Abstract Background With rapid population aging, cognitive decline has become a major public health challenge. Insulin resistance (IR) is linked to adverse neurocognitive outcomes, but reported effect sizes are generally modest and vulnerable to selection bias and attrition. We therefore examined whether three IR surrogates-Triglyceride-Glucose (TyG) index, Metabolic Score for Insulin Resistance (METS-IR), and estimated Glucose Disposal Rate (eGDR) were associated with longitudinal cognitive trajectories in a Chinese community cohort. Methods We conducted a longitudinal secondary analysis of 8,028 adults aged ≥ 45 years from the China Health and Retirement Longitudinal Study (CHARLS) with 9 years of follow-up. Global cognition was repeatedly assessed and standardized; group-based trajectory modeling identified four cognitive patterns, labeled as Group 1 (low baseline-progressive decline), Group 2 (persistently very low cognition), Group 3 (intermediate-high baseline with stable cognition), and Group 4 (sustained high cognition). Baseline TyG, METS-IR, and eGDR were the main exposures. Within each trajectory, we used Kaplan–Meier curves and Cox proportional hazards models to estimate hazard ratios (HRs) for incident cognitive impairment, and compared baseline characteristics of included versus excluded participants and trajectory-specific attrition as sensitivity analyses. Results Four distinct cognitive trajectories were identified. Lower eGDR values were modestly associated with more adverse cognitive trajectories, whereas TyG and METS-IR showed no statistically significant associations across trajectories. Baseline comparisons and attrition summaries revealed systematic differences between included and excluded participants and pronounced loss to follow-up in declining trajectories, suggesting potential selection and survivorship biases. Conclusion eGDR showed a consistent but small association with adverse cognitive trajectories, whereas TyG and METS-IR were largely unrelated to long-term cognitive outcomes. Given the near-null effect sizes, substantial missing IR data, and high attrition, these findings should be interpreted cautiously and appear to have limited utility for individual risk stratification. Assessment of glucose metabolism may still help contextualize population-level vulnerability, but causal and predictive inferences will require further longitudinal validation and strategies to mitigate selection bias.
Xie et al. (Mon,) studied this question.