Beruflich Dokumente
Kultur Dokumente
671
page 2
48
49
50
51 SCHEDULE FOR THE PROPOSED ADOPTION PROCESS OF DOCUMENT QAS/16.671:
52 GUIDELINES ON VALIDATION – APPENDIX 4
53 ANALYTICAL METHOD VALIDATION
54
55
Discussion of proposed need for revision in view of the 29 June– 56
current trends in validation during informal consultation 1 July 2015 57
on data management, bioequivalence, GMP and 58
medicines’ inspection 59
Preparation of draft proposal for revision of the main text July 2015– 60
and several appendices by specialists in collaboration April 2016 61
with the Medicines Quality Assurance Group and 62
Prequalification Team (PQT)-Inspections, based on the 63
feedback received during the meeting and from PQT- 64
Inspections, draft proposals developed on the various 65
topics by specialists, as identified in the individual 66
working documents. 67
Presentation of the progress made to the fiftieth meeting 12–16 October 2015
68
of the WHO Expert Committee on Specifications for 69
Pharmaceutical Preparations
70
Discussion at the informal consultation on good practices 4–6 April 2016
71
for health products manufacture and inspection, Geneva
72
Preparation of revised text by Ms S. Croft, member of the May 2016
73
PQT-Inspections Team, and review by Dr A.J. van Zyl,
74
participant at the above-mentioned consultation, based on
the feedback received during the informal consultation by 75
the meeting participants and members of PQT- 76
Inspections 77
Circulation of revised working document for public June 2016 78
consultation 79
Consolidation of comments received and review of August–September80
feedback 2016 81
Presentation to the fifty-first meeting of the WHO Expert 82
17–21 October 2016
Committee on Specifications for Pharmaceutical 83
Preparations 84
Any other follow-up action as required … 85
86
87
88
89
Working document QAS/16.671
page 3
90 Background information
91
92 The need for revision of the published Supplementary guidelines on good manufacturing
93 practices: validation (1) was identified by the Prequalification of Medicines Programme and a
94 draft document was circulated for comment in early 2013. The focus of the revision was the
95 Appendix on non-sterile process validation (Appendix 7), which had been revised and was
96 adopted by the Committee at its forty-ninth meeting in October 2014.
97
98 The main text was sent out for consultation as Working document QAS/15.639 entitled
99 “Guidelines on Validation” which constitute the general principles of the new guidance on
100 validation.
101
102 The draft on the specific topics, the appendices to this main text, will follow. One of them, i.e. e
103 Analytical method validation, constitutes this working document.
104
105 The following is an overview on the appendices that are intended to complement the general text
106 on validation:
107
108 Appendix 1
109 Validation of heating, ventilation and air-conditioning systems
110 will be replaced by cross-reference to WHO Guidelines on GMP for HVAC systems
111 for considerations in qualification of HVAC systems
112 (update - working document QAS/15.639/Rev.1)
113
114 Appendix 2
115 Validation of water systems for pharmaceutical use
116 will be replaced by cross-reference to WHO Guidelines on water for pharmaceutical
117 use for consideration in qualification of water purification systems
118
119 Appendix 3
120 Cleaning validation – consensus to retain
121
122 Appendix 4
123 Analytical method validation – updated text proposed in this working document
124
125 Appendix 5
126 Validation of computerized systems – (update – see working document QAS/16.667)
127
128 Appendix 6
129 Qualification of systems and equipment – update in process
130
131 Appendix 7
132 Non-sterile process validation – update already published as Annex 3, WHO Technical Report
133 Series, No. 992, 2015
Working document QAS/16.671
page 4
134 APPENDIX 4
135 ANALYTICAL METHOD VALIDATION
136
137 1. Principle
138 2. General
139 3. Pharmacopoeial methods
140 4. Non-pharmacopoeial methods
141 5. Method validation
142 6. Method verification
143 7. Method transfer
144 8. Revalidation
145 9. Characteristics of analytical procedures
146
147 1. PRINCIPLE
148
149 1.1 This appendix presents some information on the characteristics that should be considered
150 during validation of analytical methods. Approaches other than those specified in this appendix
151 may be followed and may be acceptable. Manufacturers should choose the validation protocol
152 and procedures most suitable for testing of their product.
153
154 1.2 The manufacturer should demonstrate (through validation) that the analytical procedure is
155 suitable for its intended purpose.
156
157 1.3 Analytical methods, whether or not they indicate stability, should be validated.
158
159 1.4 The analytical method should be validated by research and development before being
160 transferred to the quality control unit when appropriate.
161
162 1.5 The recommendations as provided for in good laboratory practices and guidelines for
163 transfer of technology should be considered, where applicable, when analytical method
164 validation is organized and planned.
165
166 2. GENERAL
167
168 2.1 There should be specifications for both materials and products. The tests to be performed
169 should be described in the documentation on standard test methods.
170
171 2.2 Specifications and standard test methods in pharmacopoeias (“pharmacopoeial
172 methods”), or suitably developed specifications or test methods (“non-pharmacopoeial methods”)
173 as approved by the national regulatory authority (NRA) may be used.
174
175 2.3 Well-characterized reference materials, with documented purity, should be used in
176 analysis.
177
178 2.4 The most common analytical procedures include identification tests, assay of drug
179 substances and pharmaceutical products, quantitative tests for content of impurities and limit
Working document QAS/16.671
page 5
180 tests for impurities. Other analytical procedures include dissolution testing and determination of
181 particle size.
182
183 2.5 The results of analytical procedures should be accurate, legible, contemporaneous,
184 original, reliable and reproducible. All results should be archived for an appropriate period of
185 time as defined by the laboratory and be in compliance with NRA requirements.
186
187 2.6 The procedure should become part of a continuous verification procedure to demonstrate
188 that it meets the predefined criteria over the life of the procedure.
189
190 2.7 Trend analysis and risk assessment should be considered at intervals to ensure that the
191 method is appropriate for its intended application.
192
193 2.8 Changes to methods should be managed in accordance with the authorized change control
194 procedure. The variability of reference materials and other factors such as changes in the process
195 for synthesis of the drug substance, changes in the composition of the finished product, changes
196 in the analytical procedure, when analytical methods are transferred from one laboratory to
197 another (when method transfer is not possible) or when major pieces of equipment instruments
198 change should be considered. These should be understood, controlled and, where possible,
199 reduced. Verification or revalidation should be considered where appropriate.
200
201 2.9 The scope of verification or degree of revalidation depend on the nature of the change(s)
202 and the outcome of risk assessment.
203
204 2.10 There should be evidence that the analysts, who are responsible for certain tests, are
205 appropriately qualified to perform those analyses (“analyst proficiency”).
206
207 2.11 The data obtained during method validation and verification should be considered
208 covered by good anything practices (GxP) requirements and are expected to follow the principles
209 of good data and record management practices (2). Their associated metadata are also expected
210 to be retained and subjected to good data and record management practices.
211
212 2.12 When computerized systems are used to obtain and process data relating to method
213 validation and verification, they should comply to the principles enunciated in Appendix 5 –
214 Validation of computerized systems.
215
216 2.13 Adequate attention should be paid to the method of sample preparation. The description
217 of this step should be as detailed as possible, especially if it can have a significant impact on tests
218 results (e.g. particular attention should be paid to details such as sonication time, sonication bath
219 temperature and mixing and to samples where demixing is known to occur).
220
221 2.14 Failures occurring during method validation, and how these were overcome, should be
222 included in the method validation report – it is not acceptable to present only the passing results
223 as it will give a biased imaged on the reliability of the method and on how it should be applied.
224
225
Working document QAS/16.671
page 6
272 6.3 Method verification is suitable in lieu of method validation for pharmacopoeial methods.
273
274 7. METHOD REVALIDATION
275 7.1 Methods should be maintained in a validated state over the life of the method (see point
276 2.6 above). Revalidation of an analytical procedure should be considered whenever there are
277 changes made to the method, including:
278 ‒ changes to the mobile phase (please refer to The International Pharmacopoeia and other
279 pharmacopoeias for the acceptance limits beyond which revalidation must be performed);
280 ‒ changes to the column;
281 ‒ changes to the temperature of the column;
282 ‒ changes to the concentration/composition of the sample and standards;
283 ‒ changes to the detector (change in detector type, e.g. if going from ultraviolet (UV)-
284 visible detection to fluorimetry, or wavelength of detection).
285 7.2 In case of repeated system suitability failures or when obtaining of doubtful results. In
286 such cases an investigation of the root cause should be performed, the appropriate changes made
287 and the method revalidated.
288 7.3 Periodic revalidation of analytical methods should be considered according to a period
289 that is scientifically justifiable.
290 7.4 It is acceptable for revalidation to include only the validation characteristics of relevance
291 to the particular change and method.
311 (including information to be supplied with the results, and report forms to be used, if any);
312 procedure for the handling of deviations; references; and details of reference samples (starting
313 materials, intermediates and finished products). The protocol should be authorized and dated.
314
315 8.6 In the case of independent testing by a separate entity, such as a national quality control
316 testing laboratory that is testing samples on its market, method transfer is not always possible. It
317 is not considered an obligation but may be considered as an optional step when encountering
318 difficulties in applying any particular method. See WHO guidelines on transfer of technology in
319 pharmaceutical technology (3) for further reference.
320
321 9. CHARACTERISTICS OF ANALYTICAL PROCEDURES
322 9.1 Characteristics that should be considered during validation of analytical methods include:
323
324 ‒ specificity;
325 ‒ linearity;
326 ‒ range;
327 ‒ accuracy;
328 ‒ precision;
329 ‒ detection limit;
330 ‒ quantitation limit;
331 ‒ robustness.
332
333 This list should be considered typical but occasional exceptions should be dealt with on a case-
334 by-case basis
335
336 9.1.1 Accuracy is the degree of agreement of test results with the true value, or the closeness of
337 the results obtained by the procedure to the true value. It is normally established on samples of
338 the material to be examined that have been prepared to quantitative accuracy. Accuracy should be
339 established across the specified range of the analytical procedure.
340
341 Note: It is acceptable to use a “spiked” placebo where a known quantity or concentration of a
342 reference material is used.
343
344 9.1.2 Precision is the degree of agreement among individual results. The complete procedure
345 should be applied repeatedly to separate, identical samples drawn from the same homogeneous
346 batch of material. It should be measured by the scatter of individual results from the mean (good
347 grouping) and expressed as the relative standard deviation (RSD).
348
349 9.1.2.1 Repeatability should be assessed using a minimum of nine determinations covering the
350 specified range for the procedure, e.g. three concentrations/three replicates each, or a minimum
351 of six determinations at 100% of the test concentration.
352
353 9.1.2.2 Intermediate precision expresses within-laboratory variations (usually on different days,
354 different analysts and different equipment). If reproducibility is assessed, a measure of
355 intermediate precision is not required.
Working document QAS/16.671
page 9
356
357 9.1.2.3 Reproducibility expresses precision between laboratories.
358
359 9.1.3 Robustness (or ruggedness) is the ability of the procedure to provide analytical
360 results of acceptable accuracy and precision under a variety of conditions. The results from
361 separate samples are influenced by changes in the operational or environmental conditions.
362 Robustness should be considered during the development phase and should show the reliability
363 of an analysis when deliberate variations are made in method parameters.
364
365 The verification of stability of analytical solutions is of particular importance.
366
367 Other characteristics of robustness include extraction time. In the case of liquid chromatography,
368 robustness testing may also include verification of the impact of changes in pH, temperature and
369 flow rate (see ICH Q2 – Validation of Analytical Procedures, Step 4, for further details).
370
371 9.1.3.1 Factors that can have an effect on robustness when performing chromatographic analysis
372 include:
373
374 ‒ stability of test and standard samples and solutions;
375 ‒ reagents (e.g. different suppliers);
376 ‒ different columns (e.g. different lots and/or suppliers);
377 ‒ extraction time;
378 ‒ variations of pH of a mobile phase;
379 ‒ variations in mobile phase composition;
380 ‒ temperature;
381 ‒ flow rate.
382
383 9.1.4 Linearity indicates the ability to produce results that are directly proportional to the
384 concentration of the analyte in samples. A series of samples should be prepared in which the
385 analyte concentrations span the claimed range of the procedure. If there is a linear relationship,
386 test results should be evaluated by appropriate statistical methods. A minimum of five
387 concentrations should be used.
388
389 9.1.5 Range is an expression of the lowest and highest levels of analyte that have been
390 demonstrated to be determinable for the product. The specified range is normally derived from
391 linearity studies.
392
393 9.1.6 Specificity (selectivity) is the ability to measure unequivocally the desired analyte in the
394 presence of components such as excipients and impurities that may also be expected to be
395 present. An investigation of specificity should be conducted during the validation of
396 identification tests, the determination of impurities and assay.
397
398 9.1.7 Detection limit (limit of detection) is the smallest quantity of an analyte that can be
399 detected, and not necessarily determined, in a quantitative fashion. Approaches may include
400 instrumental or non-instrumental procedures and could include those based on:
401
Working document QAS/16.671
page 10
421
422
423 Statistical analysis used to evaluate validation characteristics against predetermined acceptance
424 criteria should be appropriate for the intended evaluation. Appropriately validated software
425 should be used. An appropriate number of samples to provide adequate statistical power and
426 range should be considered.
427 9.3 System suitability testing
428
429 Note: System suitability testing is an integral part of many analytical procedures. The tests are
430 based on the concept that the equipment, electronics, analytical operations and samples to be
Working document QAS/16.671
page 11
431 analysed constitute an integral system that can be evaluated as such. System suitability test
432 parameters that need to be established for a particular procedure depend on the type of
433 procedure being evaluated, for instance, a resolution test for a high-performance liquid
434 chromatography (HPLC) procedure.
435
436 9.3.1 The suitability of the entire system should be confirmed prior to and during method
437 validation tests as well as during the test of samples.
438
439 9.3.2 System suitability runs should include only established standards or reference materials
440 of known concentration to provide an appropriate comparator for the potential variability of the
441 instrument.
442
443 9.3.3 Where a sample is used for system suitability or a trial run, written procedures should be
444 established and followed and the results of all such trial runs be included in the results and data
445 review process. A sample can be used only if it is a well characterized material. Characterization
446 in such a case should be performed prior to the use of this sample as part of system suitability
447 testing. The sample material or product under test should not be used for trial run purposes or to
448 evaluate suitability of the system (see WHO guidelines on good data and record management
449 practices (2).
450
451
452
453 References
454
455 1. Supplementary guidelines on good manufacturing practices: validation
456 (WHO Technical Report Series, No. 937, 2006, Annex 4).
457
458 2. WHO Guidelines on good data and record management practices (WHO Technical
459 Report Series, No. 996, 2016, Annex 5).
460
461 3. WHO guidelines on transfer of technology in pharmaceutical technology (WHO
462 Technical Report Series, No. 961, 2011, Annex 7).
463
464 ***
465
1.0 A Guide to Method Validation
1.1 Introduction
Virtually every aspect of society is supported in some way by analytical measurement. There are
innumerable reasons for making these measurements, for example: in-process and final inspection
or ‘quality control’ of manufactured products; supporting healthcare; checking the quality of drinking
water; metallurgical analysis of an alloy to confirm its suitability for use in aircraft construction;
forensic analysis of body fluids in criminal investigations.
The cost of carrying out these measurements is huge cost implications arise from decisions made
on the basis of the results. Analytical results may be used in evidence and challenged in a court of
law: tests showing food to be unfit for consumption my result in compensation claims; test
confirming the presence of banned drugs could result in fines, imprisonment or even, in some
countries, execution. Clearly it is important to determine the correct result and be able to show that it
is correct.
Method validation is the process of defining an analytical requirement, and confirms that the method under
consideration has performance capabilities consistent with what the application requires. Use of
equipment that is within specification, working correctly and adequately calibrated is fundamental to the
method validation process. Likewise the operator carrying out the studies must be competent in the
analysis under study and have sufficient knowledge of the method/analysis to draw conclusions from the
observations as the validation work proceeds.
Quite often method validation evolves from method development and so the two activities are often
closely tied, with the validation study employing the techniques and steps in the analysis as defined
by the method development.
A method should be validated when it is necessary to verify that its performance parameters are
adequate for use for a particular analytical problem. For example:
Certain areas of analytical practices, such as in clinical chemistry will specify validation
requirements relevant to the method. This ensures that particular validation terminology together
with the statistics used is interpreted in a manner consistent within the relevant sector. Official
recognition of a method may require characterisation using a collaborative study.
1
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
1.4 Method Development
As stated above method validation often evolves from method development. Method development
can take a number of forms. At one extreme, it may involve adapting an existing method, making
minor changes so that it is suitable for a new application.
At the othe r extreme the analytical chemist may start out with little information and apply expertise
and experience to devise a suitable method. This can involve significant innovation based on novel
exploitation of known properties of the analyte or measurand. Often a considerable amount of effort
is required and initially at least a degree of doubt as to whether the final method will be successful.
Frequently method development involves working on a number of different ideas simultaneously
and eventually picking on e of these.
In general analytical methods consist of a measurement stage which may be preceded by an isolation
stage. It is necessary to establish that the signal or reaction produced at the measurement stage is only
due to the analyte and not due to something chemically or physically similar or arising as a coincidence.
This is confirmation of identity. Whether or not other compounds interfere with the measurement of the
analyte will depend on the effectiveness of the isolation stage if it was part of the method, as well as the
selectivity/specificity of the measurement stage.
The selectivity of a method is usually investigated by studying its ability to measure the analyte of
interest in test portions to which specific interferences have been deliberately introduced (those
thought likely to be present in samples). Where it is unclear whether or not interferences are already
present, the selectivity of the method can be investigated by studying its ability to measure
compared to other independent method/techniques.
Another aspect of selectivity which must be considered is where an analyte may exist in the sample
in more than one form such as: free or complexed; inorganic or organometallic; or the possibility of a
component such as Chromium ion being present in different oxidation states such as Cr3+ or Cr6+.
2
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
1.7 Limit of Detection
In quantitative analysis, and where the analyte is present is small concentrations, it is important to know
what is the lowest concentration of the analyte or property value that can be confidently detected by the
method. Frequently the detection of the analyte does not simply cease at a threshold level, but tends to
‘fade’ from a predictable signal/concentration ratio gradually to an indiscernible response.
For qualitative measurements, there is likely to be concentration threshold below which specificity
becomes unreliable. The threshold may vary if the experiment is repeated at another time with
different reagents, fortification, spiking materials, etc. In the example shown in Table 1, positive
identification of the analyte has ceased to be 100% reliable below 100 µg.g–1.
Table 1: Qualitative Analysis – Illustration of how cut-off (i.e. threshold) concentration is determined.
The ‘limit of quantitation’ (LoQ) is the least concentration of analyte that can be determined with an
acceptable level of repeatability precision and trueness. It is also sometimes known as ‘limit of
determination’. LoQ is an indicative value and should not normally be used in decision making.
Note that neither ‘limit of detection’ LoD nor LoQ represent levels at which quantitation is
impossible. It is simply that the uncertainty of measurement and the result approach the same
magnitude in the region of the LoD.
For any quantitative method, the range of analyte concentrations (i.e. the range of concentrations or
property values in the solutions actually measured rather than in the original sam ples) over which the
method may be applied must be known. At the lower end of the concentration range the limiting factors
are the values of the limits of detection and/or quantitation. At the upper end of the concentration range
limitations may be imposed by a ‘shouldering’ of the linear range depending on the instrument response
system.
