跳到正文
EMA Scientific Guidelines·· 10 天前精选AI 评分63

VICH GL49(R2)食品动物用兽药残留消除试验分析方法验证指南草案征求意见

VICH GL49: Studies to evaluate the metabolism and residue kinetics of veterinary drugs in food producing animals: validation of analytical methods used in residue depletion studies - Revision 2

AI 导读

EMA/CVMP发布VICH GL49(R2)草案,就食品动物用兽药残留消除试验分析方法的验证要求征求意见,截止2027年4月30日。草案适用于组织、奶、蛋和蜂蜜中兽药残留分析方法,列出线性、准确度、精密度、LOD、LOQ、选择性及基质和处理样品稳定性等性能指标和接受标准。文件仍处VICH第4步征求意见阶段,按程序最终草案将在第7步推荐给欧盟、日本和美国监管机构采纳。

推荐理由

草案列出残留消除试验分析方法的线性、准确度、精密度、检出限与定量限及稳定性等验证指标和接受标准,并附估算示例。

正文 · 原文

PDF 文字版;图形和原始排版请参阅官方 PDF。

第 1 页

13 November 2026
EMA/CVMP/VICH/463202/2009
Committee for Veterinary Medicinal Products (CVMP)




VICH GL49: Studies to evaluate the metabolism and
residue kinetics of veterinary drugs in food producing
animals: validation of analytical methods used in residue
depletion studies (revision 2)


Draft agreed by VICH Steering Committee                                                                                October 2026

Adoption by CVMP for release for consultation                                                                   5 November 2026

Transmission to interested parties                                                                           13 November 2026

End of consultation (deadline for comments)                                                                            30 April 2027




Official address Domenico Scarlattilaan 6 ● 1083 HS Amsterdam ● The Netherlands

Address for visits and deliveries Refer to www.ema.europa.eu/how-to-find-us
Send us a question Go to www.ema.europa.eu/contact          Telephone +31 (0)88 781 6000             An agency of the European Union


© European Medicines Agency, 2026. Reproduction is authorised provided the source is acknowledged.

第 2 页

                                VICH GL 49 (R2) (MRK) – METHOD USED IN RESIDUE DEPLETION STUDIES
                                                                                             October 2026
                                                                                        Revision at Step 9
                                                                                 For consultation at Step 4




S TUDIES TO E VALUATE THE M ETABOLISM
  AND R ESIDUE K INETICS OF V ETERINARY
    D RUGS IN FOOD P RODUCING A NIMALS :
    V ALIDATION OF A NALYTICAL M ETHODS
    USED IN R ESIDUE D EPLETION S TUDIES



                                            Revision at step 9

                    Recommended for Consultation at Step 4 of the VICH Process
                                       in October 2026
                               by the VICH Steering Committee
                                               .



This Guideline has been developed and revised by the appropriate VICH Expert Working Group and is
subject to consultation by the parties, in accordance with the VICH Process. At Step 7 of the Process
the final draft will be recommended for adoption to the regulatory bodies of the European Union, Japan
                                                 and USA.




         Secretariat: c/o HealthforAnimals, Rue d’Idalie 9-13, Box 5, B - 1050 Brussels (Belgium)
                        e-mail : sec@vichsec.org - Website : http://www.vichsec.org




                                                                                                    Page 2 of 22

第 3 页

1. INTRODUCTION ............................................................................................................... 4
  1.1. Objective of the guidance ................................................................................................ 4
  1.2. Background ................................................................................................................... 4
2. GUIDANCE ....................................................................................................................... 4
  2.1 Purpose .......................................................................................................................... 4
  2.2. Scope ............................................................................................................................ 4
3. PERFORMANCE CHARACTERISTICS .............................................................................. 5
  3.1. Linearity........................................................................................................................ 5
  3.2. Accuracy ....................................................................................................................... 6
  3.3. Precision ....................................................................................................................... 6
  3.4. Limit of Detection .......................................................................................................... 7
  3.5. Limit of Quantitation ...................................................................................................... 7
  3.6. Selectivity ..................................................................................................................... 7
  3.7. Stability in Matrix .......................................................................................................... 7
  3.8. Processed Sample Stability .............................................................................................. 8
  3.9. Robustness .................................................................................................................... 8
4. GLOSSARY ....................................................................................................................... 9




                                                                                                                               Page 3 of 22

第 4 页

1. INTRODUCTION

1.1. Objective of the guidance
This guidance document is intended to provide a general description of the criteria that have
been found by the European Union (EU), Japan, United States of America (USA), Australia,
New Zealand and Canada to be suitable for the validation of analytical methods used in
veterinary drug residue depletion studies.

