Ambrus Et Al (2020) - Sources of Random Variation of Pesticide Residue Analytical Results

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/344961740
Sources of Random Variation of Pesticide Residue Analytical Results
Article in Journal of AOAC International · September 2020

DOI: 10.1093/jaoacint/qsaa119
CITATIONS READS
3 67
4 authors, including:
Arpád Ambrus Kata Kerekes

retired Nemzeti Élelmiszerlánc-biztonsági Hivatal
123 PUBLICATIONS 2,204 CITATIONS 21 PUBLICATIONS 184 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Sampling View project
Project BASELINE EU FP7 2013 View project
All content following this page was uploaded by Arpád Ambrus on 17 November 2022.
The user has requested enhancement of the downloaded file.

Sources of Random Variation of Pesticide Residue Analytical
Results
Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

ÁRPÁD AMBRUS
University of Debrecen, Doctoral School of Nutrition and Food Sciences, 4032 Debrecen,
Egyetem tér 1., Hungary
KATA KEREKES
National Food Chain Safety Office, System Management and Supervision Directorate, 1024
Budapest, Kis Rókus str. 15/b., Hungary
HENRIET SZEMÁNNÉ-DOBRIK
Food Chain Safety Centre Nonprofit Ltd., Pesticide Residue Analytical Laboratory, Miskolc
3526 Miskolc, Blaskovics L. str. 24., Hungary
ZSUZSANNA DOMÁK
National Food Chain Safety Office, Food Chain Safety Laboratory Directorate, Pesticide
Analytical National Reference Laboratory, 2481 Velence, Ország út 23., Hungary
Corresponding author: [email protected]
© AOAC INTERNATIONAL 2020. All rights reserved. For permissions, please e-mail:
[email protected]
1
Background: Pesticide residues are analyzed in thousands of samples yearly by national
authorities and private laboratories. Intensive research is ongoing to develop new methods
or improve existing ones concentrating on the extraction, cleanup, and detection techniques.

Little attention has been paid to the contribution of prior steps in the determination process
to the overall laboratory sampling error, though several publications demonstrated their
practical importance. Consequently, the repeatability and reproducibility of the results are
often reported based on the recovery tests alone. A few previous publications are cited in
this paper which illustrate the magnitude of random error derived from subsampling,
comminution of analytical sample and selection of small test portions.
Objectives: We aim to call the attention to the importance of considering all steps of the
laboratory sampling and analysis process in calculating the combined uncertainty of the
results and the realistic performance assessment of the methods including its long-term
intermediate precision. Methods: Validation of laboratory sampling of large fruits is used to
illustrate the recommended procedures, determination of their random error and long-term
method performance. Results: The results indicate that subsampling, comminution, and
selection of test portions can be the major contributors to the combined uncertainty of the
results. Conclusion: All these steps should be considered in estimation of random variation
(uncertainty) of measured residues.
KEY WORDS: pesticide residues, laboratory sampling, ambient and cryogenic comminution,
combined measurement uncertainty, random error and internal quality control.
2
ScholarOne Support phone: 434-964-4100 email: [email protected]
Introduction
For obtaining the desired biological effects, some residues of pesticides must remain in / on the
treated objects. Even in case of the most careful application (Figure 1), the pesticide residues are

unevenly distributed, and the residue concentrations can differ hundred-fold in individual crop
units within one treated field (1, 2). The magnitude of the field to field variation of residues in
composite samples of crops treated with the same dosage can be characterized with an average
relative standard deviation of about 70% (3). The minimum number of crop units (primary
samples) comprising in a composite sample is specified by the corresponding sampling
guidelines (4, 5). The higher the average concentration, the higher is the variability of residues as
shown in Figure 2 (3). Usually the peel contains higher residue concentrations than the pulp.
Outer leaves of cabbage (Figure 3) or lettuce, lower parts of fruits (Figure 1) or bunch of grapes
hanging on the trees or vines may contain higher residues than the internal or upper exterior part
(6, 7). When an entire fruit or vegetable is comminuted, the uneven distribution of residues
between the peel, pulp, and juice results in both compositional and distributional heterogeneity,
leading to fundamental, as well as grouping and segregation error (8).
Pesticides are generally toxic chemicals, depending on the level of exposure. Their residues are
determined in hundreds of thousands of samples by governmental and private laboratories to
assure that food offered for sale is safe for the consumers and complies with the legal maximum
residue limits (MRLs) (9, 10). The determination of pesticide residues may consist of several
steps depending on the sample material, available equipment / instruments and the analytical
methods used. The main steps of the procedure are summarized in Table 1. Steps 2-6 contribute
to the total sampling error (TSE) and Step 7 also includes total analytical error (TAE) as
described in Good Test Portions (GTP) (8). This paper demonstrates that contributions from
3
Steps 4-7 to the global estimation error (GEE) and measurement uncertainty are significant. A
procedure to estimate the error due to Steps 4-7 is described.
The Codex Guidelines, (11,12) and the European Commission Guidance document (AQC) (13)

refer to the 2nd and 3rd steps (Table 1) as sample preparation and Steps 4-6 as sample processing.
However, the terms ‘sample preparation’ and ‘sample processing’ are purposely omitted in this
paper to avoid confusion, because different authors apply them to describe different steps of the
process without clarifying their meaning. To facilitate the uniform description and
characterization of the steps of the determination of pesticide residues in food (14), the relevant
terms and definitions recommended in the Good Test Portions (GTP) guidance document (8) are
used in this paper.
Our objectives are to (i) call the attention to the importance of considering all steps of the
laboratory sampling process in calculating and reporting the performance characteristics of the
methods; (ii) provide references to illustrate the magnitude of contribution of the laboratory
determination steps to the total uncertainty of the results; (iii) provide practical examples for
estimating errors in the laboratory determinative steps including the long-term interim precision
(within-laboratory reproducibility). The sources of systematic errors and blunders are not
discussed in detail.
Selection of the portion of the commodity to be tested (Steps 2-3)
Following the Codex specifications for the portion of commodity to be analyzed (15) is critical
when determining compliance with MRLs. Any deviation from it may cause non-selection,
systematic error and may lead to an incorrect decision on the compliance of sampled commodity
with MRLs. Though these steps may greatly affect the results of the analyses, the correctness of
extraneous material removal (Step 2) and selecting the portion of commodity for analysis (Step 3)
4
cannot be validated, and their repeatability / reproducibility cannot be determined. The reason is
the large, up to hundred-fold variability of residues in individual crop units. The distribution of
residues in one laboratory sample (≥ 5 crop units) taken from a decision unit (lot or field) will

most likely be different from additional samples taken from the same site leading to different
results even for one specific crop and harvested field. Consequently, generally applicable ratios
of residues before and after extraneous material removal cannot be determined. Therefore, Steps
2-3 are not evaluated in this work. Figure 3 illustrates handling of head cabbage in Steps 2-5.
While handling of papaya fruits in Steps 3-5 is shown in Figure 1. The uniform and appropriate
performance of Steps 2-3 can be facilitated with the detailed description of the procedure in the
SOPs and proper training of the responsible staff of the laboratory (16, 17).
Calculation of laboratory sampling and analysis errors (Steps 4-7)
Assuming independence among the random errors associated with each major step of the
pesticide residue test procedure, the variance of the laboratory phase (VL) may be described as:
VL=V4+V5-6+V7 (1)
The numbers in subscript indicate the steps of the residue determination procedure (Table 1). The
combined relative random error (uncertainty) of Steps 4-7 can be expressed with their relative
standard deviations (CV):
𝐶𝑉𝐿 = 𝐶𝑉24 + 𝐶𝑉25 ― 6 + 𝐶𝑉27 (2)
The random error of the analysis (Step 7) can be determined with recovery tests (Step 8)
performed by spiking preferably blank test portions with known amounts of analyte at various
concentrations, and performing the analysis (Step 7) in the same manner as the analytical test
portions (11, 13). It should be noted that CVL cannot be estimated based on the analyses of
reference materials or proficiency test samples as they are very carefully homogenized before
5
distribution. The relative random error (CV8) includes a contribution of spiking of blank test
portion with known amount of analyte as well as Step 7 (extraction, cleanup, derivatization,
evaporation and instrumental measurement). The relative random error (CV8) includes the error

