Aims/hypothesis Glycated albumin is a measure of the mean plasma glucose concentration over approximately 2-3 weeks. We determined reference values for glycated albumin, and assessed its utility for the diagnosis of type 2 diabetes mellitus in the general population. Methods We studied 1,575 men and women (mean age, 49.9 years; range, 26-78 years) who participated in a periodic health examination in a suburban Japanese town. HbA 1c and fasting plasma concentrations of glucose (FPG) and glycated albumin were measured. Participants with FPG ≥7.0 mmol/l or HbA 1c ≥6.5% (48 mmol/mol) were diagnosed as having diabetes. In our laboratory, the glycated albumin assay had intra-assay and inter-assay CVs of 1.1% and 1.6%, respectively. Results Glycated albumin levels were significantly correlated with HbA 1c levels (r=0.766, p<0.001) and FPG (r= 0.706, p<0.001). The presence of diabetes was significantly higher in participants with glycated albumin levels between 15.0% and 15.9% (five of 276, 1.81%) than in those with glycated albumin <14% (three of 672, 0.45%) (p=0.037), and was markedly increased in those with a glycated albumin level >16% (58 of 207, 28.0%). Receiver operating characteristic curve analysis indicated that a glycated albumin level of ≥15.5% was optimal for predicting diabetes, with a sensitivity of 83.3% and a specificity of 83.3%. Conclusions/interpretation There is merit to further investigating the potential for glycated albumin to be used as an alternative measure of dysglycaemia for future research and clinical practice.