:406–460. cirrhosis: a prospective study. Hepatology 2012;55:1551–1561. [8] Jalan R, Saliba F, Pavesi M, Amoros A, Moreau R, Gines P, et al. Develop- [22] Bassetti M, Merelli M, Temperoni C, Astilean A. New antibiotics for bad ment and validation of a prognostic score to predict mortality in patients bugs: where are we? Ann Clin Microbiol Antimicrob 2013;12. with acute-on-chronic liver failure. J Hepatol 2014;61:1038–1047. [23] Fernandez J, Tandon P, Mensa J, Garcia-Tsao G. Antibiotic prophylaxis in [9] Fernandez J, Acevedo J, Wiest R, Gustot T, Amoros A, Deulofeu C, et al. cirrhosis: good and bad. Hepatology 2016;63:2019–2031. Bacterial and fungal infections in acute-on-chronic liver failure: preva- [24] Wieser A, Li H, Zhang J, Liss I, Markwardt D, Hornung R, et al. Evaluating lence, characteristics and impact on prognosis. Gut 2018;67:1870–1880. the best empirical antibiotic therapy in patients with acute-on-chronic [10] Crabb DW, Bataller R, Chalasani NP, Kamath PS, Lucey M, Mathurin P, et al. liver failure and spontaneous bacterial peritonitis. Dig Liver Dis : official Standard definitions and common data elements for clinical trials in pa- J Ital Soc Gastroenterol Ital Assoc Study Liver 2019;51:1300–1307. tients with alcoholic hepatitis: recommendation from the NIAAA alco- holic hepatitis consortia. Gastroenterology 2016;150:785–790. 1108 Journal of Hepatology 2021 vol. 74 j 1097–1108