due to selection of the test portion (Step 6) if multiple test portions are selected from a larger
mass or volume of an analytical sample containing an incurred residue.
The combined expanded uncertainty of the measured residue (CVL) should be considered (11, 16,
18), when deciding compliance of a marketed commodity with MRL [mg/kg]. The sampled lot is
accepted (5, 13) if the residue (CR, mg/kg) measured in the sample does not exceed the MRL
taking into account the expanded uncertainty calculated from CVL:
𝐶𝑅 ―(2 × 𝐶𝑉𝐿 × 𝐶𝑅) ≤ 𝑀𝑅𝐿 (3)
The default value of CVL is 25% in the European Union (13).
Obtaining test portions from analytical samples
Several publications revealed the potentially significant contribution of comminution (Step 5) to
the combined uncertainty of the results during the last 20 years (17, 19-25). The contribution of
splitting and subsampling of laboratory sample (24), the comminution to prepare the analytical
sample and stability of analytes (26, 27, 29) during Steps 4-6 to the uncertainty and accuracy of
the results are rarely tested, though they may significantly affect the reliability of the
measurements.
Most published methods concentrate mainly on the clean-up and detection techniques (analysis
phase, Steps 7-8), and unproportionally little attention is paid to the prior steps (28, 29). The
repeatability and reproducibility of the analytical methods are determined from the extraction of
spiked test portions (Step 8), and their performance characteristics are reported based on these
6
results. Biased measurements and underestimated uncertainty could lead to erroneous
interpretations of the results and trade disputes. One possible reason for paying little attention to
Steps 3-6 may be that several widely used guidance documents dealing with the estimation of

uncertainty of analytical measurements do not discuss the contributions of Steps 4-6 or assume
their contribution negligible (30-34).
Due to the heterogenic distribution of residues in crop units, the whole laboratory sample must be
processed. To get results with acceptable reproducibility [CVL ≤ 25% (13) or CV7 ≤ 20% at ≥
0.01 mg/kg residue concentrations (35)] typically ≥ 10-15 g test portions must be selected from
material comminuted at ambient temperature depending on the physical properties of the sample
material and the efficiency of the equipment (19-24). The currently most widely used
QuEChERS procedure (36) is originally based on the extraction of a 10 g test portion. It was
gradually amended to accommodate pH dependent pesticides and commodities of different water
content. It was emphasized that the smaller the test portion is, the higher the variability of the
results will be. Each laboratory should validate its procedure (37, 38). However, utilizing the high
sensitivity of the modern GC-MS/MS, LC-MS/MS instruments and aiming to reduce the amount
of extraction solvents, the test portion size has been decreased to < 1 or 2 grams (39, 40) without
testing the random error contribution of Steps 4-6. If the test portion is decreased without testing
the repeatability of the whole determination process, a major part of uncertainty of the
measurement results may remain hidden. Consequently, the performance of the method is
wrongly reported based on the repeated analyses of the spiked test portions (Step 8) (41-43).
Quality control of comminution and selection of test portion (Steps 5-7)
7
The efficiency of comminution may vary from sample to sample and day to day, as it can be
affected even in case of the same specific equipment, for instance, by the maturity and type of
sample material, sharpness of the cutting blade, mass of laboratory sample to be comminuted and

the temperature of comminution. The between laboratories reproducibility could also be greatly
influenced by the type and condition of equipment used.
Consequently, the random error of laboratory sampling and analysis of residues should be tested
regularly as part of the ongoing internal quality control. Laboratory samples containing field
incurred residues can be used for determining CV5-6 most accurately. Alternatively, crop units
(i.e. multiple tomatoes) making up the laboratory sample may be treated with pesticide standards
known to be stable during comminution by injecting an analyte standard solution into the crops
before the entire laboratory sample is comminuted (44), or by applying an analyte standard
solution to the surface of the crop, to obtain well detectable average residues of known
concentration calculated for the entire laboratory sample (27).
Based on Gy’s sampling theory (45), the variance of fundamental sampling error (FSE) can
approximately be described as:

𝐶 × 𝑑3
𝑆2𝐹𝑆𝐸 = 𝑤 (4)
where C is the sampling constant depending on the shape and physical properties of the
comminuted (chopped, ground) material, d is the diameter of the 95th percentile of the
comminuted particles, and w is the mass of the test portion. Equation 4 highlights the relationship
of particle size to fundamental sampling error. Though, Gy’s equation cannot be directly applied
for plant materials, it provides important information. For a given material, reducing the particle
size in a comminuted laboratory sample considerably reduces CV5-6 and consequently CVL. The
particle size can be substantially decreased if the comminution is performed in the presence of
8
dry ice under cryogenic conditions (21, 23, 28, 46), as illustrated in Figures 1 and 3. These
figures also show a simple and practical method for testing the effectiveness of comminution. A
small spoonful of comminuted material is spread on a Petri dish or watch glass. If the particle

size is ≤2 mm, selecting a 10-15 g test portion has been shown to be sufficient to obtain CV5-7
<0.15-0.20 for most comminuted plant matrices (19, 20-23, 26, 27). Photos can be made of the
Petri dishes and kept together with other quality control records (Figures 1, 3).
The contribution of comminution to the random error of determination of pesticide residue cannot
be quantified separately, but together with the interrelated selection of test portion (CV5-6) and
can be calculated as:
𝐶𝑉5 ― 6 = 𝐶𝑉25 ― 7 ― 𝐶𝑉27 (5)
Cryogenic processing is generally applied for the analysis of supervised pesticide residue trial
samples (6), and it is used routinely in other laboratories as well (13, 27). Compared to
comminution with dry ice, much smaller particle sizes can be obtained, and consequently smaller
test portions can be used, when the laboratory sample is comminuted with liquid nitrogen, (28) or
after freezing with liquid nitrogen (47, 48). Rousev et al. (29) demonstrated the advantages of
comminuting laboratory samples with liquid nitrogen over cryogenic processing with dry ice. The
results were confirmed by Lehotay et al (49). The authors concluded that selecting 5 g test
portions would result in an average CVL of 3-5%. It should be noted, however, that the methods,
based on one-step comminution with liquid nitrogen, presented in the latter publications are not
suitable for official control of compliance with MRLs as the sample comminuted was 500-850 g,
which is lower than the minimum mass of laboratory sample (1 kg) specified by the relevant
standards (4, 5). Consequently, they should be modified to (i) comminute the entire laboratory
sample; (ii) apply another method for comminution before liquid nitrogen comminution (2-step)
9
or (iii) perform splitting and subsampling of the entire laboratory sample and comminute two or
more representative portions.
Another advantage of comminution at temperatures < -20 C is that it can eliminate or reduce the