1.2. Background
During the veterinary drug development process, residue depletion studies are conducted to
determine the concentration of the residue or residues present in the edible products (tissues,
milk, eggs or honey) of animals treated with veterinary drugs. This information is used in
regulatory submissions around the world. Submission of regulatory methods (i.e., post approval
control methods) and the validation requirements of the regulatory methods are usually well
defined by various regulatory agencies worldwide and might even be defined by national or
regional law. However, the residue depletion studies are generally conducted before the
regulatory methods have been completed. Often times the in-house validated residue methods
provide the framework for the methods submitted for regulatory monitoring. Harmonization of
the validation requirements for methodology used during residue depletion studies and
submitted to the regulatory agencies in support of the maximum residue limits (MRLs) and
withdrawal periods should be achievable. It is the intent of this document to describe a
validation procedure that is acceptable to the regulatory bodies of the VICH regions for use in
the residue depletion studies. This validated method could continue on to become the
“regulatory method” but that phase of the process will not be addressed in any detail in these
guidelines.
A variety of validation guidelines exist for analytical methodology and many of the aspects of
those validation procedures are incorporated in this document (VICH GL1 (Validation
Definition), October 1998 and VICH GL2 (Validation Methodology), October 1998). However,
there are aspects of residue validation procedures that are addressed in this guidance document
that are not addressed in previous documents. The guidance provided here is intended to
specifically address the validation of veterinary drug residue methods.

2. GUIDANCE
2.1 Purpose
The purpose of this document is to provide a general description of procedures that can be used
for the validation of the analytical methods developed for the analysis of tissue samples obtained
in residue depletion studies.
For purposes of this guidance, “acceptable” refers to the scientific evaluation of the analytical
method in terms of the described validation criteria.

2.2. Scope
This guidance is only intended to apply to analytical procedures that have been developed for
the evaluation of veterinary drug residue methods (assays developed to determine residues in
marker residue depletion studies). It is not intended to define the criteria needed for validation
of regulatory monitoring assay procedures.
This document provides performance characteristics of the residue assays that if followed would
generally be considered acceptable by the regulatory agencies in the VICH regions. The intent

                                                                                        Page 4 of 22

第 5 页

is that methods validated according to this guidance will provide residue data that would
generally be considered acceptable by the regulatory agencies in determining appropriate
withdrawal periods.
While it is recognized that residue studies must be conducted under GLP using validated
methodology, the actual method validation experiments do not fall within the scope of the GLP
regulations. However, raw data generated as a result of a method validation should be archived
as appropriate and be available for submission to regulatory authorities upon request.
3. PERFORMANCE CHARACTERISTICS
In general, there are specific performance characteristics of a method validation. Those
performance characteristics are defined as follows:
       Linearity
       Accuracy
       Precision
       Limit of Detection
       Limit of Quantitation
       Selectivity
       Stability in Matrix
       Process Sample Stability
       Robustness
Each of the characteristics will be described below as they apply to the validation of methods
intended for use in veterinary drug residue depletion studies.

3.1. Linearity
A calibration curve should be generated in which the linear relationship is evaluated across the
range of the expected matrix (tissue, milk, egg or honey) concentrations. Calibration standard
curves can be generated in three formats depending upon the methodology: standards in
solvent/buffer, standards fortified into control matrix extract and standards fortified into control
matrix and processed through the extraction procedure. Linearity should be described by a
linear, polynomial or other (as appropriate) regression plot of known concentration vs. response
using a minimum of 5 different concentrations. Acceptability of weighting factors should be
determined by evaluation of the residuals across three runs to determine if the residuals are
randomly distributed. Evaluation of the residuals should be carried out across at least three
separate runs.
The recommended acceptance criterion for a standard curve is dependent upon the format of the
standard curve. Calibration standard curves generated by fortification of control matrix and
processed through the procedure are subject to the same acceptance criteria as the samples (see
Section 3.3. Precision). Calibration standard curves generated by standards in solvent/buffer or
by fortification of control matrix extract would require more stringent acceptance criteria
(Repeatability ≤ 15% at all concentrations except at or below LOQ where it can be ≤ 20%).
Some assays (e.g., microbiological assays) could require log transformations to achieve linearity
where other assays (e.g., ELISA, RIA) could require a more complicated mathematical function

                                                                                          Page 5 of 22

第 6 页

to establish the relationship between concentration and response. Again, acceptability of the
function selected should be verified by evaluation of the residuals generated when that function
is used.

