Correction

Abstract

Erratum 301 Acute Liver Failure in Pregnancy: How Much Is Pregnancy‐Related? Ariel Aday, Lisa Casey, Jody A. Rule and William M. Lee, University of Texas Southwestern Medical Center Background: Acute liver failure (ALF) in pregnancy is associated with significant maternal and fetal morbidity and mortality. The Acute Liver Failure Study Group (ALFSG) registry includes patients with ALF and acute liver injury (ALI, severe injury without encephalopathy). We sought to review presentations and outcomes of all pregnant women enrolled in the ALFSG registry that met standard entry criteria for ALF/ALI. Methods: Between January 1998 and November 2017, 3155 subjects were registered with ALF or ALI and 71 of these subjects (2.3%) were pregnant or postpartum at the time of enrollment. Subjects were reviewed for demographics, cause of liver disease, lab values, and outcomes at 21 days: death, liver transplant (LT) or spontaneous survival (SS). Testing for acetaminophen (APAP) protein adducts was performed on selected subjects to confirm a suspected diagnosis of APAP. Results: The median age of enrollees was 29 with 64.8% Caucasian, 23.9% African Americans, 4.2% Asian, 4.2% other and 2.8% Hispanics. Specific pregnancy‐associated acute liver diseases (PAALD; hemolysis, elevated liver enzymes, low platelets [HELLP], pre‐eclampsia, or acute fatty liver of pregnancy, [AFLP]) represented nearly 45% of cases—all PAALD patients were enrolled post‐partum indicating that obstetrical services recognize PAALD and deliver patients appropriately. APAP toxicity was observed in 30% of cases with other etiologies such as autoimmune hepatitis, drug‐induced liver injury, HSV, cancer (adenocarcinoma and lymphoma), thyroid disease, comprising the remainder; no hepatitis A, B, C, or E cases were identified. The median gestational age of presentation by etiology was PAALD 37 weeks, APAP 30 weeks, and other 30 weeks. Overall 21‐day status was 67.6 % SS, 16.9% LT, and 12.7% died. When outcomes per etiology were considered separately, PAALD: 72% SS, 13% LT, 15% died; APAP: 90% SS, 5% LT, 5% died; Other: 50% SS, 33% LT, Died 17%. Fetal outcomes also differed greatly between groups with survival rates of PAALD 96%, APAP 43%, and other 55%. Conclusion: Overall, in this population of ALF associated with pregnancy, more than half the cases were non‐PAALD, with APAP toxicity being the most frequent other cause. HSV hepatitis and other causes are associated with worse outcomes but are less common. PAALD, occurring as it does near term, is associated with very good fetal survival and fair maternal overall mortality. Disclosures: Lisa Casey – NIAAA/NIH: Grant/Research Support William M. Lee – Merck: Grant/Research Support; Conatus: Grant/Research Support; Intercept: Grant/Research Support; Synlogic: Grant/Research Support; Novo Nordisk: Grant/Research Support; Gilead: Grant/Research Support; Tobira: Grant/Research Support; Bristol Myers Squibb: Grant/Research Support; Novartis: Consulting; Sanofi: Consulting; Retros: Consulting The following people have nothing to disclose: Ariel Aday, Jody A. Rule 302 An Intrabody Against PHD2 Protects Against Acute Liver Injury in Mice Liangzhong Zhao1, Ziyu Liu1, Guoxiu Zhao1, Ying Zhang1, Haipeng Miao1, Ying Xue1, Hongli Huang1, Fang Yang1, Cynthia Ju2and Guiying Li1, (1)Key Laboratory for Molecular Enzymology and Engineering of the Ministry of Education, School of Life Sciences, Jilin University, (2)Department of Anesthesiology, Mcgovern Medical School, University of Texas Health Science Center Background: Hypoxia‐inducible factor (HIF)‐1α is a crucial transcription factor that regulates the expression of target genes involved in angiogenesis and liver repair after acute injury. Prolyl hydroxylase 2 (PHD2) is one of 2‐oxoglutarate dependent oxygenases, which dominantly hydroxylates two highly conserved proline residues of HIF‐1α to promote the degradation of HIF‐1α. Inhibition of PHD2 activity can promote the accumulation of HIF‐1α. This study was designed to construct an intrabody against PHD2 that can inhibit PHD2 activity and to investigate the effect of such intrabody on acute liver injury in mice. Methods: A human single‐chain fragment of antibody variable region (scFv) against PHD2 (INP) was isolated by phage display technique. In order to block PHD2 activity in cells, an expression plasmid pER‐INP harboring an endoplasmic reticulum (ER)‐retained scFv gene against human PHD2 (anti‐PHD2 intrabody, ER‐INP) was constructed. The plasmids were transfected into HEK293, HepG2, and RAW264.7 cells to detect the inhibitory role of ER‐INP in PHD2 activity. To investigate the role of intrabody in acute liver injury, BALB/c mice were injected with pER‐INP for 48 h and then treated with 250 mg/kg of acetaminophen (APAP). Blood samples were collected at 8, 24, 48 and 72h post‐APAP challenge to determine serum ALT and AST levels. Angiogenesis was measured by human umbilical vein endothelial cells (HUVECs) migration assay and HUVECs tube formation assay. mRNA was determined by qPCR and protein levels was measured by western blot and immunohistochemistry (IHC). Results: ER‐INP was expressed efficiently and localized in the endoplasmic reticulum of cells transfected with pER‐INP. Co‐immunoprecipitation assay showed that ER‐INP could recognize and bind to PHD2 in HEK293 cells. Compared with the control, the expression of ER‐INP significantly inhibited the hydroxylation of HIF‐1α and increased the level of HIF‐1α. After APAP challenge, mice injected with pER‐INP showed significantly lower serum ALT and AST levels, and less liver tissue damage than the control vector‐treated mice (p<0.01). We next examined the expression of various target genes of HIF‐1α from liver tissues by qPCR analysis. Compared with the control mice, ER‐INP promotes distinctly expression of a range of angiogenic factors including VEGF, ANGPTL‐2 and MMP‐2. IHC assay revealed more VEGF and CD31 positive cells were observed in liver tissues from ER‐INP‐treated mice. Co‐culture of cells transfected with pER‐INP and HUVECs induced HUVECs migration and tube formation. Conclusion: Anti‐PHD2 intrabody protects against acute liver injury in mice by blocking PHD2 activity and enhances HIF‐1α mediated angiogenesis. It suggests that anti‐PHD2 intrabody may provide a new potential strategy for the treatment of drug‐induced acute liver injury. The following people have nothing to disclose: Liangzhong Zhao, Liangzhong Zhao, Ziyu Liu, Guoxiu Zhao, Ying Zhang, Haipeng Miao, Ying Xue, Hongli Huang, Fang Yang, Guiying Li Disclosure information not available at the time of publication: Cynthia Ju 303 Circulating Peroxiredoxin‐1 As a Novel DAMP Aggravates Acute Liver Injury Via Promoting Inflammation in Mice Ying He1, Huixiang Yang1, Zhangzhe Peng2, Shenglan Li1and Lijian Tao2, (1)Department of Gastroenterology, Xiangya Hospital, Central South University, (2)Department of Nephropathy, Xiangya Hospital, Central South University Background: Damage‐associated molecular patterns (DAMPs) are initiators of sterile inflammation, which is a key contributor to acute liver injury (ALI). As a novel identified DAMP, although extracellular Peroxiredoxin‐1 (Prdx1) possesses pro‐inflammatory capability, intracellular Prdx1 has anti‐oxidative activity; therefore, the crucial role of Prdx1 in ALI remains unclear. Here, we aimed to explore the role of Prdx1 in ALI. Methods: (1) To detect the circulating Prdx1 level in ALI, we collected serum samples from normal subjects and patients with ALI. (2) To explore the role of Prdx1 in ALI in vivo, we established the ALI models induced by carbon tetrachloride (CCl4) or acetamidophenol (APAP) in wild‐type and Prdx1‐knockout mice. (3) Recombinant Prdx1 (rPrdx1) was utilized to further investigate its role in ALI induced by APAP in vivo and effect on inflammatory cytokine expression in primary peritoneal macrophages in vitro, respectively. Results: (1) We found a remarkable elevation of circulating Prdx1 in patients with ALI, which is positively correlated with the elevation of serum alanine aminotransferase. (2) Circulating Prdx1 is also significantly elevated in carbon tetrachloride (CCl4)‐ or acetamidophenol (APAP)‐induced ALI mice model; and Prdx1 deficiency protects against liver injury by reducing inflammatory cytokines (interleukin (IL)‐1β, IL‐6, and tumor necrosis factor alpha (TNF‐α)) expression in ALI models. (3) Intravenous injection with recombinant Prdx1 (rPrdx1) promotes the mortality, liver injury and inflammatory