Dynamic transitions from prediabetes to diabetes and incident cardiovascular events in middle-aged and older chinese adults: a nationwide longitudinal study.

BACKGROUND: Prediabetes is a heterogeneous metabolic state that may revert to normoglycemia, persist, or progress to type 2 diabetes mellitus (T2DM). However, the cardiovascular implications of different prediabetes transition patterns and the ability to predict diabetes progression among middle-aged and older Chinese adults remain unclear. We aimed to characterize glycemic transitions, examine their associations with incident cardiovascular events, and develop prediction models for progression from prediabetes to T2DM.

METHODS: This longitudinal cohort study used data from the China Health and Retirement Longitudinal Study. Glycemic status was defined according to American Diabetes Association criteria using fasting plasma glucose, HbA1c, self-reported diabetes, and medication use. Four-year glycemic transitions from Wave 1 to Wave 3 were assessed among 7,927 participants. Among participants with baseline prediabetes, multinomial logistic regression was used to identify factors associated with reversion to normal glucose tolerance or progression to T2DM. A Wave 3 landmark design was applied to examine subsequent incident heart disease and stroke in separate outcome-specific cohorts. Prediction models for 4-year T2DM progression were developed and internally validated using a stratified 70:30 training-test split; no external validation was performed.

RESULTS: Among 3,313 participants with baseline prediabetes, 696 (21.01%) reverted to normal glucose tolerance, 2,007 (60.58%) remained in prediabetes, and 610 (18.41%) progressed to T2DM over 4 years. Higher BMI, fasting glucose, HbA1c, dyslipidemia, smoking, and C-reactive protein, together with lower grip strength, were associated with higher odds of progression. Older age, male sex, higher fasting glucose, and higher HbA1c were associated with lower odds of reversion. Compared with stable normal glucose tolerance, prediabetes progression was associated with higher odds of incident heart disease (adjusted odds ratio [aOR], 1.53; 95% CI, 1.13-2.09), whereas the association with incident stroke was not statistically significant (aOR, 1.29; 95% CI, 0.90-1.87). Stable T2DM was associated with higher odds of both incident heart disease (aOR, 1.36; 95% CI, 1.04-1.79) and stroke (aOR, 1.44; 95% CI, 1.05-1.97). The full-variable logistic regression model achieved the highest discrimination for T2DM progression (AUC, 0.715; 95% CI, 0.674-0.754), modestly outperforming the clinical baseline model (AUC, 0.699; 95% CI, 0.656-0.739) and the evaluated tree-based models.

CONCLUSIONS: Prediabetes showed dynamic 4-year transitions within the analytic cohort. Progression from prediabetes to T2DM was associated with higher subsequent odds of incident heart disease, whereas its adjusted association with incident stroke was not statistically significant. The full-variable logistic regression model showed moderate discrimination and only limited incremental improvement over the clinical baseline model; it should therefore be considered an exploratory risk-stratification tool requiring external validation rather than a clinically deployable prediction model.