sumber : CD EC 2002 657
3. VALIDATION
Validation shall demonstrate that the analytical method complies with the criteria applicable for the relevant performance characteristics. Different control purposes require different categories of methods. The following table determines which performance characteristic shall be verified for which type of method.
Table
Classification of analytical methods by the performance characteristics that have to be determined
3.1. VALIDATION PROCEDURES
This chapter provides examples and/or references for validation procedures of analytical methods. Other approaches to demonstrate that the analytical method complies with performance criteria for the performance characteristics may be used, provided that they achieve the same level and quality of information. Validation can also be performed by conducting an interlaboratory study such as established by Codex Alimentarius, ISO or the IUPAC (12), or according to alternative methods such as single laboratory studies or in-house validation (13)(14).
This chapter provides examples and/or references for validation procedures of analytical methods. Other approaches to demonstrate that the analytical method complies with performance criteria for the performance characteristics may be used, provided that they achieve the same level and quality of information. Validation can also be performed by conducting an interlaboratory study such as established by Codex Alimentarius, ISO or the IUPAC (12), or according to alternative methods such as single laboratory studies or in-house validation (13)(14).
This part focuses on single laboratory studies (on in-house validation) using a modular approach. This approach consists of:
1. a set of common performance characteristics independent of the validation model used and
2. more specific model-dependentprocedures as described in Table 10.
1. a set of common performance characteristics independent of the validation model used and
2. more specific model-dependentprocedures as described in Table 10.
Tabel Model-independent and model-dependent performance parameters
3.1.1. Model-independent performance characteristics
Irrespective of the validation approach chosen, the following performance characteristics have to be determined. To minimise the workload, a carefully designed and statistically sound approach can be used to combine experiments performed to determine different parameters.
3.1.1.1. Specificity
For analytical methods, the power of discrimination between the analyte and closely related substances (isomers, metabolites, degradation products, endogenous substances, matrix constituents, etc) is important. Two approaches are necessary to check for interferences.
Therefore, potentially interfering substances shall be chosen and relevant blank samples shall be analysed to
detect the presence of possible interferences and to estimate the effect of the interferences:
— select a range of chemically related compounds (metabolites, derivatives, etc.) or other substances likely to be encountered with the compound of interest that may be present in the samples;
— analyse an appropriate number of representative blank samples (n ≥ 20) and check for any interferences
(signals, peaks, ion traces) in the region of interest where the target analyte is expected to elute;
— additionally, representative blank samples shall be fortified at a relevant concentration with substances that are likely to interfere with the identification and/or quantification of the analyte;
— after analysis, investigate whether:
— the presence may lead to a false identification,
— the identification of the target analyte is hindered by the presence of one or more of the interferences, or
— the quantification is influenced notably.
3.1.1.2. Trueness
In this paragraph, the determination of trueness (one component of accuracy) is described. Trueness can only be established by means of certified reference material (CRM). A CRM be used whenever available. The procedure is described in detail in ISO 5725-4 (5). An example is given below:
— analyse six replicates of the CRM in accordance with the test instructions for the method,
— determine the concentration of the analyte present in each sample of the replicates,
— calculate the mean, the standard deviation and the coefficient of variation (%) for these concentrations,
— calculate the trueness by dividing the detected mean concentration by the certified value (measured as
concentration) and multiply by 100, to express the result as a percentage.
Trueness ( %) = mean recovery-corrected concentration detected × 100/certified value.
If no CRM is available, instead of trueness, the recovery can be determined as described under 4.1.2.1 below.
3.1.1.3. Applicability/ruggedness (minor changes)
Such studies use the deliberate introduction of minor reasonable variations by the laboratory and the observation of their consequences. The pre-investigative studies have to be carried out by selecting factors of the sample pre-treatment, clean up and analysis, which may influence the measurement results. Such factors may include the analyst, the source and the age of reagents, solvents, standards and sample extracts, the rate of heating, the temperature, the pH-value as well as many other factors that may occur in the laboratory. These factors should be modified in an order of magnitude that matches the deviations usually encountered among laboratories.
— Identify possible factors that could influence the results.
— Vary each factor slightly.
— Conduct a ruggedness test using the approach of Youden (15)(16). (Other approved methods may be used at this point. The Youden approach, however, keeps the required time and effort to a minimum). The Youden approach is a fractional factorial design. Interactions between the different factors cannot be detected.
