Establishment of a machine learning predictive model for non-alcoholic fatty liver disease: A longitudinal cohort study.

BACKGROUND AND AIMS: Non-alcoholic fatty liver disease (NAFLD) is a common chronic liver disease, which lacks effective drug treatments. This study aimed to construct an eXtreme Gradient Boosting (XGBoost) prediction model to identify or evaluate potential NAFLD patients. METHODS AND RESULTS: We conducted a longitudinal study of 22,140 individuals from the Beijing Health Management Cohort. Variable filtering was performed using the least absolute shrinkage and selection operator. Random Over Sampling Examples was used to address imbalanced data. Next, the XGBoost model and the other three machine learning (ML) models were built using balanced data. Finally, the variable importance of the XGBoost model was ranked. Among four ML algorithms, we got that the XGBoost model outperformed the other models with the following results: accuracy of 0.835, sensitivity of 0.835, specificity of 0.834, Youden index of 0.669, precision of 0.831, recall of 0.835, F-1 score of 0.833, and an area under the curve of 0.914. The top five variables with the greatest impact on the onset of NAFLD were aspartate aminotransferase, cardiometabolic index, body mass index, alanine aminotransferase, and triglyceride-glucose index. CONCLUSION: The predictive model based on the XGBoost algorithm enables early prediction of the onset of NAFLD. Additionally, assessing variable importance provides valuable insights into the prevention and treatment of NAFLD.

Association of sex-specific body mass index and waist circumference trajectories with non-alcoholic fatty liver disease incidence based on growth mixture modeling.

BACKGROUND AND AIMS: Non-alcoholic fatty liver disease (NAFLD) is a common chronic liver disease. The relationship between the trajectories of obesity indicators and incident NAFLD is unknown. Therefore, this study aims to explore the sex-specific association between the trajectories of obesity indicators and the incidence of NAFLD. METHODS AND RESULTS: In total, 9067 participants were recruited for this longitudinal study. Obesity indicators use body mass index (BMI) and waist circumference (WC). The trajectory of obesity indicators was analyzed using the growth mixture modeling. The multivariate logistic regression model was used to analyze the association between obesity indicators' trajectories and incident NAFLD. Over a median follow-up of 1.82 years, 1013 (11.74%) participants developed NAFLD. We identified BMI and WC change trajectories as the stable group, increasing group, and decreasing group. After adjusting for baseline level and other confounders, multivariate logistic regression analysis showed that compared with stable group of BMI, the increasing group, and decreasing group odds ratio and 95% confidence interval of NAFLD were 2.10 (1.06-4.15), and 0.25 (0.09-0.67) in men, and 1.82 (1.08-3.04) and 0.32 (0.16-0.64) in women. Compared with stable group of WC, the increasing group was 2.57 (1.39-4.74) in men, the increasing group, and decreasing group were 2.29 (1.70-3.10) and 0.28 (0.12-0.64) in women. Sensitivity analysis showed that the results were stable. CONCLUSION: The BMI and WC changing trajectories are significantly associated with the incidence of NAFLD in men and women. Populations of real-world health examinations can be categorized based on obesity indicator changes to prevent NAFLD.

Temporal relationship between elevated serum uric acid levels and dyslipidemia: A 5-year cohort study using cross-lagged panel analysis.

BACKGROUND AND AIM: Previous studies have demonstrated an association between SUA and dyslipidemia. This study aims to explore the temporal relationship between SUA and dyslipidemia. METHODS AND RESULTS: Based on the Beijing Health Management Cohort conducted from 2013 to 2018, the data of a physical examination population was collected, including a total of 6630 study subjects. Cross-lagged panel analysis was employed to examine the temporal relationship between elevated SUA levels and dyslipidemia, indicated by either elevated TG or decreased HDL-C. The path coefficient and the 95 % CI from baseline TG to follow-up SUA were as follows: in the general population, men, women, and people with BMI ≥25 kg/m2were 0.027 (0.008-0.045), 0.024 (0.001-0.048), 0.032 (0.001-0.063) and 0.033 (0.006-0.059) (P < 0.05); however, the path coefficient from baseline SUA to follow-up TG and the 95 % CI were not statistically significant. Furthermore, the path coefficients and 95 % CIs between elevated SUA and decreased HDL-C were not statistically significant, both in the general population and in populations stratified by gender and BMI. CONCLUSIONS: We found a temporal relationship from elevated TG to elevated SUA in the general population and the populations stratified by gender and BMI (≥25 kg/m2). However, we did not observe a reverse relationship from elevated SUA to elevated TG. Additionally, we did not find a temporal relationship between decreased HDL-C and elevated SUA in both the general population and the stratified populations.

Association between nonalcoholic fatty liver disease and type 2 diabetes: A bidirectional two-sample mendelian randomization study.

OBJECTIVE: The aim of this study was to explore the mutually causal relationship between NAFLD and type 2 diabetes. METHODS: Based on the data obtained from GWAS, this study employed bidirectional two-sample MR analysis to investigate the causal relationship between NAFLD and type 2 diabetes, and also examined the causal relationship between liver fat accumulation and type 2 diabetes as well as the relationship between NAFLD and FPG, IR. RESULTS: In MR analysis of NAFLD and type 2 diabetes, when NAFLD as an exposure and type 2 diabetes as a result, the OR (95 % CI) was 1.10890 (1.00135-1.22801); in the reverse analysis, the OR value was not statistically significant. In MR analysis of NAFLD, FPG and IR, there was no statistical significance in both directions. In MR analysis of liver fat accumulation and type 2 diabetes, when liver fat as an exposure and type 2 diabetes as a result, the OR (95 % CI) was 1.17516 (1.02054-1.35321); in the reverse analysis, the OR value (95 % CI) was 1.06283 (1.02879-1.09799). CONCLUSION: There is a unidirectional causal relationship between NAFLD and type 2 diabetes. Furthermore, a bidirectional causal relationship exists between liver fat accumulation and type 2 diabetes.

