Analysis Of Risk Factors Related To Nonalcoholic Fatty Liver Disease And Establishment Of A Predictive Nomogram

Objective:To investigate and analyze the risk factors affecting the occurrence of non-alcoholic fatty liver disease(NAFLD),and further explore the value of the new clinical derived index for the identification of NAFLD disease in different populations.In addition,the prediction model of the nomogram was constructed and verified to provide reference for early prevention,control and intervention of NAFLD in the future.Methods:A total of 563 patients who met the inclusion and exclusion criteria in the First Affiliated Hospital of Xinjiang Medical University from January to April 2022 were selected as subjects.According to abdominal B-type ultrasound imaging parameters,the patients were divided into NAFLD group(n=283)and non-NAFLD group(n=280).By collecting relevant data of questionnaire survey,physical measurement and laboratory examination,the risk factors related to NAFLD were understood,and the recognition value of different clinically derived indexes for NAFLD was further analyzed by ROC curve.Based on Lasso regression combined with multiple Logistic regression,the prediction model of line graph was constructed.The differentiation,accuracy and clinical efficacy of the model were evaluated by using C-index,correction curve and decision curve.Results:1.Firstly,There were statistically significant differences in gender,dietary preference,snoring,exercise intensity,smoking amount,diabetes,hyperuricemia,familial hyperlipidemia,central obesity and BMI between NAFLD group and non-NAFLD group(P<0.05).The levels of DBP,NC,SST,WBC,FPG,TC,TG,LDL-C,TG/HDL-C,SUA,ALT,AST,ALT/AST,GGT and the values of Ty G,CMI,ABSI,BRI,LAP and VAI in NAFLD group were significantly higher than those in non-NAFLD group.While HDL-C levels were lower(P<0.05);2.Unconditioned Logistic regression analysis showed that,gender,poor dietary and lifestyle habits(dietary preference,snoring,exercise intensity,smoking amount),individual medical history(diabetes,hyperuricemia),family history of metabolic disease(family hyperlipidemia),physical measurements(DBP,NC,SST,central obesity,BMI classification),laboratory parameters(WBC,FPG,TC,TG,HDL-C,LDL-C,TG/HDL-C,SUA,ALT,AST,ALT/AST,GGT)and clinically derived index(Ty G,CMI,ABSI,BRI,LAP,VAI)were correlated with NAFLD(P<0.05);3.Among different clinically derived indexes,BRI had the largest area under ROC curve for males and BMI≥24 kg/m~2,which were 0.852(95%CI:0.802～0.893)and 0.790(95%CI:0.744～0.830),respectively.And,LAP was used to identify the largest area under ROC curve for women and BMI<24 kg/m~2,namely 0.893(95%CI:0.852～0.925)and 0.885(95%CI:0.832～0.926).4.The Lasso regression analysis combined with multivariate Logistic regression analysis showed that SST(OR=1.033),LDL-C(OR=1.587),SUA(OR=1.005),GGT(OR=1.007),Ty G(OR=1.945),BRI(OR=2.344)were independent risk factors affecting the occurrence of NAFLD;5.Based on the above six indexes,a line graph prediction model was constructed,and the calculated C index was 0.919(95%CI:0.897～0.942).After internal sampling verification with Bootstrap method,the C index was still greater than 0.9,indicating that the prediction model had good differentiation.The Hosmer-Lemeshow goodness of fit test showed that the prediction model had a good fit(P>0.05).The decision curve shows that the optional threshold probability of this model is relatively wide,which indicates that it has certain clinical application value.6.Compared with existing NAFLD prediction models(ZJU,NSS,HSI),ROC curve results showed that among the four prediction models,the area under ROC curve,sensitivity,specificity and positive likelihood ratio of the linear prediction model were the highest,which were 0.919(95%CI:0.897～0.942),86.22,85.36,5.889 and the negative likelihood ratio was the lowest(0.161,P<0.05).The result of decision curve shows that the net return rate of the line graph prediction model is higher.Conclusion:1.Gender,dietary preference,lifestyle,personal history,family history of metabolic disease,individual obesity,body glucose and lipid metabolism and clinical derived index were closely related to the occurrence of NAFLD.Clinically derived indexes(Ty G,CMI,ABSI,BRI,LAP,VAI)are of certain value in identifying NAFLD in different populations.In the prevention and treatment of NAFLD and daily health management,attention should be paid to the optimization of dietary structure,the enhancement of physical exercise,the strengthening of obesity and the monitoring of blood metabolic indicators.2.In this study,the prediction model based on a variety of clinical data has good predictive efficacy and clinical application value,which is conducive to early prediction of the risk of NAFLD.