The Guide to expression of uncertainty in measurement approach for estimating uncertainty: an appraisal.

BACKGROUND: The aim of the Guide to Expression of Uncertainty in Measurement (GUM) is to harmonize the different practices for estimating and reporting uncertainty of measurement. Although there are clear advantages in having a common approach for evaluating uncertainty, application of the GUM approach to chemistry measurements is not straightforward. In the above commentary, Krouwer suggests that the GUM approach should not be applied to diagnostic assays, because (a) the quality of diagnostic assays is to low, and (b) the GUM uncertainty intervals are too narrow to predict the outliers that occasionally trouble these methods.
METHODS: Some of the examples presented by Krouwer are reviewed. Sodium measurements are modeled mathematically to illustrate the GUM approach to uncertainty. A standardized uncertainty evaluation process is presented.
RESULTS: Modeling of sodium measurements demonstrates how the GUM uncertainty interval reflects the treatment of a bias: The width of the uncertainty interval varied depending on whether a correction for a calibrator lot bias was applied, but in both cases it was consistent with the distribution of measurement results. Expanding the uncertainty interval to include outliers runs counter to the definition of uncertainty. Used appropriately, the GUM uncertainty can be helpful in detecting outliers. In standardizing the uncertainty evaluation, the importance of the analytical imprecision and traceability was emphasized. It is problematic that manufacturers of commercial assays rarely inform about the uncertainty of the values assigned to the calibrators. As demonstrated by an example, external quality-assurance data may be used to estimate this uncertainty.
CONCLUSIONS: The GUM uncertainty should be applied to measurements in laboratory medicine because it may actually support the forces that drive the work on improving the quality of measurement procedures. However, it is important that the GUM approach is made more manageable by standardizing the uncertainty evaluation procedure as much as possible. It is essential to focus on the traceability and uncertainty of calibrators and reagents supplied by manufacturers of assays. Information about uncertainty is necessary in the evaluation of the uncertainty associated with manufacturers' measurement procedures, and in general it may force manufacturers to increase their efforts in improving the metrologic and analytical quality of their products.