ABSTRACT
The increasing need to develop quantitative chromatographic methods with upgradable multi-targeted approach, allowing flexible and reliable application on large daily workload makes the implementation of an efficient strategy of method's validation and maintenance crucial for the quality assurance policy. The expounding case of a gas chromatographic-mass spectrometric method for the urinary endogenous steroid profiling is presented to illustrate a validation strategy that combines rigorous estimation of validation parameters with highly efficient use of the collected data. The analysis of blank urine samples fortified at six concentration levels with 18 targeted steroids was replicated nine times in three working sessions along twelve days. This dataset of 54 analysis formed the groundwork on which the statistical evaluation of several validation parameters was founded, including calibration, intra- and inter-day accuracy and precision, limit of detection (LOD), limit of quantification, ion abundance ratio repeatability, selectivity, specificity, and carry-over. The preliminary comparison of the response variances at different concentration levels provided the evaluation for heteroscedasticity. Then, the most appropriate calibration model was determined for each steroid, in terms of order (linear vs. quadratic) and weighting, allowing to complete their quantitation in each solution. Intra- and inter-day accuracy and precision were calculated therefrom. LOD values were computed with the Hubaux-Vos method from the weighted linear segment of the calibration curves. Only the assessment of recovery and ionization suppression/enhancement required the execution of further independent experiments. The case study demonstrated that the application of adequate statistical testing typically produced non-homogeneous models of calibration curves, mostly arising from heteroscedastic and quadratic distribution of datasets, unlike what is reported in overly simplified approaches. The misleading information obtained from the regression coefficient R2 to evaluate linearity was evidenced. The strong dependence of calculated LOD and accuracy from the selected calibration parameters was highlighted, making the implementation of an adequate calibration maintenance policy highly advisable.
