Abstract
BACKGROUND: The aim of this study was to investigate the association between estimated glucose disposal rate (eGDR) and the risk of future diabetes development in middle-aged and elderly adults, and to construct a diabetes prediction model. METHODS: The present study comprised a total of 8,072 participants, with 6,965 drawn from the China Health and Retirement Longitudinal Study (CHARLS) cohort and 1,107 from the English Longitudinal Study of Aging (ELSA) cohort. The correlation between eGDR and the onset of diabetes was analysed by means of a logistic regression model, and subgroup analyses and restricted cubic spline (RCS) curve analyses were performed to verify the non-linear relationship. A predictive model was constructed based on multivariable variables, and model efficacy was assessed by subject operating characteristic curves (AUC) and calibration curves. RESULTS: The prevalence of diabetes mellitus was 5.87% in the CHARLS cohort and 9.94% in the ELSA cohort. It was found that eGDR was significantly lower in both cohorts of diabetic patients (P < 0.001). Furthermore, an association was observed between eGDR reduction and an increased risk of developing diabetes. The multivariable-adjusted odds ratios (OR) for Q2-Q4 in the CHARLS cohort were 0.66 (0.51-0.84), 0.36 (0.25-0.51), and 0.31 (0.20-0.47), respectively, using the eGDR quartiles (Q1 as the reference); and for the ELSA cohort, the values were 0.40 (0.23-0.70), 0.30 (0.15-0.62), and 0.06 (0.01-0.28), respectively. RCS analyses revealed no evidence of nonlinear association between eGDR and diabetes, after adjusting for confounders. A column-line graphical model, incorporating variables of heart disease, stroke, BMI, lipids, glucose and eGDR, yielded AUCs of 0.75 (0.72-0.77) and 0.85 (0.82-0.89) in the CHARLS and ELSA cohorts, respectively. Calibration curves demonstrated adequate model fit, while decision curves indicated a substantial net benefit. CONCLUSION: Reduced eGDR is an independent risk factor for the development of diabetes mellitus in middle-aged and elderly adults, and is linearly and negatively correlated with the risk of diabetes mellitus.