cytokines (IL‐1β, IL‐6, and TNF‐α) production in Prdx1‐deficiency mice in APAP‐induced ALI model in vivo; and stimulation with rPrdx1 induces the production of inflammatory cytokines (IL‐1β, IL‐6, and TNF‐α) through the activation of nuclear factor of kappa B (NF‐κB) and NOD‐like receptor family pyrin domain containing 3 (NLRP3) inflammasome signaling pathways in primary peritoneal macrophages in vitro. Conclusion: (1) We demonstrated that circulating Prdx1, as a novel DAMP, plays a detrimental pro‐inflammatory role in CCl4‐ or APAP‐induced ALI through the activation of NF‐κB and NLRP3 inflammasome signaling pathways. (2) Circulating Prdx1 may act as a promising diagnostic biomarker and a therapeutic target for ALI. Disclosures: The following people have nothing to disclose: Ying He Disclosure information not available at the time of publication: Huixiang Yang, Zhangzhe Peng, Shenglan Li, Lijian Tao 304 Etiology of Liver Injury Is More Predictive of Survival Than Race in Patients with Acute Liver Failure (ALF). Bilal Hameed, University of California, San Francisco, Elizabeth X. Zheng, Gastroenterology, University of California, San Francisco, Jyoti Arora, Public Health Sciences, Medical University of South Carolina, R. Todd Stravitz, Gastroenterology, Hepatology, and Nutrition, Virginia Commonwealth University, William M. Lee, University of Texas Southwestern Medical Center and Acute Liver Failure Study Group Background: Racial differences have been observed in the outcomes of chronic liver disease but its independence from other predictors has not been certain. These findings and their impact on outcome have not been established in acute liver failure (ALF). Methods: Using a prospectively maintained ALF Study Group database, we conducted a retrospective cohort study, evaluating differences in survival and transplant outcomes among different racial groups who were admitted with ALF. Baseline characteristics and lab data were recorded at the time of enrollment and outcomes of liver transplant and death were observed for 21 days after enrollment. Results: Among 2051 subjects (79.8% white, 15.7% African Americans {AA} and 4.5% Asian) enrolled between January 1998 to November 2017, there were no differences between the racial groups in gender ratio, years of education, or hospital length of stay. Asians were older in age compared to whites and AAs (Median (IQR) Caucasians: 40(29 – 52), AAs: 40(30 – 53), Asians: 48.5(33 – 59)). The etiology of ALF differed between races with whites presenting more frequently with favorable etiologies (grouped together: acetaminophen (APAP), ischemic liver injury, pregnancy‐induced ALF and hepatitis A) when compared to AAs or Asians (61.1% vs 34.9% vs. 30.4%; P <.001). The difference in etiology was mainly accounted for by APAP toxicity which was more frequent in whites (51%) than AA (25.5%) or Asians (22.8%), P <0.001. By 21 days, there were a total of 397 patients transplanted (19.3%) and 649 deaths (31.6%). More AAs and Asians received liver transplants compared to whites (23.3% vs 33.7% vs 18.1%; P < 0.001) but death rates were highest in AAs and Asians compared to whites (39.4% vs 31.5% vs 30.19% P = 0.005). After adjusting for age and etiology (favorable vs unfavorable), there were no significant differences in overall survival among the three races and subjects with unfavorable etiologies had significantly lower rates of transplant free survival (HR 2.93, 95% CI 2.56‐3.35, P< 0.001) and lower overall survival (HR 1.55, 95% CI 1.32‐1.82, P < 0.001). Conclusion: Once adjustment for age and favorable etiology is considered, race was no longer associated with a lower likelihood of transplant free survival or overall survival. Therefore, differences in etiology of ALF between racial groups is much more important in predicting survival than race. Disclosures: Bilal Hameed – Gilead: Grant/Research Support; Conatus: Grant/Research Support; Intercept: Grant/Research Support; Genfit: Grant/Research Support; Dova: Advisory Committee or Review Panel William M. Lee – Merck: Grant/Research Support; Conatus: Grant/Research Support; Intercept: Grant/Research Support; Synlogic: Grant/Research Support; Novo Nordisk: Grant/Research Support; Gilead: Grant/Research Support; Tobira: Grant/Research