Screening The Biomarkers Of Type 2 Diabetes Incident Risk And Development Of New Risk Models For Prediction Of Type 2 Diabetes In A Rural Chinese Population

Objectives:To develop two risk models (nonlaboratory-based and laboratory-based) for predicting type 2 diabetes (T2D) in a rural Chinese population, and investigate whether metabolic biomarkers are associated with development of T2D and improve T2D risk prediction.Methods:(1) Nonlaboratory-based risk model for predicting T2D:Data from Handan Eye Study conducted from 2006-07 to 2012-13 comprising 4132 participants aged 30+ years were analyzed. A Nonlaboratory-based risk model (simple clinical model) was derived in random two-thirds of the sample by using stepwise logistic regression and validated in the other one-third. In addition, a simple point system for T2D was built according to the procedures as dscribed in Framingham Study. (2) Clinical and laboratory-based risk model for predicting T2D:In the similar sample, an enhanced risk model (complex clinical model) was developed and evaluated improvements by adding fasting plasma glucose (FPG), triglycerides (TGs), and C-reactive protein (CRP). Similarly, a simple point system for T2D was built according to the procedures as dscribed above. (3) Associations of plasma amino acids levels and T2D incident risk: Plasma amino acids levels were measured in a case-control sample nested in the Handan Eye Study, and investigated whether levels of plasma amino acidswere associated with development of T2D. The area under the curve (AUC), net reclassification index (NRI), and integrated discrimination index (IDI) were calculated to determine if these amino acids improved T2D risk models.Results:(1) Nonlaboratory-based risk model for predicting T2D:the simple clinical model included age (8 points), BMI (3 points), waist circumference (7 points), and family history of diabetes (9 points). The score ranged from 0 to 27. The AUC was 0.686 in the validation sample. At the optimal cutoff value of 9, the sensitivity and specificity of score were 74.32% and 58.82%. (2) Clinical and laboratory-based risk model for predicting T2D:the complex clinical model included age (8 points), BMI (6 points), waist circumference (8 points), and family history of diabetes (9 points), FPG (23 points), and TGs (4 points). The score ranged from 0 to 58. FPG and TGs significantly improved the AUC (0.802) in the validation sample. At the optimal cutoff value of 27, the sensitivity and specificity of score were 70.27% and 80.83%. (3) Associations of plasma amino acids levels and T2D incident risk:the plasma concentrations of alanine, glycine, aromatic and branched-chain amino acids were significantly associated with incident T2D. Tyrosine was a particularly strong predictor of incident T2D in this rural Chinese population, even after adjustment for FPG and TGs. None of amino acids significantly improved the AUC and NRI. Only the four-amino acids combination (alanine, leucine, valine and tyrosine) significantly improved the IDI, but the net changed was small.Conclusions:(1) This is the first rural Chinese population-based cohort study to develop two risk models (nonlaboratory-based and laboratory-based) for predicting T2D. The nonlaboratory-based risk score derived from clinical informationalone is simple and can be used in rural basic health care settings and families. Another risk score derived fromclinical information combined with FPG and TGs is superior to the basic model and may be used in clinical practice. (2) Levels ofalanine, glycine, and branched chain amino acids, particularly tyrosine, are significantly associated with incident T2D, and may be potential treatment targets for T2D. However, none of these amino acids significantly improved T2D risk prediction.