— Where a factor is found to influence the measurement results significantly, conduct further experiments to
decide on the acceptability limits of this factor.
— Factors that significantly influence the results should be identified clearly in the method protocol.
The basic idea is not to study one alteration at a time but to introduce several variations at once. As an example, let A, B, C, D, E, F, G denote the nominal values for seven different factors that could influence the results, if the irnominal values are changed slightly. Let their alternative values be denoted by the corresponding lower case letters a, b, c, d, e, f and g. This results in 27 or 128 differentpossible combinations. It is possible to choose a subset of eight of these combinations that have a balance between capital and small letters (Table 11). Eight determinations have to be made, which will use a combination of the chosen factors (A-G). The results of the determinations are shown in Table 11 below as S-Z.
Irrespective of the validation approach chosen, the following performance characteristics have to be determined. To minimise the workload, a carefully designed and statistically sound approach can be used to combine experiments performed to determine different parameters.
3.1.1.1. Specificity
For analytical methods, the power of discrimination between the analyte and closely related substances (isomers, metabolites, degradation products, endogenous substances, matrix constituents, etc) is important. Two approaches are necessary to check for interferences.
Therefore, potentially interfering substances shall be chosen and relevant blank samples shall be analysed to
detect the presence of possible interferences and to estimate the effect of the interferences:
— select a range of chemically related compounds (metabolites, derivatives, etc.) or other substances likely to be encountered with the compound of interest that may be present in the samples;
— analyse an appropriate number of representative blank samples (n ≥ 20) and check for any interferences
(signals, peaks, ion traces) in the region of interest where the target analyte is expected to elute;
— additionally, representative blank samples shall be fortified at a relevant concentration with substances that are likely to interfere with the identification and/or quantification of the analyte;
— after analysis, investigate whether:
— the presence may lead to a false identification,
— the identification of the target analyte is hindered by the presence of one or more of the interferences, or
— the quantification is influenced notably.
3.1.1.2. Trueness
In this paragraph, the determination of trueness (one component of accuracy) is described. Trueness can only be established by means of certified reference material (CRM). A CRM be used whenever available. The procedure is described in detail in ISO 5725-4 (5). An example is given below:
— analyse six replicates of the CRM in accordance with the test instructions for the method,
— determine the concentration of the analyte present in each sample of the replicates,
— calculate the mean, the standard deviation and the coefficient of variation (%) for these concentrations,
— calculate the trueness by dividing the detected mean concentration by the certified value (measured as
concentration) and multiply by 100, to express the result as a percentage.
Trueness ( %) = mean recovery-corrected concentration detected × 100/certified value.
If no CRM is available, instead of trueness, the recovery can be determined as described under 4.1.2.1 below.
3.1.1.3. Applicability/ruggedness (minor changes)
Such studies use the deliberate introduction of minor reasonable variations by the laboratory and the observation of their consequences. The pre-investigative studies have to be carried out by selecting factors of the sample pre-treatment, clean up and analysis, which may influence the measurement results. Such factors may include the analyst, the source and the age of reagents, solvents, standards and sample extracts, the rate of heating, the temperature, the pH-value as well as many other factors that may occur in the laboratory. These factors should be modified in an order of magnitude that matches the deviations usually encountered among laboratories.
— Identify possible factors that could influence the results.
— Vary each factor slightly.
— Conduct a ruggedness test using the approach of Youden (15)(16). (Other approved methods may be used at this point. The Youden approach, however, keeps the required time and effort to a minimum). The Youden approach is a fractional factorial design. Interactions between the different factors cannot be detected.
— Where a factor is found to influence the measurement results significantly, conduct further experiments to
decide on the acceptability limits of this factor.
— Factors that significantly influence the results should be identified clearly in the method protocol.
The basic idea is not to study one alteration at a time but to introduce several variations at once. As an example, let A, B, C, D, E, F, G denote the nominal values for seven different factors that could influence the results, if the irnominal values are changed slightly. Let their alternative values be denoted by the corresponding lower case letters a, b, c, d, e, f and g. This results in 27 or 128 differentpossible combinations. It is possible to choose a subset of eight of these combinations that have a balance between capital and small letters (Table 11). Eight determinations have to be made, which will use a combination of the chosen factors (A-G). The results of the determinations are shown in Table 11 below as S-Z.
