Recently my colleagues at the Thermo Fishers Scientific Customer Solution Centre in Beijing encountered some challenges with the validation of new methods for the quantitation of polar pesticides in food. These challenges were outlined in a Separation Science webinar titled ‘A New Integrated, Sample-to-Result, Analytical Workflow for the Sensitive and Reliable Analysis of Polar Anionic Pesticides and Metabolites.’ Essentially, the response of an individual analytes was so variable between samples, even of the same type, accurate calibration in the absence of Isotopically Labelled Internal Standards (ILIS) was difficult.
The difficulties, mediated by matrix co-extractives prompted us to question if the current approaches to validation always truly reflect the performance of the method, or perhaps overestimate the performance of the method in routine analysis? Afterall the purpose of validation is to ensure validity and inter-laboratory equivalence of methods.
Method Validation Guidelines
The current EU SANTE 11813/2017 Guidance document on analytical quality control and method validation procedures for pesticide residues and analysis in food and feeds recommend procedures for quantitative and screening methods. For a quantitative method laboratories should obtain a bulk ‘blank’ homogenised sample and then conduct recovery experiments by spiking sub-samples before extraction; a minimum of five replicate sub-samples of the blank with all pesticides of interest at two different concentrations [five spiked samples at the Limit of Quantitation (LOQ) and five spiked samples at 2-10x the target LoQ]. The concentration of pesticides recovered in the final extract are quantified against matrix-matched standards, often prepared by spiking aliquots of extracts of sub-samples from the same matrix blank as used for recovery spikes. This is convenient because it can be costly to identify suitably blank samples, but does this approach sometimes mask matrix effects that occur when the blank matrix is used to quantify residues in different samples, even if of the same type? There are so many variables: the sample composition, the chemical properties of pesticides, the choice of extraction solvent, the pH during extraction, the clean-up, etc. Is it even possible or necessary to find a solution?
The issue of matrix effects are well documented and the easiest solution to reduce matrix effects by dilution of the extract, but this can be limited depending on the sensitivity of the instrument. The SANTE 11813/2017 guidelines permit the use of procedural standards, isotopically labelled Internal Standards (ILIS) and standard addition to correct low recovery in challenging cases, such as the analysis of polar pesticides.
The Analysis of Polar Pesticides in Wheat Flour
The optimised sample-to-results workflow is based on the EURL Quick Polar Pesticides extraction (QuPPE) method followed SPE clean-up and determination of diluted samples using high performance ion chromatography- triple quadrupole mass spectrometry. We wanted to validate the recovery, precision, limits of quantification and robustness of at low levels in a range of matrices, but for brevity of explanation I will focus on the results for wheat.
Calculation against matrix-matched calibration (sample matrix –spiked after extraction and clean-up) gave low recoveries for , Glyphosate, AMPA and N-acetyl AMPA, of 40%, 65% and 60% respectively, and with good precision ( RSDs of 5%, 7% and 3% ) . Not a surprising result since it is known that these compounds bind to the matrix. Use of ILIS corrected the recoveries to close to 100%, but in China, where the experiments were conducted, ILIS were not available for all of the compounds. The use of procedural standard calibration corrected accuracy and precision to the acceptance criteria, (mean recovery within 70-120%, RSD≤ 20%) stated in the SANTE guidelines, but only if the recovery spikes and procedural standards were prepared using the exact same sample matrix. When this approach was extended to analyse further samples the results were more variable for more analytes simply because of difference in matrix effects caused by different wheat flours from different sources. If we had accepted the validation based on a single matrix, and analysed samples containing incurred residues, alongside quality control samples (prepared using the same bulk sample blank), we would have inevitably reported some incorrect results. The validation would have provided a false sense of confidence. Laboratories with access to the ILISs will generally be able to obtain good data, but for laboratories with no access, the best option is the use of standard addition where each sample is analysed with and without spiking at an appropriate level, accepting the compromise that sample throughput will decrease.
Like what you are learning?
Interestingly the EU SANTE 11813/2017 requirements for validation of a screening method, as opposed to a quantitation method, for pesticides are different. The requirement is to analyse a minimum of 20 different samples at least twice. Once as non-spiked and then spiked at the estimated screening detection limit to determine if the response is sufficient to statistically calculate the screening detection limit. Why the difference? Well, to establish confidence for detection of the analyte at a specified level.
It is important to be aware of the pitfalls to avoid being drawn into a false sense of security when validating quantitative methods for the analysis of pesticides, and especially polar pesticides in complex matrices. Especially dry commodities with high levels of matrix co-extractives, such as cereals, spices, tea and samples containing different combinations of ingredients, such as processed baby foods, etc. A validation based on the use of different samples for the recovery experiments is more robust than a validation based on a single sample, but more time consuming. Inevitably there is a compromise between the level of accuracy and precision required the availability of ILIS, costs, and available analyst resources.
If you have experienced similar issues or have thoughts on different approaches to validation then please let us know via the comments.
If you are also interested in the analysis of GC-MS amenable pesticides at ultra-low levels in food and the possibility of diluting sample extracts to reduce matrix effects and increase robustness high sensitivity Thermo Scientific™ TSQ™ 9000 GC-MS/MS system with the Advanced Electron Ionisation (AEI) Sourcethen please:
Read the application note titled: ‘Ultra low level quantification of pesticides in baby foods using an advanced triple quadrupole GC-MS/MS system’