An artificial intelligence-generated model predicts 90-day survival in alcohol-associated hepatitis: A global cohort study.

BACKGROUND AND AIMS: Alcohol-associated hepatitis (AH) poses significant short-term mortality. Existing prognostic models lack precision for 90-day mortality. Utilizing artificial intelligence in a global cohort, we sought to derive and validate an enhanced prognostic model. APPROACH AND RESULTS: The Global AlcHep initiative, a retrospective study across 23 centers in 12 countries, enrolled patients with AH per National Institute for Alcohol Abuse and Alcoholism criteria. Centers were partitioned into derivation (11 centers, 860 patients) and validation cohorts (12 centers, 859 patients). Focusing on 30 and 90-day postadmission mortality, 3 artificial intelligence algorithms (Random Forest, Gradient Boosting Machines, and eXtreme Gradient Boosting) informed an ensemble model, subsequently refined through Bayesian updating, integrating the derivation cohort's average 90-day mortality with each center's approximate mortality rate to produce posttest probabilities. The ALCoholic Hepatitis Artificial INtelligence Ensemble score integrated age, gender, cirrhosis, and 9 laboratory values, with center-specific mortality rates. Mortality was 18.7% (30 d) and 27.9% (90 d) in the derivation cohort versus 21.7% and 32.5% in the validation cohort. Validation cohort 30 and 90-day AUCs were 0.811 (0.779-0.844) and 0.799 (0.769-0.830), significantly surpassing legacy models like Maddrey's Discriminant Function, Model for End-Stage Liver Disease variations, age-serum bilirubin-international normalized ratio-serum Creatinine score, Glasgow, and modified Glasgow Scores ( p < 0.001). ALCoholic Hepatitis Artificial INtelligence Ensemble score also showcased superior calibration against MELD and its variants. Steroid use improved 30-day survival for those with an ALCoholic Hepatitis Artificial INtelligence Ensemble score > 0.20 in both derivation and validation cohorts. CONCLUSIONS: Harnessing artificial intelligence within a global consortium, we pioneered a scoring system excelling over traditional models for 30 and 90-day AH mortality predictions. Beneficial for clinical trials, steroid therapy, and transplant indications, it's accessible at: https://aihepatology.shinyapps.io/ALCHAIN/ .

Using Bland-Altman Analysis to Identify Appropriate Clonogenic Assay Colony Counting Techniques.

OBJECTIVE: Determine the interchangeability of various methods utilized for counting colonies in clonogenic assays. METHODS: Clonogenic assays of 2 head and neck cancer cell lines were counted through 4 different counting modalities: Manual counting pen, via microscope, 1 publicly available automated algorithm, and a semiautomated algorithm presented by the authors. Each method counted individual wells (N = 24). Pen and microscopic counts were performed by 2 observers. Parameters included both low-growth (<150 colonies/well) and high-growth (>150 colonies/well) cell lines. Correlational and Bland-Altman analyses were performed using SPSS software. RESULTS: Interobserver manual pen count correlation R2 value in both growth conditions was 0.902; controlling for only low-growth conditions decreased R2 to 0.660. Correlation of microscopic versus pen counts R2 values for observers 1 and 2 were 0.955 and 0.775, respectively. Comparing techniques, Bland-Altman revealed potential bias with respect to the magnitude of measurement (P < .001) for both observers. Correlation of microscopic counts for both interobserver (R2 = 0.902) and intraobserver (R2 = 0.916) were analyzed. Bland-Altman revealed no bias (P = .489). Automated versus microscopic counts revealed no bias between methodologies (P = .787) and a lower correlation coefficient (R2 = 0.384). Semiautomated versus microscopic counts revealed no bias with respect to magnitude of measurement for either observer (P = .327, .229); Pearson correlation was 0.985 (R2 = 0.970) and 0.965 (R2 = 0.931) for observer 1 and 2. Semiautomated versus manual pen colony counts revealed a significant bias with respect to magnitude of measurement (P < .001). CONCLUSION: Counting with a manual pen demonstrated significant bias when compared to microscopic and semiautomated colony counts; 2 methods were deemed to be interchangeable. Thus, training algorithms based on manual counts may introduce this bias as well. Algorithms trained to select colonies based on size (pixels2) and shape (circularity) should be prioritized. Solely relying on Bland-Altman or correlational analyses when determining method interchangeability should be avoided; they rather should be used in conjunction.

