Glycated hemoglobin (HbA1c) has been recently adopted as a diagnostic marker of type 2 diabetes. However, its usage is currently limited to fresh blood samples. To allow retrospective HbA1c measurement in blood banks developed in large epidemic studies, here, we contribute to validate HbA1c assessment in frozen versus fresh blood samples from a cohort of diabetic/nondiabetic adult subjects. HbA1c was measured by HPLC in 237 fresh whole blood samples and on the same samples after a 12-month storage and a further 6-month-refrozen storage. Mean HbA1c ± SD in fresh, frozen, and refrozen samples was 6.9 ± 1.2, 6.6 ± 1.1, and 6.4 ± 1.0% for the Diabetes Control and Complications Trial and 52 ± 13, 49 ± 12, and 46 ± 11 mmol/mol for the International Federation of Clinical Chemistry and Laboratory Medicine reference, respectively. A significant correlation was found between fresh/frozen and fresh/refrozen (R = 0.994 and 0.993, P < 0.001) samples. HbA1c relative error ratio (%RER) between frozen/refrozen and fresh samples significantly correlated with HbA1c and depended on fresh value range, increasing in the five HbA1c classes (<6.0, 6.0-6.5, 6.5-7, 7-8, ≥8%, corresponding to <42, 42-48, 48-53, 53-64, ≥64 mmol/mol, P < 0.001). In particular, the 6.5% (48 mmol/mol) HbA1c diagnostic cutoff of fresh samples identified two classes reflecting significant differences in %RER (2.8 ± 2.0 and 3.3 ± 1.7; P < 0.05) between frozen and fresh samples. In conclusion, our results demonstrate a high correlation between data from fresh and frozen samples, with a very limited %RER between the two measurements, which increases with baseline HbA1c levels. Accordingly, when analyzing biobank frozen specimens for diagnostic purpose, the effect of the HbA1c range should be taken into account.