Accuracy Profile: A Specific Tool For Validation Diagnostic.
July 8 -12, 2012, Noorwijkerhout
J. Manceau, M. Saint-Cyr, J.G. Rolland, A. Perrin-Guyomard, M. Laurentie Anses, Fougeres Laboratory, Pharmacokinetics-Pharmacodynamics unit, (
[email protected])
BP90203, 35302 Fougères cedex, France Introduction Risk analysis assessment using toxicokinetics data needs to know concentrations with statistical confidence. Therefore, validation of analytical method was provided to document the performance of the method and verify adequation with domain of validation expected. However, several guidelines or regulations are used and some misuses are performed. For example, accuracy is used instead of trueness, concentrations under the limit of quantification (LOQ) or detection (LOD) are used to modelise drug elimination. The goal of the validation of analytical method is to prove that the method is adapted to fit the purpose. Performance criteria of the method must be assessed globally and not parameter by parameter. The main parameter is accuracy that is a combination of the trueness and the precision (repeatability, intermediate precision). Determination of two-sided 95 % confidence intervals for the total error (Hubert et al, 1999) of validation samples (accuracy profile) in the acceptance/rejection of an analytical method is presented and applied to the validation of the analytical method to allow the quantification of deoxynivalenol (mycotoxin) in pig plasma.
Material and Methods To establish the calibration curve or the accuracy profile the experimental design was: extraction and analysis of 3 series of 5 concentration levels (25-50-75-125-250 ng/ml) with 3 replicates were carried out on spiked pig plasma. Solid phase extraction was applied from plasma on Oasis ®HLB cartridge. Analytical separation was performed by HPLC on a RP18e Lichrospher® column with UV detection at 219 nm under gradient conditions
Results
Bias (%)
The accuracy profile of deoxynivalenol quantification in pig’s plasma is shown in figure. The lower LOQ was established at 25 ng/ml and the upper LOQ was 250 ng/ml. The accuracy profile is an agreement with the definition of LOQ where the accuracy was estimated. In this case, the accuracy profile allows to establish to LOQs with an accuracy less than 20 % at 25 ng/ml and less than 10 % at 250 ng/ml. Accuracy profile shows an exponential propagation of error for low concentrations and consequently the method was not adapted for concentration under 25 ng/ml. The error was associated to trueness error and to precision error both. The limit of detection was LOQ/3 (8.3 ng/ml). At this value, accuracy was estimated Concentration (ng/ml)
to close 40 %.
Figure : accuracy profile of deoxynivalenol in pig’s plasma. The red line is the bias (trueness) of the method, the blue line the 95 % tolerance interval around the bias i.e. the precision of the method; the green points are the raw data. The dashed line is the validation area defined by acceptability criteria, here ± 20 %
Conclusions Accuracy profile is a tool to investigate different source of error (trueness, precision) and also that allows to optimize the results obtained during the assessment of performance criteria when they are compared to acceptability criteria. The accuracy profile shows also that using values under LOQ or LOD are not recommended because conclusions of risk analysis will be wrong. Hubert Ph, Chiap P, Crommen J, Boulanger B, Chapuzet E, Mercier N, Bervoas-Martin S, Chevalier P, Granjean D, Lagorce P, Lallier M, Laparra Mc, Laurentie M, Nivet Jc: The SFSTP guide on the validation of chromatographic methods for drug bioanalysis: from the Washington Conference to the laboratory. Anal. Chim. Acta, 391, 135-148, 1999 Valenta H., Dänicke S. and Döll S. Analysis of deoxynivalenol and de-epoxy-deoxynivalenol in animal tissues by liquid chromatography after clean-up with an immunoaffinity column . Mycotoxin Research, Volume 19, Number 1, 51-55, 2003