Potential Roles of Activin in Head and Neck Squamous Cell Carcinoma Progression and Mortality.

BACKGROUND/AIM: Activin, a member of the TGF-ß super family of cytokines, is involved in head and neck squamous cell carcinoma (HNSCC). This study examined the constituents of the activin axis in order to further elucidate the role of activin A in HNSCC progression. MATERIALS AND METHODS: Immunohistochemistry (IHC), reverse transcription polymerase chain reaction (RT-PCR), MTT, and matrigel invasion assays, in addition to analysis of the tumor cancer genome atlas (TCGA), were employed. RESULTS: IHC in HNSCC and oral leukoplakia (OPL) lesions demonstrated increased expression of the inhibin subunit ßA (INHBA) (p<0.0001), as well as activin receptor type IB (ACVR1B) (p<0.0032) compared to normal mucosa. TCGA analysis revealed increased INHBA expression was associated with lymph node positive tumors (p=0.024), decreased overall survival (p=0.0167), and decreased promoter methylation (p<0.0001). Concomitant up-regulated expression of gene pathways strongly correlated with INHBA expression demonstrated further deleterious effects on survival (p<0.0148). CONCLUSION: Activin may be an important component of early carcinogenesis in OPL and HNSCC with unfavorable effects on clinical end-points such as survival.

MELD 3.0 adequately predicts mortality and renal replacement therapy requirements in patients with alcohol-associated hepatitis.

Background & Aims: Model for End-Stage Liver Disease (MELD) score better predicts mortality in alcohol-associated hepatitis (AH) but could underestimate severity in women and malnourished patients. Using a global cohort, we assessed the ability of the MELD 3.0 score to predict short-term mortality in AH. Methods: This was a retrospective cohort study of patients admitted to hospital with AH from 2009 to 2019. The main outcome was all-cause 30-day mortality. We compared the AUC using DeLong's method and also performed a time-dependent AUC with competing risks analysis. Results: A total of 2,124 patients were included from 28 centres from 10 countries on three continents (median age 47.2 ± 11.2 years, 29.9% women, 71.3% with underlying cirrhosis). The median MELD 3.0 score at admission was 25 (20-33), with an estimated survival of 73.7% at 30 days. The MELD 3.0 score had a better performance in predicting 30-day mortality (AUC:0.761, 95%CI:0.732-0.791) compared with MELD sodium (MELD-Na; AUC: 0.744, 95% CI: 0.713-0.775; p = 0.042) and Maddrey's discriminant function (mDF) (AUC: 0.724, 95% CI: 0.691-0.757; p = 0.013). However, MELD 3.0 did not perform better than traditional MELD (AUC: 0.753, 95% CI: 0.723-0.783; p = 0.300) and Age-Bilirubin-International Normalised Ratio-Creatinine (ABIC) (AUC:0.757, 95% CI: 0.727-0.788; p = 0.765). These results were consistent in competing-risk analysis, where MELD 3.0 (AUC: 0.757, 95% CI: 0.724-0.790) predicted better 30-day mortality compared with MELD-Na (AUC: 0.739, 95% CI: 0.708-0.770; p = 0.028) and mDF (AUC:0.717, 95% CI: 0.687-0.748; p = 0.042). The MELD 3.0 score was significantly better in predicting renal replacement therapy requirements during admission compared with the other scores (AUC: 0.844, 95% CI: 0.805-0.883). Conclusions: MELD 3.0 demonstrated better performance compared with MELD-Na and mDF in predicting 30-day and 90-day mortality, and was the best predictor of renal replacement therapy requirements during admission for AH. However, further prospective studies are needed to validate its extensive use in AH. Impact and implications: Severe AH has high short-term mortality. The establishment of treatments and liver transplantation depends on mortality prediction. We evaluated the performance of the new MELD 3.0 score to predict short-term mortality in AH in a large global cohort. MELD 3.0 performed better in predicting 30- and 90-day mortality compared with MELD-Na and mDF, but was similar to MELD and ABIC scores. MELD 3.0 was the best predictor of renal replacement therapy requirements. Thus, further prospective studies are needed to support the wide use of MELD 3.0 in AH.

