Qualitative threshold method validation and uncertainty evaluation: A theoretical framework and application to a 40 analytes liquid chromatography–tandem mass spectrometry method

Qualitative methods hold an important place in drug testing, filling central needs in screening and analyses, among others, linked to per se legislation. Nevertheless, the bioanalytical method validation guidelines do not discuss this type of method or describe method validation procedures ill‐adapted to qualitative methods. The output of qualitative methods are typically categorical, binary results, such as presence/absence or above cut‐off/below cut‐off. As the goal of any method validation is to demonstrate fitness for use under production conditions, qualitative validation guidelines should evaluate performance based on discrete, binary results instead of the continuous measurements obtained from the instrument (e.g. area). A tentative validation guideline for threshold qualitative methods was developed by in silico modelling of measurements and derived binary results. This preliminary guideline was applied to a liquid chromatography–tandem mass spectrometry method for 40 analytes, each with a defined threshold concentration. Validation parameters calculated from the analysis of 30 samples spiked above and below the threshold concentration (false negative rate, false positive rate, selectivity rate, sensitivity rate and reliability rate) showed a surprisingly high failure rate. Overall, 13 out of the 40 analytes were not considered validated. A subsequent examination found that this was attributable to an appreciable shift in the standard deviation of the area ratio on a day‐to‐day basis, a previously undescribed and unaccounted‐for behaviour in the qualitative threshold method validation literature. Consequently, the developed guideline was modified and used to validate a qualitative threshold method, based on the binary results for performance evaluation and incorporating measurement uncertainty. Validation guidelines for qualitative threshold methods were developed and applied. They evaluated the performance of a method using the binary ‘above/below threshold’ results produced and incorporating the uncertainty of measurement under the form of an unreliability zone surrounding the concentration threshold.