3.2. Accuracy
Accuracy refers to the closeness of agreement between the true value of the analyte
concentration and the mean result that is obtained by applying the experimental procedure.
Accuracy is closely related to systematic error (analytical method bias) and analyte recovery
(measured as percent recovery). Recommended accuracy for residue methods will vary
depending upon the concentration of the analyte. The accuracy should meet the range listed
below:
               Analyte Concentration* Acceptable Range for Accuracy
               < 1 µg/kg                  -50 % to +20 %
               ≥ 1 µg/kg < 10 µg/kg       -40 % to +20 %
               ≥ 10 µg/kg < 100 µg/kg -30 % to +10 %
               ≥ 100 µg/kg                -20 % to +10 %
        * µg/kg =ng/g = ppb

3.3. Precision
Precision of a method is the closeness of agreement between independent test results obtained
from homogenous test material under stipulated conditions of use. Analytical variability
between different laboratories is defined as reproducibility, and variability from repeated
analyses within a laboratory is repeatability. Single-laboratory validation precision should
include a within-run (repeatability) and between-run component.
The within- and between-run precision of the analytical method can be determined as part of the
validation procedure. There is generally not a need to determine reproducibility (between-
laboratory precision) in order to conduct a residue depletion study, because the laboratory that
is developing the method is often the same laboratory assaying the samples from the residue
study. Instead of establishing reproducibility of the assay, a between-run precision, can be
determined. Within- and between-run precision should be determined by the evaluation of a
minimum of three replicates at three different concentrations representative of the intended
validation range (which should include the LOQ) across three days of analysis.
For the purposes of the residue method validation, acceptable variability is dependent upon the
concentration of the analyte. The precision should meet the range listed below.:


                                             Acceptable    within-run Acceptable between-run
                Analyte Concentration        precision (Repeatability), precision %CV*
                                             %CV
                < 1 µg/kg                    30 %                         45%
                ≥ 1 µg/kg < 10 µg/kg         25 %                         32%
                ≥ 10 µg/kg < 100 µg/kg       15%                          23%

                                                                                      Page 6 of 22

第 7 页

                   ≥ 100 µg/kg                 10 %                         16%
              * as determined by the Horwitz equation CV = 2(1-0.5 log C) where C =
concentration                    expressed as a decimal fraction (e.g., 1 µg/kg is entered as
  -9
10 ).

3.4. Limit of Detection
The limit of detection (LOD) is the smallest measured concentration of an analyte from which
it is possible to deduce the presence of the analyte in the test sample with acceptable certainty.
There are several scientifically valid ways to determine LOD and any of these could be used as
long as a scientific justification is provided for their use. See Annexes 1, 2, and 3 for examples
of acceptable methods for determining LOD.
3.5. Limit of Quantitation
The LOQ is the smallest measured content of an analyte above which the determination can be
made with the specified degree of accuracy and precision. As with the LOD, there are several
scientifically valid ways to determine LOQ and any of these could be used as long as scientific
justification is provided. See Annexes 1, 2, and 3 for examples of acceptable methods for
determining LOQ.
3.6. Selectivity
Selectivity is the ability of a method to distinguish between the analyte being measured and
other substances which might be present in the sample being analyzed. For the methods used
in residue depletion studies, selectivity is primarily defined relative to endogenous substances
in the samples being measured. Because the residue depletion studies are well controlled,
exogenously administered components (i.e., other veterinary drugs or vaccines) could either be
known or not be allowed during the study. If it is the intent to submit the validated method as a
regulatory method, it might be prudent for the investigator to test known products used in the
animals being tested for possible interference.
A good measure of the selectivity of an assay is the determination of the response of control
samples. That response should be no more than 20% of the response at the LOQ (see Section
3.5 above). See Annex 3 for a suggested protocol for determining accuracy, precision, LOD,
LOQ and selectivity in a single study.
3.7. Stability in Matrix
Samples (tissue, milk, eggs or honey) collected from residue depletion studies are generally
frozen and stored until assayed. It is important to determine how long these samples can be
stored under the proposed storage conditions without excessive degradation prior to analysis.
As part of the validation procedure or as a separate study, a stability study needs to be conducted
to determine the appropriate storage conditions (e.g., 4°C, -20°C, or -70°C) and length of time
the samples can be stored prior to analysis.
Samples should be fortified with known quantities of analyte and stored under the appropriate
conditions. Samples should be periodically assayed at specified intervals (e.g., initially, 1 week,
1 month, 3 months). If the samples are frozen, freeze/thaw studies should be conducted (3
freeze/thaw cycles – one cycle per day at a minimum). Alternatively, incurred samples can be
used with initial assays conducted to determine the starting concentrations. The recommended
protocol for assessing stability in matrix is the analysis of two different concentrations in
triplicate near the high and low end of the validation range. Stability in matrix is considered
acceptable if the mean concentration obtained at the specified stability time point agrees with

