Prediction of the 10-year incidence of type 2 diabetes mellitus based on advanced anthropometric indices using machine learning methods in the Iranian population.

BACKGROUND: Type 2 diabetes mellitus (T2DM) is a growing chronic disease that can lead to disability and early death. This study aimed to establish a predictive model for the 10-year incidence of T2DM based on novel anthropometric indices. METHODS: This was a prospective cohort study comparing people with (n = 1256) and without (n = 5193) diabetes mellitus in phase II of the Mashhad Stroke and Heart Atherosclerotic Disorder (MASHAD) study. The association of several anthropometric indices in phase I, including Body Mass Index (BMI), Body Adiposity Index (BAI), Abdominal Volume Index (AVI), Visceral Adiposity Index (VAI), Weight-Adjusted-Waist Index (WWI), Body Roundness Index (BRI), Body Surface Area (BSA), Conicity Index (C-Index) and Lipid Accumulation Product (LAP) with T2DM incidence (in phase II) were examined; using Logistic Regression (LR) and Decision Tree (DT) analysis. RESULTS: BMI followed by VAI and LAP were the best predictors of T2DM incidence. Participants with BMI < 21.25 kg/m2 and VAI ≤ 5.9 had a lower chance of diabetes than those with higher BMI and VAI levels (0.033 vs. 0.967 incident rate). For BMI > 25 kg/m2, the chance of diabetes rapidly increased (OR = 2.27). CONCLUSIONS: BMI, VAI, and LAP were the best predictors of T2DM incidence.

Predicting the 10-year incidence of dyslipidemia based on novel anthropometric indices, using data mining.

BACKGROUND: The aim was to establish a 10-year dyslipidemia incidence model, investigating novel anthropometric indices using exploratory regression and data mining. METHODS: This data mining study was conducted on people who were diagnosed with dyslipidemia in phase 2 (n = 1097) of the Mashhad Stroke and Heart Atherosclerotic Disorder (MASHAD) study, who were compared with healthy people in this phase (n = 679). The association of dyslipidemia with several novel anthropometric indices including Conicity Index (C-Index), Body Roundness Index (BRI), Visceral Adiposity Index (VAI), Lipid Accumulation Product (LAP), Abdominal Volume Index (AVI), Weight-Adjusted-Waist Index (WWI), A Body Shape Index (ABSI), Body Mass Index (BMI), Body Adiposity Index (BAI) and Body Surface Area (BSA) was evaluated. Logistic Regression (LR) and Decision Tree (DT) analysis were utilized to evaluate the association. The accuracy, sensitivity, and specificity of DT were assessed through the performance of a Receiver Operating Characteristic (ROC) curve using R software. RESULTS: A total of 1776 subjects without dyslipidemia during phase 1 were followed up in phase 2 and enrolled into the current study. The AUC of models A and B were 0.69 and 0.63 among subjects with dyslipidemia, respectively. VAI has been identified as a significant predictor of dyslipidemias (OR: 2.81, (95% CI: 2.07, 3.81)) in all models. Moreover, the DT showed that VAI followed by BMI and LAP were the most critical variables in predicting dyslipidemia incidence. CONCLUSIONS: Based on the results, model A had an acceptable performance for predicting 10 years of dyslipidemia incidence. Furthermore, the VAI, BMI, and LAP were the principal anthropometric factors for predicting dyslipidemia incidence by LR and DT models.