Biological prediction model of insulin resistance in obese adolescents based on LASSO-Logistic regression

Abstract

To identify risk factors affecting the development of insulin resistance in obese adolescents, and to build a nomograph model for predicting the risk of insulin resistance and achieve early screening of insulin resistance. A total of 404 obese adolescents aged 10 to 17 years were randomly recruited through a weight loss camp for the detection and diagnosis of lipids and insulin resistance between 2019 and 2021, and key lipid indicators affecting the development of insulin resistance were screened by Lasso regression, nomogram model was constructed, and internal validation of the models was performed by Bootstrap method, and the area under the working characteristic curve(ROC-AUC) and clinical decision curve were used to assess the calibration degree and stability of the column line graph. The AUC was 0.825(95% CI 0.782-0.868), the internal validation result C-Index was 0.804, the mean absolute error of the column line graph model to predict the risk of insulin resistance was 0.015 and the Brier score was 0.163. The Hosmer-Lemeshow goodness-of-fit test showed that model is ideal and acceptable(χ~2=5.59, P=0.70). The nomogram model of triglyceride, low-density lipoprotein cholesterol and total cholesterol/high-density lipoprotein cholesterol based on Lasso-logistic regression can effectively predict the risk of insulin resistance in obese children and adolescents.