decomposition of analytes, which may occur when they come in contact with the plant fluids
containing enzymes or acidic compounds released by cutting the plant tissues (17, 26, 27, 49, 52).
The decomposition of residues during comminution causes biased results because it may not be
indicated by the recoveries obtained from spiked test portions.
The error of Steps 4-6 can be substantially decreased with two-step comminution (47, 48). In this
case an entire large laboratory sample is comminuted first with a suitable chopper or grinder, then
a representative portion is further comminuted a second time to obtain fine particles. Both steps
can be carried out at ambient temperature (24) or in combination with cryogenic processing
applying dry ice (21) or liquid nitrogen or only the latter two (47, 48). The random error derived
from the first step (CV5-6(1)) will be carried over to the second fine-grinding/disintegrating step
(CV5-6(2)) and the combined uncertainty of the two steps (CV5-6’’) will be
𝐶𝑉,,5 ― 6 = 𝐶𝑉25 ― 6(1) + 𝐶𝑉25 ― 6(2) (6)
𝐶𝑉,,5 ― 6 is inserted into Equation 2 to obtain the combined uncertainty of the laboratory phase
(CVL) of the determination of pesticide residues.
Ideally, the sample mass reduction and comminution should not significantly increase CVL. In
practice it means that the CV4-6 should preferably be ≤0.3 CV7 (16, 17).
To obtain repeatable results, the comminuted material should be “uniform” and the test portions
should be selected without delay by taking as many (preferably >10) increments as feasible from
various positions for minimizing the error caused by distributional heterogeneity of the sample
matrix. Segregation error can also occur if the comminuted sample is let to stand for some time
10
before the test portions are selected. The uniform comminuted sample may contain particles of
different sizes provided that they are randomly distributed throughout the entire matrix (50). The
uniform condition of a comminuted material can be tested based on the principle developed by

Wallace and Krachochwil (51). It was applied for pesticide residue analysis in various plant
materials and soil (19, 21, 24, 46, 52, 54).
Experimental
A laboratory determined pesticide residues in papaya samples from treated fields shortly after the
last pesticide application during a supervised field trial. The papaya fruits are large berries up to
45 cm long and 20 cm in diameter (Figure 1). The masses of the individual fruits range from
about 3 to 9 kgs. According to the Codex Sampling Guidelines (4) the laboratory sample must
contain a minimum of 5 large fruits. For papaya fruits the laboratory sample mass may be at or
above 25 kg, which is much larger than that which can be comminuted in one batch with regular
laboratory equipment.
The pesticide residue coded in this paper as ‘Θ’ to be determined was known to be stable during
comminution at room temperature. The analysts intended to use an analytical method that had
been fully validated for commodities of high-water content (Typical CV7: 0.12, CVL: 0.21 and
recovery 95-102%).
Recovery tests to evaluate method performance
The applicability of the selected analytical method was evaluated with testing the recovery of ‘Θ’
from spiked test portions. Untreated papaya fruits were taken from the orchard before the
treatment with pesticide ‘Θ’. One fruit (3560 g) was cut into sections and then into small cubes.
The cut pieces were placed in Stephan blender (model UM12 with stainless steel 12 L bowl,
11
rotating knives and reverse action scraper arm, 1500/3000 rpm, that can be loaded with maximum
4 kg material to obtain good disintegration) and comminuted at room temperature until visibly
uniform composition was obtained (Figure 1). Test portions (15 g) were selected from the

comminuted material and spiked in 5 replicates with analytical standard solution of ‘Θ’ at 0.2 and
1.5 mg/kg levels, respectively. The spike portions were thoroughly mixed, and the residues were
determined with the previously validated analytical method. The repeatability of the residue
determination (including extraction, clean up, partial evaporation of the extracts and gas
chromatographic determination) was calculated from the measured residues in the extracts of
spiked samples (Table 2).
The average recoveries based on 5 replicate tests, 𝑄, which is the minimum number
recommended by AQC GLs (11, 13) and their relative standard deviations, (CV8), were 95.0%
(10%) and 98.30% (8.0%) at 0.2 and 1.5 mg/kg spike levels, respectively. As the CV8 values
obtained at the two spike levels were not statistically different (based on F-test) their variances
(CV2 values) were pooled resulting in an average relative standard deviation of CV8=0.090 and
typical recovery of 96.6% for the residue range of 0.2-1.5 mg/kg. The results of recovery tests
agreed with the prior method validation data (Typical CV7: 12% and recovery 95-102%). The
results confirmed the applicability of the procedure for determination of residues in papaya
samples. However, the contribution of the laboratory sampling (sub-sampling, and selection of
test portions) to the combined uncertainty of the results had to be determined before the field trial
samples would be analyzed.
Developing method for determination of residues in large crop units
The flowchart of the procedure is shown in Figure 4. For the partial validation of the method
(Steps 3-6), the laboratory received 5 individual units of field treated papaya fruits with unit
12
masses of 7150, 5670, 5272, 7127, 4960 g (total mass 30179 g), as well as untreated ones. As the
first step, the stems (considered as extraneous material) (Step 2) of the fruits were removed. The
whole fruit, including peel and seeds were selected for analysis (Step 3). For Step 4 each of the 5

papaya fruits (containing incurred residues) were cut longitudinally into four approximately equal
wedge-shaped sections and each quarter was further split into two parts resulting in 8 sections
from each fruit (Figure 1). According to the minimum requirements for replicate tests specified
by the Codex (35) and AQC (13) guidelines for validation of a method, five subsamples (A-E)
were prepared. The subsamples contained one or two sections from each fruit to get
approximately similar mass. Their masses are shown in Table 3. The average mass of subsamples
was 3745.4 g. The difference between the expected proportion (1/8 of the total mass: 30,179 x
0.125 = 3772 g) of laboratory sample and the average mass of subsamples (3745.4 g) is 0.7%.
The relative standard deviation associated with variation of masses of the five subsamples
obtained with splitting is 0.108. The fruit sections making up one subsample were cut into small
pieces and each subsample was chopped in a Stephan blender at room temperature. Two test
portions of 15 g each (e.g. A15(1), A15(2)) were selected taking 10 or more small increments
randomly from various positions of each comminuted subsample and analyzed with the
previously validated method. The average residue obtained from 10 test portions was 0.804
mg/kg. The relative random errors of the determination of the residues (Steps 4-7) present in
analytical samples (CVL15) were calculated from the 10 individual measurements (A15(1), E15(2)
(0.70). The variability estimated was unacceptably high compared to the default value of 0.25
specified in AQC guidelines (13).
In addition, the relative random error of the average residues (0.53) in five subsamples (A-E)
estimated based on the analyses of two test portions (15 g each) selected from them was also
high. Note that the average of residues measured in replicate test portions selected from a
13
subsample gives the best estimate for the residue concentration in that subsample. These results
clearly indicated that one cannot withdraw 15 g test portions from the comminuted subsamples
(A-E) to obtain credible results.

