METHOD VALIDATION The concepts of method validation have been developed simultaneously by AOAC
5.5 METHOD VALIDATION The concepts of method validation have been developed simultaneously by AOAC
International, EURACHEM, IUPAC Working Party, and several national organiza- tions. The general criteria set by the different guidelines are similar and provide the basis for assuring reliability of the methods validated for one or a few analyte –sample matrix combinations. However, these general guidelines are not directly applicable to the methods used in pesticide residue analysis as they cannot address the specific requirements and limitations. To provide guidance on in-house method validation to analysts, national authorities, and accreditation bodies, a Guideline for Single-labora- tory Method Validation was developed and discussed at an International Workshop. 36
The Guidelines were included in the GLP GLs of CCPR. 4 The Guidelines also provide specific information for extension of the method to a new analyte and=or new sample matrix, and adaptation of a fully validated method in another laboratory.
According to the Guidelines the method validation is not a one-time, but continuous operation including the performance verification during the use of the method. Information essential for the characterization of a method may be gathered during the development or adaptation of an analytical procedure; establishment of acceptable performance; regular performance verification of methods applied in the laboratory, demonstration of acceptable performance in second or third laboratory
142 Analysis of Pesticides in Food and Environmental Samples (AOAC Peer-Verified Method), and participation in proficiency test or interlabora-
tory collaborative study. Before validation of a method commences, the method must be optimized, standard operation procedure (SOP) describing the method in sufficient detail should
be prepared, and the staff performing the validation should be experienced with the method. Parameters to be studied are: stability of residues during sample storage, sample processing, and in analytical standards; efficiency of extraction; homogeneity of analyte in processed samples; selectivity of separation; specificity of analyte detection; calibration function; matrix effect; analytical range, limit of detection, limit of quantitation (LOQ), and ruggedness of the method.
The validation should be performed in case of individual methods with the specified analyte(s) and sample materials, or using sample matrices representative of those to be tested by the laboratory; group specific methods with representative commodity(ies)* (Table 5.4) and a minimum of two representative analytes y selected from the group; MRMs with representative commodities and a minimum of 10 rep- resentative analytes. For method validation purposes, commodities should be differ- entiated sufficiently but not unnecessarily. The concentration of the analytes used to characterize a method should be selected to cover the analytical ranges of all analytes. Full method validation shall be performed in all matrices and for all compounds specified, if required by relevant legislation.
The method is considered applicable for an analyte if its performance satisfies the basic requirements summarized in Table 5.5. The repeatability and reproducibil- ity criteria given in the table are based on the Horwitz equation: RSD ¼ 2C ( 0.1505) .
In the equation, the concentration C is expressed in dimensionless mass ratios (e.g.,
6 ). Recent studies indicated that the Horwitz equation would probably overestimate the variability of the results at low concentrations (<0.1 mg=kg). 37
Therefore, the tabulated data should be considered as the upper limit of the accept- able reproducibility.
5.5.1 I NTERNAL Q UALITY C ONTROL
Based on the validation and optimization data generated, a QC scheme should be designed.
The performance of the method shall be regularly verified during its use as part of the internal quality control program of the laboratory. The internal quality control=performance verification is carried out to: monitor the performance of the method under the actual conditions prevailing during its use, and take into account the effect of inevitable variations caused by, for instance, the composition of samples, performance of instruments, quality of chemicals, varying performance of analysts, and laboratory environmental conditions; demonstrate that
* Single food or feed used to represent a commodity group for method validation purposes. A commodity may be considered representative on the basis of proximate sample composition, such as water, fat=oil, acid, sugar and chlorophyll contents, or biological similarities of tissues, and so on.
y Analyte chosen to represent a group of analytes which are likely to be similar in their behavior through a multiresidue analytical method, as judged by their physicochemical properties, for example, structure,
water solubility, K ow , polarity, volatility, hydrolytic stability, pK a , and so on.
Quality Assurance 143
TABLE 5.4 Representative Commodities for Multiresidue Methods a and Storage Stability Tests
Group Common Properties
Representative Species Plant products
Commodity Group
I High water and chlorophyll
Spinach or lettuce content
Leafy vegetables
Brassica leafy vegetables
Broccoli, cabbage, kale
Legume vegetables
Green beans, green peas
Fodder crops
II High water and low or no
Apple, pear, peach, cherry chlorophyll content
Pome fruits
Stone fruits
Strawberry
Berries, small fruits
Grape
Fruiting vegetables
Tomato, bell pepper, melon
Root vegetables
Potato, carrot, parsley
Mushroom III
Fungi
High acid content
Citrus fruits
Orange, lemon
Blueberry, current IV High sugar content
Berries, pineapple
Raisins, dates V High oil or fat
Oil seeds
Avocado, sunflower seed
Walnut, nuts, pistachios, peanut VI Dry materials
Nuts
Cereals
Wheat, rice, or maize grains
Wheat bran, wheat flour Commodities requiring
Cereal products
e.g., Garlic, hops, tea, individual test
spices, cranberry Commodities of animal origin
Mammalian meat (muscle)
Any of the major species
Poultry meat, edible offals, fat Eggs Milk
a For storage stability tests groups I and II may be combined, and crops of high protein or starch content should be considered separately.
the performance characteristics of the method are similar to those obtained during method validation, the application of the method is under ‘‘statistical control,’’ and the accuracy and uncertainty of the results are comparable to the performance characteristics established during method validation.