Within the working range there may exist a linear response range. Within the linear range signal
response will have a linear relationship to analyte concentration or property value. The extent of
this range may be established during the evaluation of the working range. Note that regression
calculations on their own are insufficient to establish linearity. To do this a visual inspection of the
line and residuals may also be necessary. Objective tests, such as ‘goodness-of-fit’ tests, are better
still. In general linearity checks require 10 or more points at different concentrations/property values.
Evaluation of the working and linear ranges will also be useful for planning the calibration required
when using the method is in routine use. It is advisable to investigate the variance across the
3
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
working range. Within the linear range, two calibration points may be sufficient, to establish the
slope of the calibration line. Elsewhere in the working range, multi-point calibration will be
necessary. The relationship of instrument response to concentration does not have to be perfectly
linear for a method to be effective but the curve should be repeatable from day to day. Note that the
working and linear range may be different for different matrices according to the effect of
interferences arising from the matrix.
1.10 Accuracy
‘Accuracy’ expresses the closeness of a result to a true value. Method validation seeks to quantify
the likely accuracy of results by assessing systematic and random effects on results. Accuracy is,
therefore, normally studied as two components: ‘trueness’ and ‘precision’. The ‘trueness’ (of a
method) is an expression of how close the mean of a set of results (produced by the method) is to
the true value. Trueness is normally expressed in terms of bias. Precision is a measure of how
close results are to one another, and is usually expressed by measures such as standard deviation,
which describe the spread of results. In addition, an increasingly common expression of accuracy is
‘measurement uncertainty’, which provides a single figure expression of accuracy. These three
different parameters will be discussed in turn.
Practical assessment of trueness relies on comparison of mean results from a method with known
values, that is trueness is assessed against a reference value (i.e. true value or conventional true
value). Two basic techniques are available: checking against reference values for a charac terised
material or from another characterised method. Reference values are ideally traceable to
international standards. Certified reference materials are generally accepted as providing traceable
values; the reference value is then the certified value of the CRM. To check trueness using a
reference material, determine the mean and standard deviation of a series of replicate test, and
compare with the characterised value for the reference material. The ideal reference material is a
certified, natural matrix reference material, closely similar to the samples of interest. Clearly, the
availability of such materials is limited. Reference materials for validation may accordingly be:
♦ Prepared by spiking typical materials with pure certified reference materials or other materials of
suitable purity and stability;
♦ Commercially available secondary standards, with certified traceability, whose preparation is
ILAB, accredited.
♦ T ypical, well-characterised materials checked in-house for stability and retained for in-house
QC.
Validation needs to fit the purpose, so the choice of reference material may be affected by the end
use. The reference material must be appropriate to the use. For regulatory work, a relevant
certified material should be used, ideally matrix matched. For methods used for long -term in-house
work, a secondary standard material or certified reference material should be used. For short- term
or non -critical work, a prepared standard or spike is often sufficient.
To check against an alternative method, compare results from the two methods for the same
sample or samples. The sample(s) may be CRMs, commercially available traceable standard, or
simply typical samples. There are advantages to using CRMs, since these have known stability and
homogeneity, and additionally give an indication of bias with respect to international standards. On
the other hand, CRMs are costly and may not be representative of typical samples.
4
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
1.11 Precision
Precision is method and concentration specific, which in prac tice can be very varied. The two most
common precision measures are ‘repeatability’ and reproducibility’. They represent the two extreme
measures of precision, which can be obtained. Repeatability (the smallest expected precision) will give an
idea of the sort of variability to be expected when a method is performed by a single analyst on one piece
of equipment over a short timescale, i.e. the sort of variability to be expected between results when a
sample is analysed in duplicate. If a sample is to be analysed by a number of laboratories for comparative
purposes then a more meaningful precision measure is reproducibility (this is the largest measure of
precision normally encountered). It may be that some in-between measure is the most useful in partic ular
cases; for example precision measured between different analysts, over extended timescales, within a
single laboratory. This is sometimes known as ‘intermediate precision’, but the exact conditions should be
stated. Precision is usually stated in terms of standard deviation or relative standard deviation. Both
repeatability and reproducibility are generally dependent on analyte concentration, and so should be
determined at a number of concentrations and if relevant, the relationship between precision and analyte
concentration should be established
1.12 Repeatability
From the repeatability standard deviation or or sr it is useful to calculate the ‘repeatability limit ‘r’’, which
enables the analyst to decide whether the difference between duplicate analyses of a sample, determined
under repeatability conditions, is significant.
1.13 Reproducibility
From the reproducibility standard deviation o R or sR it is useful to calculate the ‘reproducibility limit
‘R’, ‘which enables the analyst to decide wheth er the difference between duplicate analyses of a
sample, determined under reproducibility conditions, is significant. These calculations can be
performed directly with the built-in statistics function of the instrument, if available, or by using a
pocket calculator or a PC (Personal Computer) with a suitable software pac kage (e.g. spreadsheet
program).
Measurement uncertainty is a single parameter (usually a standard deviation with a coverage factor or
confidence interval) expressing the range of values possible on the basis of the measurement result. A
measurement uncertainty estimate takes account of all recognised effects operating on the result; the
uncertainties associated with each effect are combined according to well-established procedures.
An uncertainty estimate for analytical chemistry is often termed an ‘uncertainty budget’ and should
take into account:
Where the contribution of individual effects is important, for example in calibration laboratories, it will
be necessary to consider the individual contributions from all individual effects separately.
Note that, subject to additional consideration of effects outside the scope of a collaborative trial, the
reproducibility standard deviation forms a working estimate of a measurement uncertainty provided
that the laboratory’s bias, measured on relevant materials, is small with respect to the reproducibility
standard deviation, the in-house repeatability precision is comparable to the standard method
repeatability and the laboratory’s intermediate precision is not large than the published
reproducibility standard deviation.
1.15 Sensitivity
This is effectively the gradient of the response curve, i.e. the change in instrument response, which
corresponds, to a change in analyte concentration. Where the response has been established as linear
with respect to concentration, i.e. within the linear range of the method, and the intercept of the response
curve has been determined, sensitivity is a useful parameter to calculate and use in formulae for
quantitation. Sensitivity is sometimes used to refer to limit of detection but this use is not generally
approved.
Ruggedness is normally evaluated during method development, typically by the originating laboratory,
before collaborating with other laboratories and is a measure how well a method stands up to less than
perfect implementation. In any method there will be certain stages, which, if not carried out sufficiently
carefully, will have a severe effect on method performance, and may even result in the method not
working at all. These stages should be identified, usually as part of method development, and if possible,
their influence on method performance evaluated using ‘ruggedness tests’, sometimes also called
‘robustness tests’. This involves making deliberate variations to the method, and investigating the
subsequent effect on performance. It is then possible to identify the variables in the method, which have
the most significant effect and ensure that, when using the method, they are closely controlled. Where
there is a need to improve the method further, improvements can probably be made by concentrating on
those parts of the method known to be critical. Ruggedness tests are normally applied to investigate the
effect on either precision or accuracy.
1.17 Recovery
Analytical methods do not always measure all of the analyte of interest present in the sample. Analytes
may be present in a variety of forms in samples not all of interest to the analyst. The method might be
deliberately designed to determine only a particular form of the analyte. However, a failure to determine
all of the analyte present may reflect an inherent problem in the method. Either way, it is necessary to
assess the efficiency of the method in detecting all of the analyte present.
6
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
Because it is not usually known how much of a particular analyte is present in a test portion it is
difficult to be certain how successful the method has been at extracting it from the matrix. One way
to determine the efficiency of extraction is to spike test portions with the analyte at various
concentrations, then extract the fortified test portions and measure the analyte concentration. The
inherent problem with this is that analyte introduced in such a way will probably not be held as
strongly as that which is naturally present in the test portion matrix and so the technique will give an
unrealistically high impression of the extraction efficiency. It is however the most common way of
determining recovery efficiency, and it is recognised as an acceptable way of doing so. However
the drawback of the technique should be borne in mind. Alternatively it may be possible to carry out
recovery studies on reference materials, if suitable materials are available. Provided these have
been produced by characterisation of natural materials rather than by characterisation of synthetic
materials into which the analyte has been spiked, then the recovery study should represent the
extraction of real test portions.
(2) Sample blanks: These are essentially matrices with no analyte. They are difficult to obtain but
such materials are necessary to give a realistic estimate of interference that would be encountered
in the analysis of test samples.
(3) Samples / test materials: Test materials taken from real samples are useful because of the
information they yield on interferences etc. which could be realistically encountered in day-to-day
work. If the true analyte content of a test material is accurately know it can be used as a way of
assessing the accuracy of the method. However the true analyte content is usually difficult to
determine unless there is the possibility of using other methods which are known to show negligible
bias.
(4) Spiked material: These are material or solutions, which have been fortified with the analyte(s) of
interest. These materials or solutions may already contain the analyte of interest so care is needed
lest fortification inadvertently leads to levels outside of the range of applicability of the method.
Fortification with a known amount of analyte enables the increase in response to the analyte to be
measured and calculated in terms of the amount added (assuming 100% recovery), even though
the absolute amounts of analyte present before and after the fortification are not know. Note that
most methods of fortification add the analyte in such a way that it will not be as closely bound to the
sample matrix as it would be if it was present naturally. Therefore, recovery determinations
obtained by fortification can be expected to be over-optimistic. The nature of the spike obviously
needs to be identified.
(5) (Measurement) Standards: These are traditionally thought of as solutions of single substances
but in practice can be anything in which a particular parameter or property has been characterised
to a sufficient extent it can be used for reference or calibration purposes.
(6) Reference materials: frequently confused with certified reference materials. Reference
materials can be virtually any material used as a basis for reference, and could inc lude laboratory
reagents of known purity, industrial chemicals, or other artefacts. The property or analyte of interest
7
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
needs to be stable and homogenous but the materials does not need to have the high degree of
characterisation, traceability and certifi cation more properly associated with certified reference
materials.
(7) Certified reference materials: The characterisation of the parameter of interest in a certified
reference material is generally more strictly controlled than for a reference material, and in
addition the characterised value is certified with a stated uncertainty by a recognised institution.
Characterisation is normally done using several different methods, so that as far as possible, any
bias in the characterisation is reduced or even eliminated.
8
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
APPENDIX 4
INTRODUCTION
This case study is based on a recent validation exercise conducted by Reagecon, as part of a new
product introduction. It details below the steps taken in the validation and samples of the methods,
which were validated in the form of Test procedures. Finally an example spreadsheet shows the
data analysis, results and conclusions.
SUMMARY
The following general protocol is designed to provide traceability to primary standards and
measures of accuracy, linearity and precision for the determination of chemical raw materials: TCA,
DCA and Acetic anhydride, and active ingredients in Reagecon manufactured products by pH and
REDOX titration. In addition it is intended to provide ongoing control of the methods.
The chart ‘Validation Summary data’ is the key to the validation plan. It tabulates those Products
and Ingredients, which form the scope of the study, the procedure names, reference to the
procedures on which the methods are based, the location of raw and calculated data and a
summary of the results.
As a condensed example of the practice and evolution of this exercise, the procedures and results
of the necessary steps to validate the testing of TCA/DCM are contained in this report. These are
Validation Procedure for Sodium Hydroxide Burette (Burette 2), Validation Procedure for the assay
of TCA (the active ingredient) and Validation Procedure for TCA/DCM (the finished product).
1 CALIBRATION OF ELECTRODES
Calibration requirements:
In the case of non aqueous titrations, perform a calibration of the electrode pair: pH
Glass electrode/ Double Junction sleeve junction reference with an outer salt bridge
9
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
composed of 2 mol/l Lithium chloride in propan-2-ol, to check the electrode
parameters. Fresh buffer solutions (specified value ± pH 0.01) must be used for
this purpose. Use CALIBRATION PROCEDURE FOR pH ELECTRODE -
Reagecon document code:- CALPH.
Calibration requirements:
Each burette will be validated according to the procedure given for each Burette
Validation. This will quantify the volumetric accuracy, reproducibility and linearity of
the burette and standardise the burette against a NIST traceable reference
material.
When steps one and two have been completed accuracy of the components of the
autotitrator is ensured. The concentration of the titrant is accurately know and is
traceable to a primary standard.
For each of the raw materials and products, method development is concerned with
the specific conditions of titration. These are optimised by experiment with typical
product samples, and the optimised conditions are stored in the memory of the
Metrohm 702SM Titrino autotitrator with method names given in the validation
procedures.
The methods are now validated using the specific validation procedure with laboratory
prepared amounts of the products and ingredients. This procedure measures the mean of
ten replicate titrations, linearity in the range of interest (typically, target value, low and high
values ± 25% of target value), standard deviation and a measure of any systematic error
that may be present. It also provides similarly comprehensive data for a control material for
subsequent use in routine analysis to ensure adequate Quality Control of results obtained
on production batches.
10
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
CONTROL PROCEDURE
Guidelines
Summary
As a guideline for the preparation of standard operating procedures to check
a titration system comprising a titrator, dispensing unit, measuring chain and
possibly a sample changer, use the procedure described below. The limiting
values specified must be considered as recommendations . Specific limiting
values must be defined in the particular standard operating proc edure
regarding in -house requirements to the demanded accuracy of the
measurement system .
Test intervals
Annual . A special validation is advisable when one or more components of
the titration system are replaced.
Maintenance/Service
An indispensable requirement to assure operation conforming to GLP for all
instruments used in the laboratory is careful maintenance and cleaning.
Particular attention should also be paid to the accurate handling of such
instruments. The instructions for use supplied with the instrument should be
accessible to all workers in the laboratory.
Method
Apparatus required
Chemicals required
• Primary standard, potassium hydrogen phthalate, certified, declared
content min. 99.95 %, dried for 2 h at 120°C and then allowed to cool off
in a desiccator, where it is stored
11
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
• Fresh titrant c(NaOH) = 1 mol/L
Sodium hydroxide readily absorbs carbon dioxide from the ambient
atmosphere. Protect your titrant solution against the ingress of CO2 by
attaching a drying tube filled with CO2 absorber.
Requirements
Protect experimental setup against direct sunlight and avoid draughts. The
system must be in thermal equilibrium.
E
B Burette tip
E
Electrode
The primary standard must be dried in a flat and allowed to cool off in a
desiccator for at least 1 h. Standard substances must always be stored in a
desiccator.
Calibration requirements:
12
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
In end-point titrations (SET) to a preset pH value, a calibration is essential.
Further, it is advisable to enter the working temperature for com pensation in
the titrator or attach a Pt100 or Pt1000 sensor to the titrator. The titrant
solution should be in thermal equilibrium with the surroundings.
Procedure
1. Calculation formula for the titrant molarity
C00 * C01
Titer = RS1 = or C00 * C01/ (C02 * EP1) with 4 decimal places
C02 * EP1
Method
10 titrations are performed with the same instrument settings and diffe rent
weights of the primary standard (e.g. potassium hydrogen phthalate). The
sample size should be varied in random order and result in a consumption of
titrant of ca. 0.1 to 1.0 cylinder volume. Refilling of the cylinder should be
avoided, except between samples.
13
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
Mean value
x1 + x 2 +K+xn 1 n
∑x
Sum of the individua l values
x= = i =
n n i= 1 Number of individual values
2
n
n
∑ xi
2 ∑ xi −
2 i=1
∑( )
n n
1 i=1
Standard deviation sabs = xi − x =
n−1 i= 1 n−1
(While the limiting value of 0.3 % for the rel. standard deviation is a limit
conforming to practice and can easily be met in the normal case, under
optimum conditions rel. standard deviations of 0.1 % and lower are
obtainable.)
Accuracy
The accuracy of the results obtained depends on the content of the primary
standard guaranteed by its producer (assumption: 100.00%).
14
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
Note: In sample titrations, reproducibility and linearity (volume vs sample
size) are important. There are normally no problems with the accuracy as
long as all titrant solutions are subjected to a regular molarity determination
and the molarity and the sample are determined with the same titration
settings.
2. Systematic errors
a. Linear regression volume/sample size
The linear regression draws a line through the experimental points, which
minimises the sum of the squares of the individual deviations. The regression
line is described by the formula: y = bx + a , where a represents the
intercept on the y-axis and b is the slope of the line (see diagram below).
asys
Sample size
asys as a measure of the systematic error is calculated from the mean values
of the x values, the mean values of the y values and the regression
coefficient b (slope).
∑ xi * ∑y i
∑( )( )
n n
i= 1
xi − x y i − y ∑
i= 1
xiyi − i= 1
n
i= 1
b= =
∑ (x )
n 2
2
n
i=1
i −x
n
∑ xi
∑xi= 1
i
2
− i= 1
n
asys = y-intercept = x − b * y
15
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
Assessment
If asys > ± 0.010 mL (or ± 10 µL), it must be assumed that a systematic error
is present. A check on the titration method and other possible disturbing
influences due to the system is then imperative. If no optimisation of the
validation method is possible, the individual values of the consumption in mL
must be corrected by the value of a sys
(volume–a sys in mL) to ensure that the systematic error associated with the
method is not incorporated in the assessment of the titrator. The relevant
characteristic data for the reproducibility and the accuracy of the titration
results must then be recalculated with the corrected consumption values.
Volume Volume
Assessment
If bT/Vol is greater than ± 0.0010, a systematic error due to the method must
also be assumed here. A correction of the consumption values by asys
(volume in mL – asys in mL) and a subsequent recalculation of the molarity
shows a dramatic improvement when the regression line (molarity against
volume) is replotted.
Conclusion
If systematic errors are found, an attempt must be made to optimise the titration
method and adapt the standard operating procedure (SOP) accordingly. If no
optimisation is possible or a specified method must be used unchanged, the
relevant characteristic data must be calculated with corrected cons umption values
(volume in mL – asys in mL).
16
C op yr igh t 2 00 3 by Reag ec on D ia gn ost ics L t d. A ll R ig hts R eser ve d
METTLER TOLEDO Titrators
Res
f Lim
Con
ne
t i o n Li
b ra
Cali
f Lim
Con
VEQ V
Dear Reader
G. Reutemann H. Huber
Manager Market Support Regional Market Manager
North East Asia
Preface .................................................................................................................. 4
1 Summary .............................................................................................................. 5
2 Principle of Validation ......................................................................................... 5
3 Steps of Validation and Recommended Limits ................................................. 6
3.1 Definition of Accuracy .......................................................................................... 6
3.2 Definition of Precision .......................................................................................... 6
3.3 Systematic Errors and Linearity ............................................................................ 7
3.3.1 Definition of Systematic Errors ............................................................................. 7
3.3.2 Definition of Linearity ........................................................................................... 7
3.4 Definition of Robustness and Ruggedness ............................................................ 8
3.5 Definition of Determination Limit ........................................................................ 9
4 Practical Hints ................................................................................................... 10
4.1 Preparations and Precautions ............................................................................... 10
4.2 Titration Control Parameters ............................................................................... 10
4.3 Titration Evaluation Parameters .......................................................................... 10
4.4 Titration ............................................................................................................... 11
5 Possible Sources of Error .................................................................................. 11
6 Recommendations for Troubleshooting ........................................................... 12
7 Results not Conforming to Specifications ....................................................... 13
8 Examples ............................................................................................................ 14
8.1 Determination of Sulphuric Acid ........................................................................ 14
8.2 Titer Determination ............................................................................................. 15
8.3 Precision and Accuracy ....................................................................................... 16
8.4 Systematic Errors, Linearity ................................................................................ 17
8.5 Robustness and Ruggedness ................................................................................ 18
8.6 Determination Limit ............................................................................................ 20
8.7 Closing Remarks ................................................................................................. 21
9 Appendix 1 ......................................................................................................... 22
10 Appendix 2 ......................................................................................................... 24
10.1 Assessment of Results ......................................................................................... 24
10.2 Precision versus Accuracy ................................................................................... 24
11 Glossary .............................................................................................................. 25
12 Literature ........................................................................................................... 26
The method must be simple yet accurate. Serious consideration must be given to the possibil-
ity of the method being automated. This strategy eliminates human errors, increases produc-
tivity and reduces the time needed to release the product to consumers without compromising
method accuracy and reproducibility hence, product quality. Of course along with all the
methods simplification comes analysis cost-saving. In addition, the method must be environ-
ment-friendly via reduction, if not elimination, of the conventional use of organic solvents,
especially non-biodegradable and toxic chemical reagents.