Table 11
Experiment design for ruggedness studies (minor changes)
Experiment design for ruggedness studies (minor changes)
3.1.1.4. Stability
It has been observed that insufficient stability of the analyte or matrix constituents in the sample during storage or analysis may give rise to significant deviations in the outcome of the result of analysis. Furthermore, the stability of the calibration standard in solution should be checked. Usually analyte stability is well characterised under various storage conditions. Monitoring of the storage condition will form part of the normal laboratory accreditation system. When this is not known, examples are given below on how the stability can be determined.
Stability of the analyte in solution:
— Prepare fresh stock solutions of the analyte(s) and dilute as specified in the test instructions to yield sufficient aliquots (e.g. 40) of each selected concentration (around the minimum required performance limit for substances for which no permitted limit has been established or around the permitted limit for other
substances. Prepare both solutions of the analyte used for fortification and used in the final analysis solution,
and any other solution that is of interest (e.g. derivatised standards).
— Measure the analyte content in the freshly prepared solution according to the test instructions.
— Dispense appropriate volumes into suitable containers, label and store according to the scheme:
It has been observed that insufficient stability of the analyte or matrix constituents in the sample during storage or analysis may give rise to significant deviations in the outcome of the result of analysis. Furthermore, the stability of the calibration standard in solution should be checked. Usually analyte stability is well characterised under various storage conditions. Monitoring of the storage condition will form part of the normal laboratory accreditation system. When this is not known, examples are given below on how the stability can be determined.
Stability of the analyte in solution:
— Prepare fresh stock solutions of the analyte(s) and dilute as specified in the test instructions to yield sufficient aliquots (e.g. 40) of each selected concentration (around the minimum required performance limit for substances for which no permitted limit has been established or around the permitted limit for other
substances. Prepare both solutions of the analyte used for fortification and used in the final analysis solution,
and any other solution that is of interest (e.g. derivatised standards).
— Measure the analyte content in the freshly prepared solution according to the test instructions.
— Dispense appropriate volumes into suitable containers, label and store according to the scheme:
— The storing time could be selected as one, two, three and four weeks or longer if necessary, e.g. until the first degradation phenomena are observable during identification and/or quantification. The maximum storing
time and the optimum storing conditions have to be recorded.
— The calculation of the concentration of the analyte(s) in each aliquot should be carried out by using the
solution of the analyte freshly prepared at the time of analysis as 100 %.
Analyte Remaining (%) = Ci × 100/Cfresh
Ci= concentration at time point
Cfresh= concentration of fresh solution Stability of analyte(s) in matrix
— Whenever possible, incurred samples should be used. When no incurred material is available, matrix fortified with the analyte should be used.
— When incurred material is available, the concentration in the material should be determined while the
material is still fresh. Further aliquots of material could be taken after one, two, four and 20 weeks and the
concentrations should be determined. The tissue should be stored at least minus 20 °C or lower if required.
— If no incurred material is available, take some blank material and homogenise it. Divide the material into five aliquots. Fortify each aliquot with the analyte, which should preferably be prepared in a small quantity of
aqueous solution. Analyse one aliquot immediately. Store the remaining aliquots at least minus 20 °C or
lower if required and analyse them after one, two, four and 20 weeks.
3.1.1.5. Calibration curves
When calibration curves are used for quantification:
— at least five levels (including zero) should be used in the construction of the curve,
— the working range of the curve should be described,
— the mathematical formula of the curve and the goodness-of-fit of the data to the curve should be described,
— acceptability ranges for the parameters of the curve should be described.
When serial calibration based on a standard solution is necessary, acceptable ranges shall be indicated for the parameters of the calibration curve, which may vary from series to series.
3.1.2. Conventional validation procedures
The calculation of the parameters in accordance with conventional methods requires the performance of several individual experiments. Each performance characteristic has to be determined for each major change (see under applicability/ruggedness above). For multi-analyte methods, several analytes can be analysed simultaneously, as long as possibly relevant interferences are ruled out previously. Several performance characteristics can be determined in a similar way. So, to minimise workload, it is advised to combine experiments as much as possible (e.g., repeatability and within-laboratory reproducibility with specificity, analysis of blank samples to determine the decision limit and testing for specificity).
3.1.2.1. Recovery
If there is no CRM available, the recovery has to be determined by experiments using fortified blank matrix using, for example, the following scheme:
— select 18 aliquots of a blank material and fortify six aliquots at each of 1, 1,5 and 2 times the minimum
required performance limit or 0,5, 1 and 1,5 times the permitted limit,
— analyse the samples and calculate the concentration present in each sample,
time and the optimum storing conditions have to be recorded.