The analyst decided to apply next day a two-step comminution (sequential mass reduction step)
procedure by first selecting a 200 g portion from the re-homogenized remaining part (kept in
freezer during the night) of each previously comminuted subsample (A-E). The portions were
composited from small increments collected from various positions of the bowl containing the
freshly blended material. Each 200 g portion (PA-PE) was transferred to Warring blender (2
speed, 1L stainless steel container), 10% w/w distilled water was added (to improve the
efficiency of disintegration of the sample material) and then the mixture was blended at room
temperature until all peel particles were smaller than 2 mm (Figure 1). Duplicate 15 g test
portions (e.g. PA15(1), PA15(2)) were then taken with several increments from randomly selected
positions of the blended 200 g analytical samples for extraction and quantitation of residue. The
test results are shown in Table 3. The relative random errors of the determination of the residues
(Steps 4-7) present in the analytical samples (CVL200/15) were calculated from the residues
measured in test portions (PA15(1)- PE15(2)) (12.4%).
The average residue adjusted for the 10% added water in the 15(200) g test portions was 0.807
mg/kg. The difference between the average residues measured in 15 g test portions (0.804 mg/kg)
is 0.42%, The two sets of measures compare favorably.
Internal quality control during the long-term use of the method
ISO/IEC 17025:2017 Standard (53) recommends testing portions retained from previously
analyzed samples together with the freshly received samples to determine the long-term within
laboratory intermediate precision (within laboratory reproducibility) of Steps 4-7. In our study,
14
15 g test portions were selected, as described above, from the 200 g 2-step comminuted
subsamples and placed in deep freezer. Later, when the samples were received from the field
trial, the test portions stored in deep freezer were analyzed in the same batch with the incoming

samples. The residues measured in the retained test portions were compared to the residue
determined in the corresponding analytical sample during the performance verification of the
method. The results obtained in this study are listed in Table 4.
The intermediate precision is calculated with range statistics (55, 56). For one test portion, the
CVLi is calculated from the residues (R) measured during analyzes of supervised trial samples
and in the retained test portion analyzed later together with the trial samples:
𝑅𝑚𝑎𝑥 ― 𝑅𝑚𝑖𝑛
𝐶𝑉′𝐿𝑖 = 1.128 × 𝑅
(7)
Since the average residues in ‘n’ replicate measurements are different, the CVLi values are pooled
instead of the standard deviations.
∑𝐶𝑉′2𝐿𝑖
𝐶𝑉′𝐿𝑝 = 𝑛
(8)
The pooled CV’L200/15 calculated with Equations 7 and 8 is 0.155 (Table 4), which is statistically
not different, based on the F-test, from the CV’L200/15 of 0.132 obtained during method
development (Table 3). Thus, the internal quality control confirmed the results of validation of
Steps 4-7.
Results and discussion
The combined uncertainty (12.4%) obtained with two step sequential sampling is acceptable in
view of the guidance reproducibility value of 0.25 specified in EU AQC GL (13) because the
predicted among laboratories reproducibility is about 1.5-2 times higher than the within
laboratory reproducibility (57). Therefore, the procedure based on two-step comminution is
15
considered suitable for future analysis of residues in large crops. Nevertheless, it is worthwhile to
examine possibilities for improvement.
One-way ANOVA test performed from the replicate measurements confirmed the null hypothesis

(the sample means are not different) with high probability (P values 0.533 and 0.387) that the
average residues in the subsamples and in the 200 g analytical samples are not significantly
different (Table 5).
The contribution of the steps of laboratory sampling and analysis to the random variation of ‘Θ’
residues determined in papaya fruits can be calculated from their variances (Table 6). It is
assumed that the random error of analysis (Step 7) is equal to the repeatability of recovery tests
(CV7=CV8), and the best estimate for the residue contents of subsamples (Step 4) is the average
of residues determined in 15 g test portions (TP). The total variance (VT=VL) is calculated from
the repeated measurements of test portions. The variance (V5-6) of Steps 5-6 is calculated after
rearranging Equation 1:
V5-6=VL-V4-V7 ((9)
The calculated contribution of the steps is summarized in Table 7.
The contribution of subsampling (V4) to the total variance of laboratory sampling is 56.9%, if 15
g test portions are selected from the comminuted subsamples, while the comminution and test
portion selection (Steps 5-6) is 41.4% of the total variance . These results indicate that the
comminution of subsample and selection of test portions must be improved to obtain CVL
conforming to current quality standards. In case of two-step comminution (sequential mass
reduction step including the subsampling and selection 200 g analytical sample, V5-6) and
analysis (V8) contribute with 11.7% and 46.9%, respectively, to total variance of laboratory phase
of determination of  (VL=VT=0.01138), which is about 28 times lower than VT of subsamples.
16
These results further confirm that the two-step sequential sampling procedure developed is
suitable for the analysis of supervised trial samples according to current international standards.

We examined the possibility of further improvement. The percentage contribution of the steps of
determination process is interrelated. It can be conveniently illustrated with their relative standard
deviations (Equ. 2). Some examples are given in Table 8. If we could halve CV4 and CVA (option
2) then the CVL would be reduced by nearly 42% compared to option 1, while reducing all
components to half would similarly decrease CVL (option 3). However, reducing CV5-6 to 2.25%
alone the CVL will only be slightly lower (12.6%) (option 4). CVL of 6-7% (options 3, 5) satisfies
all practical requirements and its further improvement with extra expenses is not justified.
Nevertheless, applying internal standards for calibration as well as dry ice and or liquid nitrogen
for comminution would substantially improve CVL=CVT (options 5) enabling the selection of
smaller test portions as shown in several publications (27, 28, 46, 47, 48, 49), which would
reduce the cost of extraction solvents and disposal of used chemicals and may also speed up the
whole process.
To achieve optimum performance the CV4 should be reduced. Our further experiments, which are
not discussed in this article, revealed that subsampling random error can be reduced by selecting
the sections of individual fruits systematically when large crops are analyzed. The principle is to
represent the individual crops as well as possible by taking sections from their opposite side. For
example, coding of the sections is illustrated in Figure 5 and their selection for replicate samples
is shown in Table 9.
Conclusions and recommendations
17
CV4 is affected by the sample material and the method used for splitting and selection of
subsamples. CV5-6 is largely affected by the sample material, the operation condition and type of
equipment used for comminution. CV7 depends on the test portion mass, the method used,

including instrumental analysis. Therefore, CV4-7 is laboratory specific and its determination
should be part of the method validation and regular performance verification carried out by
individual laboratories. For this purpose, the analysis of retained test portions in one batch
together with the fresh samples is a powerful quality control check and should be performed
regularly. Individual test portions should be retained for quality control purposes and ready for
extraction after thawing together with the extracting solvent. Deep-freezing larger portions for
reanalyzes would require defrosting the whole mass before taking the test portion, which would
take extra time and could result in the loss of labile or volatile compounds.
Certified reference materials or remaining parts of proficiency test materials cannot be used for
determining CV5-7 because the homogeneity of these materials is rigorously tested and verified
before distribution.
The CVL reflecting the combined uncertainty of laboratory sampling (subsampling, comminution
selection of test portions) and analyses of residues should be reported based on the replicate
analyses of samples containing incurred residues. Results of recovery studies do not indicate the
true uncertainty of determination of analytes of interest.
If possible, an entire laboratory sample should be comminuted before selection of the test portion.
It is not permissible to take a few natural units (primary samples) from the laboratory sample for
comminution and keep the rest as reserve. The adverse effects of the uneven distribution of
residues should be considered and controlled as much as possible during laboratory sampling
operations to minimize the random and systematic errors.
18
The correct performance of splitting and subsampling of large crops is critical for obtaining
accurate results, which should be verified by selecting and analyzing minimum 2 replicate
subsamples and reporting the average of residue values obtained.