Accepted protocols/SOP's followed during method validation work are generally universal,
irrespective of country, industry, company or product category. The only thing that is differ-
ent between them is the accuracy required, with the strictest limits applied to those products
prepared for human consumption, e.g., food and medicine.
In this brochure, METTLER TOLEDO summarizes the general method development protocols.
These include accuracy, reproducibility, linearity, ruggedness and limit of determination. Each
method must pass all these tests, just like the consumer products for which the method will be
used.
A detailed method is applied to obtain correct results. This method describes all the different
steps from the sampling to the result. Whether correct results can be obtained or not with a
certain method has to be validated. Validation of a method comprises tests for accuracy,
precision, linearity, systematic errors, robustness/ruggedness and detection limit/determina-
tion limit. So the validation of a method proves, whether or not the instruments used for this
purpose fulfill the specific requirements.
In the context of validation a variety of expressions is used. Please refer to the glossary
(chapter 11) for a short definition of the expressions used in this brochure.
2 Principle of Validation
Accuracy, Precision, Linearity, Systematic Errors, Robustness/Ruggedness and Determina-
tion limit are checked, considering the complete analytical procedure from taking the sample
to result calculation and documentation.
The following application serves as a guideline, showing how a titration method can be vali-
dated. As an example, the method for the determination of sulphuric acid was validated.
Recommended limits for accuracy, reproducibility and linearity are subject to the tested method.
Other methods e.g. analysis of foods and drugs may require much stricter limits.
Recommendation:
Results obtained should not deviate from the true value by more than 0.3%.
Systematic errors of a titration are for example disturbing influences due to the method itself
or to solvent blank values.
In the linear regression according to chapter 3.3 systematic errors show up as a significant
deviation of the y-axis intercept a of the regression line from the zero point coordinates (see
graph 1), i.e. asys is clearly different from zero.
graph 1:
volume
asys
sample size
Recommendation:
The systematic error asys should be smaller than 15 µL. If it can not be eliminated by optimizing
the method or the reagents, it has to be corrected for in the results calculation of the titration
method.
Linearity expresses whether a certain method produces correct results over the interesting
concentration range [4]. In titration the analyte concentration depends on the sample concen-
tration, on the sample size and on the solvent volume added for the analysis. By varying the
sample size and thereby the analyte concentration, the linearity of a titration method may be
detected in the range of interest.
A) The regression coefficient (R2) of the linear regression described in graph 1 must be
better than a given limit, depending on the demanded accuracy for the specific determi-
nation:
i.e. R2 > 0.995
B) A significant positive or negative slope b (resp ∆R/∆V) of the regression line in graph
2 (results of the titration versus sample size) indicates a non-linearity of the titration
method, meaning that the result depends on the sample size.
graph 2:
result
result
non linear
linear
Recommendation:
If ∆R/∆V is greater than 0.1%, a systematic non-linearity has to be assumed.
The RUGGEDNESS describes the correctness of the results obtained under disturbed experi-
mental (analytical) conditions such as different matrices (e.g. solutions, reagents etc.), other
temperatures of the analyte solution or elsewise deviating conditions. To determine rugged-
ness, the same sample is titrated with and without exposure to relevant disturbances. If the
results are the same, the method is considered to be rugged against this specific influence.
The determination limit is determined by titrating sample series, each with a continuously
reduced amount of sample. The determination limit is the smallest amount of substance (mmol)
or sample, which can be titrated with a good precision (RSD) of ≤ 0.3%. It can be evaluated
by intrapolation of the graph “amount of substance versus relative standard deviation”.
determination limit
rel. stand. dev. [RSD]
0.3%
graph 3:
amount of substance [mmol]
In normal routine titration the determination limit is not a problem, since one can simply
enlarge the sample size to get a better response from the electrode. However, this can be
different, if the analyte has a very low concentration in the sample.
• The primary standard must be dried in a drying oven (e.g. 2 h at 105 °C, depending on the
type of primary standard) and cooled to ambient temperature in a desiccator for at least 1
hour. It should always be stored in a desiccator.
• For acid/base endpoint titrations, it is necessary to calibrate the pH electrode. Certified
buffers from METTLER TOLEDO may be used for this purpose.
• The experimental setup must be protected from direct sunlight and should be in thermal
equilibrium with the environment.
• The analytical balance must have a vibration free standing and should be calibrated regu-
larly. METTLER TOLEDO balances of the MT, AT and PR series offer FACT (Fully
Automatic Calibration Technology), which automatically executes a calibration whenever
needed. All steps to ensure proper weighing must be observed [5].
• Samples should be titrated immediately after weighing and dissolution. Enough solvent
must be added to cover the sensor.
• When performing a series of titrations, the interval time between samples should be kept
to a minimum.
• In sample series, the electrode as well as stirrer and temperature sensor should be rinsed
between two measurements.
• Temperature compensation is essential for pH endpoint titrations.
• Sample size/Balance balance not accurate, air humidity too high or too low, contami-
nated balance, temperature changes or gradient from titration ves-
sel to balance, careless weighing, sample weight, concentration or
volume too low or too high, sample inhomogeneous, improper
sampling.
• Dispensing unit tube connections not tight, contaminated burette cylinder (visible
corrosion marks), leaky piston (liquid film or crystals below the
piston), leaking burette tip, air in tubing system, three-way stop-
cock leaking.
• Solvent impure (blank value), poor solubilising power, not stable, contami-
nated (e.g. by CO2, O2), wrong pH value or ionic strength.
• Titration parameters unsuitable titration mode, wrong measure mode parameters, titra-
tion rate too fast or too slow, unsuitable evaluation procedure.
All non-conforming values must be reported and commented on in the validation record and
the subsequent procedure noted and explained.
If relevant deviations are found, the sections “Possible Sources of Error” and “Recommen-
dations for Troubleshooting” must be checked carefully and the disturbing influences elimi-
nated. It is essential to repeat the validation afterwards.
The titrators of METTLER TOLEDO have undergone various tests during development and
manufacturing. Furthermore, they have been time tested by numerous users in different ap-
plications all over the world and considered to be okay. If irregular results are obtained,
primary consideration should be given to the working technique of the operator or to wrong
or accidentally altered titration parameters.
Titrant . . . . . . . . . . . . . NaOH
Printer HP Deskjet Concentration [mol/L] . . . . . . 1.0
Sensor . . . . . . . . . . . . . . DG111-SC
Unit of meas. . . . . . . . . . . As installed
Accessories: DT120 T-sensor Titration mode . . . . . . . . . . EQP
Predispensing 1 . . . . . . . . mL
Volume [mL] . . . . . . . . 2
Indication: DG111-SC Titrant addition . . . .
∆E(set) [mV] . . . . .
.
.
.
.
.
.
DYN
8.0
Limits ∆V . . . . . . . . . Absolute
∆V(min) [mL] . . . . . . 0.05
∆V(max) [mL] . . . . . . 0.3
Measure mode . . . . . . . . . EQU
∆E [mV] . . . . . . . . . . 1.0
∆t [s] . . . . . . . . . . . 1.0
t(min) [s] . . . . . . . . . 3.0
t(max) [s] . . . . . . . . . 15.0
Threshold . . . . . . . . . . . 3.0
Maximum volume [mL] . . . . . . 10.0
Termination after n EQPs . . . Yes
n = . . . . . . . . . . . . 1
Evaluation procedure . . . . . Standard
Rinse
Auxiliary reagent . . . . . . . . H2O
Volume [mL] . . . . . . . . . . . 10.0
Calculation
Result name . . . . . . . . . . . H2SO4 Conc.
Formula . . . . . . . . . . . . . R=Q*C/U
Constant . . . . . . . . . . . . . C=M/z
Result unit . . . . . . . . . . . g/L
Decimal places . . . . . . . . . . 5
Record
Output unit . . . . . . . . . . . Printer
Raw results last sample . . . . . Yes
Table of values . . . . . . . . . Yes
E - V curve . . . . . . . . . . . Yes
Conditioning
Interval . . . . . . . . . . . . . 1
Time [s] . . . . . . . . . . . . . 10
Rinse . . . . . . . . . . . . . . Yes
Auxiliary reagent . . . . . . . H2O
Volume [mL] . . . . . . . . . . 10.0
Statistics
Ri (i=index) . . . . . . . . . . . R1
Standard deviation s . . . . . . . Yes
Rel. standard deviation srel . . . Yes
Outlier test . . . . . . . . . . . Yes
Record
Output unit . . . . . . . . . . . Printer
All results . . . . . . . . . . . Yes
The titer of this NaOH (1.0 mol/L) was determined against the primary standard potassium
hydrogen phthalate (dried for 2 h at 150 °C). The following results were obtained:
Results:
Titer
9 1.73847 1.0024 1.010
10 1.32274 1.0011
1.005
11 1.63289 1.0008
1.000
12 1.17670 1.0011
0.995
13 0.64051 1.0018
14 1.52071 1.0017 0.990
Number of samples: 22
Mean value x: 1.0012
Standard deviation: 9.06 • 10-4
Relative Standard Deviation: 0.0905 %
Comment:
The standardization is highly reproducible and linear. Results do not depend on the sample
weight.
945
Number Size Result Temperature:
y = 4.9106 - 1.030e-4x R =21
0.43 °C
940 Theoretical Content: 4.9030 g/L
[mL] [g/L] 935 No. of Samples: 12
Mean Value: 4.90529 g/L
930
Stand. Dev.: 0.004752 g/L
1 75 4.90143 925
Rel. Stand. Dev.: 0.0968%
920
2 34 4.90828
H2SO4 [g/L]
1234567890123456789012345678
915
3 44 4.89803 1234567890123456789012345678
910
1234567890123456789012345678
4 60 4.90046 905 1234567890123456789012345678
1234567890123456789012345678
5 35 4.90870 900 1234567890123456789012345678
895
6 44 4.90672
890
7 77 4.89912 885
8 32 4.90273 880
Number of samples: 12
Theoretical value: 4.9030 g/L
Mean value found: 4.90529 g/l
Deviation to theoretical: 0.00229 g/L
Relative deviation to theoretical: 0.0467%
Standard deviation: 0.04752 g/L
Relative standard deviation: 0.0968%
Comment:
Precision as well as accuracy are excellent. The requirements are easily met.
43 4.3972 4.90672
4
77 7.6831 4.89912
32 3.1953 4.90273 3
52 5.2001 4.90995
2
91 9.0925 4.90584
31 3.1008 4.91109 1
42 4.2011 4.91117
0
0 20 40 60 80 100
4.90
The second way to determine the linearity is to
plot the results of the titration (H2SO4 in g/L) 4.89
against the sample size, as one can see in the graph
nearby. Then a linear regression is performed on 4.88
The results show a systematic error (SE) and non-linearity (NL). Presumable causes are
pipetting errors when preparing the samples.
SE as well as NL are very small and well below the recommended limits.
The uptake of carbon dioxide CO2 from ambient air is the major threat of alkaline titrants.
CO2 reacts to CO32-. Carbonate precipitation and reduction of the strength of the titrant are
the consequences.
Other analytical influences as well as the robustness (see chap. 3.4) can be checked in a
similar way.
The ruggedness of the sulphuric acid method was evaluated by exposing the titrant to air and
thereby also to CO2. Batches of NaOH titrants were exposed to air for 1, 2, 3, 4, 5, 6, 7 days.
The CO32- content of each sample was determined by titration with sulphuric acid.
Result CO32- [mg/L]
CO32--Content 30000
1 2526
2 5026
3 8793 10000
4 14422
6 20684
7 24568
0
0 2 4 6 8
Air exposure [days]
Each NaOH batch was then standardised against potassium hydrogen phthalate. Then it was
used as the titrant to determine the H2SO4 concentration. The following table shows the cor-
responding results.
Comment:
The method was found not to be rugged at all against exposure to air. Even the NaOH sample
exposed to air for only one day did not allow correct determinations any more.
Reason: When titrating strong acids with NaOH that contains CO32-, a typical double
jump of the titration curve is found, caused by the following reactions:
However, this double jump does not occur when titrating weak acids such as potassium hy-
drogen phthalate, which is mainly used for the titer determination. Therefore, the carbonate
error cannot be compensated for by frequent standardization of the titrant. It is advisable to
periodically check the carbonate content by a specific titration and dispose of the titrant if a
significant amount of carbonate is found.
The determination limit was examined using 0.005 mol/L NaOH. Series of 5 to 6 samples
were run, in order to check the reproducibility (RSD) with a low amount of sample.
The results show, that in the sample with less than 0.01 mmol sulphuric acid the Relative
Standard Deviation RSD increases drastically and continuously, while the absolute standard
deviation s remains more or less constant. The uptake of CO2 from the air is very severe in
this concentration range. Therefore the titrant has to be protected from air intake with an
absorption tube filled with NaOH on a carrier. Even then, it remains usable only for one day.
Comment: 2
The limits for different parameters or the set up of priorities in the validation process have to
be adapted by the user depending on the method and the specifications for a given task (e.g.
other tests for the determination of ruggedness).
Anyway the basic course of a method validation remains the same. Thus, this example may
well serve as a guideline for further validations.
. * 1) These substances can not be acchieved as guaranteed primary standard substances from MERCK.
So the highest aviable quality is indicated.
Errors
Deviations from the correct or expected value
Correctness
high precision
(small RSD)
low precision
(big RSD)
This application bulletin represents selected, possible application examples. These have been tested with all possible
care in our lab with the analytical instrument mentioned in the bulletin. The experiments were conducted and the
resulting data evaluated based on our current state of knowledge.
However, the application bulletin does not absolve you from personally testing its suitability for your intended methods,
instruments and purposes. As the use and transfer of an application example are beyond our control, we cannot accept
responsibility therefore.
When chemicals and solvents are used, the general safety rules and the directions of the producer must be
observed.
The IHCP carries out research to improve the understanding of potential health risks posed by
chemical, physical and biological agents from various sources to which consumers are exposed.
European Commission
Joint Research Centre
Institute for Health and Consumer Protection
Contact information
Address: DG JRC, IHCP TP 260, I-21020 Ispra
E-mail: Catherine.simoneau@jrc.ec.europa.eu
Tel.: 39.0332.785889
Fax: 39.0332.785707
http://ihcp.jrc.ec.europa.eu/
http://www.jrc.ec.europa.eu/
Legal Notice
Neither the European Commission nor any person acting on behalf of the Commission is responsible for
the use which might be made of this publication.
(*) Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.
A great deal of additional information on the European Union is available on the Internet.
It can be accessed through the Europa server http://europa.eu/
JRC 53034
EUR 24105 EN
ISBN 978-92-79-14483-7
ISSN 1018-5593
DOI 10.2788/49046
Printed in Italy
2
Executive Summary
Test methods for materials and articles in contact with foodstuffs are required to
determine the concentration of residues of monomers in the materials themselves or to
determine the concentration of individual or groups of substances in food (or food
simulants) which have migrated from the food contact materials.
The Community Reference Laboratory and National Reference Laboratories for food
contact materials (FCM) prepared the present Guidelines to illustrate the required
performance criteria for the analytical methods applied in the laboratories for FCM and
provide procedures for method validation in order to estimate their performance
characteristics. The scope of these guidelines is to provide rules for the performance of
the analytical methods to be used in the verification of compliance with the migration
limits defined in Directive 2002/72/EC,as amended, and in accordance with Directive
82/711/EEC, as amended, and others defined in the European legislation, in order to
ensure the quality and comparability of the analytical results.
The document presents 4 approaches, according to the different purpose of
performance assessment.
These guidelines are intended as a dynamic document and they will evolve and
expand into further editions. This is the first edition. These guidelines have been
endorsed by the European Union official Network of National Reference Laboratories
and approved by the EU Commission competent service DG SANCO.
This work also highlights an important deliverable for the Network of NRLs. In
particular, the members of the task force “Method Performance” that have dedicated
time and effort to provide input into the development of these guidelines. They are
gratefully acknowledged here for their contribution: NRL-BE (Fabien Bolle, Tina n’Goy),
NRL-DE (Oliver Kappenstein), NRL-DK (Jens Petersen), NRL-ES (Juana Bustos),
NRL-FR1 (Patrick Sauvegrain), NRL-EL (Timokleia Togkalidou), NRL-IT (Maria
Rosaria Milana), NRL-NL (Durk Schakel, Dita Kalsbeek-van Wijk), NRL-PL (Kazimiera
Cwiek-Ludwicka), NRL-SI (Viviana Golja), NRL-UK (Emma Bradley). Special thanks
are extended to Emma Bradley for her contribution to the editing of the document.
Acknowledgement:
This work was performed thanks to the CRL mandate:
SANCO/2007/FOODSAFETY/0043-Contact Materials
SANCO/2008/FOODSAFETY/0047-Contact Materials
SANCO/2008/FOODSAFETY/055-Contact Materials
3
4
Table of Contents
1 INTRODUCTION ........................................................................................................................7
2 SCOPE OF THIS DOCUMENT .................................................................................................8
3 GLOSSARY - DEFINITIONS.....................................................................................................8
4 METHOD VALIDATION PLAN ...............................................................................................9
4.1 CHOICE OF VALIDATION SCHEME ...........................................................................................9
4.1.1 “Full” single laboratory validation protocol applicable to the field of food contact
materials and articles. ..................................................................................................................10
4.1.2 “Standard level” of single laboratory validation applicable to the field of food contact
materials and articles ...................................................................................................................10
4.1.3 “Basic level” of single laboratory validation applicable to the field of food contact
material and articles.....................................................................................................................10
4.1.4 Method verification ........................................................................................................11
5 “FULL” SINGLE-LABORATORY METHOD VALIDATION PROCEDURE..................15
5.1 PERFORMANCE CHARACTERISTICS OF THE MIGRATION PART ...............................................16
5.2 PERFORMANCE CHARACTERISTICS OF THE DETERMINATION (ANALYTICAL) PART ...............18
5.2.1 Selectivity/specificity ......................................................................................................19
5.2.2 Ruggedness.....................................................................................................................19
5.2.3 Calibration range, assessment of the calibration function.............................................21
5.2.4 Working concentration range.........................................................................................32
5.2.5 Limit of detection (LOD) / method detection limit (MDL)..............................................32
5.2.6 Limit of quantification ....................................................................................................35
5.2.7 Accuracy.........................................................................................................................35
5.2.8 Uncertainty.....................................................................................................................49
5.3 REPORTING OF RESULTS .......................................................................................................55
5.4 INTERPRETATION OF RESULTS ..............................................................................................55
5.5 VALIDATION/VERIFICATION DOCUMENTATION ....................................................................55
5.6 MONITORING OF THE VALIDATED METHOD ..........................................................................56
6 “STANDARD LEVEL” VALIDATION SCHEME FOR FCM METHODS........................57
6.1 SELECTIVITY/SPECIFICITY ....................................................................................................57
6.2 RUGGEDNESS .......................................................................................................................57
6.3 CALIBRATION RANGE, ASSESSMENT OF CALIBRATION FUNCTION .........................................57
6.3.1 Calibration range...........................................................................................................57
6.3.2 Basic of calibration and quantification ..........................................................................57
6.3.3 Assessment of the linearity model...................................................................................58
6.4 WORKING CONCENTRATION RANGE .....................................................................................58
6.5 LOD ....................................................................................................................................58
6.6 LOQ ....................................................................................................................................58
6.7 PRECISION............................................................................................................................59
6.8 TRUENESS ............................................................................................................................60
6.9 UNCERTAINTY .....................................................................................................................60
7 “BASIC LEVEL” VALIDATION SCHEME FOR FCM METHODS..................................61
7.1 SELECTIVITY/SPECIFICITY ....................................................................................................61
7.2 RUGGEDNESS .......................................................................................................................61
7.3 ASSESSMENT OF CALIBRATION FUNCTION............................................................................61
7.3.1 Calibration range...........................................................................................................61
7.3.2 Basic of calibration and quantification ..........................................................................61
7.3.3 Assessment of the linearity model...................................................................................61
7.4 WORKING CONCENTRATION RANGE .....................................................................................62
7.5 LOD ....................................................................................................................................62
7.6 LOQ ....................................................................................................................................62
5
7.7 PRECISION............................................................................................................................62
7.8 TRUENESS ............................................................................................................................63
7.9 UNCERTAINTY .....................................................................................................................64
8 REFERENCES ...........................................................................................................................64
6
1 INTRODUCTION
The European Framework Regulation (EC) No. 1935/2004 [29] is the basic
Community legislation that covers all food contact materials and articles. It
empowers the European Commission to set requirements for specific materials.
Specific requirements for such materials can include limits on the overall migration
and on the specific migration of certain constituents or groups of constituents into
foodstuffs. These limits have been defined for some substances in plastic materials
and articles.
Test methods for materials and articles in contact with foodstuffs are required to
determine the concentration of residues of monomers in the materials themselves or
to determine the concentration of individual or groups of substances in food (food
simulants) which have migrated from the food contact materials or to determine
overall migration from food contact materials.
The determination of migration from materials and articles intended to come into
contact with foodstuffs is quite unlike any other measurement tasks in ensuring food
safety and quality. Reliable measurements depend upon more than simply having
validated analytical methods for measuring chemical concentrations in foods. The
Directives allows, as an alternative to the analysis of the foodstuff itself, migration
testing to be carried out with food simulants applied under conditions which simulate
actual use of the material or article with food. This introduces additional potential
sources of variability in the final migration value.
It is necessary to ensure the quality and comparability of the analytical results
generated by laboratories for enforcement purposes, for compliance purposes, and
for the creation of data for risk assessment purposes. This should be achieved by
using quality assurance systems and specifically by applying methods that have
been validated according to common procedures and that meet defined performance
criteria, and by ensuring traceability to common standards or standards that are
commonly agreed upon.
Commission Regulation (EC) No. 882/2004 [30] on official controls performed to
ensure the verification of compliance with feed and food law, animal health and
animal welfare rules, which also incorporates food contact materials, requires official
control laboratories to be accredited according to EN ISO/IEC 17025 [21]. Moreover,
approved laboratories must prove their competence by regular and successful
participation in adequate proficiency testing schemes recognised or organised by the
national or Community reference laboratories.
A network of Community Reference Laboratory and National Reference Laboratories
for food contact materials (FCM) operates under Commission Regulation (EC) No.
882/2004 [30] to enhance coordination.
In the field of materials and articles in contact with food numerous chemicals are
used in the manufacturing processes and it is not possible to prepare standard test
methods for all. Therefore the concept of routine methods and reference methods
should be superseded by a criteria approach, in which performance criteria and
procedures for the validation of screening and confirmatory methods are defined.
It is necessary to determine common criteria for the interpretation of test results of
official control laboratories in order to ensure a harmonised implementation of
Commission Regulation (EC) No. 882/2004 [30].
7
2 SCOPE OF THIS DOCUMENT
The scope of these guidelines is to provide rules for the performance of the
analytical methods to be used in the verification of compliance with the migration
limits defined in Directive 2002/72/EC [22], as amended, and in accordance with
Directive 82/711/EEC [23], as amended, and 85/572/EEC [24], as amended, and the
further provisions set out in Annex 1 of Commission Directive 2002/72/EC [22].
The verification of compliance with the migration limits is made using methods that:
(a) are documented in test instructions, preferably according to ISO 78-2 [20];
(b) comply with the performance criteria defined in these guidelines;
(c) have been validated according to the procedures described in these
guidelines.
The quality of the results of the analysis of samples for verification of the compliance
with the migration limits should be ensured according to Chapter 5.9 of ISO 17025
[21].
3 GLOSSARY - DEFINITIONS
Abbreviations Explanation
The closeness of agreement between a test result and the accepted reference value. It is
Accuracy
determined by determining trueness and precision.
The substance, and any derivatives emerging during its analysis, that has to be detected,
Analyte
identified and/or quantified.
Bias The difference between the expectation of the test result and an accepted reference value
A device for measurements that represents the quantity of substance of interest in a way that ties
Calibration standard
its value to a reference base
Certified reference A material that has had a specified analyte content (or for food contact materials a migration
material (CRM) value) assigned to it
Analysing the same sample by the same method to determine the performance characteristics of
Collaborative study
the method. The study covers random measurement error and laboratory bias
Compositional limit (QM) The maximum permitted amount of the named substance in the material or article.
Methods that provide full or complementary information enabling the substance to be
Confirmatory method
unequivocally identified and if necessary quantified at the level of interest.
Food Simulant A medium intended to simulate (‘mimic’ or ‘model’) the essential characteristics of a foodstuff.
Fortified sample material A sample enriched with a known amount of the analyte to be detected
Organisation, performance and evaluation of tests on the same sample by two or more
Interlaboratory study
laboratories in accordance with predetermined conditions to determine testing performance.
(comparison)
According to the purpose the study can be classified as collaborative study or proficiency study.
A substance not contained in the sample with physical-chemical properties as similar as possible
Internal standard (IS) to those of the analyte that has to be identified and which is added to each sample as well as to
each calibration standard.
The concentration of a substance or analyte in a sample that is significant to determine its
Level of interest
compliance with legislation.
The mass of material transferred to the food simulant or test media as determined by the relevant
Overall migration
test method.
Functional quality that can be attributed to an analytical method. This may be for instance
Performance
specificity, accuracy, trueness, precision, repeatability, reproducibility, recovery, detection
characteristic
capability and ruggedness.
Requirements for a performance characteristic according to which it can be judged that the
Performance criteria
analytical method is fit for the purpose and generates reliable results.
The closeness of agreement between independent test results obtained under stipulated
(predetermined) conditions. The measure of precision usually is expressed in terms of
Precision
imprecision and computed as standard deviation of the test result. Less precision is determined
by a larger standard deviation.
Analysing the same sample allowing laboratories to choose their own methods, provided these
Proficiency study methods are used under routine conditions. The study has to be performed according to ISO
guide 43-1 [31] and 43-2 [32] and can be used to assess the reproducibility of methods.
8
Range (working or
Means a set of values for which a measure is intended to lay within specified uncertainty limits
measuring)
The percentage of the true concentration of a substance recovered during the analytical
Recovery
procedure. It is determined during validation, if no certified reference material is available.
A material of which one or several properties have been confirmed by a validated method, so that
Reference material
it can be used to calibrate an apparatus or to verify a method of measurement.
Conditions where independent test results are obtained with the same method on identical test
Repeatability conditions items in the same laboratory by the same operator using the same equipment and short interval
of time.
Precision under repeatability conditions (r) - the value below which the absolute difference
Repeatability between 2 single test results obtained under repeatability conditions, may be expected to lie
within a specific probability (typically 95 %) and hence r = 2,8 x sr.
Conditions where test results are obtained with the same method on identical test items in
Reproducibility conditions
different laboratories with different operators using different equipment.
Precision under reproducibility conditions (R) - the value below which the absolute difference
Reproducibility between 2 single test results obtained under reproducibility conditions, may be expected to lie
within a specific probability (typically 95 %) and hence R = 2,8 x sR.
Residual content The mass of the substance present in the final material or article.
The susceptibility of an analytical method to changes in experimental conditions which can be
expressed as a list of the sample materials, analytes, storage conditions, environmental and/or
sample preparation conditions under which the method can be applied as presented or with
Ruggedness
specified minor modifications. For all experimental conditions which could in practice be subject
to fluctuation (e.g. stability of reagents, composition of the sample, pH, temperature) any
variations which could affect the analytical result should be indicated.
Sample blank The complete analytical procedure applied to a test portion taken from a sample from which the
determination analyte is absent.
Methods that are used to detect the presence of a substance or class of substances at the level
of interest. These methods have the capability for a high sample throughput and are used to sift
Screening method
large numbers of samples for potential non-compliant results. They are specifically designed to
avoid false compliant results.
Single laboratory study An analytical study involving a single laboratory using one method to analyse the same or
(in-house validation) different test materials under different conditions over justified long time intervals.
The maximum permitted level of a group of named substances migrating from the final material or
SML(T) article into food or food simulants, expressed as total of chemical moiety or substance(s)
indicated.
Specific Migration The mass of the substance transferred to the food/simulant as determined in the test method.
Specific Migration Limit The maximum permitted level of a named substance migrating from the final material or article
(SML) into food or food simulants.
Ability of a method to distinguish between the analyte being measured and other substances.
Specificity This characteristic is predominantly a function of the measuring technique described, but can
vary according to class of compound or matrix.
A procedure in which the test sample is divided in two (or more) test portions. One portion is
analysed as such and known amounts of the standard analyte are added to the other test
portions before analysis. The amount of the standard analyte added has to be between two and
Standard addition
five times the estimated amount of the analyte in the sample. This procedure is designed to
determine the content of an analyte in a sample, taking account of the recovery of the analytical
procedure.
The closeness of agreement between the average value obtained from a large series of test
Trueness
results and an accepted reference value. Trueness is usually expressed as bias.
The confirmation by examination and the provision of effective evidence that the particular
Validation
requirements of a specific intended use are fulfilled.
Precision obtained in the same laboratory under stipulated (predetermined) conditions
Within-laboratory
(concerning e.g. method, test materials, operators and environment) over justified long time
reproducibility
intervals.
The detailed design and the correct execution of method validation studies should,
as far as possible, provide a realistic assessment of the number and range of effects
operating during normal use of the method, as well as covering the working
9
concentration range(s) and sample types that fall within the scope of the method.
The applicable working concentration range is an important part of the validation or
verification of the analytical method. It will often save both time and effort to choose
the area of application on the basis of laboratory or Regulatory needs, instead of
validating the whole range of possibilities.
It is also important that the validation and verification report describes which sample
matrices have been used. In fact, the performance of some methods may be also
dependent on the matrix. In such cases, it is very important to ensure that the entire
stated area of application is included in the validation.
Validation or verification of an already validated method must always be performed
before it is used in the laboratory for official control purposes. This work must be
repeated partly or fully if the result of the first validation makes it necessary to modify
the method.
In the following paragraphs four different approaches are presented:
This scheme represents the conditions specifically developed and agreed by the
official control laboratories for FCM which should be applied as a working standard
for use in the field of FCM. This validation scheme represents the minimum
requirements to establish non-compliance of a material or article intended for food
contact.
Experienced laboratories, who have already implemented more sophisticated
procedures, compliant with consolidated standards of validation, can continue to use
them provided they respect at least the minimum requirements set by the agreed
level described in these guidelines.
This scheme represents the base level that must be met by all laboratories. This
level does not fulfil all of the legal and official requirements, but it is considered as
10
the starting point from which a harmonised level of control (and results) in all
European countries can be achieved. From this starting point the laboratories will
have to update and improve their procedures and achieve the level foreseen for the
“Standard level” validation by January 2011.
This “Basic level” validation may also be used for emergency or occasional cases
and/or for a reduced number of samples. In such cases this approach may be used
beyond 2011.
Experienced laboratories, who have already implemented more sophisticated
procedures, compliant with consolidated standards of validation, can continue to use
them provided they respect at least the minimum requirements set by the agreed
level described in these guidelines.
Figure 1-4 show the respective flow charts for the four validation and verification
processes described above.
11
FULL VALIDATION
Worked examples
Specificity
At least 3 blanks (from
reproducibility)
Robustness
8 repetitions in at least
duplicate of the method
Linearity
6 calib. curves (from
reproducibility)
Working
range From calibration curve
Based on validation
From reprod. and bias
Uncertainty
12
FCM STANDARD SCHEME
Worked example
Linearity
3 calib. curve (from
reproducibility)
Reproducibility
3 repetitions of 0.2*LL, 1*LL and 2*LL
repeated for 3 different days by different
OR operators
Repeatability +
Accuracy Reproducibility
ANOVA (using a
combination of table 10)
Bias
Trueness
From CRM, RM, collaborative
study, PT, or spiked samples
Recovery
u Rw + ubias
2 2
Uncertainty uc= From reproducibility and recovery
Figure 2: Flow chart of the effective standard working validation scheme for FCM
13
BASIC LEVEL SCHEME
Worked example
Linearity
3 calib. curve (from
reproducibility)
Recovery
VERIFICATION
Worked example
Accuracy
From CRM, RM, collaborative
study, PT, or spiked samples
Bias
Trueness
LOD, LOQ
From blank or calib. curve
14
5 “FULL” SINGLE-LABORATORY METHOD
VALIDATION PROCEDURE
In the validation of test methods in the field of FCM the following cases could be
encountered:
• An overall migration test for the determination of the total quantity of all non-
volatile substances that have migrated from the FCM test specimen to a food
simulant
• A specific migration test for the determination of the substances in food
simulants that have migrated from a food contact material or article
• The determination of the substance(s) in food, that have migrated from a food
contact material or article
• The determination of residues of monomers in food contact materials and
articles intended to come into contact with foodstuffs.
In an overall migration test the total quantity of all of the substances that have
migrated from the test specimen of a food contact material or article to a food
simulant is determined gravimetrically using the procedures outlined in European
Commission Directive 2002/72/EC [22], as amended, and the detailed methods in
the EN 1186 series of standards.
In a specific migration test the quantity of an individual substance (monomer,
additive, etc.) or group of substances is determined in a food simulant following the
exposure of a food contact specimen to the food simulant for the prescribed period of
time at the prescribed temperature using an appropriate analytical method.
Both of these migration tests consist of two parts - the exposure of the test specimen
to the food simulant (or food for specific migration) and the analytical determination
part. The performance characteristics of both a specific migration test and an overall
migration test are a combination of factors from the migration part and the
determination part of the test. The outcome of the exposure part (and therefore the
test result) is furthermore dependent on the material tested e.g. degree of
homogeneity and interaction with the food or food simulant and the test conditions
applied.
The performance characteristics for the determination of substances in food (when
the material is already in contact with the food) and the determination of residues of
monomers in food contact materials are just those of the determination part
(analytical method).
The methods described allow compliance with the following legislative limits (LL) to
be demonstrated:
• QM= Maximum permitted quantity of the ‘residual’ substance in the material or
article;
• QM(T)= Maximum permitted quantity of the ‘residual’ substance in the
material or article expressed as total of moiety or substance(s) indicated.
• QMA= Maximum permitted quantity of the ‘residual’ substance in the finished
material or article expressed as mg per 6 dm2 of the surface in contact with
foodstuffs.
15
• QMA(T) = Maximum permitted quantity of the ‘residual’ substance in the
material or article expressed as mg of total of moiety or substance(s) indicated
per 6 dm² of the surface in contact with foodstuffs.
• SML = Specific migration limit in food or in food simulant, unless it is specified
otherwise.
• SML(T) = Specific migration limit in food or in food simulant expressed as total
of moiety or substance(s) indicated.
• OML= Overall migration limit is the maximum permitted total amount of
substances that may be released from a material or article into food or food
simulant.
In the following paragraphs the term Legislative Limit (LL) will be used for all the
above type of limits
According to the CEN TR 15330 tests performed with short times and high
temperatures will be expected to have a lower precision than tests performed with
16
long times and low temperatures, due to difficulties in controlling/replicating exposure
conditions, particularly temperature. Most high temperature tests are performed with
exposure times of less than 24 hours and special care is needed to control the
heating-up and cooling-down stages of the exposure of the specimen to the hot
simulant.
Another potential source of variability in the migration contact stage is the
conventional assumption that the fat simulant D - olive oil, and its substitute solutions
sunflower oil and the synthetic triglyceride HB307 - are equivalent. This is unlikely to
be strictly true since these oils differ both in their chemical composition (e.g. chain-
length) and physical composition (e.g. viscosity). It can be expected that a laboratory
using one of the choices of simulant D will, under some circumstances, obtain a
different test result, for SM or OM, than a laboratory using one of the other choices of
simulant D.
As a result of all mentioned above in the validation of test methods in the FCM field
the performance characteristics of one migration method should be determined at
strictly controlled migration conditions.
Accuracy and its components precision and trueness, as well as uncertainty as a
general term assessing accuracy are the performance characteristics which could be
assessed for the migration part.
Two main approaches to the estimation of uncertainty of the migration part could be
described: an empirical approach and a modelling approach.
The empirical approach uses repeated migration procedures and analysis, under
various conditions, to quantify the effects caused by variability of conditions to
quantify uncertainty (and usually some of its component parts). Uncertainty of
measurements as a common expression of the most important performance
characteristics can be considered to be generated by broad sources of error. These
four sources are the random errors and the systematic errors arising from the
methods of both the migration and the analysis. These errors have traditionally been
quantified as the migration precision, analytical precision, migration bias and the
analytical bias respectively, as shown in Table 2.
Precision Trueness
(Random effects) ( Systematic effects)
Migration
Migration variability Migration bias
(exposure) part
Determination
Analytical variability Analytical bias
(Analytical) part
The modelling approach uses a predefined model that identifies each of the
component parts of the uncertainty, making estimates of each component, and sums
them in order to make an overall estimate. This can be done using specific
softwares.
The contribution of the migration test to the total uncertainty could be small but most
probably it represents the dominant effect. Therefore to reduce the total uncertainty
of the method the exposure phase should be well controlled in order to achieve
fitness for purpose.
Such Guidelines can never be fully comprehensive, and although there are details of
some of the statistical techniques employed and references to more detailed advice,
17
there will often be a need for expert judgment for more complex situations.
If errors from these four sources (migration variability, migration bias, analytical
variability and analytical bias) could be quantified, separately or in combinations, it is
possible to estimate the uncertainty of the measurements that these migration
methods could produce.
Methods for the estimation of three of the four errors are established. Analytical
precision and analytical bias can be estimated by one of the approaches described
in the paragraph (6.7), concerning analytical part of the method (Table 2).
Migration precision can be estimated by replication of the migration procedure.
s2 migration = s2 total - s2 analytical
where
s migration is the standard deviation of the migration part;
s analytical is standard deviation of the analytical part;
s total is the standard deviation of the whole procedure (migration + analytical part).
Precision Trueness
(Random effects) ( Systematic effects)
Hence the focus of interest in validation study of the migration part will be exclusively
the precision aspect. As there are no CRM for specific migration, it is not possible to
assess bias in migration validation. It is though possible to assess bias of the overall
migration, in this case some CRM are available.
