Wednesday, Mar 03, 2010, 1:30 PM - 2:55 PM
Hemoglobin A1c and Prediction of Diabetes Incidence in the Multi-Ethnic Study of Atherosclerosis (MESA)
Grand Ballroom B
Wednesday, Mar 03, 2010, 1:55 PM - 2:10 PM
+EPI - Diabetes/Metabolic Syndrome
Type 2 Diabetes; Hemoglobin; Biomarkers
Julie K Bower
, Pamela J Schreiner, James S Pankow, Elizabeth R Seaquist, Univ of Minnesota, Minneapolis, MN; Alain G Bertoni, Wake Forest Univ, Winston-Salem, NC; David A Bluemke, Natl Insts of Health, Bethesda, MD; Moyses Szklo, Johns Hopkins Univ, Baltimore, MD; Michael W Steffes, Univ of Minnesota, Minneapolis, MN
BACKGROUND: Hemoglobin A1c (HbA1c), a marker for chronic glucose exposure, has been recently proposed as a better measure for diabetes risk (compared to fasting glucose). Epidemiologic studies suggest that HbA1c levels in non-diabetic populations may vary between population subgroups; HbA1c as a predictor of future risk in ethnically diverse populations is not well established. The current study used logistic regression and receiver-operator characteristic (ROC) curves to assess the association between HbA1c levels and cumulative incidence of diabetes in the MESA cohort, and to evaluate potential racial/ethnic differences in the prediction of incident diabetes. METHODS: MESA is a community-based study that enrolled 6814 participants aged 45-84 years and free of clinical CVD at baseline (2000-2002) from six centers across the US. After excluding individuals with treated or untreated diabetes at exam 2 (2002-2004, when HbA1c levels were measured), 2042 European Americans, 542 Chinese Americans, 1196 African Americans, and 952 Hispanic Americans had data on HbA1c and diabetes status through the exam 4 (2005-2007). Diabetes status was defined as self-reported diagnosis, use of diabetes medications, or measured values at exam 4 (fasting glucose ≥126 mg/dL). RESULTS: The mean and standard deviation (SD) for HbA1c was 5.44±.43% and ranged from 3.5-8.6%, with race/ethnic-specific means of 5.3-5.6%. A total of 233 incident cases of diabetes were identified. The area under the ROC curve (AUC) for predicting cumulative incidence of diabetes was 0.84 for fasting glucose compared to 0.83 for HbA1c; these AUCs did not differ significantly (Mann-Whitney U-statistic=-.03, p=.12). The strength of association between HbA1c and incident diabetes differed by race/ethnic group (Wald χ2 = 8.71, 3df, p<.05); the AUCs were .87 for European Americans, .82 for Chinese Americans, .76 for African Americans, and .84 for Hispanic Americans. For every 1-SD increment in HbA1c (0.43% based on the data from all ethnic groups combined) the OR (95% CI) for incident diabetes was 3.25 for European Americans (2.77-3.80), 3.34 for Chinese Americans (2.87-3.90), 3.16 for African Americans (2.72-3.67), and 3.31 for Hispanic Americans (2.84-3.86) (after adjustment for age, clinic site, and BMI). In tests of interaction, each race/ethnic group was statistically significantly (p<0.05) different from the African American group; sex, educational attainment, smoking, alcohol use, family history of diabetes, and physical activity did not significantly alter the observed associations in this sample. CONCLUSION: In conclusion, the efficacy of HbA1c as a predictor of incident diabetes was comparable to fasting glucose, but the strength of association between HbA1c and diabetes incidence varied by race/ethnic group. Future analyses will explore potential underlying causes of these observed race/ethnic differences.
Download the Abstracts to Your Ebook Reader
The Online Abstract Submission and Invitation System
© 1996 - 2013 Coe-Truman Technologies, Inc. All rights reserved.