LBODP027 Detection Of Metabolic Syndrome In Adolescent With Explainable Artificial Intelligence: Which Place For Mean Blood Pressure?

Abstract Metabolic syndrome (MetS) definition in adolescents is complicated by the physiologic evolution with age and sexual maturation of the components of this syndrome. While screening for MetS in youth should be useful to assess current complications and prevent future ones, it is not simple in clinical practice due to the number of different definitions of MetS in adolescents, which are generally based for most of them on percentiles. The aim of this study was to check the validity of artificial intelligence (AI)-based scores in screening for MetS in adolescents, using new parameters identified with AI techniques and without using percentiles. This study included 1,086 adolescents (559 girls and 527 boys) aged 12 to 18 with BMI 21.2±3.9 kg/m2. All had anthropometric measurements taken and had blood tests. Mean blood pressure (MBP), and triglyceride glucose index (TyG) were calculated. AI techniques are characterized by their black-box nature. Explainable AI methods are used to extract the learned function. "Gini importance" techniques were tested and used to build new scores for the detection of MetS among children and adolescents. We used 2007 IDF and Cook definitions of MetS to test the validity of these scores. MetS prevalence was 0.9% and 2.2% according to IDF and Cook definitions, respectively. The most accurate AI scores (S-IDF, S-Cook) for the detection of MetS according to these definitions include age, waist circumference, MBP and TyG index: S-IDF= - 0.52×age + 0.48×WC - 0.51×TyG + 0.27×MBP; S-Cook= - 0.52×age + 0.48×WC - 0.51×TyG + 0.29×MBP. With cut-off levels of 51.35/51. 06, these AI scores offer AUC 0.91/0.93, specificity 81%/75%, and sensitivity 90%/100% for MetS detection. Furthermore, with a cut-off point of 92 mmHg, MBP alone offers a better specificity 85%/83% and sensitivity 80%/88%, AUC 0.89, than TyG index alone (100%/46% for specificity, 30%/88% for sensitivity) to detect MetS based on IDF and Cook definitions, and slightly lower performance than AI-scores. MBP was associated with the occurrence of type 2 diabetes (T2D) and cardiovascular disease among the general population; cardiovascular disease and cardiovascular mortality among patients with T2D; and a discriminatory ability as strong as for SBP and DBP in predicting T2D. However, we did not find in the literature similar studies in the pediatric population. TyG index has been assessed mainly in adults and in a few studies in adolescents, and recently proposed as an alternative marker of insulin resistance and a marker of high risk for diabetes and cardiovascular complications or events in both adult and pediatric populations. In this study, AI techniques have proven their ability to extract knowledge from data and come up with new scores based on simplified criteria in adolescents. It allowed us to reveal the MBP as a simple alternative marker for MetS screening in adolescents. Presentation: No date and time listed