The results of internal quality control provide essential information for the con- firmation and refinement of performance characteristics established during the initial
validation, and extension of the scope of the method. Some key components of the QC scheme are summarized later.
The correct preparation of analytical standards should be verified by comparing its analyte content to the old standard, or preparing the new standard in duplicate at
144 Analysis of Pesticides in Food and Environmental Samples
TABLE 5.5 Acceptable within Laboratory Performance Characteristics of a Method a
Concentration Repeatability
Trueness b mg=kg
Reproducibility
CV A % c CV L % d CV A % c CV L % d Range of Mean Recovery, % 0.001
>1 10 14 16 19 70 –110 a With multiresidue methods, there may be certain analytes where these quantitative performance criteria cannot be strictly met. The acceptability of data produced under these conditions will depend on the
purpose of the analyses, for example, when checking for MRL compliance the indicated criteria should be fulfilled as far as technically possible, while any data well below the MRL may be acceptable with the higher uncertainty.
b These recovery ranges are appropriate for multiresidue methods, but strict criteria may be necessary for some purposes, for example, methods for single analyte.
c CV A : Coefficient of variation of analysis excluding sample processing. d CV L : Overall coefficient of variation of a laboratory result, allowing up to 10% variability of sample processing.
the first time. A balance with 0.01 mg sensitivity should not be used to weigh <10 mg standard material. The dilutions of standard solutions should be made independently based on weight measurement except the last step for which an
A-grade volumetric flask should be used. 12 Weighted regression (WLR) should be applied for evaluation of the linear cali-
bration function for GLC and HPLC measurements especially at the lower third of the calibrated concentration range. The confidence limits at the middle and upper cali- brated range are about the same with the weighted and ordinary (OLR) regression calculation (Figure 5.2).
The goodness of the calibration should be characterized with the standard deviation of the relative residuals (S rr ), as it is much more sensitive indicator than the regression coefficient (see Figure 5.2 and Table 5.6). The relative residuals’ (residuals=predicted response Dy i ¼y i
Y rel,i ¼ Dy i =^y) standard deviation (S rr ) is calculated as
where y i is the response of standard x i and ^y is the corresponding response on the regression line. Where the calibration points are spread over the analytical sequence, an S rr of 0.1 and 0.6 may be considered acceptable for GC and HPLC methods, respectively.
Quality Assurance 145
y = 48,383x ⫺ 50,583 y = 52,203x ⫺ 67,737 R 2 =1
R 2 =1 400,000
S rr = 0.06 S rr = 0.1 300,000
Linear (Y
Linear (Y
0 i ) 0 5 10 15 0 5 10 15
Injected amount [pg] Injected amount [pg] FIGURE 5.2 Evaluation of calibration with weighted (WLR) and ordinary (OLR) linear
regression. For the most effective internal quality control, analyze samples concurrently
with quality control check samples. For checking acceptability of individual recov- ery results, the initial control charts is constructed with the average recovery (Q) of representative analytes in representative matrices and the typical within laboratory reproducibility coefficient of variation (CV typ ) of analysis. The warning and action
Atyp 3Q Atyp 3Q , respectively. At the time of the use of the method, the recoveries obtained for individual analyte=sample matrices are plotted in the chart.
Based on the results of internal quality control tests, refine the control charts at regular intervals if necessary. If the analyte content measured in the quality control check samples is outside the action limits, the analytical batch (or at least the analysis
and occasionally detected analytes, respectively) may have to be repeated. When the
TABLE 5.6 Comparison of the Standard Deviation of the Relative Residuals and the Regression Coefficient
S rr
146 Analysis of Pesticides in Food and Environmental Samples
FIGURE 5.3 Illustration of long-term reproducibility of a MRM with different pesticide sample combinations.
results of quality control check samples fall repeatedly outside the warning limits (1 in 20 measurements outside the limit is acceptable), the application conditions of the method have to be checked, the sources of error(s) have to be identified, and the necessary corrective actions have to be taken before the use of the method is continued.
The differences of the replicate measurements of test portions of positive sam- ples can be used to calculate the overall within laboratory reproducibility of the method (CV Ltyp calculated with Equation 5.4). The CV Ltyp will also include the uncertainty of sample processing, but will not indicate if the analyte is lost during the process.
The long-term reproducibility of the MRM can be demonstrated by plotting on the control chart all recovery values of compounds, that can be characterized with the same typical average recovery and CV A , obtained during the use of the method. Figure 5.3 shows the quality control chart including 394 recoveries of 35 GC amenable residues in 21 commodities at spiking levels of 0.01 –1 mg=kg over 1 year (F. Zakar, personal communication, 2000).
The applicability of the method for the additional analytes and commodities shall
be verified as part of the internal quality control program. All reported data for a specific pesticide matrix combination should be supported with either validation or performance verification performed on that particular combination.