Support; Bristol Myers Squibb: Grant/Research Support; Novartis: Consulting; Sanofi: Consulting; Retros: Consulting The following people have nothing to disclose: Elizabeth X. Zheng Disclosure information not available at the time of publication: Jyoti Arora, R. Todd Stravitz 305 Automated Liver Lesion Segmentation with Convolutional Neural Networks Sean Sall, Jesse Lieman‐Sifry, Felix Lau and Daniel Golden, Machine Learning, Arterys Background: MRI is commonly used to assess patients with known or suspected pathologies of the liver. LI‐RADS, which is the current clinical standard of care for liver lesions, relies on measurements of the longest linear diameter (LLD) to quantify lesion size and measure growth over time 1. However, LLD measurements may provide a misleading quantification of the size of irregularly shaped lesions 2. Volumetric measurements give a more complete impression of the overall lesion size as illustrated in Figure 1, which shows axial slices of lesions with similar LLD (all within 5% of each other) but with very different volumes (up to 2x difference). Convolutional neural networks have been successfully used to segment liver lesions in abdominal CT, but very few studies have explored their application to segmenting liver lesions in abdominal MRI3. In this work, we propose a deep‐learning‐based method for volumetric segmentation of lesions in abdominal MRI. We show that the accuracy of LLD calculated from the segmentations is viable for clinical usage under current guidelines. Methods: We utilize 1,312 annotated lesions from 607 unique abdominal MRI studies and 393 unique patients. DICOM images and lesion locations were collected as part of standard clinical care. Identified lesions were segmented manually on multiphasic contrast‐enhanced series by annotators trained in liver lesion segmentation. The median lesion diameter was 15.8 mm, while the median lesion volume was 1.469 mL. We designate an 80%‐10%‐10% split between the training, validation, and testing sets. We report all metrics on the test set. We use a fully‐convolutional neural network that operates on 3D patches, centered on the lesion of interest. The network is a variant of the ENet segmentation architecture4. Results: Our automated contouring method achieved a median volume error of 0.277 mL and a median LLD error of 2.01 mm. The median LLD error is significantly below the LI‐RADS “threshold growth” threshold (5.0mm). Figure 2presents boxplots for the relative error of the two metrics of interest, binned by lesion size, and demonstrates that our model performance is similar across lesions of different sizes. Figure 3shows predicted contours for the axial slices encompassing a single lesion.Conclusion: Our automated lesion segmentation method yields a low median LLD error, demonstrating that our estimates may be used as part of a semi‐automated clinical workflow in which the clinician may review and modify the segmentations. Additionally, we demonstrate that automated volumetric estimates are feasible from MRI and, with further validation, may provide a viable method of tracking tumor volume over time. Disclosures: Sean Sall – Arterys: Employment Disclosure information not available at the time of publication: Jesse Lieman‐Sifry, Felix Lau, Daniel Golden 306 Characterization of a Novel Mouse Model of Type a Hepatic Encephalopathy Elaina Williams1,2,3, Stephanie Grant1,3, Gabriel Frampton1, Anca Petrescu1,3, Matthew McMillin4and Sharon DeMorrow1,4, (1)Texas A&M College of Medicine, (2)University of Mary Hardin‐Baylor, (3)Central Texas Veterans Healthcare System, (4)Central Texas Veterans Health Care System Background: Type A hepatic encephalopathy (HE) describes the neurological complications caused by acute liver failure. Research into this type of HE has been hampered as existing models cause severe and rapid HE, with a very limited therapeutic window and high mortality, that do not reflect all features observed in patients. The aim of this study was to establish and characterize an alternative mouse model of Type A HE that has a more protracted timeline of pathology. Methods: Male C57Bl/6 mice were fed a standard rodent chow that was enriched with 0.1% 3,5‐diethoxycarbonyl‐1,4‐dihydrocollidine and 10% ammonium