The applicability of point-of-care testing (POCT) glucometers for monitoring blood glucose concentrations has been demonstrated. However, their use in diagnosing type 2 diabetes is still debated. Therefore, a new statistical procedure for estimating discordance rates (DRs) was applied in comparing a well-established laboratory method (Ebio) with another laboratory method (Cobas Integra 700) and with several POCT glucometers (Accu-Chek, Accutrend, Elite, HemoCue, Omni) in detecting glucose intolerance states. All procedures led to parallel glucose concentration patterns in capillary blood, venous plasma, and venous blood during oral glucose tolerance tests. However, the mean concentrations differed more or less. The Ebio and Integra results agreed within a maximal deviation of 3%. In blood samples, the HemoCue and Accutrend results were closest to the laboratory procedures (Ebio and Integra) and the highest differences were obtained with the Elite. Comparing whole blood values with those obtained in the aqueous blood compartment (Omni), even greater differences were observed. When all procedures were referred to the same glucose standard, the Ebio, Integra, Accutrend, and Omni results remained almost unchanged, whereas the Elite "moved" toward the Ebio results, and the Accu-Chek results toward the Omni results. Thus, traceability to an aqueous standard was observed with the Ebio, Integra, Accutrend, and Elite in all three sample systems. The Accu-Chek was only traceable in the presence of albumin, and HemoCue was not traceable at all. The clinical relevance of the differences observed between Ebio and POCT glucometers was tested by comparing the relative number of discordant classifications. The highest DRs were observed in the fasting state. They were higher in capillary blood than in the other sample systems. The DRs were found higher with POCT glucometers than with the other established laboratory procedure (Integra). Thus, at least in the fasting state, all POCT glucometers were less reliable than the established laboratory procedures and above the chosen criteria of clinical acceptability (DR < or = 5%). After transforming all glucometer results with a regression function (bias correction), the DRs were less than 5% if compared with the Ebio procedure in all sample systems. In conclusion, the WHO recommendation not to use POCT glucometers for diagnosing type 2 diabetes must be supported. However, after proper recalibration, the tested systems were acceptable. Therefore, manufacturers should reconsider their calibration procedure. Those POCT procedures should be preferred that can be referred to aqueous glucose solutions.