— The calculation of the concentration of the analyte(s) in each aliquot should be carried out by using the
solution of the analyte freshly prepared at the time of analysis as 100 %.
Analyte Remaining (%) = Ci × 100/Cfresh
Ci= concentration at time point
Cfresh= concentration of fresh solution Stability of analyte(s) in matrix
— Whenever possible, incurred samples should be used. When no incurred material is available, matrix fortified with the analyte should be used.
— When incurred material is available, the concentration in the material should be determined while the
material is still fresh. Further aliquots of material could be taken after one, two, four and 20 weeks and the
concentrations should be determined. The tissue should be stored at least minus 20 °C or lower if required.
— If no incurred material is available, take some blank material and homogenise it. Divide the material into five aliquots. Fortify each aliquot with the analyte, which should preferably be prepared in a small quantity of
aqueous solution. Analyse one aliquot immediately. Store the remaining aliquots at least minus 20 °C or
lower if required and analyse them after one, two, four and 20 weeks.
3.1.1.5. Calibration curves
When calibration curves are used for quantification:
— at least five levels (including zero) should be used in the construction of the curve,
— the working range of the curve should be described,
— the mathematical formula of the curve and the goodness-of-fit of the data to the curve should be described,
— acceptability ranges for the parameters of the curve should be described.
When serial calibration based on a standard solution is necessary, acceptable ranges shall be indicated for the parameters of the calibration curve, which may vary from series to series.
3.1.2. Conventional validation procedures
The calculation of the parameters in accordance with conventional methods requires the performance of several individual experiments. Each performance characteristic has to be determined for each major change (see under applicability/ruggedness above). For multi-analyte methods, several analytes can be analysed simultaneously, as long as possibly relevant interferences are ruled out previously. Several performance characteristics can be determined in a similar way. So, to minimise workload, it is advised to combine experiments as much as possible (e.g., repeatability and within-laboratory reproducibility with specificity, analysis of blank samples to determine the decision limit and testing for specificity).
3.1.2.1. Recovery
If there is no CRM available, the recovery has to be determined by experiments using fortified blank matrix using, for example, the following scheme:
— select 18 aliquots of a blank material and fortify six aliquots at each of 1, 1,5 and 2 times the minimum
required performance limit or 0,5, 1 and 1,5 times the permitted limit,
— analyse the samples and calculate the concentration present in each sample,
— using the equation below, calculate the recovery for each sample,
— calculate the mean recovery and CV from the six results at each level,
— % Recovery = 100 × measured content/fortification level.
This conventional method for the determination of recovery is a variant of the standard addition method
described in 3.5, when:
— the sample is considered as a blank sample instead of a sample to be analysed,
— itis considered thatyield (1) and recovery (2) are similar for the two test portions,
— the test samples have the same masses and the test portion extracts the same volumes,
— the amount of the calibration standard that is added to the second (spiked) test portion is noted xADD. (xADD = ρA.VA),
— x1 is the measured value for the blank and x2 the measured value for the second (spiked) test portion,
— then, % Recovery = 100 (x2 – x1)/xADD. When any of the above conditions is (or is supposed) not to be achieved, then the complete procedure for determination of the recovery by mean of the standard addition method as described in 3.5 has to be performed.
3.1.2.2. Repeatability
— Prepare a set of samples of identical matrices, fortified with the analyte to yield concentrations equivalent to 1, 1,5 and 2 times the minimum required performance limit or 0,5, 1 and 1,5 times the permitted limit.
— At each level the analysis should be performed with at least six replicates.
— Analyse the samples.
— Calculate the concentration detected in each sample.
— Find the mean concentration, standard deviation and the coefficient of variation (%) of the fortified samples.
— Repeat these steps on at least two other occasions.
— Calculate the overall mean concentrations and CVs for the fortified samples.
3.1.2.3. Within-laboratory reproducibility
— Prepare a set of samples of specified test material (identical or different matrices), fortified with the analyte(s) to yield concentrations equivalent to 1, 1,5 and 2 times the minimum required performance limit or 0,5, 1 and 1,5 times the permitted limit.
— At each level the analysis should be performed with at least six replicates.