acetate (DDC+NH4). Neurobehavioral indices and neuromuscular deficits were assessed by open field test, rotarod, grip strength test and gait analysis. Serum, liver and brain tissue were collected after 13 days of DDC+NH4 feeding. Liver damage was assessed by serum chemistry and H&E staining. Serum and cortical ammonia and total bile acid content were assessed by using colorimetric assays. Cerebral edema was assessed using the wet weight/dry weight method. Microglia activation was assessed by Iba1 immunofluorescence. The expression of proinflammatory cytokines were assessed by qPCR and EIA. Results: DDC+NH4 feeding caused significant neurological and neuromuscular deficits, with impaired locomotor activity and motor co‐ordination, reduced grip strength, and the presence of ataxia commencing after 5‐7 days of feeding. After euthanasia, the liver damage observed was similar to a cholestatic hepatitis phenotype of drug‐induced liver injury in humans in that AST, ALT, alkaline phosphatase and total bilirubin were elevated. Histologically, there was evidence of intrahepatic cholestasis, drug‐induced bile duct blockage, ductular reaction and immune cell infiltration, with low incidence of hepatocyte necrosis and little to no discernible fibrosis. Serum and cortical ammonia and total bile acid levels were elevated, and there was increased brain water content, indicating cerebral edema in DDC+NH4‐fed mice. Furthermore, there was evidence of microglia activation and increased proinflammatory cytokine expression in the cortex of DDC+NH4 group compared to control diet‐fed mice. Conclusion: DDC+NH4 feeding generated key features of Type A HE with a longer timeline of pathology than that observed in other mouse models of this form of HE. With further characterization, this model may prove an effective and preferable model for pre‐clinical studies to identify the pathogenesis or possible treatment options for Type A HE. Disclosures: The following people have nothing to disclose: Elaina Williams, Stephanie Grant, Matthew McMillin, Sharon DeMorrow Disclosure information not available at the time of publication: Gabriel Frampton, Anca Petrescu 307 Characterization of Human Umbilical Vein Endothelial Cells Endothelializing the Intact Decellularized Rat Liver Scaffolds Jun Li1, Qian Zhou1, Jing Jiang1, Jiaojiao Xin1, Jun Li2, Longyan Jiang1, Dongyan Shi1, Beibei Guo1and Xingping Zhou1, (1)State Key Laboratory for Diagnosis and Treatment of Infectious Diseases, Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, the First Affiliated Hospital, Zhejiang University School of Medicine, (2)Department of Pathology, the First Affiliated Hospital, Zhejiang University School of Medicine Background: Creation of functional blood vessels is a major bottleneck for construction of liver organoids, and endothelial cells play a critical role in vascularization. Here we focus on establishing an integrated endothelialized scaffold, and characterizing the detail biological process of endothelialization. Methods: We performed endothelialization of intact rat decellularized liver scaffolds (DLSs) via multistep infusions of 1.5×107 human umbilical vein endothelial cells (HUVECs) through the portal vein, followed by 3 weeks’ continuous perfusion with self‐developed bioreactor. DLS samples and endothelialized scaffolds at conditioned time points were identified and evaluated from histology and proteomics aspects. Results: On the dynamic condition, the reseeded HUVECs survived to more than 3 weeks. During the period, these cells gradually adhered to vessel walls and the matrix around vessels was increasingly dense. To elucidate the biological process of endothelialization, mass spectrometry analysis was performed. Clustering analysis of the 165 human‐derived proteins that up‐regulated on day 1 revealed the processes of cell adhesion, hydroxylation, mRNA processing, transcription regulation, protein transport and secretion, which indicated that the reseeded HUVECs started to synthesize proteins 1 day post‐reseeding. Besides, 388 human‐derived matrisome proteins were identified from endothelialized scaffolds 3 w post‐reseeding. Compare to the first day, 11 proteins, including annexin A1, A2, A5, and A7, protein glutamine gamma‐glutamyltransferase 