Points to be considered during laboratory sampling operations and analysis of pesticide residues.
 The analytical sample must be selected from the whole laboratory sample after removal of
extraneous materials and selection of portions to be analyzed according to the relevant
regulations and purpose of the test.
 Each fruit or single increment included in the laboratory sample should be proportionally
represented in the subsample and or analytical sample.
 Since the surface residue concentrates on the lower part of the fruits hanging on the trees
(Figure 1), or laying on the soil, where subsampling is required, longitudinal sections
must be cut and their surface/pulp ratio should be the same as that of fruit.
 Wedge-shaped sections of similar size should be cut if a subsampling step is necessary.
Slices or cross sections must not be cut to prepare the subsample. Moreover, the entire
laboratory sample should not be cut into small cubes and “mixed” before an appropriate
subsample or analytical samples are selected, as proper mixing is impossible for sample
materials with hard peel and soft pulp. Exception is when the whole pre-cut laboratory
sample is deep-frozen before cryogenic comminution.
 The homogeneity of the comminuted analytical sample can be improved by adding dry ice
in small proportions to the sample material during chopping until the entire sample gets
deep-frozen. Liquid nitrogen could also be used considering the safety precautions.
Attention should be paid to keep the lid of the bowl open as short time as possible to limit
19
condensation of moisture from the air. If the analytes are labile or volatile, the whole
laboratory sample or representative part of it should be deep-frozen before comminution.
 Special attention is required to select the test portions for extraction where the test portion

is only few grams. The test portion should be selected by taking as many small increments
(preferably ≥10) with suitable tool from random positions in the vessel immediately after
comminution to represent the average composition of analytical sample bearing in mind
the compositional (peel and pulp) and distributional heterogeneity (residue levels can
widely differ among natural units included in the laboratory sample), segregation of
components of different density changing by time, etc. If needed, the blending can be
repeated while collecting the increments.
 The abovementioned factors can be different for various material types and analytes
therefore adequate mass and number of increments should be uniquely established for
each selection process. When the parameters of laboratory sampling process are
optimized and validated for “difficult” materials (e.g having hard peel and juicy flesh,
such as tomato), they can be applied for the easier ones (e.g. orange, apple) in order to
rationalize the method development and validation workload (13).
Acknowledgement
The authors are indebted to Eloisa Dutra Caldas and Kit Chen for cooperation in field trials,
Bálint Mátyásné, Vanda Glócz and Victor Juhász for performing the laboratory tests, and Nancy
Thiex and Jo Mary Cook for useful advices during the preparation of the manuscript.
References
20
1. Horváth, Zs., Ambrus, Á., Mészáros, L., & Braun, S. (2013) J. Environ. Sci. Health, Part
B. 48, 615-625. doi:10.1080/03601234.2013.777277
2. European Food safety Authority (2005) EFSA J. 177, 1-61. doi: 10.2903/j.efsa.2005.177

3. Ambrus, Á., Horváth, Zs., Farkas, Zs., Szabó, I.J., Dorogházi, E., & Szeitzné-Szabó, M.
(2014) J. Environ. Sci. Health, Part B. 49, 229-244. doi:10.1080/03601234.2014.868272
4. Codex Alimentarius Commission. (1999) Recommended methods of sampling for the
determination of pesticide residues for compliance with MRLs. CAC GL 33/1999.
http://www.fao.org/fao-who-codexalimentarius/sh-
proxy/en/?lnk=1&url=https%253A%252F%252Fworkspace.fao.org%252Fsites%252Fco
dex%252FStandards%252FCXG%2B33-1999%252FCXG_033e.pdf (accessed May 8
2020)
5. Commission Directive 2002/63/EK (2002) Establishing community methods of sampling
for the official control of pesticide residues in and on products of plant and animal origin
and repealing Directive 79/700/EEC. OJ. 187, 30–43.
6. Food and Agriculture Organization. (1982- ) Pesticide Residues in Food — Reports, FAO
Plant Production and Protection Papers series http://www.fao.org/agriculture/crops/core-
themes/theme/pests/jmpr/jmpr-rep/en/ (accessed May 7, 2020)
7. Tiryaki, O., Gozek, K., Yucel, U., & Ilim, M. (1996) Toxicol. Environ. Chem. 53, 227–
233. doi: 10.1080/02772249609358287
8. Association of American Feed Control Officials. (2018) GOOD Test Portions
http://www.aafco.org/Publications/GOODTestPortions (accessed May 7, 2020)
21
9. Ambrus, Á., Szenczi-Cseh, J., Tamás, G., Kerekes, K., Miklós, G., Szigeti, T., &
Vásárhelyi A. (2020) Journal of Food Investigation, 66, 2773-2801. https://hu.wessling-
group.com/fileadmin/user_upload/wessling_hu/Egyeb/EVIKO/2020_1/EVIK_2020-1.pdf

(accessed May 7, 2020)
10. Humphrey, P., Margerison, S., van der Velde-Koerts, T., Doherty M.A., & Rowland, J.
(2017) in Food Safety Assessment of Pesticide Residues, Á. Ambrus & D. Hamilton,
(Eds), World Scientific: New Jersey, pp. 37-111.
11. Codex Alimentarius Commission (1993) Guidelines on Good Laboratory Practice in
Pesticide Residue Analysis, CAC/GL 40-1993. http://www.fao.org/fao-who-
codexalimentarius/sh-
dex%252FStandards%252FCXG%2B40-1993%252Fcxg_040e.pdf (accessed May 7,
2020)
12. Codex Alimentarius Commission (2006) Guidelines on Estimation of Uncertainty of
Results CAC/GL 59-2006. http://www.fao.org/fao-who-codexalimentarius/sh-
dex%252FStandards%252FCXG%2B59-2006%252Fcxg_059e.pdf (accessed May 7,
2020)
13. SANTE/12682/2019 (2020) Method Validation & Quality Control Procedures for
Pesticide Residues Analysis in Food & Feed, European Commission Directorate-General
for Health and Food Safety https://www.eurl-
pesticides.eu/userfiles/file/EurlALL/AqcGuidance_SANTE_2019_12682.pdf (accessed
August 11, 2020)
22
14. Thiex, N., Paloetti, C., & Esbensen K.H. (2015) J. AOAC Int. 98, 259-263.
doi:10.5740/jaoacint.14-290
15. Codex Alimentarius Commission (2010), Portion of commodities to which Codex MRLS

apply and which is analyzed. CAC/GL 41-1993. http://www.fao.org/fao-who-
codexalimentarius/sh-
dex%252FStandards%252FCXG%2B41-1993%252FCXG_041e.pdf (accessed May 7,
2020)
16. Ambrus, Á. (2004) Accred. Qual. Assur. 9, 288-304. doi:10.1007/s00769-004-0781-6
17. Ambrus, Á., & Suszter, G. (2019) in Analysis of pesticides in Food and Environmental
Samples, 2nd ed., J. Toledo, (Ed), CRC Press, Roca Baton, Fl. pp. 135-173.
18. US FDA, Manual of Quality Policies for ORA Regulatory Laboratories, ORA-LAB-QM,
Rev. 3 08/13/2019 https://www.fda.gov/media/73734/download (accessed May 7, 2020)
19. Ambrus, Á., Solymosné, M.E., & Korsós, I. (1996) J. Environ. Sci. Health. Part B, 31,
443-450. doi: 10.1080/03601239609373005
20. Young, S.J.V., Parfitt Jr., C.H., Newell, R.F., & Spittler, T.D. (1996) J. AOAC Int. 79,
976-980. doi:10.1093/jaoac/79.4.976
21. Maestroni, B., Ghods, A., El-Bidaoui M., Rathor, N., Jarju, O.P., Ton, T., & Ambrus, Á.
(2000) in Principles of Method Validation, A. Fajgelj, Á. Ambrus, (Eds), Royal Society
of Chemistry, Cambridge, UK, pp. 49-58.
23
22. Hill, A.R.C., Harris, C.A., & Warburton, A.G. (2000) in Principles and Practices of
Method Validation, A. Fajgelj, Á. Ambrus, (Eds), Royal Society of Chemistry,
Cambridge, UK, pp. 41-48.