— Repeat these steps on at least two other occasions with different operators and different environmental
conditions, e.g. different batches of reagents, solvents etc., different room temperatures, different instruments, etc. if possible.
— Analyse the samples.
— Calculate the concentration detected in each sample.
— Find the mean concentration, standard deviation and the coefficient of variation ( %) of the fortified samples.
3.1.2.4. Reproducibility
When reproducibility has to be verified, laboratories should participate in collaborative studies according to ISO 5725-2 (5).
3.1.2.5. Decision Limit (CCα)
The decision limit has to be established according to the requirements for identification or identification plus
quantification as defined under ‘Performance criteria and other requirements for analytical methods’ (part 2).
— calculate the mean recovery and CV from the six results at each level,
— % Recovery = 100 × measured content/fortification level.
This conventional method for the determination of recovery is a variant of the standard addition method
described in 3.5, when:
— the sample is considered as a blank sample instead of a sample to be analysed,
— itis considered thatyield (1) and recovery (2) are similar for the two test portions,
— the test samples have the same masses and the test portion extracts the same volumes,
— the amount of the calibration standard that is added to the second (spiked) test portion is noted xADD. (xADD = ρA.VA),
— x1 is the measured value for the blank and x2 the measured value for the second (spiked) test portion,
— then, % Recovery = 100 (x2 – x1)/xADD. When any of the above conditions is (or is supposed) not to be achieved, then the complete procedure for determination of the recovery by mean of the standard addition method as described in 3.5 has to be performed.
3.1.2.2. Repeatability
— Prepare a set of samples of identical matrices, fortified with the analyte to yield concentrations equivalent to 1, 1,5 and 2 times the minimum required performance limit or 0,5, 1 and 1,5 times the permitted limit.
— At each level the analysis should be performed with at least six replicates.
— Analyse the samples.
— Calculate the concentration detected in each sample.
— Find the mean concentration, standard deviation and the coefficient of variation (%) of the fortified samples.
— Repeat these steps on at least two other occasions.
— Calculate the overall mean concentrations and CVs for the fortified samples.
3.1.2.3. Within-laboratory reproducibility
— Prepare a set of samples of specified test material (identical or different matrices), fortified with the analyte(s) to yield concentrations equivalent to 1, 1,5 and 2 times the minimum required performance limit or 0,5, 1 and 1,5 times the permitted limit.
— At each level the analysis should be performed with at least six replicates.
— Repeat these steps on at least two other occasions with different operators and different environmental
conditions, e.g. different batches of reagents, solvents etc., different room temperatures, different instruments, etc. if possible.
— Analyse the samples.
— Calculate the concentration detected in each sample.
— Find the mean concentration, standard deviation and the coefficient of variation ( %) of the fortified samples.
3.1.2.4. Reproducibility
When reproducibility has to be verified, laboratories should participate in collaborative studies according to ISO 5725-2 (5).
3.1.2.5. Decision Limit (CCα)
The decision limit has to be established according to the requirements for identification or identification plus
quantification as defined under ‘Performance criteria and other requirements for analytical methods’ (part 2).
In the case of substances for which no permitted limit has been established, CCα can be established:
— either by the calibration curve procedure according to ISO 11843 (17) (here referred to as critical value of the net state variable). In this case blank material shall be used, which is fortified at and above the minimum required performance level in equidistant steps. Analyse the samples. After identification, plot the signal against the added concentration. The corresponding concentration at the y-intercept plus 2,33 times the standard deviation of the within-laboratory reproducibility of the intercept equals the decision limit. This is applicable to quantitative assays only (α = 1 %),
— or by analysing at least 20 blank materials per matrix to be able to calculate the signal to noise ratio at the
time window in which the analyte is expected. Three times the signal to noise ratio can be used as decision
limit. This is applicable to quantitative and qualitative assays.
In the case of substances an with established permitted limit, CCα can be established:
— either by the calibration curve procedure according to ISO 11843 (17) (here referred to as critical value of the net state variable). In this case blank material shall be used, which is fortified around the permitted limit in equidistant steps. Analyse the samples. After identification, plot the signal against the added concentration.
The corresponding concentration at the permitted limit plus 1,64 times the standard deviation of the
within-laboratory reproducibility equals the decision limit (α = 5 %),
— or by analysing at least 20 blank materials per matrix fortified with the analyte(s) at the permitted limit. The concentration at the permitted limit plus 1,64 times the corresponding standard deviation equal the decision limit( α = 5 %).
