Abstract
This paper describes the validation of an LC-MS/MS-based method for the quantification of > 500 secondary microbial metabolites. Analytical performance parameters have been determined for seven food matrices using seven individual samples per matrix for spiking. Apparent recoveries ranged from 70 to 120% for 53-83% of all investigated analytes (depending on the matrix). This number increased to 84-94% if the recovery of extraction was considered. The comparison of the fraction of analytes for which the precision criterion of RSD ≤ 20% under repeatability conditions (for 7 replicates derived from different individual samples) and intermediate precision conditions (for 7 technical replicates from one sample), respectively, was met (85-97% vs. 93-94%) highlights the contribution of relative matrix effects to the method uncertainty. Statistical testing of apparent recoveries between pairs of matrices exhibited a significant difference for more than half of the analytes, while recoveries of the extraction showed a much better agreement. Apparent recoveries and matrix effects were found to be constant over 2-3 orders of magnitude of analyte concentrations in figs and maize, whereas the LOQs differed less than by a factor of 2 for 90% of the investigated compounds. Based on these findings, this paper discusses the applicability and practicability of current guidelines for multi-analyte method validation. Investigation of (apparent) recoveries near the LOQ seems to be insufficiently relevant to justify the enormous time-effort for manual inspection of the peaks of hundreds of analytes. Instead, more emphasis should be put on the investigation of relative matrix effects in the validation procedure. Graphical abstract.