Development and validation of machine learning models for nonalcoholic fatty liver disease.

BACKGROUND: Nonalcoholic fatty liver disease (NAFLD) had become the most prevalent liver disease worldwide. Early diagnosis could effectively reduce NAFLD-related morbidity and mortality. This study aimed to combine the risk factors to develop and validate a novel model for predicting NAFLD. METHODS: We enrolled 578 participants completing abdominal ultrasound into the training set. The least absolute shrinkage and selection operator (LASSO) regression combined with random forest (RF) was conducted to screen significant predictors for NAFLD risk. Five machine learning models including logistic regression (LR), RF, extreme gradient boosting (XGBoost), gradient boosting machine (GBM), and support vector machine (SVM) were developed. To further improve model performance, we conducted hyperparameter tuning with train function in Python package 'sklearn'. We included 131 participants completing magnetic resonance imaging into the testing set for external validation. RESULTS: There were 329 participants with NAFLD and 249 without in the training set, while 96 with NAFLD and 35 without were in the testing set. Visceral adiposity index, abdominal circumference, body mass index, alanine aminotransferase (ALT), ALT/AST (aspartate aminotransferase), age, high-density lipoprotein cholesterol (HDL-C) and elevated triglyceride (TG) were important predictors for NAFLD risk. The area under curve (AUC) of LR, RF, XGBoost, GBM, SVM were 0.915 [95% confidence interval (CI): 0.886-0.937], 0.907 (95% CI: 0.856-0.938), 0.928 (95% CI: 0.873-0.944), 0.924 (95% CI: 0.875-0.939), and 0.900 (95% CI: 0.883-0.913), respectively. XGBoost model presented the best predictive performance, and its AUC was enhanced to 0.938 (95% CI: 0.870-0.950) with further parameter tuning. CONCLUSIONS: This study developed and validated five novel machine learning models for NAFLD prediction, among which XGBoost presented the best performance and was considered a reliable reference for early identification of high-risk patients with NAFLD in clinical practice.

Atherogenic index of plasma combined with waist circumference and body mass index to predict metabolic-associated fatty liver disease.

BACKGROUND: Early identification of metabolic-associated fatty liver disease (MAFLD) is urgent. Atherogenic index of plasma (AIP) is a reference predictor of obesity-related diseases, but its predictive value for MAFLD remains unclear. No studies have reported whether its combination with waist circumference (WC) and body mass index (BMI) can improve the predictive performance for MAFLD. AIM: To systematically explore the relationship between AIP and MAFLD and evaluate its predictive value for MAFLD and to pioneer a novel noninvasive predictive model combining AIP, WC, and BMI while validating its predictive performance for MAFLD. METHODS: This cross-sectional study consecutively enrolled 864 participants. Multivariate logistic regression analysis and receiver operating characteristic curve were used to evaluate the relationship between AIP and MAFLD and its predictive power for MAFLD. The novel prediction model A-W-B combining AIP, WC, and BMI to predict MAFLD was established, and internal verification was completed by magnetic resonance imaging diagnosis. RESULTS: Subjects with higher AIP exhibited a significantly increased risk of MAFLD, with an odds ratio of 12.420 (6.008-25.675) for AIP after adjusting for various confounding factors. The area under receiver operating characteristic curve of the A-W-B model was 0.833 (0.807-0.858), which was significantly higher than that of AIP, WC, and BMI (all P < 0.05). Subgroup analysis illustrated that the A-W-B model had significantly higher area under receiver operating characteristic curves in female, young and nonobese subgroups (all P < 0.05). The best cutoff values for the A-W-B model to predict MAFLD in males and females were 0.5932 and 0.4105, respectively. Additionally, in the validation set, the area under receiver operating characteristic curve of the A-W-B model to predict MAFLD was 0.862 (0.791-0.916). The A-W-B level was strongly and positively associated with the liver proton density fat fraction (r = 0.630, P < 0.001) and significantly increased with the severity of MAFLD (P < 0.05). CONCLUSION: AIP was strongly and positively associated with the risk of MAFLD and can be a reference predictor for MAFLD. The novel prediction model A-W-B combining AIP, WC, and BMI can significantly improve the predictive ability of MAFLD and provide better services for clinical prediction and screening of MAFLD.

Patients with different types of arthritis may be at risk for major depression: results from the National Health and Nutrition Examination Survey 2007-2018.

