BACKGROUND: Determining the stability of stored samples for assays that were not available at the time of original collection is problematic. To assess sample stability for a relatively new assay of glycated albumin (GA), we first measured GA in fresh samples and in samples stored for 19-23 years. We then compared the regression of the contemporaneous glycohemoglobin (Hb A 1c) values against the GA results from fresh vs stored samples, reasoning that similar slopes and intercepts would provide strong, albeit indirect, support for the stability of the stored samples for GA measurements. METHODS: We assayed 90 samples frozen for 19-23 years and 90 fresh samples from participants in the Diabetes Control and Complications trial cohort for GA. Hb A1c was measured contemporaneously in fresh samples at each time period. A single normal-errors linear model regressed the Hb A1c values on the GA, with an additional effect for collection period (fresh vs stored for GA) and the interaction of period and GA. RESULTS: Analysis of the regressions lines between GA and Hb A1c revealed intercepts (3.69 and 2.97 for the fresh and stored samples, respectively) and slopes (0.198 vs 0.187) that were not significantly different (P = 0.182 and P = 0.639, respectively). CONCLUSIONS: This simple approach can be used to assess the stability of stored samples in new assays. Samples stored for as long as 23 years are suitable for theGA assay.