Abstract 19462: Improvement of 10-Year Cardiovascular Risk Prediction Using Hemoglobin A1c in Diabetic Women Compared to Classification of Diabetes as a Cardiovascular Risk Equivalent
Background: Current guidelines suggest that diabetes should be treated as a cardiovascular risk equivalent (20% 10 year risk of an event) for the purposes of risk classification. However, this classification system has not been compared to scores which include a term for hemoglobin A1c (HbA1c) levels in diabetics.
Methods: Overall prediction scores were developed in the 24,674 initially healthy participants of the Women's Health Study and a median of 10.18 years of follow-up for cardiovascular disease. Standard risk models were based on the Framingham covariates (age, systolic blood pressure, smoking, and total and HDL cholesterol). The 685 women with baseline diabetes were assigned to the higher of 1) their predicted 10-year risk from the model or 2) 20% risk. Extended risk models, also generated in the entire study, included an additional term for HbA1c in the diabetic women only. The resulting predicted risks for the standard and extended models were compared in the diabetic women using discrimination and reclassification measures.
Results: There were 125 cardiovascular events in the diabetic women over the study follow-up. All models were well calibrated in the whole cohort, and the predicted 10-year cardiovascular risk in the diabetic women using the Framingham covariates plus HbA1c ranged from 0.5% to 78.6%, with 71.5% of the women having a predicted 10-year risk below 20%. In comparison to the standard Framingham covariates with diabetes as a cardiovascular risk equivalent, the addition of HbA1c levels resulted in improved discrimination, as measured by the c-statistic (0.692 vs 0.631, p < 0.001). Using HbA1c also improved reclassification measures, with a 26% Net Reclassification Improvement (p =0.001) and an Integrated Discrimination Improvement of 0.032 (p < 0.001). Use of HbA1c also showed improvement in reclassification compared to a dichotomous term for diabetes (AUC 0.680 vs 0.692, p=297; NRI 11.5%, p=0.037; IDI 0.002, p=0.68)
Conclusions: Incorporation of HbA1c levels suggests better risk classification for diabetic women than classification as a cardiovascular risk equivalent.
- © 2010 by American Heart Association, Inc.