In the new lowered diagnostic discriminator for diabetes mellitus (DM) from the American Diabetes Association (ADA), fasting peripheral venous plasma glucose (f-vPG) of 7.0 mmol/l is identical to the 99.9 centile of f-vPG (7.05 mmol/l, 95%CI: 6.91-7.20 mmol/l) in a low-risk reference population. We investigated its diagnostic concordance with other diagnostic discriminators. As no index test is available for DM we used the ADA discriminator as gold standard. We isolated a low-risk reference population (n = 424) from a randomised general population (n = 726) by ruling out of all cases with clinical and biochemical risk indicators for DM. We based our analysis on measurements traceable to primary standard concentration, a bias of < 1.5% and CV% < 2.5. The distribution of the fasting capillary whole blood glucose (f-CBG; mmol/l) in the reference population was in Gaussian with the 99.9 centile of 6.62 mmol/l (95% CI 6.47-6.77 mmol/l) and the 97.5 centile of 5.92 mmol/l (5.82-6.02 mmol/l). The 6.1 mmol/l f-CBG WHO limit corresponds approximately to the 97.6 centile, and this limit is thus not traceable to the ADA discriminator, which corresponds to f-CBG of 6.4 mmol/l. This is the case in groups only, as recalculation will introduce unpredictable errors. Thus, in our general population a varying number of subjects will be at risk of DM as a mere consequence of different limits. The f-CBG limit of 6.1 mmol/l will thus lead to 2.4% false-positive diagnoses or, in EU, to around 44 x 10(6) adults being diagnosed. The number of cases at risk of DM vary from 5.4 x 10(6) to 44 x 10(6) in EU. We conclude that application of different diagnostic limits results in highly variable number of diagnosed DM cases, and therefore one diagnostic discriminator is needed to provide reproducible diagnoses.