                                                                                         Page 7 of 22

第 8 页

the initial assay results or freshly fortified control sample assay results within the accuracy
acceptance criteria established in Section 3.2.
3.8. Processed Sample Stability
Often, the samples are processed one day and assayed on a second day or because of an
instrument failure are stored additional days, e.g. over a weekend. The stability of the analyte
in the process sample extract might be examined as necessary to determine stability under
processed sample storage conditions. Examples of storage conditions would be 4 to 24 hours at
room temperature and 48 hours at 4°C. Other storage conditions might be investigated
consistent with the method requirements. The recommended protocol for assessing processed
sample stability is the analysis of two different concentrations in triplicate near the high and low
end of the validation range. Processed sample stability is considered adequate if the mean
concentration obtained at the specified stability time point agrees with the initial assay results
or with freshly fortified and processed control sample assay results within the accuracy
acceptance criteria established in Section 3.2.
3.9. Robustness
Evaluation of the robustness of regulatory methods is of major importance. Evaluation of
robustness for residue methodology is less of a concern for residue methods as these are usually
conducted within a single laboratory using the same instrument. However, robustness should
still be evaluated particularly for areas of the method that could undergo changes or
modifications over time. These might include reagent lots, incubation temperatures, extraction
solvent composition and volume, extraction time and number of extractions, solid phase
extraction (SPE) cartridge brand and lots, analytical column brand and lots and HPLC elution
solvent composition. During the development, validation or use of the assay, method sensitivity
to any or all of these conditions can become apparent and variations in the ones most likely to
affect the method performance should be evaluated.




                                                                                          Page 8 of 22

第 9 页

4. GLOSSARY
Accuracy – The accuracy of an analytical procedure expresses the closeness of agreement
between the true value of the analyte concentration and the mean result that is obtained by
applying the analytical procedure. This is generally expressed as % recovery or % bias.
Between-run Precision – Between-run precision expresses within-laboratory between-run
variations.
Control sample – Tissue, milk, egg or honey from an animal that has not been treated with
the veterinary drug under investigation.
Incurred sample – Tissue, milk, egg or honey from an animal treated with the veterinary drug
under investigation that has a residue concentration of the analyte of interest.
Limit of Detection – The limit of detection of an individual analytical procedure is the lowest
amount of analyte in a sample that can be detected with acceptable certainty but not quantitated
as an exact value.
Limit of Quantitation – The limit of quantitation of an individual analytical procedure is the
lowest amount of analyte in a sample that can be quantitatively determined with acceptable
precision and accuracy.
Linearity – The linearity of an analytical procedure is its ability (within a given range) to
obtain test results that are directly proportional to the concentration (amount) of analyte in the
sample.
Marker residue – The residue whose concentration is in a known relationship to the
concentration of total residue in an edible tissue.
Matrix – The matrix is basic edible animal products (tissue, egg, milk or honey) that contains
or could contain the residue of interest.
Precision – The precision of an analytical procedure expresses the closeness of agreement
between a series of measurements obtained from multiple sampling of the same homogenous
sample under prescribed conditions. The precision of an analytical procedure is usually
expressed as the variance, standard deviation or coefficient of variation of a series of
measurements.
Processed Sample – A processed sample is a sample that has been extracted or otherwise
processed to remove the analyte from much of the original sample matrix.
Repeatability – Repeatability expresses the precision under the same operating conditions
over a short interval of time.
Reproducibility – Reproducibility expresses the precision between laboratories.
Residue – Veterinary drug (parent) and/or its metabolite.
Robustness – The robustness of an analytical procedure is a measure of its capacity to remain
unaffected by small variations in method parameters and provides an indication of its reliability
during normal usage.
Selectivity – Selectivity is the ability to assess the analyte in the presence of components
(endogenous materials, degradation products, other veterinary drugs) that might be expected
to be present.

                                                                                        Page 9 of 22

第 10 页

Within-run Precision – Within-run precision expresses within-laboratory within-run
variations.




                                                                         Page 10 of 22

第 11 页

                                          Annex 1
                  Examples of Methods for Determining LOD and LOQ
One commonly used approach is referred to as the IUPAC definition. 1 In that procedure the
LOD is estimated as mean of 20 control sample (from at least 6 separate sources) assay results
plus 3 times the standard deviation of the mean. The LOQ then becomes the mean of the same
results plus 6 or 10 times the standard deviation of the mean. Testing of the accuracy and
precision at the estimated LOQ will provide the final evidence for determination of the LOQ.
If the %CV for the repeatability measurement at that concentration is less than or equal to the
accuracy and precision acceptance criteria (Section 3.2 and 3.3), then the estimated LOQ is
acceptable.