Disproportionate increases in alcohol-associated hepatitis incidence in women and individuals of low socioeconomic status: A population-based study using the Rochester epidemiology project database.

BACKGROUND: Alcohol-associated hepatitis (AH) is among the deadliest liver diseases, but its incidence is poorly defined. The aim of our study was to define the incidence of AH meeting the National Institute on Alcohol Abuse and Alcoholism criteria and to identify risk factors for AH. METHODS: We conducted a retrospective cohort study using the Rochester epidemiology project database on adult patients hospitalized with AH between January 1, 2000 and December 31, 2018. Patients were screened using ICD-9 codes and then included if they met the National Institute on Alcohol Abuse and Alcoholism criteria on manual chart review. Baseline demographics, comorbidities, access to care, liver-related complications, and outcomes were obtained. The HOUsing-based index of SocioEconomic status index was used to measure socioeconomic status. Incidence rates were calculated in cases per 100,000 person-years of follow-up. RESULTS: Among 204 patients, the cumulative AH incidence was 6.8 per 100,000 person-years. Between 2000-2004 and 2015-2018, AH incidence among males increased from 8.4 to 14.7 per 100,000 py, whereas AH incidence among females increased by 7-fold from 0.8 to 5.9 per 100,000 py. Such increases among females were accompanied by increases in comorbid depression and anxiety. The proportion of patients with AH in the lower socioeconomic status quartiles increased from 62.9% between 2000 and 2004 to 73.3% between 2015 and 2019. CONCLUSIONS: The incidence of AH is increasing rapidly, especially among females and individuals of lower socioeconomic status. There are areas of unmet need in preventative measures and treatments for comorbid psychiatric disorders in patients at high risk of AH.

Machine Learning Techniques Differentiate Alcohol-Associated Hepatitis From Acute Cholangitis in Patients With Systemic Inflammation and Elevated Liver Enzymes.

OBJECTIVE: To develop machine learning algorithms (MLAs) that can differentiate patients with acute cholangitis (AC) and alcohol-associated hepatitis (AH) using simple laboratory variables. METHODS: A study was conducted of 459 adult patients admitted to Mayo Clinic, Rochester, with AH (n=265) or AC (n=194) from January 1, 2010, to December 31, 2019. Ten laboratory variables (white blood cell count, hemoglobin, mean corpuscular volume, platelet count, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, total bilirubin, direct bilirubin, albumin) were collected as input variables. Eight supervised MLAs (decision tree, naive Bayes, logistic regression, k-nearest neighbor, support vector machine, artificial neural networks, random forest, gradient boosting) were trained and tested for classification of AC vs AH. External validation was performed with patients with AC (n=213) and AH (n=92) from the MIMIC-III database. A feature selection strategy was used to choose the best 5-variable combination. There were 143 physicians who took an online quiz to distinguish AC from AH using the same 10 laboratory variables alone. RESULTS: The MLAs demonstrated excellent performances with accuracies up to 0.932 and area under the curve (AUC) up to 0.986. In external validation, the MLAs showed comparable accuracy up to 0.909 and AUC up to 0.970. Feature selection in terms of information-theoretic measures was effective, and the choice of the best 5-variable subset produced high performance with an AUC up to 0.994. Physicians did worse, with mean accuracy of 0.790. CONCLUSION: Using a few routine laboratory variables, MLAs can differentiate patients with AC and AH and may serve valuable adjunctive roles in cases of diagnostic uncertainty.