2, collagen alpha‐1 (VIII), collagen alpha‐3 (VI), collagen alpha‐5 (VI), collagen alpha‐2 (V) , fibrillin‐2 and nidogen‐1 were identified to be differentially expressed with fold‐change>2.0, p‐value<0.5. Functional classification of these proteins revealed the main biological processes of cell adhesion and proliferation, anticoagulation, basement membrane organization and collagen fibril organization, et al., which played important roles in endothelialization. Conclusion: Our self‐developed bioreactor was successfully developed for the long‐term survival of the reseeded HUVECs. Exposure of HUVECs to continuous fluid flow stimulated the alignment of the cells and the formation of an integrated endothelium. 11 significantly differentially expressed proteins were identified to play important roles in endothelialization, which were promising candidate proteins for further construction of mature vessels. Disclosures: The following people have nothing to disclose: Jun Li, Qian Zhou Disclosure information not available at the time of publication: Jing Jiang, Jiaojiao Xin, Jun Li, Longyan Jiang, Dongyan Shi, Beibei Guo, Xingping Zhou 308 Prognostic Markers in Pediatric Acute Liver Failure with Advanced Hepatic Encephalopathy Seema Alam, Bikrant Bihari Lal, Vikrant Sood and Rajeev Khanna, Pediatric Hepatology, Institute of Liver and Biliary Sciences Background: Pediatric acute liver failure (PALF) with advanced hepatic encephalopathy (HE) has been seen as a group with relatively poor prognosis. Early detection of patients with poor prognosis can improve the outcome of the PALF in this subgroup. This study was undertaken to identify the prognostic markers in PALF with advanced HE. Methods: All cases of PALF presenting to this centre during January 2012 to January 2018, with HE grade III and IV were included in the study. PALF was defined as per the PALF study group. HE was classified as per the age appropriate West Haven Classification. Results: In the study period 156 children with PALF were admitted to this centre, of which 78 had HE grade III and IV. Median age of the children was 9.05 years (IQR 4.2 – 13). Forty –four ( 56.4%) had a viral etiology. Of the 78 PALF with HE grade III & IV, 26 survived with native liver, 10 were transplanted and 42 died. Overall survival and survival with native liver were 46.15% and 33.33% respectively. On logistic regression analysis (LRA), poor prognosis (death or need for liver transplant) was associated with a higher Pediatric End Stage Liver Disease (PELD) score at 72 hours (Poor prognosis: 39.72 ± 10.5 vs Survivors: 17.42 ± 9.2, OR1.24; 95%CI: 1.08 to 1.44; p‐value 0.003), and longer jaundice to encephalopathy interval (JEI) (Poor prognosis: 20 ± 14.5 vs Survivors: 2.88 ± 1.81 days, OR 1.43; 95%CI:1.02 to 2.01; p‐value 0.038). Viral etiology was protective (OR 0.055; 95%CI: 0.003 to 0.857; p‐value 0.038). The model predicted 86% of the variation in the outcome (R square 0.858) and correctly classified it in 93.2%. The ROC analysis revealed PELD score at 72 hours ≥ 28 (Sensititvity 93.9% and Specificity 84.6%), and JEI of > 6 days (Sensititvity 75.8% and Specificity 100%), to be the best discriminators with an AUROC of 94% and 84% respectively (figure 1). Conclusion: Outcome in a case of PALF with advanced HE can be predicted correctly in 93.2% cases by a PELD score at 72 hours of ≥28, JEI of >6 days and viral etiology. Disclosure information not available at the time of publication: Seema Alam, Seema Alam, Bikrant Bihari Lal, Vikrant Sood, Rajeev Khanna 309 Artificial Neural Network‐Based Risk Model to Predict 28‐ and 90‐ Day Mortality in Acute on‐Chronic Hepatitis B Liver Failure Hou YIXIn1, Qianqian Zhang2, Fangyuan Gao3, Dewen Mao4, Jun Li5, Zuojiong Gong6, Xinla Luo7, Guoliang Chen8, Yong Li9, Zhiyun Yang10, Kewei Sun2and Xianbo Wang3, (1)Capital Medical University, Beijing, China., (2)Department of Hepatology, The First Hospital Affiliated to Hunan University of Chinese Medicine, Changsha, China., (3)Center of Integrative Medicine, Beijing Ditan Hospital, Capital Medical University, (4)Department of Hepatology, The First Affiliated Hospital of Guangxi University of Chinese Medicine, Guangxi, China., (5)Center of Integrative Medicine,, 302 Military Hospital of