Bidirectional relationship between nonalcoholic fatty liver disease and serum creatinine-to-body weight ratio as a proxy for skeletal muscle mass index.

BACKGROUND AND AIM: Although an association between skeletal muscle mass index and nonalcoholic fatty liver disease (NAFLD) has previously been demonstrated, the causal direction remains unclear. Herein, we investigated the directional association between NAFLD and the serum creatinine-to-body weight ratio (sCr/bw), a surrogate marker of the muscle mass index, using longitudinal data. METHODS: We recruited 9662 participants in 2017 and performed follow-up over 4 years. We evaluated whether sCr/bw was related to NAFLD development (Analysis I) and whether NAFLD was associated with a low sCr/bw incidence (Analysis II) using logistic regression models. Furthermore, a random intercept cross-lagged panel model was applied to evaluate the bidirectional association between sCr/bw ratio and NAFLD (Analysis III). RESULTS: Analysis I demonstrated an association between sCr/bw and incident NAFLD (odds ratio [OR] = 0.160, 95% confidence interval [CI]:0.107-0.232). Analysis II indicated a relationship between NAFLD and subsequent low sCr/bw ratio (OR = 1.524, 95% CI: 1.258-1.846). Analysis III indicated that the standard regression coefficient from sCr/bw to subsequent hepatic steatosis (HS) was -0.053 for ßsCr/bw2017 â†’ HS2019 and -0.060 for ßsCr/bw2019 â†’ HS2021 and the coefficient from HS to subsequent sCr/bw was -0.093 for ßHS2017 â†’ sCr/bw2019 and -0.112 for ßHS2019 â†’ sCr/bw2021 (all P < 0.05). CONCLUSIONS: This study indicated mutual causality between sCr/bw and NAFLD. Considering that sCr/bw is a surrogate marker of muscle mass index, the findings emphasize that NAFLD and low muscle mass form a vicious cycle, which should be taken seriously in clinical practice.

Serum uric acid to creatinine ratio and metabolic syndrome in middle-aged and elderly population: Based on the 2015 CHARLS.

BACKGROUND AND AIMS: Serum uric acid to creatinine ratio (SUA/Cr) may be associated with metabolic syndrome (MS). Here, we investigated the correlation between SUA/Cr and MS in Chinese residents aged ≥ 45 years. METHODS AND RESULTS: Data were obtained from the 2015 China Health and Retirement Longitudinal Study (CHARLS) database. MS was diagnosed using the Chinese Diabetes Society 2017 criteria. We grouped the population according to SUA/Cr quartiles and compared the index differences between groups. We used spearman correlation analysis and binary logistic regression. The possible dose-response association of SUA/Cr with MS were analyzed using restricted cubic spline model. Of 12,946 included participants, 3370 (26.0%) had MS, and 1900 (56.4%) were female. After adjusting for multiple confounders, binary logistic regression analysis showed that compared with Quartile 1, the odds ratio (95% confidence interval) of the MS risk was 1.29 (1.09-1.52), 1.47 (1.25-1.74), and 1.80 (1.53-2.12) in Quartiles 2, 3, and 4, respectively. The restricted cubic spline model indicated a significant nonlinear dose-response association (Poverall < 0.001, Pnon-linearity = 0.029) between SUA/Cr and strength of MS prevalence association; MS risk began increasing when SUA/Cr > 6.22. CONCLUSIONS: A significant positive correlation existed between SUA/Cr and MS risk in Chinese individuals aged ≥ 45 years, which may be a new predictive marker for MS risk.

Causal relationship between serum uric acid and abnormal blood pressure based on the panel model study: A 5-year cohort study.

BACKGROUND AND AIMS: To investigate the relationship between elevated serum uric acid (SUA) levels and blood pressure (BP). METHODS AND RESULTS: Based on the Beijing Health Management Cohort, 5276 health examination people were enrolled. Cross-lagged model was used to explore the relationship between SUA levels and blood pressure. The results showed: (1) increased SUA and increased systolic blood pressure (SBP): ① The path coefficients from baseline SUA to follow-up SBP were statistically significant in both the general population (ß = 0.034, P < 0.05) and men (ß = 0.048, P < 0.05). The path coefficients from baseline SBP to follow-up SUA were not statistically significant in either the general population (ß = 0.010, P > 0.05) or men (ß = 0.011, P > 0.05). ② The path coefficients from baseline SUA to follow-up SBP and from baseline SBP to follow-up SUA were not statistically significant in women with BMI ≥ 25 kg/m2 and BMI < 25 kg/m2. (2) Increased SUA and diastolic blood pressure (DBP): ① There was no statistical significance between the path coefficients from baseline DBP to follow-up SUA and the path coefficients from baseline SUA to follow-up DBP. ② In men and women, BMI ≥ 25 kg/m2 and BMI < 25 kg/m2, the path coefficients from baseline DBP to follow-up SUA and from baseline SUA to follow-up DBP were not statistically significant. CONCLUSIONS: SUA can increase blood pressure in the general male population; no reverse time sequence relationship was found. The temporal relationships between SUA levels and SBP abnormalities were different in the sex and BMI subgroups. No bidirectional causal temporal relationship was found between SUA elevation and DBP abnormality.