Improved diabetes control in the elderly delays global cognitive decline

Abstract

Objectives: To examine whether improved diabetes control is related to better cognitive outcomes.

Design: Randomized control trial.

Setting: A randomized trial of telemedicine vs. usual care in elderly persons with type 2 diabetes.

Participants: Participants were 2169 persons 55 years and older with type 2 diabetes from New York City and Upstate New York.

Intervention: The diabetes case management intervention was implemented by a diabetes nurse, via a telemedicine unit in the participant's home, and in coordination with the primary care physician.

Measurements: Hemoglobin A1c (HbA1c), systolic blood pressure (SBP), and low density lipoprotein cholesterol (LDL), were measured at a baseline visit and at up to 5 annual follow-up visits. Global cognition was measured at those visits with the Comprehensive Assessment and Referral Evaluation (CARE).

Result: In mixed models the intervention was related to slower global cognitive decline in the intervention group (p = 0.01). Improvements in HbA1c (p = 0.03), but not SBP or LDL, mediated the effect of the intervention on cognitive decline.

Conclusion: Improved diabetes control in the elderly following existing guidelines through a telemedicine intervention was associated with less global cognitive decline. The main mediator of this effect seemed to be improvements in HbA1c.

Figure 1

Comparison of changes in adjusted means in the Comprehensive Assessment and Referral Evaluation (CARE)-Diagnostic Scale between the intervention and control from mixed models (adjusted for clustering) during a maximum follow-up of 5 years.

Figure 2

Path diagram depicting the direct and indirect effects of the telemedicine intervention through Hemoglobin A1c on cognition (n = 2169)^a. ^a SAS PROC Mixed with a compound symmetry covariance structure was used in the cognition analyses with an adjustment for clustering within PCP. HgA1c was treated as a time-varying covariate in the cognition analyses. SAS PROC Mixed was used to predict HgA1c and included adjustments for clustering and heterogeneity of variances in group and residual variances and exponential terms to model the nonlinear distribution of HgA1c over time. A first order auto-regressive covariance structure was used for the HgA1c analyses. Up to six waves of data (baseline plus five follow-up) were included in the analyses.