We calculated critical differences of fasting plasma glucose and HbA1c and compared them with differences between clinical decision limits for diagnosis (WHO recommendation) and therapy (ADA recommendation).We derived from obtained data the required precision and bias for both measurements.For a reliable discrimination of glucose decision limits (6,1 and 7,0 mmol/l) there is required precision 3%, if maximal bias is 2% or precision 2%, if maximal bias is 2,5% or 4% if bias is near to zero. After calculation of combined uncertainty of decision limits with using introduced precision and bias values we ascertained no coverage of their values for 68% confidence interval but not for 95% confidence interval. We also applied the same procedure on the HbA1c measurement. Here we ascertained necessity of precision < 3%, while bias may be maximally 2%. Precision 5% serving as certification criterion for manufacturers is not sufficient even if bias has zero value. The same conclusion like for glucose uncertainty is valid for HbA1c. No coverage of both decision limits (7 and 8%) can be observed for 68% confidence interval only, but not yet if 95% confidence interval is considered.

        Key words: glucose, HbA1c, critical differences, decision limits, uncertainty.