3.1.2.6. Detection capability (CCβ)
The detection capability should be determined according to the requirements for screening, identification or
identification plus quantification as defined (see part 2). In the case of substances for which no permitted limit has been established, CCβ can be established by:
— the calibration curve procedure according to ISO 11843 (17) (here referred to as minimum detectable value of the net state variable). In this case representative blank material shall be used, which is fortified at and below the minimum required performance level in equidistant steps. Analyse the samples. After identification, plot the signal against the added concentration. The corresponding concentration at the decision limit plus 1,64 times the standard deviation of the within-laboratory reproducibility of the mean measured content at the decision limit equals the detection capability (β = 5 %),
— analysing at least 20 blank materials per matrix fortified with the analyte(s) at the decision limit. Analyse the samples and identify the analytes. The value of the decision limit plus 1,64 times the standard deviation of
the within-laboratory reproducibility of the measured content equals the detection capability (β = 5 %),
— where no quantitative results are available, the detection capability can be determined by the investigation of fortified blank material at and above the decision limit. In this case the concentration level, where only ≤ 5% false compliant results remain, equals the detection capability of the method. Therefore, at least 20
investigations for at least one concentration level have to be carried out in order to ensure a reliable basis for
this determination. In the case of substances for which a permitted limit has been established, CCβ can be established:
— either by the calibration curve procedure according to ISO 11843 (17) (here referred to as minimum
detectable value of the net state variable). In this case representative blank material shall be used, which is
fortified around the permitted limit in equidistant steps. Analyse the samples and identify the analyte(s).
Calculate the standard deviation of the mean measured content at the decision limit. The corresponding
concentration at the value of the decision limit plus 1,64 times the standard deviation of the within-laboratory
reproducibility equals the detection capability (β = 5 %),
— or by analysing at least 20 blank materials per matrix fortified with the analyte(s) at the decision limit. The
value of the decision limit plus 1,64 times the corresponding standard deviation equals the detection
capability (β = 5 %).
Recovery
— repeatability per concentration level (sir),
— within-laboratory reproducibility per concentration level (sir),
— decision limit(CC α),
— detection capability (CCβ),
— power curve (β-error rate versus concentration (see 3.1.3.2),
— ruggedness of major changes; ruggedness to minor changes can be determined according to paragraph
3.1.1.3,
— 16 sample-related calibration curves,
— one overall calibration curve,
— prediction interval of the overall calibration curve,
— matrix-induced deviations (smat),
— run-induced deviations (srun),
— effect of the individual factors on the measurement results.
These performance characteristics allow the comprehensive evaluation of the performance level of the method, since not only the influence of the individual factors is investigated, but also the relevant combinations of these factors. With the help of this experiment design it is possible to decide if one or the other of the selected factors shall be excluded from the overall calibration curve, because it significantly deviates from the standard deviations of the other factors.
3.1.3.2. Power curve
The power curve provides information about the detection capability of the method within the chosen
concentration range. It refers to the β-error risk when applying the investigated method. The power curve allows to calculate the detection capabilities for the respective categories (screening, confirmation) or types (qualitative or quantitative) of methods for a certain β-error (e.g. 5 %)..
— either by the calibration curve procedure according to ISO 11843 (17) (here referred to as critical value of the net state variable). In this case blank material shall be used, which is fortified at and above the minimum required performance level in equidistant steps. Analyse the samples. After identification, plot the signal against the added concentration. The corresponding concentration at the y-intercept plus 2,33 times the standard deviation of the within-laboratory reproducibility of the intercept equals the decision limit. This is applicable to quantitative assays only (α = 1 %),
— or by analysing at least 20 blank materials per matrix to be able to calculate the signal to noise ratio at the
time window in which the analyte is expected. Three times the signal to noise ratio can be used as decision
limit. This is applicable to quantitative and qualitative assays.
In the case of substances an with established permitted limit, CCα can be established:
— either by the calibration curve procedure according to ISO 11843 (17) (here referred to as critical value of the net state variable). In this case blank material shall be used, which is fortified around the permitted limit in equidistant steps. Analyse the samples. After identification, plot the signal against the added concentration.
The corresponding concentration at the permitted limit plus 1,64 times the standard deviation of the
within-laboratory reproducibility equals the decision limit (α = 5 %),
— or by analysing at least 20 blank materials per matrix fortified with the analyte(s) at the permitted limit. The concentration at the permitted limit plus 1,64 times the corresponding standard deviation equal the decision limit( α = 5 %).