23. Fussell, R.J., Hetmanski, M.T., MacArthur, R., Findlay, D., Smith, F., Ambrus, Á., &
Brodesser, P.J. (2007) J. Agric. Food Chem. 55, 1062-1070. doi: 10.1021/jf0623743
24. Yolci Omeroglua, P., Ambrus, Á., Boyacioglu, D., & Solymosne-Majzik, E. (2013) Food
Addit. Contam., Part A, 30, 116-126. doi: 10.1080/19440049.2012.728720
25. Lehotay, S.J., & Cook, J.M. (2015) J. Agric. Food Chem. 63, 4395−4404.
doi:10.1021/jf5056985
26. Fussell, R.J., Addie, K.J., Reynolds, S.L., & Wilson, M.F. (2002) J. Agric. Food Chem.
50, 441-448. doi: 10.1021/jf010852y
27. Ambrus, Á., Buczkó, J., Hamow, K.Á., Juhász, V., Solymosné-Majzik, E., Szemánné-
Dobrik, H., & Szitás, R. (2016). J. Agric. Food Chem, 64, 6071-6081. doi:
10.1021/acs.jafc.5b05779
28. Han, L., Lehotay, S.J., & Sapozhnikova, Y. (2018) J. Agric. Food Chem. 66, 4986−4996.
doi:10.1021/acs.jafc.7b04359
29. Roussev, M., Lehotay, S.J., & Pollaehne, J. (2019) J. Agric. Food Chem. 67,
9203−9209. doi:10.1021/acs.jafc.9b04006
30. Taverniers, I., De Loose, M., & Van Bockstaele, E. (2004) TrAC, 23, 480–
490. doi:10.1016/s0165-9936(04)00733-2
31. Ellison, S.L.R., & Barvick V.J. (1998) Analyst, 123, 1387–1392. doi:10.1039/a706946d
24
32. Ellison, S.L.R., Williams, A. (Eds). (2012) EURACHEM/CITAC Guide: quantifying
uncertainty in analytical measurement. 3rd ed. pp. 141.
https://www.eurachem.org/index.php/publications/guides/quam (accessed May 9, 2020)

33. JGCM-100. (2008) Evaluation of measurement data – Guide to expression of uncertainty
of measurement. Joint Committee for Guides in Metrology. Bureau International des
Poids et Mesures.
ttps://www.bipm.org/utils/common/documents/jcgm/JCGM_100_2008_E.pdf (accessed
May 9, 2020).
34. Gustavo González, A., & Ángeles Herrador, M. (2007). TrAC, 26, 227–
238. doi:10.1016/j.trac.2007.01.009
35. Codex Alimentarius Commission. (2017) Guidelines on performance criteria for
methods of analysis for the determination of pesticide residues in food and feed CAC/GL
90-2017. http://www.fao.org/fao-who-codexalimentarius/sh-
dex%252FStandards%252FCXG%2B90-2017%252FCXG_090e.pdf (accessed May 7,
2020)
36. Anastassiades, M., Lehotay, S.J., Štajnbaher, D., & Schenck F.J. (2003) J. AOAC Int. 86,
412–31. doi:10.1093/jaoac/86.2.412
37. Anastassiades, M., Scherbaum, E., Tasdelen, B., & Stajnbaher, D. (2007) in Pesticide
Chemistry: Crop Protection, Public Health, Environmental Safety, H. Ohkawa, H.
Miyagawa, P.W. Lee (Eds), WILEY-VCH Verlag GmbH & Co. pp 439-458.
https://doi.org/10.1002/9783527611249.ch46
25
38. EN 15662:2018 (2018) Foods of plant origin – Multimethod for the determination of
pesticide residues using GC- and LC based analysis following acetonitrile extraction /
partitioning and clean-up by dispersive SPE – Modular QuEChERS method. European

Committee for Standardization, Brussels, Belgium
39. Gorji, S., Biparva, P., Bahram, M., & Nematzadeh, G. (2019) Food Anal. Methods, 12,
394−408. doi:10.1007/s12161-018-1371-2
40. Xue, J., Zhang, D., Wu, X., Pan, D., & Hua, R. (2019) Chromatographia. 82, 695−704.
doi:10.1007/s10337-018-3678-y
41. Prodhan, M.D.H., & Papadakis, E.N., Papadopoulou-Mourkidou, E. (2018) J. Sci. Food
Agric. 98, 2277-2284. doi:10.1002/jsfa.8716
42. Botitsi, H., Economou, A., & Tsipi, D. (2007) Anal. Bioanal. Chem. 389, 1685–1695.
doi:10.1007/s00216-007-1452-3
43. Vázquez, P., Hakme, E., Uclés, S., Cutillas, V., Martínez Galera, M., Mughari, A.R., &
Fernández-Alba, F. (2016) J. Chromatogr A. 1463, 20–31.
doi:10.1016/j.chroma.2016.08.008
44. Fussell, R.J., Hetmanski, M.T., Colyer, A., Caldow, M., Smith, F., & Findlay D. (2007)
Food Addit. Contam. Part A. 24, 1247−1256. doi:10.1080/02652030701317319
45. Gy, P. (1999) Sampling for Analytical Purposes, John Wiley & Sons, Ltd, Chichester,
UK, p.71.
46. Suszter, G., Ambrus, Á., Schweikert-Turcu, M., & Klaus P.M. (2006) J. Environ. Sci.
Health, Part B. 41, 1-22. doi:10.1080/03601230600701668
47. Riter, L.S. & Wujcik, C.E. (2017) J. AOAC Int. 101, 1-9. doi:10.5740/jaoacint.17-0317
26
48. Riter, L.S., Lynn, K.J., Wujcik, C.E., & Buchholz, L.M. (2015). J. Agric. Food Chem.
63, 4405–4408. doi:10.1021/jf505249y
49. Lehotay, S. J., Michlig, N., & Lightfield, A. R. (2020) J. Agric. Food Chem, 68, 1468-

1479. doi:10.1021/acs.jafc.9b07685 (accessed July 6 2020)
50. Ingamells, C.O., & Switzer, P. (1973) Talanta 20, 547-568. doi:10.1016/0039-
9140(73)80135-3
51. Wallace, D., & Kratochvil, B. (1987) Anal. Chem. 59, 226-232.
doi:10.1021/ac00129a003
52. Maestroni, B., Ghods, A., El-Bidaoui, M., Rathor, N., Ton, T., & Ambrus A. (2000) in
Principles of Method Validation, A. Fajgelj, Á. Ambrus, (Eds), Royal Society of
Chemistry, Cambridge, UK, pp. 59–74.
53. ISO/IEC 17025: 2017 (2017) General requirements for the competence of testing and
calibration laboratories. International Standard Organization, Geneva, Switzerland
54. Tiryaki, O., & Baysoyu, D. (2006) Accredit. Qual. Assur. 10, 550−553.
doi:10.1007/s00769-005-0070-z
55. Youden, W.J. (1967) Statistical techniques for collaborative tests, AOAC, Washington
DC
56. Anderson, L.R. (1987) Practical statistics for analytical chemists. Appendix D12.Van
Nostrand Reinhold Company: New York. NY.
57. Horwitz, W. (2000) in Principles of Method Validation, A. Fajgelj, Á. Ambrus, (Eds),
Royal Society of Chemistry, Cambridge, UK. pp. 1-8.
27

