Study On The Correlation And Prediction Model Between Non-invasive Anthropometric Indexes And Metabolic Syndrome

Purpose:1.Analyze the association between traditional obesity evaluation indexes,including BMI,WC,HC,WHR,WHt R,and new anthropometric indexes such as BRI,BAI,CUN＿BAE,AVI,C＿Index,and PBF measured by BIA,with Met S and its components.2.Compare the ability of different anthropometric indexes to predict the risk of Met S,calculate appropriate cutoff values for diagnosing Met S and its components using different anthropometric indexes.3.Use variable selection techniques to screen non-invasive human body measurement indicators and establish a simple metabolic diagnostic model in the Chinese population.Evaluate the model’s performance through discrimination and calibration and assess the model’s efficiency through internal,internal-external,and external validation.Methods:1.Trained staff members conducted standardized questionnaire surveys to obtain information on demographic data,personal and family medical histories,and lifestyle related to diabetes,hypertension,dyslipidemia,cardiovascular events,education levels,and occupations.During training,detailed explanations of the research questionnaire were provided to surveyors,and clinical staff members were trained to measure blood pressure,weight,height,waist circumference,and blood samples according to standard methods.2.Preliminary analysis of CNDMDS data was conducted to understand the prevalence of metabolic syndrome(Met S).Quartiles of Met S diagnostic indicators,including SBP,DBP,FBG,TG,and HDLC,were used to compare the distribution of different anthropometric indexes.Single-factor and multiple-factor logistic regression analyses were performed using quartiles of each anthropometric index,with age,sex,region,economic development level,education level,smoking history,alcohol consumption history,physical exercise,and family history of metabolic disorders as covariates.The association between different anthropometric indexes and Met S was analyzed.Receiver operating characteristic(ROC)curves were used to evaluate the diagnostic effectiveness of different anthropometric indexes for Met S and its components,and the optimal cutoff values were determined using the Youden index.3.A metabolic prediction model for Met S was developed and validated using CNDMDS data as the training set,CNDMDS Shaanxi cohort data as validation set 1,and NHANES 1999-2004 data as validation set 2.4.LASSO regression and random forest models were used to screen predictive variables,and ROC and calibration curves were used to evaluate the discrimination and calibration of the model.The AUC and Brier score were calculated.10-fold internal cross-validation and multi-center internal-external cross-validation were performed in the training set,and external validation was performed in the validation set.Finally,the Delong’s test was employed to compare the predictive performance between the model and anthropometric indexes.Results:1.According to the JIS Asian population standard,7,601 people have Met S,with a crude prevalence rate of 32.70%.The crude prevalence rate of Met S in males is 32.20%,and in females is 33.00%.2.There are significant differences(P<0.05)between different anthropometric index groups in different quartiles of SBP,DBP,FBG,and TG,and as the quartiles of SBP,DBP,FBG,and TG increase,different anthropometric indexes also tend to increase.HDLC does not show a significant decreasing trend for some anthropometric indexes in each quartile group.3.Single-factor logistic regression and multiple-factor logistic regression with age,gender,education level,smoking history,physical activity,and family history of metabolic disorders as covariates show that compared with the Q1 group reference of different anthropometric indexes,the OR values of Q2,Q3,and Q4 groups have a statistically significant association with Met S and its components,and the OR values increase as the quartiles increase(trend test P<0.05).After further adjusting for WC,HC,and BMI,the OR values of Q2,Q3,and Q4 groups also show a statistically significant association with Met S and its components,and the OR values increase as the quartiles increase(trend test P<0.05).The OR values of the included anthropometric indexes associated with hypertension are not statistically significant,and the OR values of some indexes associated with high blood sugar,elevated TG,and decreased HDLC are not statistically significant.4.ROC curve analysis for discriminating Met S and its components: After stratification by gender,different anthropometric indexes have a statistically significant AUC for discriminating Met S in males,females,and the overall population.Overall,WHt R,BRI,WC,and AVI have a better performance in discriminating Met S.CUN＿BAE,WHt R,and BRI have a better performance in discriminating Met S components other than abdominal obesity.However,the performance of WHt R and BRI,WC and AVI in discriminating Met S and its components is comparable.5.Finally,SBP,DBP,PBF,CUN＿BAE,WC,and WHt R were included as six variables to fit logistic regression as the final prediction model.The original model does not have any strong influential points,and the variance inflation factors of each variable are within an acceptable range.The AUC of the model in the training set is 0.879(95%CI:0.874-0.873),the predicted probability cutoff value judged by the maximum Youden index is 0.3,and the Brier score is 0.133.After 10-fold internal cross-validation,the average AUC of the prediction model is 0.878,and the average Brier score is 0.133.After multi-center internal-external cross-validation,the average AUC is 0.877,and the evaluation Brier score is 0.133,indicating that the model has good internal validation results.6.The AUC in validation set 1 is 0.852(95%CI: 0.829-0.874).The calibration curve is close to the diagonal line,and the Brier score is 0.155,indicating that the model performs well in validation set 1.The AUC in validation set 2 is 0.806(95%CI: 0.788-0.825),indicating good discrimination,but the calibration curve deviates greatly from the diagonal line,and the calibration in validation set 2 is poor.WC and WHt R,which performed the best in the first part of the study,have lower ability to predict Met S than our final model.7.An We Chat microprogram was developed based on this prediction model for ease of use.Conclusions:1.PBF,WHR,WHt R,BRI,CUN＿BAE,and AVI have a strong association with Met S and its components and have good predictive performance for Met S and its components.2.The prediction model developed based on non-invasive anthropometric indexes has good predictive ability,and the We Chat mini-program developed based on this model facilitates individuals or clinical doctors to calculate the risk of Met S.