18
5.2.1 Selectivity/specificity
Selectivity is the degree to which the method can quantify the target analyte in the
presence of other analytes, matrices, or other potentially interfering materials. This is
usually achieved by isolation of the analyte through selective solvent extraction,
chromatographic or other phase separations, or by application of analyte-specific
techniques such as biochemical reactions (enzymes, antibodies) or instrumentation
(nuclear magnetic resonance spectroscopy, infrared or mass spectrometry).
In the validation study, at least 3 “processed” blanks have to undergo the whole
procedure before being analysed. This allows any interfering substances derived
from the analytical process to be identified. For chromatographic methods then there
should be no peak(s) at the same retention time as the analyte of interest. In case
this is not possible a limit should be defined during the validation study which
guarantees that the interfering peak does not have an impact on the results.
The analysis of the processed blanks is the same as for the within-laboratory
reproducibility assessment (5.2.7.1.3) i.e. at least 1 blank should be processed for
each day that the method validation is performed.
Moreover, to prove that the method is selective enough, a check could be performed
by spiking any potential interference substances to verify that what the method is
measuring is the target substance 5.2.7.2.2.6
5.2.2 Ruggedness
19
small but deliberate variations in the method do not have any effect on the final
results, providing an indication of the method's or procedure's suitability and
reliability during normal use.
Robustness traditionally has not been considered as a validation parameter in the
strictest sense because usually it is investigated during method development, once
the method is at least partially optimised. Nevertheless, for the completeness of the
present study it could be useful to evaluate also this parameter, to provide better and
more specific information on the method use, in case a full validation of such a
method will be carried on.
In Table 4 the analytical scheme (according to Youden ruggedness trial) is reported:
the symbol + refers to the original parameters (A-G), while the symbol – refers to the
parameter that has been changed (a-g). The scheme foresees 8 repetitions of the
method, changing each time some of the seven parameters for which robustness is
being assessed.
If all seven parameters are changed, an additional repetition of the method
performed with the original parameters has to be carried out, to serve as a blank for
reference.
It is possible to change also less than 7 parameters. In this case 8 experiments have
in any case to be carried out, using the original parameters, and the experiment(s)
carried out with the original parameters will represent the “blank”.
TRIAL
+ - 1 2 3 4 5 6 7 8
A a + + + + - - - -
B b + + - - + + - -
C c + - + - + - + -
D d + + - - - - + +
E e + - + - - + - +
F f + - - + + - - +
G g + - - + - + + -
Worked example
Calibration curve: A total of 12 analyses had to be performed (5 concentrations+ blank each one analysed 2
times). If the robustness assessment in carried on in the same day of repeatability or reproducibility assessment
no new calibration curve has to be prepared.
=> 12 analyses
For each of the 8 experiments one concentration value will be calculated (y1, y2, y3,
…y8).
These values will be combined, according to the sign shown in Table 4 and divided
by half of the number of the experiments, to give a reference value for each
parameter (Ei):
EA=(y1+ y2+y3+y4- y5-y6-y7-y8)/4
20
EB=(y1+ y2-y3-y4+y5+y6-y7-y8) /4
EC=(y1- y2+y3-y4+y5-y6+y7-y8) /4
ED=(y1+ y2-y3-y4-y5-y6+y7+y8) /4
EE=(y1-y2+y3-y4-y5+y6-y7+y8) /4
EF=(y1+ y2+y3+y4+y5+y6+y7+y8) /4
EG=(y1+ y2+y3+y4+y5+y6+y7+y8) /4
Ei ⋅ n
t=
SD ⋅ 2
If the calculated t-value results are lower than the t critical value for ν=n − 1 degrees
of freedom (tcrit = 2.45, ν=6, 97.5% confidence level), then the method is robust with
respect to that change in the procedure, otherwise the method should be considered
sensitive for that parameter and in the procedure particular care should be taken to
ensure that the method is followed exactly.
The calibration range for any method in the field of FCM should be carefully chosen
with regards to the relevant legislative limits (LL) as follows:
21
The calibration range of the analytical method for fatty food simulants (simulant D)
must be extended beyond 2*SML according to the reduction factor applicable to the
foodstuff (up to 5 times) or the sample should be diluted prior to instrumental
analysis to fit within the calibration range.
1
if this is not possible the lowest point in the calibration should be at least equal to the LOQ,
providing that LOQ+ULOQ < SML.
5.2.3.1.3.1 Substances where the SML(T) is expressed as one equivalent analyte (e.g.
as acrylic acid, as tin etc)
In this case the analytical method targets one final analyte rather than different
species and therefore the requirements can be as those for a single substance SML
as described above.
22
The target calibration range should contain at least five points equally distributed in
the range from 0.2 * QM 1 up to 2 * QM. When it is necessary to expand the
calibration range, this should be done with the inclusion of extra points at levels
above 2 * QM.
1 if this is not possible the lowest point in the calibration should be at least equal to the LOQ,
providing that LOQ + ULOQ) < QM.
Usually the calibration function in the chosen working range is a priori supposed to
be linear, but this should be verified in the course of the validation study.
23
In the linear model the three main parameters (a, b, R2) are defined by the
equations:
y = a+bx + εI
Where:
a is the intercept;
b is the slope;
x is the concentration
y is the measurement response
εi is the difference between the measured value and the expected value
)
ε i = yi - y i
( )( )
2
⎡ ⎤
and
R =⎢
2 ∑ xi − x yi − y ⎥
⎢
⎣⎢ ( 2
) ( 2 ⎥
∑ xi − x ∑ yi − y ⎦⎥ )
where
In case the adequateness of the linear model cannot be proven, the non-linear
second order calibration function model should be applied. In this case the graphical
representation of the data is a parabola.
In the non-linear model the main parameters (a, b, c) are defined by the equation:
y=a+bx+cx2
Where:
x is the concentration
y is the area value
a is a constant term, it controls the height of the parabola, more specifically, it is
the point were the parabola crosses the y-axis. It is a constant term.
b is the linear coefficient
c is the quadratic coefficient, the declivity of the parabola as it crosses the y-
axis
The coefficients b and c together control the axis of symmetry of the parabola (also
the x-coordinate of the vertex).
24
The data from the 6 calibration curves derived from the 2 repetitions on the 3
different days in the reproducibility assessment can be used for the assessment of
the linearity model (see Figure 1 – flow chart of full validation scheme and worked
example from paragraph 5.2.7.1.3)
• the residue calculated for the lowest level of calibration curve (at LOQ level) <
20%;
• the residues calculated for all other levels of calibration < 15%, expressed with
the following Figure 5
residue, %
25
5.2.3.4 Assessment of the adequateness of the linearity model by statistical
tests
To verify the linearity of the model a number of different statistical tests can be used.
The most commonly used are described in the following paragraphs.
The test for adequateness of the linearity model allows the validity of the regression
model and the chosen working range to be verified. The ANOVA lack of fit model is
based on the comparison of the tabulated F of Fisher values with the observed F of
Fischer calculated on the basis on the experimental results, and on the sums of
squares.
To verify the linearity, the sum of the squares of the difference between any
measured value y ij and the general mean y is calculated from the sum of squares
of the linear regression and the sum of the squares due to the error of the model
(and therefore a non linearity) , as follows:
∑∑ ( y
i j
ij − y ) 2 = ∑∑ ( yˆ ij − y ) 2 + ∑∑ ( y i − yˆ ij ) 2 + ∑∑ ( y ij − y i ) 2
i j i j i j
1≤ i ≤ p
Where:
)
yij is the response predicted by the model;
SSD total = SSD cal.function + SSD residual = SSD cal.function + SSD lack of fit+ SSD pure
error
where:
SSD total = ∑∑ ( y
i j
ij − y )2
SSD residual = ∑∑ ( y
i j
ij − yˆ i ) 2 = ∑∑ (ε ij ) 2
i j
26
SSD lack of fit represents the sum of the squares due to the error of the model (non-
linearity)
SSD residual represents the sum of squares due to the distance between the average
calculated in the sample considered representative of the population (working range,
calibration curve range) and the real average value that should be calculated in the
total population (considering an infinite number of calibration points)
The observed F values relative to the calibration function (Fcal.function) and to the lack
of fit (Flack of fit) are calculated on the basis of the ratio between the variance due to
the linearity and that due to non linearity, and the variance due to the residue, as
follows:
s calfunctio n l ( y ) (y)
2 2
s lackoffit
Fcalfunctio n = and Flackoffit =
s 2pureerror ( y ) s 2pureerror ( y )
Where:
SSDcal. function ( y )
. function ( y ) =
2
s cal
1
SSDlackoffit ( y )
.of . fit ( y ) =
2
s lack
p−2
SSD pure.error ( y )
s 2pure.error ( y ) =
p(n − 1)
Where:
n is the number of replicates, which means 1 for each day of reproducibility
study, for a total of 3. Each one of these 3 values corresponds to a mean of 2 values.
p is the number of calibration levels (at least 5 plus 1 blank)
(p-2) represents the number of degrees of freedom relative to the error of the
model (non-linearity)
p(n-1) represents number of degrees of freedom relative the residual.
s2 is the variance
For the calculation of degrees of freedom associated with any of the expressions, the
following formula can be used:
Total = 1+ Error of the model (non-linearity) + Residual
In Table 5, the representative values that have to be calculated for the evaluation of
the linearity are reported.
27
Variation Sum of Degree of Variance F of Fisher F1 Fisher 5% F2 Fisher 1%
source squares freedom (variance ratio)
. function ( y )
SSDcal . function ( y ) 2
s cal
Calibration Ftab (1, (p*(n-1), Ftab (1, , (p*(n-1),
SSD cal.function 1 Fcal . function =
function 1 s 2
pure.error (y) 95%) 99%)
.of . fit ( y )
SSDlack .of . fit ( y ) 2
s lack Ftab (p-2, , (p*(n- Ftab (p-2, , (p*(n-
Lack of fit SSD lack of fit Flack .of . fit = 2
p−2 s pure.error ( y )
(p-2)
1), 95%) 1), 99%)
SSD pure.error ( y )
Pure error SSD pure error (p*(n-1))
p(n − 1)
Total SSD total (p*n-1)
Table 5: ANOVA table to compare lack of fit and pure error under the assumption of
proportional residual standard deviation
The data from the calibration curves coming from the 2 repetitions in the 3 different
days (reproducibility assessment) can be used for the assessment of linearity,
following the ANOVA lack to fit approach.
The plot of the residuals ε versus the corresponding concentration (x) values or the
)
fitted value y i is a powerful tool to verify the two assumptions of linearity and of
constant residual standard deviation.
If these two assumptions are confirmed, then the figure should display a plot of
randomly distributed points centred on zero (as shown in Figure 6).
0.3
0.2
Residuals
0.1
0
0.00 5.00 10.00 15.00 20.00 25.00
-0.1
-0.2
X Variable 1
If the model is not linear, a systematic pattern between the residuals and the
concentration values can be seen (as shown in Figure 7).
28
X Variable 1 Residual Plot
0.1
0.05
Residuals
0
0.00 5.00 10.00 15.00 20.00 25.00
-0.05
-0.1
X Variable 1
29
Cl c = t(P,n-3) * sc
Where:
CI c confidence interval of the quadratic coefficient
t(P,f) Student's t value for level of statistical confidence P and degrees of freedom f
t(P,n-3)
sc is the standard deviation of the quadratic coefficient c, based on the different
calibration curves calculated.
If c ± CIc includes 0: the quadratic coefficient is not significant, i.e. there is no
significantly better fit using the quadratic regression, so the linear regression (of the
type: y = a + b*x) can be accepted.
The data from the calibration curves coming from the 2 repetitions in the 3 different
days (reproducibility assessment) can be used for the assessment of linearity,
following the present approach.
In addition to the linear regression model the data also have to be interpolated with a
quadratic regression, y = a + b*x + c*x2. The standard deviation (sc) and confidence
interval (CI) of the 6 values of the quadratic coefficient c obtained have to be
calculated and the statistical test described above has to be applied.
The test according to Mandel [10] is based on the comparison of a linear regression
and a quadratic regression to interpolate the data coming from the analysis of the
calibration curve.
In this case, a test is performed to determine whether the quadratic function
represents a significantly better model than the linear function, by comparing the
residual variances of both regressions.
For this purpose a value (Fcalc) the effects of the residual variance of the linear
regression (sy12) and the residual variance of the quadratic regression (sy22) are
combined and compared with the tabulated F of Fisher (for a level of statistical
confidence P (generally 95%) and degrees of freedom 1 and n-3).
Thus:
Where:
30
n is the number of the calibration curves used for this linearity verification
sy12 is the residual variance of the linear regression
sy22 is the residual variance of the quadratic regression
The data from the calibration curves coming from the 2 repetitions in the 3 different
days (reproducibility assessment) can be used for the assessment of linearity,
following this approach.
Calculations for residual variance are performed as described in 5.2.3.4.1.
The homogeneity of variances over the whole range (homoscedasticity, the variance
of the lowest level is comparable with the variance of the highest level) is a
prerequisite for a non-weighted linear regression [11]. Thus, if the model has
demonstrated that the model is linear, but the responses are not normally distributed
or the variances within the calibration range are not homogeneous and the working
range is over more than two orders of magnitude, there is the need to decide
whether to apply the weighting factor to the linear regression.
The homogeneity can be verified by calculating the standard deviation of repeated
measurements (n = 6 to 10) at the lowest and at the highest levels of the required
linear range. There are two possibilities:
The standard deviation is not significantly different: weighing is not necessary.
The standard deviation is significantly different: weighting is necessary and the
weighting factor should be selected.
To evaluate this, the statistical F-Test is applied and the F value relative to the
standard deviations of the lowest and the highest calibration levels have to be
calculated and compared with the tabulated F of Fisher for probability P (generally
95%) and degrees of freedom (n-1, n-1). To calculate Fcalc the following formula
should be used:
Fcalc= s2low/s2high
s2 are the variances of the replicates at the low and the high calibration levels
If Fcalc < F: The standard deviation is not significantly different: weighting is not
necessary.
If Fcalc > F: The standard deviation is significantly different: weighting is necessary
and the weighing factor should be selected.
If the standard deviation is significantly different, the interval of the calibration curve
could be reduced, or split into narrower calibration ranges (for example for calibration
range of 0 and 1000 could be split into three different ranges of 0-10, 10-100 and
31
100-1000). Otherwise if weighting is shown to be necessary, as the interval cannot
be reduced or split, the concentration values should be calculated by using different
weighing factors.
The most commonly used factors are:
1/x;
1/x2;
1/y or
1/y2
Where:
x is the response of the instrument
y is the concentration calculated based on the calibration curve
The results should be calculated as % recovery (calculated value / real value
*100%). As well as the % recovery the relative error (the calculated difference
between the % recovery and 100%) should be determined for all the levels of the
calibration curve and all weighting factors. Then the sum of all of the absolute
relative errors at all levels of the calibration curve for each weighing factor should be
calculated.
The calibration curve calculated with the weighting factor that provides the smallest
summarised relative error should be used.
The data from the calibration curves coming from the 2 repetitions in the 3 different
days (reproducibility assessment) can be used for the decision if weighting has to be
applied to the calibration curves.
The working concentration range is the range in which the method is validated and
which gives an acceptable trueness and precision. It should be distinguished from
the calibration range in which the regression model for calibration (most frequently
linear one) is established and verified.
The lowest limit of the working range is the lowest limit of the calibration range.
However the upper limit of the working range could be not only the highest point in
the calibration curve, for which the regression model is validated, but a higher
concentration at which acceptable trueness can be proven (e.g. by comparing with t-
test the results of a diluted upper limit concentration with a concentration inside the
calibration range).
32
acceptable certainty.
When the level of interest or the legislative limit (LL) is much higher than the
detection limit, the determination of the detection limit is of no value. The limit of
detection does not need to be calculated (and thus reported in the validation report)
when:
• the lowest concentration point of the calibration curve should be 10 times
lower than the LL, if possible or at least 5 times lower than the LL,
corresponding to LOQ level;
• the signal to noise ratio for concentration level corresponding to the LL is
higher than 30:1.
The calculation of detection limit is very important when the validation is performed
for substances that have to be “non detectable” with a defined detection limit, as the
capability of the method to reach the target detection limit should be proven.
In this case a blank sample has to be spiked with the concentration considered as
limit of detection and the verification that it can be clearly distinguished from the
blank has to be performed. This means that 3 repeated analyses have to be
performed and if the mean response of the spiked samples (at the limit of detection
level) is greater than 3 times the maximum blank value the limit of detection is
verified.
It is important to distinguish between the instrument detection limit (IDL) and the
method detection limit (MDL). The IDL is an instrument parameter and can be
obtained from measurements of pure analyte, in contrast to the MDL (LOD) which is
based on measurements of blank real sample or a low-level spiked sample that has
been taken through all the steps of the method.
The limit of detection can be calculated in several ways and the most used are
reported in the next paragraphs.
x BL is the mean concentration calculated from the area of a noise peak for at
least 6 analyses of a blank sample. In case the noise peak in the blank sample is not
measurable, the peak relative to the lowest concentration of the calibration curve can
be used, provided its value is comparable to LOD. In this case the LOD will be
simply 3 times the standard deviation of the lowest concentration level, as x BL is
considered as zero.
sBL is the standard deviation of the analysis.
33
The blank samples used to calculate the LOD are the same as those prepared in the
reproducibility assessment. No additional analyses have to be performed.
The LOD should be confirmed by spiking a blank sample at this level to see if it is
really visible with S/N of at least 3.
This approach is followed for analytical methods producing a baseline response. For
this reason it is widely applied for chromatographic methods. The signal to noise
ratio (S/N) is determined by comparison between the signals obtained from samples
containing known low concentrations of the analyte and the signal obtained from a
blank sample. The lowest analyte concentration level that can be determined with
acceptable reliability is calculated. The limit of detection is defined as the
concentration of the analyte at a signal/noise ration S/N=3 after blank correction
If calibrations are used to quantify the analyte, the intercept can be regarded as an
extrapolation of a blank determination.
The detection limit can then be defined as 3 times the standard deviation of the
intercept (the b value in the equation y=ax+b), according to the formula:
LOD = 3.sb
Where:
sb is the standard deviation of the b value of the intercept
The LOD values obtained with the formula above are then converted to
concentration using the calibration curve.
The calibration curves used to calculate the LOD in this way are the same as those
prepared in the reproducibility assessment. No additional analyses have to be
performed.
This approach is also based on the regression analyses of the calibration curves, so
only data from the calibration curves are used.
These calculations are performed by the instrument software (Chemstation etc.),
LIMS or through others specific software (DINTEST, Valistat, Effichem, MVA3, etc.).
In DIN 32645 [35] it is remarked, that the calculation of LOD/LOQ with the calibration
function should based on calibration points close to the LOD/LOQ. If it is done from
10 times the LOD or using the calibration from working range, it results in false
findings.
34
5.2.6 Limit of quantification
The limit of quantification represents the lowest concentration of analyte that can be
determined with an acceptable level of uncertainty e.g. the lowest reportable result. It
should be established by using an appropriate reference material or sample. It
should be not determined by extrapolation, although by conventions it is usually
taken as a fixed multiple (2-5) of the detection limit. This approach gives an
approximate value; its variation with the type of sample should be taken in
consideration, so a verification of the calculated LOQ is required.
For the present method validation scheme the limit of quantification xLOQ should be
calculated as:
xLOQ = 2 xLOD
5.2.7 Accuracy
Accuracy expresses the quality of the method. Method validation seeks to quantify
the likely accuracy of results by assessing both systematic and random effects on
the results. Accuracy is, therefore, normally studied as two components - trueness
and precision.