28
Figures
Figure 1. Main steps of determination of pesticide residue in papaya fruits. From top left:
pesticide application; field sampling; packed fruits, split sections (1/8 of portions) of

individual fruits in laboratory sample; subsample comminuted with UM12 Stephan chopper;
200 g analytical sample further homogenized with Warring blender.
Fruiting vegetables & cucurbits (n=128, N=2296)

6
4
ln(Variance of residues)
2
0
-7 -6 -5 -4 -3 -2 -1 -2 0 1 2 3
-4
-6
-8
-10
-12
-14
ln(mean residue)
Figure 2. Relationship of the logarithm of average and variance of residues in fruiting
vegetables and cucurbits observed in supervised field trials. Each data point represents the
variance of residues measured in composite samples (number of primary samples ≥12) of one
crop species treated with the same pesticide at similar dosage rate in ≥ 8 independent trials; n
= the number of crop-pesticide combinations; N = number of residue data. The entire
laboratory samples were comminuted and their representative portions analyzed. Details of
trials are given in reference 3.

Figure 3. Example for laboratory sampling (Steps 2-5) of head cabbage. Upper left: head
cabbage with outer leaves; upper right: part of outer leaves; middle left: part of cabbage to be
used for testing compliance with MRLs; middle right: sections of cabbage head after

The Journal of AOAC INTERNATIONAL Page 32 of 45
1
2
3 quartering; bottom: comminution at ambient temperature (left) and with dry ice (right)
4
5
6 applying Stephan UM 12 chopper.
7
8
9
10
11

12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Laboratory sample
Steps: 1-3; see Table 1
Splitting, subsampling &

comminution; Steps: 4, 5
Subsamples
A B C D E
15 g Tp selection & analyses A15(1) B15(1) C15(1) D15(1 E15(1)

CVL(15)
Steps 5,6,7 )
A15(2) B15(2) C15(2) D15(2 E15(2)
) )
200 g analytical sample selection & PE

PA PB PC PD
comminution;
Steps 5,6,7
15 g Tp selection & analyses PA15(1 PB15(1 PC15( PD15(1) PE15(

CVL(200/15)
))) )) 1) 1)
Steps 5,6,7 PA15(2 PB15(2 PC15(2 PD15(2) PE15(2
)) ) ) )
Figure 4. Flowchart of laboratory sampling of large fruits

I
I
/ IV/1 I/1
1
IV I
IV/2 I/2

III/1 II/1
III II
III/2 II/2
I
Figure 5. Coding of sections of large crops

Table 1. Steps of the determination of pesticide residues in plant materials
Steps Description Comment, examples
Laboratory Sample

1 Receipt and verification of compliance E.g. verify product identity, packing,
with sampling protocol labelling, minimum mass and number of
units.
2. Extraneous materials removal E.g. loose, withered soiled leaves of
cabbage (Figure 2); gentle rinsing of
adhering soil from root vegetables; plant
materials from cereal grains (CACGL-41-
1993).13
3 Commodity portion selection (whole E.g. for testing compliance with MRL:
commodity, edible portion, etc.) whole units unless specified otherwise by
regulations. (CACGL-41-1993)
For dietary intake calculation: peeling,
shelling, juicing, etc. to prepare the edible
portion.
4 Splitting and subsampling (only if E.g. watermelons, heads of cabbages, jack
necessary) fruit and other very large food commodities
may require subsampling to obtain a
representative portion that can be handled
with the choppers, grinders etc. available in
the laboratory.

5. Comminution E.g. chopping, grinding, blending to obtain
a well-mixed material with smallest

particles possible.
Analytical Sample
6. Test portion selection E.g. a mass reduction step conducted by

selecting and combining multiple small
increments up to the total test portion mass.
(Smaller test portion masses may result in
larger measurement uncertainty. A fit-for-
purpose test portion mass should be
selected.)
Test Portion
7. Analysis All analytical operations required to
measure the concentration of the analyte
(measurand) e.g. extraction, evaporation,
cleanup, identification and quantitative
determination of the analyte.
8 Recovery test with spiked blank test Blanks, spikes encompassing the
portions concentration of residues expected in the
samples. (Blank sample material is
prepared similarly to the test portion.)
9 Reporting test result Residue conc±SD
10 Ongoing quality control Reanalysis of retained test portions.
Reference material testing, etc..

Note: Codex and EU documents often refer to steps 2 & 3 as sample preparation and steps 4-6 as
sample processing.

Table 2. Results* of recovery tests
Spike level [mg/kg] 0.2 1.5
Residues measured 0.172 1.62
0.196 1.43
0.209 1.37
0.205 1.37
0.168 1.58
Ave. residue 0.190 1.47
Recovery [%] 95.0 98.3
CV8 [%] 10.0 8.0
Ave. rec. % 96.6
Pooled CV8 [%] 9.0
*: The numbers given in the table are rounded, but the calculations were performed with Excel
without rounding.

Table 3. Results of the laboratory sampling and analyses of papaya fruits
Subsample A B C D E Ave CV
Subsample mass [g] 3188 4008 3758 4231 3542 3745 0.108

Residues1 [mg/kg]
in 15g TP 0.61 0.78 0.16 0.04 1.31 0.93 1.40 0.61 1.80 0.38 0.804 0.703a
Average residue2 0.695 0.103 1.12 1.01 1.09 0.804 0.531b
Residues [mg/kg]
in 200/15 TP3 0.69 0.58 0.86 0.64 0.72 0.86 0.65 0.75 0.74 0.78 0.723 0.125c
Residues in 200/15
g TP + H2O 4 0.76 0.65 0.95 0.72 0.80 0.96 0.72 0.83 0.82 0.86 0.807
Average residue5 0.705 0.833 0.882 0.776 0.842 0.807 0.085e
Notes: The numbers given in the table are rounded, the calculations were performed with Excel without rounding.
1: Residues [mg/kg] measured in 15g test portions (TP) taken from subsamples (A-E); 2: Average residue [mg/kg] in
duplicate 15 g TP taken from subsamples A-E; 3: Residues [mg/kg] measured in 15 g TP selected from 200 g
comminuted subsamples (A11-E11; A12-E12); 4; Residues [mg/kg] in 200/15 g TP corrected for 10% added water; 5:
Average residues in 200/15 TP corrected for water.

a: CVL15 calculated from the residues measured in 10 times 15 g test portions (A15(1)-E15(2))
b: CV’15 of average residues in 15 g test portions; c: CVL200/15 calculated from 15 g TP taken from the 200 g portions;
d: CVR200 of average residues corrected for added water in 200 g analytical samples.

Table 4. Results of re-analyses of retained test portions
Residues [mg/kg] measured 1
Date of analyses1 Analytical sample2 Average residue3 in TP200/15 Residue in retained TP4

Day 0 PA 0.705 0.663
Day 1 PA 0.705 0.534
Day 2 PA 0.705 0.872
Day 5 PA 0.705 0.816
Day 10 PB 0.833 0.764
Day 16 PB 0.833 0.989
CVL200/15 [%] 15.5
1: Retained test portion were analyzed together with the samples received from the trial site.
2: Analytical sample from which the test portions were selected for future analyses
3: Residue measured during development of test procedure.
4: Residues measured in retained TP at T1-T6 occasions.

