The discordance rate, a new concept for combining diagnostic decisions with analytical performance characteristics. 2. Defining analytical goals applied to the diagnosis of type 2 diabetes by blood glucose concentrations.

A novel statistical approach has recently been described to estimate diagnostic "errors" in method comparison studies. This procedure requires a population-based probability, pp, which describes the spread of values within the population under study, and an analytical probability, Pa, quantifying the risk of errors at a decision limit due to replacing one method by the other. The population probability was derived from 207 subjects who were submitted to an oral glucose tolerance test due to suspicion of type 2 diabetes. The new concept was then applied in a reverse mode by determining the analytical variability of glucose concentrations for a fixed discordance rate diagnosing type 2 diabetes. If a combined discordance rate (sum of positive and negative discordances) of 5% is allowed, a maximal imprecision of 3.7% can be tolerated in the absence of bias. In the presence of a 3.0% bias, the allowable imprecision must be reduced to 2.8%. The relationship between bias and imprecision followed a complex function and not a simple linear model. These allowable limits were achieved with venous plasma in the fasting state. The allowable analytical specifications were slightly more stringent with capillary blood. After a 2 h glucose challenge, higher error rates could be tolerated, indicating that post-load glucose concentrations have a higher diagnostic efficiency than fasting levels. The new concept has the advantage that it is derived from patient's samples in relation to diagnostic requirements.