                                                                                    Page 11 of 22

第 12 页

                                            Annex 2
       U.S. Environmental Protection Agency Method for Determining LOD and LOQ


The procedure described below is a slight modification of a procedure used by USDA’s
Interregional Project No. 4 program, which is published in 40 CFR Part 136, Appendix B. 2 This
modified procedure can be found in Appendix 1 of the U.S. Environmental Protection Agencies
document entitled “Assigning Values to Non-detected/Non-quantified Pesticide Residues in
Human Health Food Exposure Assessments. 3 The procedure is provided below with minor
modifications making it more representative of a tissue marker residue assay procedure
example.
In this procedure, the estimation of the LOD and LOQ of a specific method for a specific analyte
in a specific matrix can be done in the following two steps.
   •     The first step is to produce a preliminary estimate of the LOD and LOQ and to verify
         that a linear relationship between concentration and instrument response exists. These
         preliminary estimates correspond to what some term the IDL (Instrument Detection
         Limit) and IQL (Instrument Quantitation Limit), respectively. The matrix of interest
         will be fortified (spiked) at the estimate LOQ in the next step for the actual estimation
         of LOD and LOQ of the method.
   •     The second step is to use the initial estimate of the LOD and LOQ determined in Step 1
         to estimate the method detection limit and the method quantitation limit in the matrix of
         interest.
An illustrative example follows:


Step 1. The analyst derives a standard curve for the method of interest. In this particular
instance, the analyst prepares the standard solution in buffer or water with the following
concentrations of the analyte of interest: 0.005, 0.010, 0.020, 0.050 and 0.100 µg/mL. For each
concentration in the sample solution, the following instrument responses (measure peak height)
are recorded:
                                                       Instrument Response (peak
                          Concentration (µg/mL)
                                                                height)
                                   0.100                         206,493
                                   0.050                         125,162
                                   0.020                          58,748
                                   0.010                          32,668
                                   0.005                          17,552


In order to verify that a linear response is seen throughout the tested range, the instrument
response is plotted as a function of injected concentration. The results (and associated statistics)
are shown in Figure 1. Note from these results that the instrument response appears to be

                                                                                         Page 12 of 22

第 13 页

adequately linear throughout the range of tested concentrations (0.005 to 0.100 µg/mL) with a
high R2 value of 0.99 (see the “Summary of Fit” table in Figure 1). The residual standard
deviation (root mean square error) is 8986.8. The equation which describes this relationship
(provided in the “Parameter Estimates” table in Figure 1) is as follows:
        Y = 15,120 + 1,973,099 * (Concentration)
Where Y is the instrument response (peak height)
The estimated LOD and LOQ are calculated as follows (assuming these values are set to 3 and
10 standard deviations above the blank response, respectively):
   1. The Peak Height at the LOD (YLOD) is calculated at 3 times the standard deviation while
       the Peak Height at the LOQ (YLOQ) is calculated at 10 times the standard deviation
                      YLOD = 15120 + 3 * (8987) = 42,081
                      YLOQ = 15120 + 10 * (8987) = 104,990
   2. These values (peak height at LOD and peak height and LOQ) are then used to calculate
      the concentrations associated with these peak heights as follows:
                      Y = 15,120 + 1,973,099 * (Concentration)
     Rearranging,
                      Concentration = (Y – 15,120) / 1,973,099
     Therefore,
        LOD = YLOD - 15,120 / 1,973,099 = (42,081 - 15,120) / 1,973,099 = 0.014 µg/mL
        LOQ = YLOQ - 15,120 / 1,973,099 = (104,990 - 15,120) / 1,973,099 = 0.046 µg/mL
   Thus, the initial estimated LOD and LOQ are 0.014 and 0.046 µg/mL, respectively which
   correspond to the IDL and IQL.
   These estimated LODs (or IDLs) and LOQs (or IQLs) are expressed in terms of the solution
   concentration and not in terms of the matrix concentration. At this stage, the solution
   concentration (µg/mL solution) should be converted to the effective concentration in the
   matrix (e.g., µg/g of matrix).




                                                                                   Page 13 of 22

第 14 页

Figure 1. Statistical Results




                                Page 14 of 22

第 15 页

Step 2. With the initial estimate of LOD (or IDL) and LOQ (or IQL) obtained and linearity
verified, Step 2 involves estimating the LOQ and LOD in spiked matrix samples. This
procedure uses the estimated instrumental LOQ and the procedure detailed in 40 CFR Part 136,
Appendix B2 to provide a better estimate of LOQ and verifies that method recoveries are
acceptable.
The method calls for the analysis of 7 or more untreated control samples spiked at the estimated
LOQ. The standard deviation of these samples is measured and the LOD and LOQ are
determined as follows:
    LOD = t0.99 * S
    LOQ = 3 * LOD
where       t = one-tailed t-statistic at the 99% confidence level for n-1 replicates
            S = Standard Deviation of n sample spikes at the estimated LOQ
The following is a set of t-values for use in the above equation:


    # of                                                 # of
                       Degrees of                                        Degrees of
  Replicates                               t0.99       Replicates                         t0.99
                     Freedom (n-1)                                     Freedom (n-1)
     (n)                                                  (n)
        3                   2             6.965            13                 12         2.681
        4                   3             4.541            14                 13         2.650
        5                   4             3.747            15                 14         2.624
        6                   5             3.365            16                 15         2.602
        7                   6             3.143            17                 16         2.583
        8                   7             2.998            18                 17         2.567
        9                   8             2.896            19                 18         2.552
        10                  9             2.821            20                 19         2.539
        11                  10            2.764            21                 20         2.528
        12                  11            2.718            22                 21         2.518


In this example, the analyst prepared 7 untreated control samples spiked at the above estimated
LOQ of 0.05 µg/g. The following results were obtained:




                           Concentration detected
                                                                % Recovery
                                  (µg/g)


                                                                                        Page 15 of 22

第 16 页

                                0.0397                       79.4
                                0.0403                       80.6
                                0.0400                       80.0
                                0.0360                       72.0
                                0.0498                       99.6
                                0.0379                       75.8
                                0.0388                       77.6


Average Concentration: 0.0404 µg/g
Standard Deviation: 0.0044 µg/g
Average Recovery: 80.7%


Given that recoveries are adequate at the LOQ (average = 80.7%, range = 72.0% to 99.6%), the
LOD and LOQ for the method are estimated as follows:


    LOD = t0.99 * S (for 7-1 = 6 degrees of freedom)
        = 3.143 * 0.0044 µg/g
        = 0.0138 µg/g


    LOQ = 3 * LOD
        = 3 * 0.0138 µg/g
        = 0.0414 µg/g




                                                                                  Page 16 of 22

第 17 页

                                          Annex 3
                          Protocol for Residue Method Validation
Selectivity, LOD and LOQ are all interrelated and are affected by endogenous interferences that
might be present in the matrix being assayed. LOD is often difficult to determine, particularly
in LC/MS assays where control samples actually provide zero response at the retention time of
the analyte. Without a response, it is impossible to calculate a standard deviation (SD) and
therefore impossible to determine the LOD based on the mean plus 3 times the SD of the mean.
Even if a mean plus 3 times the SD of the mean can be determined, it is often related to the
instrument limit of detection rather than the method limit of detection. The following protocol
is designed to determine accuracy, precision, selectivity, LOD, and LOQ in one study.
   1. Collect drug-free matrix from 6 separate sources (animals) and screen for any possible
      analyte contamination.
   2. From the 6 sources, randomly select at least 3 samples to fortify (spike) at the estimated
      LOD (eLOD, determined during preliminary assay development), at the estimated LOQ
      (eLOQ, 3x the eLOD), and at 3 other concentrations that cover the expected
      concentration range. Repeat the fortification process for runs on Day 2 and Day 3 using
      a second and third set of 3 randomly selected samples per fortification level. Ensure that
      each source is represented at least once per fortification level across the 3 days. See
      Table 1 for an example.

Table 1 – Example of a minimal study design with six sources (animals) A through F
Fortification level                   Day/Run 1           Day/Run 2           Day/Run 3
0 (Control)                             B, F, D            A, C, C              B, E, F
eLOD                                    B, C, E             D, F, F             A, B, E
eLOQ                                    C, C, E             A, B, E             D, F, D
Lower part of validation range          A, B, E            A, C, D              B, E, F
Middle of validation range              B, C, E             C, E, F             A, D, F
Upper part of validation range         A, B, B              D, F, F             A, C, E


   3. Assay the 18 samples each day and calculate the concentrations from the calibration
      standard curve.
   4. Accuracy can be determined by comparison of the concentration found to the
      concentration added. Within- and between-run precision (i.e., intra- and inter-day
      precision) can be calculated as the %CV at each fortification level. Precision estimates
      incorporate variability due to the matrix, since the protocol uses 6 different sources of
      matrix. Acceptance criteria for accuracy and precision are provided in Sections 3.2 and
      3.3, respectively.
   5. If accuracy and precision are acceptable, plot the concentration found against the
      concentration added for all 3 days of assays. Whether variability remains constant across
      fortification levels (homoscedasticity) or increases (heteroscedasticity) can be assessed
      visually from the plot or by examining the residuals of the regression.


