Blood HbA1c determination is a powerful tool for the evaluation and management of patients with diabetes mellitus. Many HbA1c analytical methods demonstrate bias in samples from patients with hemoglobinopathies. This study evaluated the analytical performance of Roche Diagnostics' 1st and 2nd generation HbA1c assays in patients with or without hemoglobinopathies whose HbA1c levels were elevated or normal, respectively. Boronate-affinity high performance liquid chromatography (HPLC) served as the reference method. Whole blood samples were collected from 80 patients with HbS or HbC whose group mean HbA1c value was elevated and also from 80 patients without hemoglobinopathy whose HbA1c values were in the well-controlled range. Each sample was assayed for HbA1c by the Primus boronate-affinity HPLC technique and by Roche's 1st and 2nd generation immunoassays using a Cobas Integra 800 analytical system. Results by the HPLC technique were compared with the results of both Roche assays by linear regression and Bland-Altman analysis. The 1st and 2nd generation assays yielded regression lines and correlation values vs HPLC assay of y = 1.43x - 1.59; R(2) = 0.83, and y = 0.94x + 0.10; R(2) = 0.92, respectively, in the 80 patients with hemoglobinopathies. The mean difference and the +/-2SD range were greater in the 1st than in the 2nd generation assay (2.68, +/-2.07 vs -0.54, +/-0.86, respectively). The 2nd generation assay also showed better performance than the 1st generation assay in samples from the 80 patients without hemoglobinopathy. In conclusion, this study validates the accuracy of Roche's 2nd generation assay, which is substantially improved over Roche's 1st generation HbA1c assay.