INTRODUCTION: Hemoglobin A1c (HbA1c) measurement in epidemiology studies could be increased if reliability of measurements in frozen stored samples was known. In the Reasons for Geographic And Racial Differences in Stroke, a longitudinal study of 30,239 Black and White U.S. adults, we investigated reliability of HbA1c measurements for two types of samples stored at -80°C for up to 14 years. METHODS: Among 917 participants without diabetes, HbA1c was measured in 2017 in frozen packed cells from the first visit (2003-2007) and in frozen whole blood samples from the second visit (2013-2016). To study reliability, associations between HbA1c and glycemia-related characteristics were examined. RESULTS: Each 10 mg/dL greater fasting glucose was associated with 0.08% (95% CI: 0.05%-0.11%) greater HbA1c in frozen packed cells (Visit 1) and 0.10% (95% CI: 0.08%-0.12%) greater HbA1c in whole blood (Visit 2). HbA1c was also similarly higher with both methods with increasing age, gender, systolic blood pressure, body mass index, high-density lipoprotein, triglycerides, C-reactive protein, and hemoglobin. Using both methods, ≤ 3.5% would be classified with diabetes based on HbA1c ≥ 6.5%. CONCLUSIONS: In REGARDS participants without diabetes, HbA1c measurement appeared reliable in frozen packed cells or whole blood under long-term storage, suggesting acceptability for study of the epidemiology of HbA1c.