                                                                                     Page 17 of 22

第 18 页

6. With homoscedastic data, the LOD and LOQ can be calculated with the approach in
   Sections 6.3 and 7.3 of VICH GL2, respectively. A linear regression of concentration
   found vs concentration added is fitted, and then LOD is estimated as 3.3σ/s and LOQ as
   10σ/s, where σ is the regression’s residual SD and s is the slope. The approach can also
   be used with heteroscedastic data, but σ and s must be estimated from a weighted
   regression in which the residuals are scaled by their weights prior to calculating σ. In
   this protocol, σ is the residual SD of the regression. The alternative VICH GL2 methods
   to estimate σ (SD of blanks or SD of y-intercepts) are not applicable because control
   samples may show zero or near-zero responses in sensitive methods, and only 3 runs are
   available for calculations in this scenario.
7. Alternatively, LOD and LOQ can be derived from the prediction intervals around the
   regression 4, as shown in Figure 1. With homoscedastic data, ordinary regression can be
   used, but with heteroscedastic data, weighted regression should be used.5 The upper
   prediction interval is calculated to control the false positive rate (α), while the lower
   prediction interval is calculated to control the false negative rate (β). Typically, both α
   and β are 5%, so the prediction intervals are estimated to cover 90% of measurements,
   from the 5th to 95th percentile. The decision limit (YC) is where the upper interval crosses
   the y-axis. The corresponding point on the x-axis (LC, also called the critical value or
   CCα) represents the concentration where there is a 50% probability of correctly
   identifying the analyte as present. At the concentration where the decision limit
   intersects the lower prediction interval (also known as CCβ), the probability of detecting
   the analyte when it is present is 1-β. This concentration is taken as the LOD. The LOQ
   is derived similarly. A determination limit (YQ) is established by multiplying the
   decision limit (YC) by 3, which is a commonly accepted ratio between LOD and LOQ
   (see e.g., VICH GL2). Where the determination limit intersects the lower prediction
   interval is taken as the LOQ.
8. The selectivity of the assay is considered sufficient if the response of the blank, control
   samples is less than 20% of the LOQ, as noted in Section 3.6. If this criterion is not met,
   the LOQ should be adjusted upward to ensure adequate selectivity.




                                                                                    Page 18 of 22

第 19 页

Figure 1 – estimating LOD and LOQ from the prediction intervals around the regression of
concentration found vs concentration added. In this example, the intervals widen with increasing
concentration, indicative of heteroscedasticity.
Example:
A validation procedure based on the above methodology was conducted on an LC-MS/MS milk
assay procedure.
Control bovine milk samples obtained from 6 different animals were each fortified with the
analyte at 0, 4.2 (eLOD), 14.0 (eLOQ), 35, 140 and 400 ng/mL, giving a total of 36 samples. For
each run, milk samples from 3 of the 6 animals were randomly chosen for each fortification level,
ensuring that each of the 6 animals was selected at least once per level. This results in 18
measurements per day and a total of 54 measurements across 3 runs on 3 different days.
The following parameters were evaluated: accuracy, within-run precision (repeatability),
between-run precision, LOD and LOQ. The raw data are listed in Table 2.




         Table 2 – Concentration of analyte (ng/mL) in fortified control milk across
                                   three days of analysis
                        Run 1                   Run 2                  Run 3
          Conc.    Animal     Conc.       Animal     Conc.       Animal      Conc.
         added       ID       found         ID       found         ID        found
                      B        0.494         A        0.233         B         0.154
             0        F        0.654        C         0.012         E         0.120
                      D        0.588        C         0.117         F         0.313
                      B         4.38        D          4.97         A          3.80
            4.2       C         4.13         F         3.85         B          4.12
                      E         4.33         F         4.41         E          3.67
                      C         13.2         A         11.1         D          11.8
           14.0       C         13.5         B         12.0         F          10.5
                      E         11.9         E         12.8         D          11.7
                      A         31.5         A         31.0         B          27.3
           35.0       B         32.7        C          33.2         E          29.4
                      E         34.4        D          32.9         F          25.5
                      B         131         C          137          A          118
           140        C         147          E         124          D          106
                      E         127          F         131          F          118
                      A         396         D          396          A          335
           400        B         394          F         390          C          316
                      B         384          F         373          E          344




                                                                                      Page 19 of 22

第 20 页

At each fortification level, the mean concentration found was calculated and then the accuracy
was expressed in percent recovery. To assess precision, the within- and between-run variances
were estimated; these variances can be derived directly or with the aid of models, like mixed
models or analysis of variance (ANOVA).6 Here, a separate ANOVA was conducted for each
fortification level, with “concentration found” as dependent variable and “run” as a discrete,
independent variable. An ANOVA’s mean square error represents the within-run variance at
the given fortification level. Criteria for between-run precision are from the Horwitz equation
(see Section 3.3) and are based on the total between-run variance, which is the sum of the within-
run variance and the ‘pure’ between-run variance. The ‘pure’ between-run variance represents
the variability due to run-to-run differences, beyond that within each run. It can be determined
from a balanced ANOVA as the difference between the mean square for ‘run’ and the mean
square error, divided by the number of measurements per group (3 per run in this example).6
Equivalently, it is the between-run random effect variance in a mixed model. Precision
estimates were expressed as CVs, calculated by dividing the square root of the variance by the
mean and multiplying by 100.




