These studies are used by groups
intending to produce analytical methods to be used in a broader environment,
e.g., international standards for a commercial interface, or methods approved by
regulatory bodies, official methods produced by reference committees, or other
types of international organisations.
One of the basic aspects of this kind of experiment is that all participants use the same analytical method during the standardisation phases of their work. Differently from other applications, the final results desired are the precision curves that represent the performance of the method in terms of repeatability and reproducibility.
When the methods approved are to be used as official methods, or as reference methods, it is important that they be compared to existing methods, in terms of precision and trueness (or relative bias).
InterLab2.0 provides estimates of repeatability and reproducibility, following the algorithms in ISO 5725-2 and ISO/TC 102/TCR 5. Further, it produces the curves obtained by the regression of standard deviations against concentrations, using the mathematical models described in ISO 5725-2, making it easy for the group convener or statistician to choose the best fit for the data at hand.
After the curves are approved, they can be plotted together with the Scheffeé-Hotelling confidence hyperbolas for the regression lines. This provides a graphical test for the comparison of different methods used for analysing the same analytes. A plot of the precision curve (just as an example, the repeatability standard deviation against concentrations) with its confidence hyperbola, may be overlaid by another precision curve obtained from another analytical method. When the area defined inside the hyperbolas for the two methods coincide for a reasonable extent, there is no evidence that the two methods being compared produce results with different precision.
If the samples used in the standardisation experiments are certified reference materials, InterLab2.0 uses the certified values to compare the means obtained by the analytical method being evaluated with those references, using the algorithm in ISO 5725-4. It also calculates if the size of the experiment is good enough for taking conclusions. If necessary, it indicates the necessity of using additional participants, or doing more replicate determinations.
Additionally, the same comparison may be accomplished by means of regression analysis with formal hypothesis testing for evaluating all samples together. The approach used by InterLab2.0 is that published by Riu and Rius (1996), with a simultaneous test for the significance of both the fixed bias and the proportional bias. This comparison method is equivalent to the one proposed by Ripley and Thompson (1987), but the hypothesis test gives better significance levels than the isolated tests proposed by the latter authors. The approach in Galea-Rojas et al. (2003) is being implemented now, as it was included in the new version of TCR5. The complete reference for these papers is given here.
The same regression analysis structure provided by InterLab2.0 is taken advantage of to provide comparisons between two participants in the same test, and for comparing two different methods, when they are evaluated in the same samples (that may be certified reference materials or not).
With these functionalities at hand, it is easy to adapt the use of InterLab2.0 for internal validation of methods in a single laboratory. A new method may be tested against an old one, with known performance and already validated and approved by the laboratory. Comparison of precision, as well as deriving estimates for the relative biases (fixed or proportional) between both, and their statistical significance, may be provided with the same ease as when approving international methods.
All this bearing in mind that the procedures used are very well accepted worldwide by the technical community involved in laboratory accreditation, quality assurance and standards development.
Questions or comments about this website, please contact us at: