Prentice JC, Health Care Financing and Economics, VA Boston Helathcare System; Pizer SD, Health Care Financing and Economics, VA Boston Healthcare System; Conlin PR, VA Boston Healthcare System;
Objectives:
Hemoglobin A1c (A1c) levels have an inconsistent relationship with long-term health outcomes. Emerging evidence in both cellular and clinical studies imply an independent effect of A1c variability on these outcomes. This study is the first to use nationwide data to characterize the relationship between A1c variability and health outcomes.
Methods:
This retrospective study combined VA and Medicare utilization claims for veterans diagnosed with Type 2 diabetes (n = 50,861). The study population included veterans taking metformin who then initiated a second type of diabetes medication (e.g. sulfonylureas). Patients could enter the sample starting January 1, 2003 and were followed until they experienced their first outcome or December 31, 2010. A1c variability, categorized into quartiles, was measured during a three-year baseline period. Specific measures included standard deviation, coefficient of variation and an adjusted standard deviation that accounted for the number and average time between A1c tests. Cox proportional hazard models predicted mortality, ambulatory-care sensitive condition (ACSC) hospitalization, and acute myocardial infraction (AMI) or stroke. Control variables included demographics, average A1c levels, body mass index, serum creatinine, blood pressure, cholesterol, type of second drug initiated, diabetes severity, comorbidities (physical and mental) and provider process quality.
Results:
For all measures, greater A1c variability significantly predicted each outcome. For example, using the adjusted standard deviation measure, the hazard ratio when predicting mortality in the third quartile is 1.18 (95% confidence interval (C.I.) 1.08, 1.29) and 1.49 in the fourth quartile (95% C.I. 1,35, 1.63). These numbers are 1.12 (95% C.I. 1.05, 1.19) and 1.13 (95% C.I. 1.05, 1.21) for ACSC hospitalization and 1.26 (95% C.I. 1.10, 1.43) for AMI or stroke. Higher baseline A1c levels independently predict the likelihood of experiencing each outcome
Implications:
In a nationwide sample, A1c variability is associated with an increased risk of mortality, ACSC hospitalization and AMI or stroke even when controlling for A1c levels.
Impacts:
Type 2 diabetes complications may be delayed or prevented if A1c variability is minimized. Future research should use quasi-experimental research designs to determine if the relationship between A1c variability and health outcomes is causal. When managing Type 2 diabetes, providers should pay attention to A1c variability and A1c levels.