BACKGROUND: Arthritis is one of the common causes of physical pain and disability, which often makes patients fall into major depression. However, the correlation between arthritis and major depression, and how different types of arthritis correspond to major depression remain to be explored. The purpose of this study is to explore the relationship between arthritis and major depression. METHODS: Arthritis status was reported by participants themselves, and the Patient Health Questionnaire-9 in National Health and Nutrition Examination Survey (NHANES) was used to evaluate major depression, logistic regression was used to evaluate the relationship between arthritis and major depression. RESULTS: We analyzed the data of 25,990 adults who participated in the NHANES from 2007 to 2018. Participants with major depression were more likely to be female, Hispanic, smoker, less educated, less recreational activities, poverty-to-income ratio <5, coronary heart disease, stroke, cancer or malignant tumor, diabetes, hypertension and higher body mass index (BMI). Arthritis was significantly correlated with major depression (25.4% vs. 44.9%; P<0.001), even after adjusting for gender, age, race, BMI, PIR, education, marriage, moderate recreational activities, smoking, history of coronary heart disease, stroke, cancer or malignant tumor, diabetes, and hypertension (OR =2.30, 95% CI, 2.06-2.56, P<0.001). Subgroup analysis showed that compared with degenerative arthritis, rheumatoid arthritis (RA), or other arthritis, psoriatic arthritis (PsA) had the greatest influence on major depression patients. CONCLUSIONS: All patients with arthritis, especially PsA, may have the risk of major depression. Psychological intervention necessary for patients with arthritis.

Untargeted metabolomics characteristics of nonobese nonalcoholic fatty liver disease induced by high-temperature-processed feed in Sprague-Dawley rats.

BACKGROUND: Nonalcoholic fatty liver disease (NAFLD) has become one of the most common chronic liver diseases in the world. In our early clinical data and questionnaire analysis of NAFLD, it was found that the body mass index of some patients did not meet the diagnostic criteria for overweight or obesity. The consumption of high-temperature-processed foods such as fried food, hot pot and barbecue is closely related to the occurrence of nonobese NAFLD. Reducing the intake of this kind of food can reduce disease severity and improve prognosis. AIM: To explore the untargeted metabolomics characteristics of nonobese nonalcoholic fatty liver disease in Sprague-Dawley rats induced by high-temperature-processed feed. METHODS: Fifty-four male Sprague-Dawley rats were divided into three groups: The control group received a standard diet; the nonfried soybeans (NDFS) group received 60% NDFS and 40% basic feed and the dry-fried soybeans (DFS) group received 60% DFS and 40% basic feed. Six rats were sacrificed at week 4, 8, and 12 in each group. The food intake, body weight, Lee's index, liver index, serological index and hepatic histopathology were assessed. Untargeted metabolomics characteristics were used to analyze the changes in liver metabolites of rats at week 12. Correlations between metabolites and pathology scores between the DFS and control groups and between the DFS and NDFS groups were analyzed. We selected some of the metabolites, both within the pathway and outside of the pathway, to explain preliminarily the difference in liver pathology in the three groups of rats. RESULTS: There were no statistically significant differences in the food intake, body weight, Lee's index or serological index between the DFS group and the control group (P > 0.05). At week 8 and week 12, the steatosis scores in the DFS group were significantly higher than those in the other two groups (P < 0.05). At week 12, the liver index of the DFS group was the lowest (NDFS group vs DFS group, P < 0.05). The fibrosis score in the DFS group was significantly higher than those in the other two groups (P < 0.05). The correlation analysis of the liver pathology score and differential metabolites in the DFS and NDFS groups showed that there were 10 strongly correlated substances: Five positively correlated substances and five negatively correlated substances. The positively correlated substances included taurochenodeoxycholate-3-sulfate, acetylcarnitine, 20a,22b-dihydroxycholesterol, 13E-tetranor-16-carboxy-LTE4 and taurocholic acid. The negatively correlated substances included choline, cholesterane-3,7,12,25-tetrol-3-glucuronide, nicotinamide adenine dinucleotide phosphate, lysoPC [16:1 (9Z)] and glycerol 3-phosphate. The correlation analysis of the liver pathology score and differential metabolites in the DFS and control groups showed that there were 13 strongly correlated substances: Four positively correlated substances and 9 negatively correlated substances. The positively correlated substances included 4-hydroxy-6-eicosanone, 3-phosphoglyceric acid, 13-hydroxy-9-methoxy-10-oxo-11-octadecenoic acid and taurochenodeoxycholate-3-sulfate. The negatively correlated substances included lysoPC [16:1(9Z)], S-(9-hydroxy-PGA1)-glutathione, lysoPC [20:5 (5Z, 8Z, 11Z, 14Z, 17Z)], SM (d18:1/14:0), nicotinamide adenine dinucleotide phosphate, 5,10-methylene-THF, folinic acid, N-lactoyl-glycine and 6-hydroxy-5-methoxyindole glucuronide. CONCLUSION: We successfully induced liver damage in rats by using a specially prepared high-temperature-processed feed and explored the untargeted metabolomics characteristics.