3.1.2.6. Detection capability (CCβ)
The detection capability should be determined according to the requirements for screening, identification or
identification plus quantification as defined (see part 2). In the case of substances for which no permitted limit has been established, CCβ can be established by:
— the calibration curve procedure according to ISO 11843 (17) (here referred to as minimum detectable value of the net state variable). In this case representative blank material shall be used, which is fortified at and below the minimum required performance level in equidistant steps. Analyse the samples. After identification, plot the signal against the added concentration. The corresponding concentration at the decision limit plus 1,64 times the standard deviation of the within-laboratory reproducibility of the mean measured content at the decision limit equals the detection capability (β = 5 %),
— analysing at least 20 blank materials per matrix fortified with the analyte(s) at the decision limit. Analyse the samples and identify the analytes. The value of the decision limit plus 1,64 times the standard deviation of
the within-laboratory reproducibility of the measured content equals the detection capability (β = 5 %),
— where no quantitative results are available, the detection capability can be determined by the investigation of fortified blank material at and above the decision limit. In this case the concentration level, where only ≤ 5% false compliant results remain, equals the detection capability of the method. Therefore, at least 20
investigations for at least one concentration level have to be carried out in order to ensure a reliable basis for
this determination. In the case of substances for which a permitted limit has been established, CCβ can be established:
— either by the calibration curve procedure according to ISO 11843 (17) (here referred to as minimum
detectable value of the net state variable). In this case representative blank material shall be used, which is
fortified around the permitted limit in equidistant steps. Analyse the samples and identify the analyte(s).
Calculate the standard deviation of the mean measured content at the decision limit. The corresponding
concentration at the value of the decision limit plus 1,64 times the standard deviation of the within-laboratory
reproducibility equals the detection capability (β = 5 %),
— or by analysing at least 20 blank materials per matrix fortified with the analyte(s) at the decision limit. The
value of the decision limit plus 1,64 times the corresponding standard deviation equals the detection
capability (β = 5 %).
Recovery
— repeatability per concentration level (sir),
— within-laboratory reproducibility per concentration level (sir),
— decision limit(CC α),
— detection capability (CCβ),
— power curve (β-error rate versus concentration (see 3.1.3.2),
— ruggedness of major changes; ruggedness to minor changes can be determined according to paragraph
3.1.1.3,
— 16 sample-related calibration curves,
— one overall calibration curve,
— prediction interval of the overall calibration curve,
— matrix-induced deviations (smat),
— run-induced deviations (srun),
— effect of the individual factors on the measurement results.
These performance characteristics allow the comprehensive evaluation of the performance level of the method, since not only the influence of the individual factors is investigated, but also the relevant combinations of these factors. With the help of this experiment design it is possible to decide if one or the other of the selected factors shall be excluded from the overall calibration curve, because it significantly deviates from the standard deviations of the other factors.
3.1.3.2. Power curve
The power curve provides information about the detection capability of the method within the chosen
concentration range. It refers to the β-error risk when applying the investigated method. The power curve allows to calculate the detection capabilities for the respective categories (screening, confirmation) or types (qualitative or quantitative) of methods for a certain β-error (e.g. 5 %)..
Figure 1 shows an example of the graphical establishment of detection capability (CCβ) of an analytical method. This particular method has a remaining risk of taking a false decision of 5 % at a concentration of 0,50 μg/kg. At a concentration of 0,55 μg/kg the risk of taking a false compliant decision decreases to 1 %.
3.1.3.3. Reproducibility The determination of a method's reproducibility by the single laboratory studies (in-house validation) concept requires repeated participation in proficiency studies in accordance with ISO guide 43-1 (3) and 43-2 (4). The laboratories are allowed to choose their own methods, provided these methods are used under routine conditions. The standard deviation of the laboratory can be used to assess the reproducibility of the method
3.1.3.3. Reproducibility The determination of a method's reproducibility by the single laboratory studies (in-house validation) concept requires repeated participation in proficiency studies in accordance with ISO guide 43-1 (3) and 43-2 (4). The laboratories are allowed to choose their own methods, provided these methods are used under routine conditions. The standard deviation of the laboratory can be used to assess the reproducibility of the method
.
Tidak ada komentar:
Posting Komentar