A Comparative Study On The Relationship Between Metabolic Disorders,body Mass Index And Insulin Resistance(IR),and The Construction And Application Of IR Prediction Model

Background:Obesity is closely related to a variety of metabolic abnormalities,but patients with the same degree of obesity have obvious differences in metabolic disorder phenotypes in clinical work,and the reasons for this difference are still unclear.Studies have found that insulin resistance(IR)is also one of the important causes of metabolic disorders,and is closely related to obesity.However,no studies have compared the relationship between body mass index(BMI)and IR and metabolic disorders.IR is closely related to metabolic disorders,so evaluating IR is of great significance for the prevention and management of metabolic disorders.Currently the most commonly used IR assessment method is the Homeostatic Model Assessment of IR(HOMA-IR),developed based on fasting blood glucose(FBG)and fasting insulin(FINS).However,HOMA-IR cannot be used to assess IR in large epidemiological studies such as the UK biobank(UKB)due to the lack of FINS.Currently,the triglyceride glucose index(Ty G)developed using only FBG and triglyceride(TG)has been proposed to evaluate IR,but it has not been recommended for large-scale epidemiological investigations.At present,there is still a lack of IR assessment methods that can be widely used in clinical work and large-scale epidemiological investigations.Objects:1.Comparative analysis of the relationship between BMI and HOMA-IR and metabolic disorders in overweight and obese samples;2.In overweight and obese samples,a random forest(RF)model was constructed to predict IR based on body measurement parameters and metabolic indicators through machine learning;3.Using the RF model to identify the IR population and analyze the risk of metabolic disorders among the IR population in large-scale epidemiological survey samples both domestically and internationally;Methods:1.In the overweight and obese samples,the metabolic indicators and the prevalence of metabolic disorders were compared between different BMI groups and different HOMA-IR groups.The independent correlations of BMI and HOMA-IR with metabolic indicators and metabolic disorders were tested by partial correlation analysis and binary logistic regression analysis.The BMI or HOMA-IR subgroup samples matched by other factors were extracted using matched sampling method and the differences in the prevalence of metabolic disorders between groups were compared.2.In overweight and obese samples,an IR prediction model based on commonly used body measurement parameters and metabolic indicators was constructed by using the RF algorithm in machine learning.3.The IR prediction score was obtained by applying the RF model in the Hunan Province samples,normal weight subgroup and overweight and obese subgroup of "China Chronic Disease and Nutrition Surveillance" in 2018,and the samples were divided into non-insulin resistance(n IR)group and IR group according to the score distribution.The metabolic indicators and the prevalence of metabolic disorders were compared between the n IR group and the IR group,and the risk of metabolic disorders in IR patients was analyzed by binary logistic regression model.4.The IR prediction score was obtained by applying the RF model in UKB cohort samples,normal weight subgroup and overweight and obese subgroup.According to the score distribution,the samples with baseline undiagnosed metabolic disorders were divided into n IR group and IR group.The probability of metabolic disorders in n IR group and IR group was compared,and the risk of metabolic disorders in IR patients was analyzed by competing risk analysis model.Results:1.In overweight and obese samples,the prevalence of metabolic disorders and metabolic indicators increased with the increase of BMI and HOMA-IR.Partial correlation analysis and logistic regression analysis showed that the risk of hypertension and abnormal glucose metabolism was independently correlated with BMI and HOMA-IR,but it was HOMA-IR rather than BMI that was independently correlated with the risk of dyslipidemia,and hyperuricemia BMI,but not HOMA-IR,was independently associated with the risk of comorbidity.The results of the sampling analysis matched for age,sex and BMI or HOMA-IR showed that when the age,sex and BMI distributions were similar,there were significant differences in the prevalence of dyslipidemia but not in the prevalence of hyperuricemia among different HOMA-IR groups.Significant difference;when age,sex,and HOMA-IR distributions were similar,there was no significant difference in the prevalence of dyslipidemia but a significant difference in the prevalence of hyperuricemia between different BMI groups.2.In overweight and obese samples,an IR prediction model based on commonly used body measurement parameters and metabolic indicators was constructed by the RF algorithm in machine learning.The performance of this model in diagnosing IR is good and better than that of Ty G index.3.The optimized RF model was used to divide the samples into n IR group and IR group in the samples of "China Chronic Disease and Nutrition Survey".The comparison found that the prevalence of metabolic disorders in the IR group was higher than that in the n IR group in the total sample,normal weight subgroup and overweight and obesity subgroup.The results of binary logistic regression analysis showed that in the normal weight subgroup,the association between IR and the risk of dyslipidemia could not be judged.In the total sample,normal weight subgroup and overweight and obese subgroup,IR was an independent risk factor for the occurrence of metabolic disorders.4.An optimized RF model was applied to samples from the UKB cohort with undiagnosed metabolic disorders at baseline to divide the samples into n IR and IR groups.Comparing the occurrence probability of metabolic disorders in n IR group and IR group,it was found that in the total sample,normal weight subgroup and overweight and obese subgroup,the probability of metabolic disorder in IR group was higher.Competing risk analysis occurred in the total sample,the normal weight subgroup and the overweight and obese subgroup in the IR group had a higher risk of metabolic disorders.Conclusion:In overweight and obese people,IR is an independent risk factor for hypertension,abnormal glucose metabolism and dyslipidemia,while BMI is an independent risk factor for hypertension,abnormal glucose metabolism and hyperuricemia.An RF model for predicting IR was constructed in this population,and the model was successfully applied to identify IR populations with high risk of metabolic disorders in large epidemiological data.