BACKGROUND: The aim of the study is to develop a risk score model for identifying postprandial hyperglycemia without oral glucose tolerance tests (OGTT) in Chinese population, and minimize the number of subjects needing further OGTT. METHODS: Multivariable stepwise logistic regression was used to develop risk score models in a derivation cohort (7953 participants without known diabetes). The developed models were verified in a validation cohort (another 1455 subjects without known diabetes). All subjects had completed questionnaires, physical examination and OGTT. Performances of the risk score models were estimated using receiver operating characteristic curves. RESULTS: Two risk score models for screening postprandial hyperglycemia were developed. The simple model used non-invasive risk factors (age, height, weight, waist, systolic blood pressure, pulse, hypertension, dyslipidemia and family history of diabetes mellitus), and the full model contained additional variables [fasting blood glucose (FBG), triglyceride/high density lipoprotein cholesterol] obtainable by invasive laboratory tests. The area under receiver operating characteristic curve (AUC) of simple model was similar to FBG and glycated haemoglobin. The full model has the largest AUC [0.799 (0.789-0.809) and 0.730 (0.702-0.758)] in both derivation and validation cohorts (p < 0.001 compared with simple model, FBG, and glycated haemoglobin). At a cutoff point of 80, the sensitivity, specificity and percentage that needed subsequent OGTT were 75.97, 67.56 and 48.38%, respectively. CONCLUSIONS: We developed a risk score model for screening postprandial hyperglycemia based on routine clinical information. It could effectively identify patients at high-risk for postprandial hyperglycemia and remarkably reduce the number of subjects requiring OGTT.
BACKGROUND: The aim of the study is to develop a risk score model for identifying postprandial hyperglycemia without oral glucose tolerance tests (OGTT) in Chinese population, and minimize the number of subjects needing further OGTT. METHODS: Multivariable stepwise logistic regression was used to develop risk score models in a derivation cohort (7953 participants without known diabetes). The developed models were verified in a validation cohort (another 1455 subjects without known diabetes). All subjects had completed questionnaires, physical examination and OGTT. Performances of the risk score models were estimated using receiver operating characteristic curves. RESULTS: Two risk score models for screening postprandial hyperglycemia were developed. The simple model used non-invasive risk factors (age, height, weight, waist, systolic blood pressure, pulse, hypertension, dyslipidemia and family history of diabetes mellitus), and the full model contained additional variables [fasting blood glucose (FBG), triglyceride/high density lipoprotein cholesterol] obtainable by invasive laboratory tests. The area under receiver operating characteristic curve (AUC) of simple model was similar to FBG and glycated haemoglobin. The full model has the largest AUC [0.799 (0.789-0.809) and 0.730 (0.702-0.758)] in both derivation and validation cohorts (p < 0.001 compared with simple model, FBG, and glycated haemoglobin). At a cutoff point of 80, the sensitivity, specificity and percentage that needed subsequent OGTT were 75.97, 67.56 and 48.38%, respectively. CONCLUSIONS: We developed a risk score model for screening postprandial hyperglycemia based on routine clinical information. It could effectively identify patients at high-risk for postprandial hyperglycemia and remarkably reduce the number of subjects requiring OGTT.