Precision is a measure of how close results are to one another and is usually
expressed by measures such as standard deviation (SD) or as relative standard
deviation (RSD) sometimes called the coefficient of variation (CV), which describe
the spread of results. Precision is generally dependent on analyte concentration, and
so should be determined at a number of concentrations and if relevant, the
relationship between precision and analyte concentration should be established. In
that case relative standard deviation is more useful. Precision is normally determined
for specific circumstances, which in practice can be very varied. The two common
precision measures are repeatability and reproducibility.
Trueness of a method is an expression of how close the mean of a set of results
(produced by the method) is to the true value. Trueness is usually expressed in
terms of bias, the difference between the true value (calculated on the basis of a
certified reference material or a reference material) and the mean of the results.
When certified reference materials or reference materials are not available, trueness
can be expressed in terms of recovery, after spiking.
35
Uncertainty of measurement is another common expression of accuracy. It is
calculated combining bias and its relative standard deviation (calculated for the
assessment of trueness) and the standard deviation of the reproducibility
assessment.
When certified reference materials or reference materials are not available no real
bias can be assessed. Further for single laboratory validation only a within laboratory
reproducibility can be determined. Thus, in these cases uncertainty cannot be
calculated based on bias and reproducibility, and it has to be estimated through the
combination of all possible sources of uncertainty associated with the method
(combined uncertainty).
Figure 8 represents a graphical representation of the accuracy components and on
how to assess them.
Accuracy
Trueness Precision
Repeatability Reproducibility
Repeatability Reproducibility
36
5.2.7.1 Precision
5.2.7.1.1 Definitions
Repeatability means precision under repeatability conditions (r), which means the
closeness of agreement between mutually independent test results obtained under
repeatability conditions, i.e. using the same method on identical test material in the
same laboratory by the same operator using the same equipment within short
intervals of time, often called for that short-term repeatability.
The repeatability limit value obtained in the validation study is intended as a
reference for the general use of the method: each laboratory using the method
during their daily work can use the established repeatability limit to verify the
appropriateness of their analyses.
Reproducibility means precision under reproducibility conditions (R), which means
the distribution of measurement results obtained under reproducibility conditions. It
may be measured by means of collaborative studies (identical test materials,
identical methods, different persons, different instruments and different laboratories).
It is expressed as standard deviation and it represents the overall precision of the
method.
For a single laboratory validation (SLV) scheme, real reproducibility cannot be
assessed, but only within-laboratory reproducibility (intermediate precision) is
determined. It is expressed as within-laboratory standard deviation or CV (%) of
different batches of samples analysed in the single laboratory by different operators,
using different instruments over long period of time, varying as much as possible the
factors influencing the analytical result and representing the real fluctuation in
performing a method in the laboratory. It represents the precision of the method but
only for the laboratory that performed the validation.
37
performed.
For spiked samples, at least 6 replicate analyses per sample spiked at a minimum of
3 concentration levels within the working range (e.g. 0.2 LL, 1*LL and 2*LL) have to
be performed under the same conditions, by the same person and with the same
equipment, representing minimum variability (repeatability conditions).
If a CRM or RM is available at least 6 replicate analyses for each available
concentration level have to be performed under the same conditions as for spiked
samples. The number of the concentration levels in this case depends on the
number of available CRM or RM concentrations (normally only one). Thus, the
method can be validated only in a range very close to the available concentrations.
In case a broader range of concentrations is necessary, additional studies on spiked
samples have to be performed.
Samples: A total of 24 analyses have to be performed (6 times 3 concentrations + blank each one) by
one operator.
=> 24 analyses in day 1
38
5.2.7.1.2.3 Processing of experimental data
After the evaluation of the consistency of the data sets, for each analysed
concentration level (series of n data obtained), the following calculations had to be
performed:
• Mean value
• Estimation of the standard deviation (sr), and/or
• Coefficient of variation (CVr) or repeatability relative standard deviation RSDr.
r = t * 2 * sr
where t = 2.45 for 6 replicates for s calculation and n=2 (for difference between two
results).
In Table 2, in Annex 1 the student t-values with the relative effective degrees of
freedom are reported.
39
repeated analyses for each available concentration level have to be performed under
reproducibility conditions. The number of the concentration levels in this case
depends on the number of available CRM, RM or QC samples concentrations
(normally only one). Thus, the method can be validated only in a range very close to
the available concentrations. In case a broader range of concentrations is necessary,
additional studies on spiked samples have to be performed.
For spiked samples, the 3 concentration levels should be: lowest calibration point as
defined according to paragraph 5.2.3.1, LL, and 2-5*LL. At least 6 replicate analyses
should be performed for each of the 3 concentration levels have to be performed
under the reproducibility conditions.
Samples: In addition to the samples analysed in the assessment of the mehod repeatability another two
sets of analyses have to be carried out (not in sequence but on another 2 different days). The same
samples described in the repeatability assessment should be prepared on the two additional days.
Calibration curve day 2 and day 3: Another two sets of calibration samples should be prepared (not in
sequence but on another 2 different days), for a total of 24 new analyses (5 concentrations + blank for 2
batches of samples -1 set was already analysed to assess repeatability) per day, by two different operators.
40
• The compatibility of the various means via the single-factor variance analysis
(ANOVA), has to be verified for a significance level α equal to 0.05, as
indicated in [18] and UNICHIM Manual No. 179/2, sections 9.1 and 9.2 [6].
If there are incompatibilities between the mean values, a Scheffè test has to be
performed for a significance level α equal to 0.05, as indicated in [19] and the
UNICHIM Manual No. 179/2, section 10 [6] and to eliminate data series for which the
mean values are significantly different from the others.
ANOVA has to be applied with the hypothesis (H0) that the reproducibility conditions
have no influence on the final results.
In ANOVA F is the statistic value that represents the ratio of two variance estimates:
If H0 is the true F value calculated on the basis of the experimental results then it has
to be lower than the F critical value, calculated by ANOVA.
If F calculated, is higher than F critical, then the reproducibility conditions have an
influence on the performance of the method and thus the method cannot be
considered reproducible.
If the null hypothesis (H0) is rejected by ANOVA this implies that at least one of the
means isn't the same as the other means, so the reason should be found. For this
the Scheffè test is used, testing if pairs of means are different one from the other.
sWR = s r2 + sbb
2
where:
sr is within batch standard deviation representing repeatability s :
p
∑s 2
r ,i
sr = i =1
p
where p is the number of batches,
sbb is the between batch standard deviation
41
− sr2
sbb = s( xi ) − 2
n
where n is the number of replicates within a batch.
− sr2
In case when s ( xi ) 2 − < 0, then
n
sWR = sr
WRLAB = t * 2 * sWR
where
t is the Student’s t-value and its values are: t=2.11 for 18 replicate for S calculation
and n=2 (for difference between two results).
In case of problems with the analyses resulting in only 17 replicates (i.e. one of the
replicates was excluded as an outlier) then t=2.12 should be used. All t-values
should correspond to a 97.5% probability value.
The “within laboratory reproducibility limit” WR enables the analyst to decide whether
the difference between two analyses of sample (Δx), determined under
reproducibility conditions within the same laboratory, is significant.
Once the WR value is fixed based on the results of the validation study, then for all
the subsequent applications of the method it should always be higher than the
difference between duplicate analyses. This will allow the analysts that will use the
validated method to verify the consistency of their results.
42
The equation is:
RSD (%) predicted = 2(1-0.5logC) = 2*C(-0.5log2) = 2*C-0.15
Where:
C is the mass fraction expressed as a power (exponent) of 10 (e.g. 1 mg/g = 10-3).
A more contemporary model based on results from PT schemes has shown that the
relationship is best represented if three equations are used to cover from high to low
concentrations.
For concentration lower than 1.2 x 10-7 and higher then 0.138 Thompson [17]
introduced a correction in the Horwitz formula:
s (%) predicted = 0.22*C, if C < 1.2x10-7
s (%) predicted = 0.01*C0.5 if C > 0.138
For analyses carried out under short term repeatability conditions, the repeatability
standard deviation sr would typically be between one half and two thirds of the above
values.
5.2.7.2 Trueness
5.2.7.2.1 Definitions
43
Reference value Uncertainty of the reference value
XCRM (CRM) UREF = UCRM/2 if in the certificate the uncertainty of the reference value is
expanded uncertainty with coverage factor of 2
if not UREF = UCRM/ SQRT(3)
XRM (Reference material) UREF = URM/2 if in the certificate the uncertainty of the reference value is
expanded uncertainty with coverage factor of 2
if not UREF = URM/ SQRT(3)
X Interlab comparison UREF = U IL provided by the organisers of an ILC or
UREF = s interlab reproducibility/SQRT(p); p=number of laboratories
XRM ( Reference method) UREF = s RM from the n results obtained with the reference method/ SQRT(n)
XRL ( Reference Laboratory) UREF = SRL from the n results obtained by the reference laboratory/ SQRT(n)
Table 7: Different calculation of the uncertainty based on different reference value
type
When CRMs, reference methods or collaborative study or proficiency test results are
not available the results of the spiking/recovery experiments can be used (see
section 5.2.7.2.2.6 ).
bias − 0 bias
t calc = = ≤ t crit
1 u bias
s bias ⋅
n
Where:
bias is the difference between the mean calculated concentration and the certified
44
reference value,
n is the number of replicates (6 in this case)
sbias is the standard deviation of the six repeated analyses.
If tcalc > tcrit the hypothesis is rejected and the bias is significantly different than “0”. A
decision should be taken based on pre-defined acceptability criteria whether the bias
is acceptable or not.
Worked example
Samples: One set of 6 analyses per each available CRM has to be carried out (+ blank).
i.e. 7 analyses on day 1 by operator A or B
5.2.7.2.2.3 Methods for which organised collaborative studies and proficiency testing
results are available
The trueness of the analytical method is examined by participating in a proficiency
testing scheme for samples equivalent to the type of sample the method will be used
for. The documented trueness only applies to the concentration range and the
matrices which are included in the proficiency testing scheme.
If such testing does not exist, smaller comparisons with a few laboratories, (at least
4) after rejection of outliers, can in some cases provide valuable information.
45
A t value is calculated from the 2 mean values obtained from the analyses performed
with the 2 different methods as follows:
⎛ ⎞
⎜ xi − x ref ⎟
t calc = ⎝ ⎠ ≤t
1 1
s combined ⋅ +
n m
Where:
x is the mean value obtained with the tested method,
−
x ref is the mean value obtained with the reference method,
n is the number of replicate analyses performed for the reference method (6 in this
example)
m is the number of replicate analyses performed for the tested method (6 in this
example),
s is the combined standard deviation calculated according to the following formula:
1
s= * [(n − 1) * s x2 + (m − 1) s ref
2
]
m+n−2
If the calculated t-value is lower than t critical for P probability level (generally 95%),
and m+n-2 degrees of freedom, there is no significant difference between the two
sets of results and the tested method can be considered comparable to the
reference method and thus the trueness of the reference method can be used as
trueness of the tested method.
Worked example
Samples: 6 analyses for each of the 3 concentration levels have to be performed according to the
reference method (+ blank).
24 analyses on day 1 by operator A or B
Calibration curve: a new calibration curve has to be established for the reference method (5
concentration + blank, analyses in duplicate)
46
results may not be a proof that the method completely behaves like a fully validated
one, significant bias on any material remains a cause for investigation.
No new analyses have to be performed by the laboratory.
−
( x − x ) *100
Ri % = i 0
xi spike
Where:
−
x i is the mean concentration determined in the spiked sample;
x0 is the concentration of the analyte in the sample before spiking, normally blank
sample, or real sample for the method of standard addition;
x spike is the spiked concentration
Strictly, recovery studies as described here only assess bias due to effects operating
on the added analyte; the same effects do not necessarily apply to the same extent
to the native analyte, and additional effects may also apply to the native analyte.
Spiking/recovery studies are accordingly very strongly subject to the observation that
while good recovery is not a guarantee of trueness; poor recovery is certainly an
indication of lack of trueness.
No new analyses are required
The samples, calibration curves and relative analyses are the same as for
reproducibility assessment. No additional analyses have to be performed.
A significance test is used to determine whether the mean recovery for the working
range is significantly different from 1.0. The test statistic t is calculated using the
following equation:
47
_ _
1 − ( R/ 100) 1 − ( R/ 100)
t calc = = ≤ t crit
1 u Re c
s Re c ⋅
n
Where:
R = (∑ Ri ) / n is the mean recovery for each of the 3 concentration levels
s
u− = is the uncertainty of the mean recovery value
R n
s is the standard deviation of the mean recovery in the working range
This value of t (calculated) is compared with the 2-tailed critical value tcrit, for kp -1
degrees of freedom at 95% confidence (where kp is the number of results used to
estimate R ). If it is greater than or equal to the critical value tcrit then R is significantly
different from 1.
Moreover, in case the calculated t value is higher than the t critical valuea decision
should be made, based on pre-defined acceptability criteria, as to whether the
recovery is acceptable or not.
In cases when bias is significantly different from “0” and/or recovery is significantly
different from “1 or 100%”, then an assessment should be performed to establish
whether or not the bias/recovery are acceptable according to the fixed pre-defined
maximum acceptable criteria
For the field of FCM the following maximum acceptable criteria for recovery are laid
down based on convention:
48
maximum acceptable recovery in Table 8. The laboratory should decide whether to
report the results corrected for recovery or to include the low recovery in the
estimation of the uncertainty of the method. In any case it is fundamental to state
clearly if the results are corrected for the recovery or if they are not.
In cases when significant bias is accepted based on the above maximum acceptable
bias in Table 9, it should be taken into consideration when calculating the uncertainty
of the method and of the reported result.
5.2.8 Uncertainty
X= x ±U
where:
X= real value
x= measured value
U= expanded uncertainty
49
5.2.8.1 Horwitz approach
When none of the other approaches are possible, the following formula can be used
to estimate the uncertainty:
U (%) predicted = 2(1-0.5logC) = 2*C(-0.5log2) = 2*C-0.15
Where:
C is the mass fraction expressed as a power (exponent) of 10 (e.g. 1 mg/g = 10-3).
Where:
Uexp is the expanded uncertainty;
Uc is the combined uncertainty;
Srepeat is the standard deviation obtained from at least 6 samples analyses (those
from repeatability study) integrating all uncertainty sources due to the sample
treatment;
utype B is given by the combination of all possible sources of uncertainty which are not
∑u
2
included in srepeat: i
Combined relative uncertainty (uc) is the uncertainty deriving from the combination of
all associated uncertainties (u) of all the different stages of the method, such as
uncertainty on volumes (such as those related to pipettes, volumetric flasks, etc.),
weighing, measures and the uncertainty due to errors deriving from the calibration
curve. For chemical measures it can be calculated by a general formula:
u&c ( y ) = c = ⎜⎜ ⎟⎟ + ⎜ ⎟ +⎜ ⎟ + ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟
y ⎝ y ⎠ ⎝ c ⎠ ⎝ v ⎠ ⎝ p ⎠ ⎝ R ⎠ ⎝ M ⎠
Where:
∑ (x )
n 2
i −x
is the relative uncertainty due to repeatability, Srepeat
( )=
i =1
u yA s n −1
=
y x⋅ n x⋅ n
50
Where: s is the Standard deviation, an estimate of the dispersion
of the data (n) around the mean value x ,
x is the mean value,
xi is the measured values,
n is the number of analyses.
( )
u cB
=
s
. is the uncertainty due to errors on calibration curves.
cs cs ⋅ q
Where: cs
is the mean value of all standard concentration
values of for each point of the calibration curve
q is the number of samples for repeatability analyses.
∑s i
Where:
⎛ n ⎞
⎜ ∑ [y − yˆ i ] ⎟
2
⎜
i
⎛ ⎞⎟
⎜
( ) ⎟⎟
i =1
⎜
2
n−2 ⎜ 1 1 yj − y ⎟⎟
⎜ ⋅ + +
⎜m n ⎟⎟
( )
n
⎜ b
b 2 ∑ ci − cc
2
⎜ ⎜ ⎟⎟
⎝ i =1 ⎠
⎜ ⎟
si = ⎝ ⎠
Where:
yj is the value given by the instrument for each measure
of the repeatability assessment (peak area)
yi is the value given by the instrument for each measure
of all points of the calibration curve (peak area)
y is the mean value of all values given by the
instrument for each measure of all points of the
calibration curve (peak area)
ŷ i = a+bc is the value calculated with calibration curves
i
parameters for each point of calibration (peak area)
ci is the concentration of standards for each point of the
calibration curve
cc is the mean value of all standard concentration values
of for each point of the calibration curve
b is the angular coefficient
n is the number of points of the calibration curve
m is the number of analyses for each sample
(repetitions) for repeatability study.
( )
u vB
=
u v s + u v r + u v ΔT
2 2 2
is the uncertainty due to volumes, the used volume
v v
being v .
51
Where: uv s
u v s = v ⋅ 3 The first term is the uncertainty declared by
the supplier.
uS is the value taken from manufacturer certificate.
sv
u v r = v ⋅ nv = uncertainty due to repeatability in filling.
sV is the standard deviation due to variation in filling and
it is calculated performing a repeatability experiment
on nv fillings and weightings.
C e ⋅ v ⋅ ΔT
u v ΔT = v is the uncertainty due to the
difference in temperature between the moment when
the volumetric glassware was calibrated by the
supplier and the moment it is used. Ce is the
coefficient of volume expansion and T is the
difference in temperature between calibration of
glassware and analyses.
52
The general expression of uncertainty has to be adapted to the considered method,
eventually adding all necessary additional terms for the calculation of all components
of uncertainty or deleting the terms that are not necessary as already included in
other estimations.
The numbers of freedom degrees νeff can be calculated by:
ν eff =
[u& (y )]
c
4
⎡ (u& (x )) ⎤ ⎡ (u& (x )) ⎤
4 4
⎢ ⎥+⎢ ⎥
r m
⎢⎣ νr ⎥⎦ ⎢⎣ νc ⎥⎦
Where:
νr is the number of analyses performed to calculate repeatability
νc is the number of calibration curve points
( )
u& x r is the relative uncertainty on repeatability
u Rw + ubias
2 2
uRw
u Rw + ubias
2 2
uc =
ubias
53
DETERMINATION OF RANDOM ERRORS
Within-laboratory reproducibility uRw
Suitable CRM,
RM, quality uRw is the standard deviation of the reproducibility
control samples study (obtained with as much variation in the
available, or conditions as possible)
spiked samples
NO
∑bias
2
i
i =1
+ uref
2
n
NO
∑bias
experiments 2
i
i =1
+ uadd
2
54
5.3 Reporting of results
The reporting of the analytical results depends on how the limits or restrictions
specified in the legislation are expressed. Thus, the number of significant figures
used when expressing an analytical result has to be consistent with what is
expressed in the legislation.
In cases where legislation already provides clear guidance on the number of
significant figures to be specified (e.g. when maximum levels are expressed as 2.00,
15.0, or 0.040) then the analyst should report the result to the same number of
significant figures as indicated in the specification. In other cases, and certainly in
cases where it is appropriate for the precision of the result, the analyst should use
one significant figure more than indicated in the specification (assuming that the
analyst is using an appropriate method).
The analytical result has to be reported as x +/- U, where x is the analytical result
(the best estimate of the true value) and U is the expanded measurement
uncertainty. The result is reported with uncertainty of measurement when it is
relevant to the validity or application of the test results, when a customer’s instruction
so requires, or when the uncertainty affects compliance to a specification limit [21].