            Table 3 – Accuracy and precision estimates for each fortification level*
Concentration n           Mean             Mean                       Precision
    added               concentration     recovery
   (ng/mL)             found (ng/mL)         (%)          Within-run           Between-run
                                                       Variance     CV     Variance**       CV
                                                                    (%)                     (%)
      4.2          9        4.18            99.6         0.128     8.56       0.167         9.77
       14          9        12.1            86.1         0.657     6.72        1.03         8.42
       35          9        30.9            88.2         2.45      5.07        10.8         10.6
      140          9         127            90.4         67.4      6.49        168          10.2
      400          9         370            92.4         129       3.08       1180          9.30
*Recovery and CV calculated directly from the means and variances in the table may be slightly
different than the stated values due to rounding. Deriving means and variances from ANOVA
or mixed models, or conducting weighted regression to estimate prediction intervals (see below),
can be done with statistical software (e.g., R or SAS).
**The sum of the within-run variance and the ‘pure’ between-run variance.

Method accuracy, within-run precision and between-run precision were all acceptable (Table
3).




                                                                                        Page 20 of 22

第 21 页

Plotting concentration found vs concentration added showed that variability increased with
concentration. Given this heteroscedasticity, weighted regression is appropriate. Commonly,
each data point is weighted by the inverse of its variance (1/variance), although other weighting
functions (e.g., 1/x, 1/x2) may be suitable depending on the data. Here, weights were inverse
variances, and to estimate them, the relationship between the empirical SD and concentration
added was modelled. This relationship is often linear5, and with the example dataset, it was SD
= 0.297 + 0.076 * concentration. Squaring the SD yielded by this weighting function gave
variance estimates, and inverting them yielded the weights 11.3, 2.64, 0.541, 0.115, 0.008, and
0.001 respectively for the 6 fortification levels that were then used in the weighted regression.
The weighted regression’s slope was 0.893 and the intercept was 0.292 ng/mL. The SD of the
weight-adjusted residuals was 0.906, so according to VICH GL2, LOD and LOQ were estimated
as 3.35 ng/mL and 10.1 ng/mL, respectively.
The alternative approach of deriving LOD and LOQ from the weighted regression’s prediction
intervals is demonstrated in Figure 2. The prediction intervals were calculated according to
Zorn et al.4 with α and β set at 0.05 and using df = n - 4 to account for the parameters of both
the weighted regression (2 df) and the weighting function (2 df). The decision limit was 0.763
ng/mL, the determination limit was 2.29 ng/mL, the critical value (Lc) was 0.528 ng/mL, the
LOD was 1.21 ng/mL, and the LOQ was 3.16 ng/mL. Consistent with this estimated LOQ,
accuracy and precision were acceptable at the lowest fortification level of 4.2 ng/mL (Table 3).
Regarding selectivity, the maximum response of blank, control samples (0.654 ng/mL) was less
than 20% of the LOQ estimated with GL2, but it was about 20% of the LOQ estimated from
prediction intervals. So, in this case, taking a slightly higher LOQ would ensure that background
noise is not considered quantifiable analyte.




              Figure 2 – derivation of LOD and LOQ from the example dataset.




                                                                                      Page 21 of 22

第 22 页

1
  Codex Alimentarius Procedural Manual, 15th Ed., Twenty-eight Session of the Codex Alimentarius
Commission, Rome, 2005, p 81.
2
  U.S. Code of Federal Regulations, Title 40: Protection of Environment , Part 136 – Guidelines Establishing Test
Procedures for the Analysis of Pollutants, Appendix B to Part 136 – Definition and Procedure for the
Determination of the Method Detection Limit – Revision 1.11.
3
  U.S. Environmental Protection Agency, Office of Pesticide Programs, March 23, 2000, “Assigning Values to
Non-detected/Non-quantified Pesticide Residue in Human Health Food Exposure Assessments” Appendix 1, A-
1 through A-8.
4
  Zorn ME, Gibbons RD, Sonzogni WC. Weighted Least-Squares Approach to Calculating Limits of Detection
and Quantification by Modeling Variability as a Function of Concentration, Anal Chem 1997, 69, 3069-3075.
5
  ISO 11843-2: 2000, Capability of detection – Part 2: Methodology in the linear calibration case.
6
  ISO 5725-2: 2019, Accuracy (trueness and precision) of measurement methods and results – Part 2: Basic
method for the determination of repeatability and reproducibility of a standard measurement method.




                                                                                                     Page 22 of 22

来源:EMA Scientific Guidelines · ema.europa.eu