The Mortality Index for Alcohol-Associated Hepatitis: A Novel Prognostic Score.

OBJECTIVE: To develop a new scoring system that more accurately predicts 30-day mortality in patients with alcohol-associated hepatitis (AH). METHODS: A cohort of consecutive adults diagnosed with AH at a single academic center from January 1, 1998, to December 31, 2018, was identified for model derivation. Multivariate logistic regression was used to create a new scoring system to predict 30-day mortality. External validation of this score was performed on a multicenter retrospective cohort. RESULTS: In the derivation cohort of 266 patients, the 30-day mortality rate was 19.2%. The following variables were found to be significantly associated with mortality on multivariate analysis: age (P=.002), blood urea nitrogen (P=.003), albumin (P=.01), bilirubin (P=.02), and international normalized ratio (P=.001). A model incorporating these variables, entitled the Mortality Index for Alcohol-Associated Hepatitis (MIAAH), achieved a C statistic of 0.86. Comparison of the accuracy of the MIAAH to existing prognostic models, including the Model for End-Stage Liver Disease and Maddrey Discriminant Function, showed that the highest concordance was achieved by the MIAAH and that this difference was significant. In the validation cohort of 249 patients, the MIAAH C statistic decreased to 0.73 and was found to be significantly superior to the Maddrey Discriminant Function but not to the Model for End-Stage Liver Disease. CONCLUSION: The MIAAH competes with the current prognostication models and is at a minimum as accurate as these existing scores in identifying patients with AH at high risk of short-term mortality. Furthermore, the MIAAH demonstrates advantageous performance characteristics in its ability to increasingly accurately dichotomize patients into those at highest risk of death and those likely to survive.

Development of the AI-Cirrhosis-ECG Score: An Electrocardiogram-Based Deep Learning Model in Cirrhosis.

INTRODUCTION: Cirrhosis is associated with cardiac dysfunction and distinct electrocardiogram (ECG) abnormalities. This study aimed to develop a proof-of-concept deep learning-based artificial intelligence (AI) model that could detect cirrhosis-related signals on ECG and generate an AI-Cirrhosis-ECG (ACE) score that would correlate with disease severity. METHODS: A review of Mayo Clinic's electronic health records identified 5,212 patients with advanced cirrhosis ≥18 years who underwent liver transplantation at the 3 Mayo Clinic transplant centers between 1988 and 2019. The patients were matched by age and sex in a 1:4 ratio to controls without liver disease and then divided into training, validation, and test sets using a 70%-10%-20% split. The primary outcome was the performance of the model in distinguishing patients with cirrhosis from controls using their ECGs. In addition, the association between the ACE score and the severity of patients' liver disease was assessed. RESULTS: The model's area under the curve in the test set was 0.908 with 84.9% sensitivity and 83.2% specificity, and this performance remained consistent after additional matching for medical comorbidities. Significant elevations in the ACE scores were seen with increasing model for end-stage liver disease-sodium score. Longitudinal trends in the ACE scores before and after liver transplantation mirrored the progression and resolution of liver disease. DISCUSSION: The ACE score, a deep learning model, can accurately discriminate ECGs from patients with and without cirrhosis. This novel relationship between AI-enabled ECG analysis and cirrhosis holds promise as the basis for future low-cost tools and applications in the care of patients with liver disease.

An active surveillance program in oral preneoplasia and translational oncology benefit.

OBJECTIVE: We desired to establish an active surveillance clinic for head and neck cancer. In this review we examined. METHODS: We examined the natural history of human oral carcinogenesis, the types of preneoplastic lesions, and efforts at oral chemoprevention over the past decades for presentation here. RESULTS: We established a clinic and program for patients with oral premalignant lesions approximately over 15 years ago based on an unmetneed for this service. We have completed over 4000 outpatient visits for this cohort and have a place for referrals of difficult oral lesions. We have leveraged this population for multiple federally funded trials on oral cancer prevention as well as specimen banking. CONCLUSION: There is need for routine active surveillance for oral preneoplastic conditions in patients at high risk for conversion to cancer. There are no effective durable treatment or preventions for these individuals and we have attempted to fill this unmet need with our program.