55
• An unambiguous conclusion concerning which analytical tasks the examined
method is suitable for.
• If the method is found to be unsuitable for some matrices or concentrations
this must be evident in the report.
56
6 “STANDARD LEVEL” VALIDATION SCHEME FOR
FCM METHODS
All the theoretical considerations and description of the different parameters are
already extensively presented in the previous chapter. Thus the present chapter will
concentrate mostly on the practical part of the analytical work that has to be
performed by the laboratory.
All information reported in paragraphs 5.3, 5.4, 5.5, 5.6 should also be applied also
for the standard FCM validation scheme.
6.1 Selectivity/specificity
6.2 Ruggedness
It can be performed according to 5.2.3.2 for full validation with some small changes
in the experimental plan.
In order to simplify the procedure, when possible, the methods are divided into two
main categories (for inorganic and organic analyses). Inorganic and organic
analyses present different requirements and are assessed in two different ways:
• Inorganic analyses: 3 concentration levels plus one blank have to be analysed
and the procedure has to be repeated 3 times (12 analyses in total).
• Organic analyses: 5 concentration levels plus one blank have to be analysed
and the procedure has to be repeated 3 times (18 analyses in total).
57
6.3.3 Assessment of the linearity model
For the present validation scheme the calibration function has to be assessed initially
by visual inspection of the plotted data generating a calibration curve and then by
statistical evaluation.
The correlation coefficient, r, is not a measure of linearity. It indicates the extent of
the relationship between the instrument response and analyte concentration.
For the assessment of the regression model for the calibration function it is important
to carry out first the visual inspection of the plot of residuals (difference between the
measured value and the expected value) versus the corresponding concentration (x)
values.
Using the data from the calibration curves of the reproducibility assessment,
estimation and examination of the residuals has to be performed and the
requirements in paragraph 5.2.3.3 should be met.
A statistical treatment of data is necessary. For this purpose ANOVA lack of fit or
Mandel’s test or another equivalent test has to be performed (see paragraph 5.2.3.4
for all details regarding the tests).
In some cases where the calibration range is very wide it is important to establish the
homogeneity of the variance of the method across the working range according to
the procedure described in 5.2.3.5.
As related to “homoscedasticity” a distinction between spectrometric and
chromatographic methods has to be made:
• for spectrometric methods the test for homogeneity of the variance
(“homoscedasticity”) is suggested only if necessary (in cases with a very large
working range);
• for chromatographic methods, as the working range is usually narrow,
calibration without checking for “homoscedasticity” is accepted and calibration
without weighting is suggested .
6.5 LOD
The calculation of LOD is not necessary, except for the cases when LL=ND
6.6 LOQ
58
A distinction between spectrometric and chromatographic methods has to be made:
• For chromatographic methods estimation of the LOQ should be based on the
lowest level of the calibration curve and should be equal to the concentration
at which the ratio between the signal to noise (S/N) is at least 6.
• For spectrometric methods estimation of LOQ should be based on the blank
and should be equal to the average response of the blank + at least 6 times
the standard deviation of the blank or the lowest point if the blank does not
give a signal.
It is necessary to perform a verification of the quality of the value estimated as LOQ
following paragraph 5.2.6.1.
All data necessary for the calculation of LOQ are those collected for the assessment
of precision (lowest point). Thus no new analyses have to be performed for the
calculation of LOQ.
6.7 Precision
59
Day 2: at least 3 replicates at 3 concentration levels (0.2*LL, 1*LL and 2*LL) and
calibration curves should be analysed.
Day 3: at least 3 replicates at 3 concentration levels (0.2*LL, 1*LL and 2*LL) and
calibration curves should be analysed.
Basic statistical evaluation of the data for verification of the presence of anomalous
data should be performed by applying Grubbs or Dixon’s tests as described in
5.2.7.1.2.2
Calculation of within-laboratory reproducibility as described in 5.2.7.1.3.3.
Acceptability criteria for precision, described in 5.2.7.1.4 should be met.
Samples: A total of 21 analyses had to be performed. First day - 1 concentration (LL) in 7 replicates, two other
concentration levels ( 0.2*LL and 2*LL) in 3 replicates; two other different days the 3 concentration levels ( 0.2*LL,
1.0*LL and 2.0*LL) in 3 replicates by different operator.
=> 13 analyses in day 1, 9 analyses in day 2, 9 analyses in day 3
3 calibration curves:
For inorganic analyses: A total of 12 analyses had to be performed (3 times 3 concentrations+ blank) in different
days by three different operators
For organic analyses: A total of 18-21 analyses had to be performed (3 times 5-6 concentrations+ blank) in
different days by different operators.
The scheme of analyses is the same as for the previous case, but the numbers vary according to the combination
chosen from table 6 at 3 different concentration levels.
6.8 Trueness
6.9 Uncertainty
60
7 “BASIC LEVEL” VALIDATION SCHEME FOR FCM
METHODS
All the theoretical considerations and description of the different parameters are
already extensively presented in the previous chapters 5 and 6. Thus the present
chapter will concentrate mostly on the practical part of the analytical work that has to
be performed by the laboratory.
All information reported in paragraphs 5.3, 5.4, 5.5 and 5.6 should also be applied
also to the basic validation scheme.
7.1 Selectivity/specificity
7.2 Ruggedness
For the assessment of the regression model for the calibration function it is important
to carry out first the visual inspection of the residual plot. There should be no
curvature in the residuals plot as described in paragraph 5.2.3.4.2 and as shown in
Figures 6 and 7;
61
No statistic treatment of data is necessary.
As related to the “homoscedasticity” distinction between spectrometric and
chromatographic methods has to be made:
• for spectrometric methods the test for homogeneity of the variance
(“homoscedasticity”) is suggested only if necessary (in cases with a very large
working range);
• for chromatographic methods as the working range is usually narrow.
calibration without checking for “homoscedasticity” is accepted and calibration
without weighting is suggested.
The working concentration range is the range in which the method is validated and
which gives an acceptable trueness and precision
As the basic level of method validation is focusing only on the LL, a narrow working
range around that LL is considered to be acceptable.
7.5 LOD
The calculation of LOD is not necessary except for the cases when LL=ND.
7.6 LOQ
7.7 Precision
62
ANOVA can be used to calculate both repeatability and reproducibility: repeatability
is represented by the within batch standard deviation and reproducibility is
represented by square root of the sum of the squares of within batch standard
deviation and between batches standard deviation.
Within-laboratory reproducibility standard deviation must be determined by at least 6
degrees of freedom.
For laboratories not familiar with ANOVA data treatment, the following scheme can
be used:
For repeatability: at least 5 replicated samples for legislative level in the working
range in food/food simulant, plus calibration curve.
For reproducibility: 3 replicates plus calibration curves in another 2 different days
(repeatability analyses can be considered to be day 1)
Basic statistical evaluation of the data for verification of the presence of anomalous
data should be performed by applying Grubbs or Dixon’s tests as described in
5.2.7.1.2.2.
Acceptability criteria for precision, described in 5.2.7.1.4 should be met.
Samples: A total of 11 analyses had to be performed (1 concentration - 5 times the first day, 3 times on day 2 and
3 times on day 3) by different operator.
=> 5 analyses on day 1, 3 analyses on day 2, 3 analyses on day 3
3 calibration curve:
For inorganic analyses: A total of 12 analyses had to be performed (3 times 3 concentrations+ blank) by different
operators in different days.
For organic analyses: A total of 18 analyses had to be performed (3 times 5 concentrations+ blank) by one
operator.
7.8 Trueness
63
7.9 Uncertainty
8 REFERENCES
4. ISO Standard 5725 1-6: Accuracy (trueness and precision) for measurement
methods and results.
10. DIN 38 402 (Teil 51): Deutsche Einheitsverfahren zur Wasser-, Abwasser
und Schlammuntersuchung. Allgemeine Angaben (Gruppe A) Kalibrierung
von Analysenverfahren, Auswertung von Analysenergebnissen und lineare
Kalibrierfunktionen für die Bestimmung von Verfahrenskenngrößen (A51)
(1986) 5.1.3 Linearitätstest.
12. ICH Q2B Validation of analytical procedures: text and methodology, Q2(R1),
November 2005.
64
13. CIPAC Guidelines for Collaborative Study Procedures for Assessment of
Performance of Analytical Methods, 1989.
20. ISO Standard 78-2: Chemistry – Layouts for Standards – Part 2: Methods of
Chemical Analysis (Second Edition, 1999).
23. 82/711/EEC: Council Directive of 18 October 1982 laying down the basic
rules necessary for testing migration of the constituents of plastic materials
and articles intended to come into contact with foodstuffs.
24. 85/572/EEC: Council Directive of 19 December 1985 laying down the list of
simulants to be used for testing migration of constituents from plastic
materials and articles intended to come into contact with foodstuffs.
65
28. ISO/IEC Guide 98:1995 - Guide to the expression of uncertainty in
measurement (GUM).
34. Commission Directive 97/48/EC of 29 July 1997 amending for the second
time Council Directive 82/711/EEC laying down the basic rules necessary for
testing migration of the constituents of plastic materials and articles intended
to come into contact with foodstuffs.
35. DIN 32645:2008-11 - Chemical analysis - Decision limit, detection limit and
determination limit under repeatability conditions - Terms, methods,
evaluation.
66
ANNEX 1
67
Effective 75% 80% 85% 90% 95% 97.5% 99% 99.5% 99.75% 99.9% 99.95%
degrees of
freedom ( eff)
1 1.000 1.376 1.963 3.078 6.314 12.71 31.82 63.66 127.3 318.3 636.6
2 0.816 1.061 1.386 1.886 2.920 4.303 6.965 9.925 14.09 22.33 31.60
3 0.765 0.978 1.250 1.638 2.353 3.182 4.541 5.841 7.453 10.21 12.92
4 0.741 0.941 1.190 1.533 2.132 2.776 3.747 4.604 5.598 7.173 8.610
5 0.727 0.920 1.156 1.476 2.015 2.571 3.365 4.032 4.773 5.893 6.869
6 0.718 0.906 1.134 1.440 1.943 2.447 3.143 3.707 4.317 5.208 5.959
7 0.711 0.896 1.119 1.415 1.895 2.365 2.998 3.499 4.029 4.785 5.408
8 0.706 0.889 1.108 1.397 1.860 2.306 2.896 3.355 3.833 4.501 5.041
9 0.703 0.883 1.100 1.383 1.833 2.262 2.821 3.250 3.690 4.297 4.781
10 0.700 0.879 1.093 1.372 1.812 2.228 2.764 3.169 3.581 4.144 4.587
11 0.697 0.876 1.088 1.363 1.796 2.201 2.718 3.106 3.497 4.025 4.437
12 0.695 0.873 1.083 1.356 1.782 2.179 2.681 3.055 3.428 3.930 4.318
13 0.694 0.870 1.079 1.350 1.771 2.160 2.650 3.012 3.372 3.852 4.221
14 0.692 0.868 1.076 1.345 1.761 2.145 2.624 2.977 3.326 3.787 4.140
15 0.691 0.866 1.074 1.341 1.753 2.131 2.602 2.947 3.286 3.733 4.073
16 0.690 0.865 1.071 1.337 1.746 2.120 2.583 2.921 3.252 3.686 4.015
17 0.689 0.863 1.069 1.333 1.740 2.110 2.567 2.898 3.222 3.646 3.965
18 0.688 0.862 1.067 1.330 1.734 2.101 2.552 2.878 3.197 3.610 3.922
19 0.688 0.861 1.066 1.328 1.729 2.093 2.539 2.861 3.174 3.579 3.883
20 0.687 0.860 1.064 1.325 1.725 2.086 2.528 2.845 3.153 3.552 3.850
21 0.686 0.859 1.063 1.323 1.721 2.080 2.518 2.831 3.135 3.527 3.819
22 0.686 0.858 1.061 1.321 1.717 2.074 2.508 2.819 3.119 3.505 3.792
23 0.685 0.858 1.060 1.319 1.714 2.069 2.500 2.807 3.104 3.485 3.767
24 0.685 0.857 1.059 1.318 1.711 2.064 2.492 2.797 3.091 3.467 3.745
25 0.684 0.856 1.058 1.316 1.708 2.060 2.485 2.787 3.078 3.450 3.725
26 0.684 0.856 1.058 1.315 1.706 2.056 2.479 2.779 3.067 3.435 3.707
27 0.684 0.855 1.057 1.314 1.703 2.052 2.473 2.771 3.057 3.421 3.690
28 0.683 0.855 1.056 1.313 1.701 2.048 2.467 2.763 3.047 3.408 3.674
29 0.683 0.854 1.055 1.311 1.699 2.045 2.462 2.756 3.038 3.396 3.659
30 0.683 0.854 1.055 1.310 1.697 2.042 2.457 2.750 3.030 3.385 3.646
40 0.681 0.851 1.050 1.303 1.684 2.021 2.423 2.704 2.971 3.307 3.551
50 0.679 0.849 1.047 1.299 1.676 2.009 2.403 2.678 2.937 3.261 3.496
60 0.679 0.848 1.045 1.296 1.671 2.000 2.390 2.660 2.915 3.232 3.460
80 0.678 0.846 1.043 1.292 1.664 1.990 2.374 2.639 2.887 3.195 3.416
100 0.677 0.845 1.042 1.290 1.660 1.984 2.364 2.626 2.871 3.174 3.390
120 0.677 0.845 1.041 1.289 1.658 1.980 2.358 2.617 2.860 3.160 3.373
0.674 0.842 1.036 1.282 1.645 1.960 2.326 2.576 2.807 3.090 3.291
Table 2: Factor k values (k=tp, where tp is the Student’s t), for different confidence
levels
68
ANNEX 2
n
Relative intensity EI-GC-MS CI-GC-MS, GC-MS ,
n
(% of base peak) (relative) LC-MS, LC-MS (relative)
Table 1. Maximum permitted tolerance for relative ion intensities using a range of
mass spectrometric techniques.
When full-scan UV/VIS detection id used, the maximum absorption in the spectrum
of the analyte shall be at the same wavelengths as those of the calibration standard
69
within a margin determined by the resolution of the detection system. For diode array
detection, this is typically within ± 2 nm. The spectrum of the analyte above 220 nm
shall, for those parts of the two spectra with a relative absorbance ≥ 10 %, not be
visibly different from the spectrum of the calibration standard. This criterion is met
when firstly the same maxima are present and secondly when the difference in
absorbance between the two spectra is at no point observed greater than 10%. In
the case computer-aided library searching and matching are used, the comparison of
the spectral data in the test samples to that of the calibration solution has to exceed
a critical match factor. This factor shall be determined during the validation process
for every analyte on the basis of spectra for which the criteria described above are
fulfilled. Variability in the spectra caused by the sample matrix and the detector
performance shall be checked.
If absorbance ratios are used the absorbance ratios for the samples should agree to
within ±10% of ratios for the standards (at concentrations as close as possible to
each other).
LC with UV/VIS detection (single wavelength) is not suitable on its own for use as a
confirmatory method.
70
Annex 3 Quality control and updates
These guidelines will be reviewed at regular intervals. The CRL-NRL Network for
FCM is charged with this review and the undertaking of any necessary updates.
Updates of text
Versions Dates Chapters Action
st
1 edition 2009 09.09.2009 all creation
71
Annex 4. NRL Network
72
European Commission
EUR 24105 EN – Joint Research Centre – Institute for Health and Consumer Protection
Title: Guidelines for performance criteria and validation procedures of analytical methods
used in controls of food contact materials
Abstract
Test methods for materials and articles in contact with foodstuffs are required to determine
the concentration of residues of monomers in the materials themselves or to determine the
concentration of individual or groups of substances in food (or food simulants) which have
migrated from the food contact materials.
The Community Reference Laboratory and National Reference Laboratories for food contact
materials (FCM) prepared the present Guidelines to illustrate the required performance
criteria for the analytical methods applied in the laboratories for FCM and provide procedures
for method validation in order to estimate their performance characteristics. The scope of
these guidelines is to provide rules for the performance of the analytical methods to be used
in the verification of compliance with the migration limits defined in Directive 2002/72/EC,as
amended, and in accordance with Directive 82/711/EEC, as amended, and others defined in
the European legislation, in order to ensure the quality and comparability of the analytical
results.
The document presents 4 approaches, according to the different purpose of performance
assessment.
These guidelines are intended as a dynamic document and they will evolve and expand into
further editions. This is the first edition. These guidelines have been endorsed by the
European Union official Network of National Reference Laboratories and approved by the EU
Commission competent service DG SANCO.
This work also highlights an important deliverable for the Network of NRLs. In particular, the
members of the task force “Method Performance” that have dedicated time and effort to
provide input into the development of these guidelines. They are gratefully acknowledged
here for their contribution: NRL-BE (Fabien Bolle, Tina n’Goy), NRL-DE (Oliver Kappenstein),
NRL-DK (Jens Petersen), NRL-ES (Juana Bustos), NRL-FR1 (Patrick Sauvegrain), NRL-EL
(Timokleia Togkalidou), NRL-IT (Maria Rosaria Milana), NRL-NL (Durk Schakel, Dita
Kalsbeek-van Wijk), NRL-PL (Kazimiera Cwiek-Ludwicka), NRL-SI (Viviana Golja), NRL-UK
(Emma Bradley). Special thanks are extended to Emma Bradley for her contribution to the
editing of the document.
The Publications Office has a worldwide network of sales agents. You can obtain their
contact details by sending a fax to (352) 29 29-42758.
73
LB-NA-24105-EN-C
The mission of the JRC is to provide customer-driven scientific and technical support
for the conception, development, implementation and monitoring of EU policies. As a
service of the European Commission, the JRC functions as a reference centre of
science and technology for the Union. Close to the policy-making process, it serves
the common interest of the Member States, while being independent of special
interests, whether private or national.
74
IOSR Journal Of Pharmacy
(e)-ISSN: 2250-3013, (p)-ISSN: 2319-4219
www.iosrphr.org Volume 5, Issue 10 (October 2015), PP. 07-19
Abstract: When analytical method is utilized to generate results about the characteristics of drug related
samples it is essential that the results are trustworthy. They may be utilized as the basis for decisions relating to
administering the drug to patients. Analytical method validation required during drug development and
manufacturing and these analytical methods are fit for their intended purpose. To comply with the requirements of
GMP pharmaceutical industries should have an overall validation policy which documents how validation will be
performed. The purpose of this validation is to show that processes involved in the development and manufacture
of drug, production and analytical testing can be performed in an effective and reproducible manner. This review
article provides guidance on how to perform validation characteristics for the analytical method which are
utilized in pharmaceutical analysis.
Keywords: Analytical method validation, Pharmaceutical analysis, Specificity, Precision, Accuracy.
I. INTRODUCTION
It may be defined that Analytical chemistry is the study of separation, quantification and chemical
components identification of natural and artificial materials constituted with one or more compounds or
elements. Analytical chemistry is separated into two main categories, qualitative analysis that is to say the
identification with regard to the chemical components exist in the sample, where as quantitative
analysis estimates the amount of certain element or compound in the substance i.e., sample.
Pharmaceutical analysis [1-3] plays a very outstanding role in the examination of pharmaceutical
formulations and bulk drugs regarding the quality control and assurance. Rapid increase in pharmaceutical
industries [4] and production of drug in and around the world bring forward a rise in inevitable demand to seek
novel and systematic analytical techniques in the pharmaceutical industries. As a consequence, analytical
method development has become the basic activity of analysis.