China, Beijing, China., (6)Department of Infectious Diseases, Renmin Hospital of Wuhan University, Wuhan, China., (7)Department of Hepatology, Hubei Provincial Hospital of TCM, Wuhuan, China., (8)Department of Hepatology,, Xiamen Hospital of TCM, Xiamen, China., (9)Department of Hepatology, Shandong Provincial Hospital of TCM, Shandong, China., (10)Center of Integrative Medicine,, Beijing Ditan Hospital Capital Medical University, Beijing, China. Background: Hepatitis B virus (HBV) infects approximate 240 million people worldwide and 93 million in China . HBV has become one of the leading causes for acute‐on‐chronic liver failure (ACLF), mainly characterized as a rapid deterioration of liver function with a high short‐term mortality .Liver transplantation is currently the main effective therapeutic option for HBV‐ACLF. However, due to shortage of liver donors and also some socioeconomic problems, liver transplantation is limited in clinics . To decrease mortality of HBV‐ACLF, it is vital to accurately identify patients with poor prognosis so as to take treatment as early, including organ allocation from limited liver donors. At present, there are actually no ideal models to predict short‐term outcomes of patients with HBV‐ACLF yet. Model of end‐stage liver disease (MELD) and the modified type‐‐MELD integrating sodium (MELD‐Na)‐‐have been gradually proposed to predict the prognosis of patients with ACLF, with moderate accuracy . However, due to limited predictive accuracy, these scoring systems are still unsatisfactory . More accurate prognostic models are urgent. It is important to combine more parameters to construct a model to predict short‐term mortality of ACLF. The artificial neural networks (ANN), structurally and functionally mimic biological neural systems, have been widely using to manage nonlinear complex biological systems. It has been revealed that the ANN models are more accurate than multiple logistic regression and multiple linear discriminant analysis models . In this study, we established an ANN‐based model to predict 28‐ and 90‐ day mortality of HBV‐ACLF. Methods: Sixty hundred and eight‐four cases of consecutive patients with HBV‐ACLF treated at multiple medical centers from January 2008 to May 2016 were retrospectively reviewed. Four hundred and twenty‐three cases were used for training an ANN model to predict 28‐ and 90‐day mortality, and the remaining 261 for validating the established model. Predictors associated with mortality were determined by univariate analysis and were then included in a ANN model for predicting prognosis. The receiver operating characteristic curve analysis was used to evaluate the predictive performance of the ANN model in comparison with various current prognostic models. Results: Seven independent risk factors, including hepatic encephalopathy, serum sodium, prothrombin activity, γ‐glutamyltransferase, hepatitis B e antigen, alkaline phosphatase and total bilirubin, were eventually used to establish an ANN model. This model showed more predictive values when compared to CLIF‐ACLF, Model for End‐stage Liver Disease (MELD), and MELD‐sodium. Conclusion: The established ANN model could accurately predict the short‐term mortality risk of patients with HBV‐ ACLF than the other current prediction models. Disclosures: The following people have nothing to disclose: Hou YIXIn, Fangyuan Gao, Jun Li, Zhiyun Yang, Kewei Sun, Xianbo Wang Disclosure information not available at the time of publication: Qianqian Zhang, Dewen Mao, Zuojiong Gong, Xinla Luo, Guoliang Chen, Yong Li 310 Hepatotoxicity from Iron Overdose: Report of 6 Cases in Adolescents and Adults Vishnu Vardhan Reddy1, Tarun Joseph1and Harshad Devarbhavi2, (1)Department of Gastroenterology, St. John’s Medical College Hospital, Bangalore, India, (2)Department of Gastroenterology, St John’s Medical College Hospital Background: Iron overdose is a rare but important cause of predictable hepatotoxicity, reported mainly in children. With the widespread availability of iron in various formulations including its presence in multivitamins, the prospect of overdose is high. Iron overdose pathogenic mechanisms include direct caustic effect on the GI mucosa; systemic complications including hepatotoxicity is from circulator