Identification of optimal therapeutic window for steroid use in severe alcohol-associated hepatitis: A worldwide study.

BACKGROUND & AIMS: Corticosteroids are the only effective therapy for severe alcohol-associated hepatitis (AH), defined by a model for end-stage liver disease (MELD) score >20. However, there are patients who may be too sick to benefit from therapy. Herein, we aimed to identify the range of MELD scores within which steroids are effective for AH. METHODS: We performed a retrospective, international multicenter cohort study across 4 continents, including 3,380 adults with a clinical and/or histological diagnosis of AH. The main outcome was mortality at 30 days. We used a discrete-time survival analysis model, and MELD cut-offs were established using the transform-the-endpoints method. RESULTS: In our cohort, median age was 49 (40-56) years, 76.5% were male, and 79% had underlying cirrhosis. Median MELD at admission was 24 (19-29). Survival was 88% (87-89) at 30 days, 77% (76-78) at 90 days, and 72% (72-74) at 180 days. A total of 1,225 patients received corticosteroids. In an adjusted-survival-model, corticosteroid use decreased 30-day mortality by 41% (hazard ratio [HR] 0.59; 0.47-0.74; p <0.001). Steroids only improved survival in patients with MELD scores between 21 (HR 0.61; 0.39-0.95; p = 0.027) and 51 (HR 0.72; 0.52-0.99; p = 0.041). The maximum effect of corticosteroid treatment (21-30% survival benefit) was observed with MELD scores between 25 (HR 0.58; 0.42-0.77; p <0.001) and 39 (HR 0.57; 0.41-0.79; p <0.001). No corticosteroid benefit was seen in patients with MELD >51. The type of corticosteroids used (prednisone, prednisolone, or methylprednisolone) was not associated with survival benefit (p = 0.247). CONCLUSION: Corticosteroids improve 30-day survival only among patients with severe AH, especially with MELD scores between 25 and 39. LAY SUMMARY: Alcohol-associated hepatitis is a condition where the liver is severely inflamed as a result of excess alcohol use. It is associated with high mortality and it is not clear whether the most commonly used treatments (corticosteroids) are effective, particularly in patients with very severe liver disease. In this worldwide study, the use of corticosteroids was associated with increased 30-day, but not 90- or 180-day, survival. The maximal benefit was observed in patients with an MELD score (a marker of severity of liver disease; higher scores signify worse disease) between 25-39. However, this benefit was lost in patients with the most severe liver disease (MELD score higher than 51).

Retinoids Augment Thiazolidinedione PPARγ Activation in Oral Cancer Cells.

BACKGROUND/AIM: Head and neck squamous cell carcinoma affects nearly 500,000 people annually. Augmenting PPARγ functional activation is linked with multiple anti-carcinogenic processes in aerodigestive cell lines and animal models. PPARγ/RXRα heterodimers may be key partners in this activation. MATERIALS AND METHODS: CA 9-22 and NA cell lines were treated with the PPARγ agonist ciglitazone and/or the RXRα agonist 9-cis-retinoic acid. PPARγ functional activation, cellular proliferation, apoptosis activity, and phenotype were subsequently analyzed. RESULTS: Ciglitazone and 9-cis-retinoic acid independently activated PPARγ and down-regulated the carcinogenic phenotype in vitro. Combination treatment significantly augmented these effects, further decreasing proliferation (p<0.0001), and increasing PPARγ functional activation (p<0.0001), apoptosis (p<0.05), and adipocyte differentiation markers (p<0.0001). CONCLUSION: The efficacy of the combination of ciglitazone and 9-cis-retinoic acid afforded lowering treatment concentrations while maintaining desired therapeutic outcomes, optimistically supporting the feasibility and practicality of this novel treatment option.