Table 5. One-way ANOVA table for results of analyses of 2 test portions
SUMMARY: 15g Test portions taken from A-E subsamples*
Groups Count Sum Average Variance
Row 1 2 1.389 0.695 0.0134

Row 2 2 0.206 0.103 0.0067
Row 3 2 2.247 1.123 0.0721
Row 4 2 2.018 1.009 0.3132
Row 5 2 2.182 1.091 1.0144
ANOVA
Source of VAR SS df MS F P-value F crit.
Between Groups1 1.4597 4 0.3649 1.285 0.3870 5.192
Within Groups2 1.4199 5 0.2840
Total 2.8796 9
SUMMARY: 15g Test portions taken from PA-PE 200 g analytical samples*
Groups Count Sum Average Variance
Row 1 2 1.410 0.705 0.0065
Row 2 2 1.666 0.833 0.0277
Row 3 2 1.764 0.882 0.0122
Row 4 2 1.552 0.776 0.0059
Row 5 2 1.683 0.842 0.0008
ANOVA
Source of VAR SS df MS F P-value F crit.
Between Groups3 0.0377 4 0.0094 0.886 0.534 5.192

Within Groups4 0.0532 5 0.0106
Total 0.0909 9
*: The 7 digit values shown in Excel output tables are rounded for clarity
1: Average residues in A-E subsamples

2: Residues measured in 15 g duplicate test portions taken from A-E subsamples
3: Average residues in PA-PE 200 g analytical samples
4: Residues measured in duplicate test portions taken from 200 g portions (PA15(1)- PE15(1) and PA15(2)-
PE15(2))

Table 6. CV values and variances of steps of laboratory sampling of papaya fruits*
CV% V
Residues [mg/kg] in 15g TP from SS 70.3a 0.3199

Average residues in subsamples (A-E) 53.1b 0.1821
Residues in 15g Tp1 12.4c 0.0114
Average residues in 200 g anal.
samples2 8.5d 0.0047
CVA 9.0 0.0053
*: Values reported are rounded, but the calculations were made with Excel with all digits
1: Tp: test portion [PA15(1)-PE15(2)] corrected for added water
2: Estimated from duplicate 15 g test portions selected from the comminuted analytical samples
a: CV-s calculated from 15 g test portions A1--E2; b: CVL15 of average residues in 15 g test
portions; c: CVLp200/15 calculated from 15 g TP taken from the 200 g portions; d: CV of average
residues corrected for added water in 200 g analytical samples.

Table 7. Contribution of laboratory sampling steps to combined uncertainty1
Subsample2 Analytical.sample3
V % V %
VL=VT 0.3199 100% 0.0114 100%

Step 4 (V4) 0.1821 56.9% 0.0047 41.4%
Steps 5-6 (V5-6) 0.1326 41.4% 0.0013 11.7%
Analysis (V8) 0.0053 1.7% 0.0053 46.9%
1: Values reported are rounded, but the calculations were made with Excel with all digits
2: Selecting 15 g test portions from subsamples

3: Selecting 15 g test portions from 200 g analytical samples
Table 8. Examples for the relationship of CV-s [%] of steps 4, 5-6 and 7, and CVL*
Options CVL(200/15) CV4 CV5-6 CVA
1 13.2 8.5 4.5 9.0
2 7.7 4.3 4.5 4.5
3 6.6 4.3 2.3 4.5
4 12.6 8.5 2.3 9.0
5 5.4 4.0 2.0 3.0
*: Figures reported are rounded, but the calculations are performed with Excel using all digits.

Table 9. Allocation of sections of large crops to replicate subsamples
Sections representing 1/8 Sections representing 1/4
portion of fruits portion of fruits
AI/1 AIII/2 AI AIII

BII/1 BIV/2 BII BIV
CIII/1 CI/2 CIII CI
DIV/1 DII/2 DIV DII
EI/1 EIII/2 EI EIII
A-E code for individual crop units; Roman number: quarter of crop;
Arabic number: split quarters (1/8 portion of crop)
View publication stats

Ambrus Et Al (2020) - Sources of Random Variation of Pesticide Residue Analytical Results

Uploaded by

Copyright:

Available Formats

Ambrus Et Al (2020) - Sources of Random Variation of Pesticide Residue Analytical Results

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ambrus Et Al (2020) - Sources of Random Variation of Pesticide Residue Analytical Results

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Sources of Random Variation of Pesticide Residue Analytical Results

Article in Journal of AOAC International · September 2020

Arpád Ambrus Kata Kerekes

SEE PROFILE SEE PROFILE

Sampling View project

Project BASELINE EU FP7 2013 View project

The user has requested enhancement of the downloaded file.

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

Egyetem tér 1., Hungary

Budapest, Kis Rókus str. 15/b., Hungary

3526 Miskolc, Blaskovics L. str. 24., Hungary

Analytical National Reference Laboratory, 2481 Velence, Ország út 23., Hungary

Corresponding author: [email protected]

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

comminution of analytical sample and selection of small test portions.

intermediate precision. Methods: Validation of laboratory sampling of large fruits is used to

(uncertainty) of measured residues.

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

samples) comprising in a composite sample is specified by the corresponding sampling

leading to fundamental, as well as grouping and segregation error (8).

determined in hundreds of thousands of samples by governmental and private laboratories to

procedure to estimate the error due to Steps 4-7 is described.

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

used in this paper.

Selection of the portion of the commodity to be tested (Steps 2-3)

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

Calculation of laboratory sampling and analysis errors (Steps 4-7)

standard deviations (CV):

𝐶𝑉𝐿 = 𝐶𝑉24 + 𝐶𝑉25 ― 6 + 𝐶𝑉27 (2)

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

mass or volume of an analytical sample containing an incurred residue.

taking into account the expanded uncertainty calculated from CVL:

𝐶𝑅 ―(2 × 𝐶𝑉𝐿 × 𝐶𝑅) ≤ 𝑀𝑅𝐿 (3)

The default value of CVL is 25% in the European Union (13).

Obtaining test portions from analytical samples

Several publications revealed the potentially significant contribution of comminution (Step 5) to

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

their contribution negligible (30-34).

gradually amended to accommodate pH dependent pesticides and commodities of different water

Quality control of comminution and selection of test portion (Steps 5-7)

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

influenced by the type and condition of equipment used.

concentration calculated for the entire laboratory sample (27).

approximately be described as:

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

can be calculated as:

𝐶𝑉5 ― 6 = 𝐶𝑉25 ― 7 ― 𝐶𝑉27 (5)

more representative portions.

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

indicated by the recoveries obtained from spiked test portions.

𝐶𝑉,,5 ― 6 = 𝐶𝑉25 ― 6(1) + 𝐶𝑉25 ― 6(2) (6)

(CVL) of the determination of pesticide residues.

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

materials and soil (19, 21, 24, 46, 52, 54).

Recovery tests to evaluate method performance

Downloaded from https://academic.oup.com/jaoac/advance-article/doi/10.1093/jaoacint/qsaa119/5900660 by guest on 04 September 2020

spiked samples (Table 2).

samples would be analyzed.

Developing method for determination of residues in large crop units