Development in scientific and concrete analytical methods has been resulted from the advancements of
analytical instruments. The improvements of the analytical method development and analytical instruments have
reduced the time and cost of analysis [5] and enhanced precision and accuracy. Techniques pertaining to analysis
are developed and validated for active pharmaceutical ingredients, excipients, related substances, drug products,
degradation products and, residual solvents, etc. Resulting which become an integral part of the required
necessities for regulatory organization[6].
Analytical method development finally results in official test methods[7]. Consequently quality control
laboratories used these methods to check the efficacy, identity, purity, safety as well as performance of products
of the drug. Regulatory authorities give utmost importance on analytical methods in manufacturing. Drug
approval by regulatory authorities requires the applicant to prove control of the entire process of drug
development by using validated analytical methods[8].
The numerous novel drugs are being introduced and are constantly growing day by day. Therefore it is
absolutely imperative to evolve novel methods and introduced them for controlling their quality. Modern
pharmaceutical analysis needs the following requirements.
7
A Review on Step-by-Step Analytical…
The concept of quality control is intended to examine and identify a genuine and right product by series
of measures designed to avoid and get rid of errors at varied stages in production. To take a decision to release
or discard a product is based on one or more sorts of control actions. Providing simple and analytical process for
various complex formulations is a subject matter of utmost importance. Rapid increase in pharmaceutical
industries and constant production of drug in various parts of the world has brought a quick rise in demand for
new analytical techniques in the pharmaceutical industries as a consequence, analytical method development has
become the basic activity of analysis in a quality control laboratory.
The reasons for the development of novel methods of drug analysis are:
b) When there is no decorous analytical process for the existing drug in the literature due to patent
regulations.
c) When there are no analytical methods for the formulation of the drug due to the interference caused
by the formulation excipients.
d) Analytical methods for the quantitation of the analyte in biological fluids are found to be unavailable.
e) The existing analytical procedures may need costly reagents and solvents. It may also involve
burdensome extraction and separation procedures.
8
A Review on Step-by-Step Analytical…
b) If there are no past suitable methods available to analyze the analyte to be examined.
3.8. Optimization
While performing optimization, one parameter is changed at a time and a set of conditions are isolated,
before utilizing trial and error approach. The said work need to be accomplished basing on a systematic
methodical plan duly observing all steps and documented with regard to dead ends.
9
A Review on Step-by-Step Analytical…
1. System suitability
1. Specificity
2. Linearity
3. Precision
4. Accuracy
5. LOD
7. LOQ
8. Robustness
The parameters used in the system suitability tests (SST) report are as follows:
Resolution (Rs).
(2)
The above said capacity factor sometimes is called as a retention factor which has no dimension and
independent from flow rate of mobile phase as well as column dimensions which is the measure of extent of
retention relating to an analyte relative to an un-retained peak. Where tR implies retention time of the sample
peak and retention time of an un-retained peak is tM.
k' = 0 means no compound is left in the column. Generally the value of k' is > 2.
11
A Review on Step-by-Step Analytical…
(3)
= Relative retention.
= Retention time calculated from point of injection.
= Unretained peak time (Retention time (tR) of an inert component not retained by
the column).
= the retention time from the point of injection of reference peak defined. (Suppose no reference peak is
found, value would be zero).
Resolution (Rs)
Resolution is the capability of the column to separate 2 drugs in 2 individual peaks or chromatographic
zones and it is improved by enhancing column length, reduction of particle size and rising temperature, altering
the eluent or stationary phase. It can be told in terms of ratio of separation of the apex of two peaks by the
tangential width average of the peaks. By using the following formula resolution is calculated.
(4)
tR1 and tR2 are the retention times for the two peaks of components.
tw1 and tw2 = At the baseline lies between tangents drawn to the sides of the peaks. (Tangents are drawn at 0.6
times the peak height). If the peaks are correctly symmetric, provided the valley between the two peaks should
touch the baseline Rs is 1.5. Generally good value of resolution is Rs ≥2 should be adequate and preferred
normally.
(5)
As = B/A (6)
Where:
As = peak asymmetry factor.
12
A Review on Step-by-Step Analytical…
Acceptance criteria (limits) of system suitability parameters are shown in the following Table 1.1.
4.2. Specificity
One of the significant features of HPLC is its ability to generate signals free from interference.
Specificity refers to the ability of the analytical method to differentiate and quantify the analyte in complex
mixtures. An investigation of specificity is to be conducted during the determination of impurities and validation
of identification tests.
An ICH guideline defines specificity as ability to assess unequivocally the analyte in the presence of
other compounds that may be likely to be present. Typically these might be impurities, degradants, matrix, etc.
The definition has the following implications:
Identification test: Identification tests should be able to differentiate compounds of closely related
structure which are expected to be present i.e., to assure identity of an analyte.
Purity test: To ensure that the analytical procedure performed allows an accurate statement of
content of the impurity of an analyte i.e. related substances, residual solvents content, heavy metals,
etc.
Assay: To arrive at an accurate result, this permits a correct report on the potency or content of analyte
in a sample.
13
A Review on Step-by-Step Analytical…
4.4. Precision
The precision of an analytical procedure represents the nearness of agreement between a series of
measurements got from multiple sampling of the same homogenous sample under the similar analytical
conditions and it is divided into 3 categories.
Repeatability: precision under same operating conditions, same analyst over a short period of time.
Intermediate precision: method is tested on multiple days, instruments, analysts etc.
Reproducibility: inter-laboratory studies.
The ICH guidelines suggest that repeatability should be conformed duly utilizing at least 9
determinations with specified range for the procedure (e.g., three concentrations / three replicates each) or a
minimum of 6 determinations at 100 % of the test concentration.
4.5. Accuracy
The accuracy of a measurement is defined as the closeness of the measured value to the true value. In
a method with high accuracy, a sample (whose “true value” is known) is analyzed and the measured value is
identical to the true value. Typically, accuracy is represented and determined by recovery studies. There are
three ways to determine accuracy:
1. Comparison to a reference standard.
2. Recovery of the analyte spiked into blank matrix.
3. Standard addition of the analyte.
It should be clear how the individual or total impurities are to be determined.
A number of approaches are recommended by the ICH for determining the detection limit of sample,
depending on instrument used for analysis, nature of analyte and suitability of the method. The acceptable
approaches are
Visual evaluation.
Signal-to-noise ratio.
Standard deviation of the response.
Standard deviation of the slope of linearity plot.
14
A Review on Step-by-Step Analytical…
Similar to LOD, ICH recommends the following four methods for estimation of LOQ. The acceptable
approaches are
Visual evaluation.
Signal-to-noise ratio.
Standard deviation of the response.
Standard deviation of the slope of linearity plot.
4.8. Robustness
Robustness is defined by the measure of the capability of an analytical method to stay unchanged by
small deliberate changes in method parameters. The variable method parameters in HPLC technique may
involves flow rate, column temperature, sample temperature, pH and mobile phase composition.
The purpose of carrying out a determination is to achieve a valid estimate of a 'true' value. When one
considers the criteria as said by which an analytical procedure is selected; precision and accuracy are generally
the primary prominent point to come to mind for application. Precision and accuracy together determine the
error of an individual determination. They are among the most significant critical parameters for judging the
analytical procedures by their achieved results.
5.1. Precision
Precision may be defined as the concordance of a series of measurements (n) of the same quantity.
Precision expresses the „reproducibility‟ of a measurement. One of the most common statistical terms employed
for precision is the sample standard deviation which is shown in the following equation.
(9)
The square of standard deviation is called Variance (S2). RSD or % RSD is the absolute value of the
CV (Coefficient of Variation) and it is often stated as a % which is used to juxtapose the uncertainty among
different measurements of varying total magnitude.
5.2. Accuracy[26]
Accuracy normally refers to the difference between the mean „x‟ of the set of results and the true or
most probable value for the quantity measured. As stated by IUPAC, accuracy relates to the difference between
result (or) mean and the true value. For analytical methods, there are two feasible ways of determining the
accuracy: absolute method and comparative method.
Absolute method
The test of the accuracy of the method under consideration is carried out by taking amounts of the
constituents and proceeds as said by specified instructions to perform the test for accuracy of the method. The
difference between the means of an adequate number of results and amount of constituent in fact present,
usually expressed as parts per hundred (%) which is termed as % error.
15
A Review on Step-by-Step Analytical…
The constituent in question will be determined in the presence of other substances, and it will therefore
be necessary to know the effect of the determination. This will require testing the influence of a large number of
probable compounds in the chosen samples of each varying amounts. In a few instances, the accuracy of the
method is controlled by separations (usually solvent extraction or chromatography technique) involved.
Comparative method
In the analysis of pharmaceutical formulations (or solid synthetic samples of desired composition), the
content of the constituent to be sought has been determined by one or more accurate analysis methods. Basically
if several distant (dissimilar) methods of analysis for a given constituent are available, e.g. gravimetric,
titrimetric, spectrophotometric, or chromatographic, the agreement between at least two methods of essentially
different character can generally be accepted as indication of the absence of an appreciable systematic error in
either method.
Recovery experiments
A known amount of the constituent being estimated is added to the sample, which is analyzed for the
total amount of constituent present. The difference between the analytical results for samples with and without
the added constituent gives the recovery of the amount of the added constituent. If the recovery is satisfactory,
our confidence in the accuracy of the procedure is improved.
Student t-test
Student t-test is used to compare the means of two related (paired) samples analyzed by reference and
test methods. It gives answer to the correctness of the null hypothesis with a certain confidence such as 95 % or
99 %. If the number of pairs (n) is less than 30, the condition normality of x or at least the normality of the
difference (di) is required. If this is the case the quantity has a student t-distribution with (n -1) degrees of
freedom.
di
t (10)
sd / n
Where di = xR (Reference method) – xT. (Test method) and sd is the standard deviation.
F-test
By the F-test we can test the significance of the difference in variances of reference and test methods.
Let us suppose that one carried out n1 replicate measurements by test methods and n2 replicate measurements by
using reference method. If the null hypothesis is true then the estimates S T2 (variance of the test method) and SR2
(variance of reference method) do not differ very much and their ratio should not differ much from unity. In
fact, one uses the ratio of variances.
It is conventional to calculate the F - ratio by dividing the larger variance by the smaller variance in
order to attain a value equal or larger than unity. If the calculated F - value is smaller than F - value from the F-
table, one can conclude that the procedures are not significantly different in precision at a given confidence
level.
5.4. Calibration
A very essential part of all analytical procedures is the calibration and standardization process.
Calibration determines the relationship between the analytical response (Absorbance, peak area, peak height
etc.,) and the analyte concentration. Generally this is accomplished by the use of chemical standards. A good
precision and accuracy can only be obtained when an excellent calibration procedure is used. In pharmaceutical
analysis the following calibration procedures are normally employed.
16
A Review on Step-by-Step Analytical…
Standard addition
In standard addition, known quantities of drug are added to the unknown. From the increase in signal
and then figure out how much analyte existed in the original unknown. The standard addition method gets a
linear response to analyte.
Least-squares regression analysis can be used to express the relationship between response (y) and
concentration (x). The relationship can be represented by the general function.
We adopt the convention that the x values relate to the controlled or independent variable (e.g. the
concentration of a standard) and the y values are related to the dependent variable (the response measurements).
This means that the X values have no error. On the condition that the errors made in preparing the standards are
significantly smaller than the measuring error (which is usually the case in analytical problems). The values of
the unknown parameters a, b1, ....bm must be estimated in such a way that the model fits the experimental data
points (xi, yi) as closely as possible. The true relationship between x and y is considered to be given by a straight
line. The relation between each observation pair (Xi, Yi) can be represented as
Yi = + Xi + ei (12)
The signal yi is composed of deterministic component predicted by linear model and a random
component ei. One must now find the estimates of a and b of the two values and . This can be done by
calculating values a and b for which ei2 is minimal. The component ei represent the differences between the
observed yi values and the predicted yi values by the model. The ei are called the residuals, a and b are the
intercept and slope respectively.
n n n
n xi yi
xi yi
b i 1 i 1 i 1
2
n
n (13)
n x xi
2
i
i 1 i 1
17
A Review on Step-by-Step Analytical…
n n n n
yi xi2 xi xi yi
a i 1 i 1 i 1 i 1
2
n
n
n xi2 xi
i 1 i 1
n
( y
(14)
Se i yi ) 2 /( n 2)
i 1
yiand yi , are the observed and predicted values, respectively. Standard deviations on slopes (S b)
and intercepts (Sa) are quoted less frequently, even though they are used to evaluate proportional differences
between or among methods as well as to compute the independent variables such as concentration etc. It is
important to understand how uncertainties in the slope are influenced by the controllable properties of the data
set such as the number and range of data points and also how properties of data sets can be designed to optimize
the confidence in such data.
n 1
( yi yi ) 2 n (15)
Sb i 1
(n 2)
* ( xi xi ) 2
i 1
Where
n 1 n
( yi yi ) 2 n x i
2
Here
Sa i 1
(n 2)
* (x x )
i 1
i i
2
* i 1
n
(16)
xi is the
arithmetic mean of xi values
Correlation coefficient (r)
In order to ascertain whether there is a linear relationship between two variables „x‟ and „y‟, the
correlation coefficient (r) is used. To achieve a correlation coefficient, the co-variance is divided by the product
of the standard deviation of x and y.
18
A Review on Step-by-Step Analytical…
n
( xi x )( y i y ) /( n 1)
r
n 2 (17)
( xi x ) 2
( yi y ) /( n 1)
2
VI. CONCLUSION
This article provides an idea how to perform validation process to prove that the method is apt for its
intended purpose and to assure the capabilities of the test method. The definitions of method validation
parameters are well explained. Although the requirements of validation have been clearly documented by
regulatory authorities, the approach to validation is varied and opened to interpretation, and validation
requirements differ during the development process of pharmaceuticals. Validation is an important procedure in
the pharmaceutical industry and it is utilized to ensure that quality is built in to the processes supporting drug
development and manufacture.
REFERENCES
[1] G. David Watson, Pharmaceutical Analysis (3rd Ed., Churchill Livingstone, London: Harcourt Publishers Limited, Essex CM 20
2JE, 2012).
[2] A.H. Beckett, and J.B. Stenlake, Practical Pharmaceutical Chemistry (4th Ed., Vol. I & II. CBS Publishers and Distributors, New
Delhi: 2007).
[3] T. Higuchi, and Brochman-Hansen, Pharmaceutical Analysis, (3rd edition, CBS Publishers and Distributors pvt. Ltd., New
Delhi:1997).
[4] G. Oliver, R. Gerrit, and VZ. Maxmilian, Leading Pharmaceutical Innovation, „Trends and drivers for Growth in the
pharmaceutical industry, (2nd Ed., Springer, 2008)12-15.
[5] Br. Jay, J. Kelvin, and B. Pierre, Understanding and Implementing Efficient Analytical Methods Development and Validation,
2003.
[6] R.M. Christopher, and W.R. Thomas, Quality Systems approach to Pharmaceutical cGMP Development and validation of
Analytical Methods, (1st Ed., 2005) 147-152.
[7] R. Lloyd Snyder, J. Joseph Kirkland and L. Joseph Glajah, Practical HPLC method development (2nd Ed., 1997) 179-184.
[8] B.K. Sharma, Instrumental method of chemical analysis (29th Ed., Meerut, Chromatography, HPLC, Goel Publishing House,
2013) 286-385.
[9] H.H. Willard, L.L. Merrit, J.A. Jr. Dean, and F.A. Jr. Settle, Instrumental Methods of Analysis (CBS Publishers, New Delhi:
1986).
[10] R.A. Day, and A.L, Underwood, Quantitative Analyses, (5th Ed., Prentice Hall, New Delhi: 1986).
[11] Macek and Karel, Pharmaceutical Applications of Thin Layer and Paper Chromatography, 62(6) 1972,1032.
[12] G. Ramana Rao, S.S.N. Murthy, and P. Khadgapathi, Gas Chromatography to Pharmaceutical Analysis, Eastern Pharmacist,
30(353), 1987, 35.
[13] G. Ramana Rao, S.S.N. Murthy, and P. Khadgapathi, High Performance Liquid Chromatography and its Role in Pharmaceutical
Analysis, Eastern Pharmacist, 29(346), 1986,53.
[14] C.S.P. Sastry, T.N.V. Prasad, and E.V. Rao, Recent applications of High Performance Liquid chromatography in Pharmaceutical
analysis, Indian J. Pharm. Education, 21 (37),1987.
[15] Ravisankar P, Gowthami S, and Devala Rao G, A review on analytical method development, Indian journal of research in
pharmacy and biotechnology, 2(3), 2014, 1183-1195.
[16] Ravisankar P, Rajyalakshmi G, Devadasu Ch, and Devala Rao G, Instant tips for right and effective approach to solve HPLC
trouble shooting, Journal of chemical and pharmaceutical sciences. 7(3), 2014, 259-274.
[17] Jay Breaux, Kevin Jones, and Pierre Boulas, Development services analytical method development and validation
„Pharmaceutical technology, 27(1), 2003, 6-13.
[18] E. Michael Swartz, and Iras Krull, Analytical method development and validation, CRC press, Marcel dekker, Inc., Madison
Avenue, New York: 1997.
[19] K. Yuri, and LB. Rosario, HPLC for pharmaceutical scientists, (John Wiley and Sons, Inc., Hoboken, New Jersey, 2007) 92-98.
[20] G. P. Carr, and J. C. Wahlichs. „A practical approach to method validation in pharmaceutical analysis‟, J. Pharm, Biomed.
Anal,8,1990,613-618.
[21] United States Pharmacopoeia,24, National Formulary 19, section <1225> „Validation of compendial methods‟. US
Pharmacopoeial convention, Rockville, Validation of analytical procedures text and methodology Q2 (R1), November, 2000:
2005.
[22] International conference on harmonization (ICH) of technical requirements for registration of pharmaceuticals for human use,
Validation of analytical procedures: Text methodology, (Q2 (R1) Geneva, 2005) 6-13.
[23] Draft guidance analytical procedures and method validation, US food and drug administration, Centre for drugs and biologics,
Department of Health and Human Services. http://www.fda.gov/cder/guidance/2396 dft.htm#111, 2000.
[24] Orr J.D, Krull I.S and Swartz M.E, Validation of impurity methods Part II, (LC North Am; 21, 2003) 1146-1152.
[25] ICH Q2 (R1), Validation of analytical procedures (definitions and terminology); 2005; pp. 9-10.
[26] D.L. Massart, B.G.M. Vandeginste, S.N. Deming, Y. Michotte and L. Kaufman, Chemometrics: A Text Book, Elsevier,
Amsterdam: 1988.
[27] M.M. Kiser and J.W. Dolan, Selecting the best curve fit LC-GC (Europe: 2004)138-143.
[28] R.E. Moore, Interval analysis (Englewood Cliffs, NJ: Prentice-Hall, 1966).
19
Department of Physical Chemistry ARS DOCENDI PUBLISHING HOUSE
4-12 Regina Elisabeta Blvd, District 3, Bucharest Sos. Panduri 90, District 5 Bucharest
phone: +40-21-3143508; fax: +40-21-3159249 phone/fax: +40-21-4102575
ISSN: 1844-0401 e-mail: arsdocendi@yahoo.com
Introduction
The possibility of quantitative determination of Cl- from drinking and surface water using
direct potentiometry with Cl––selective electrode constitutes an opportunity for routine
measurements / tests.
Analytical method validation is a “necessary evil” for the quality assurance and production
processes safety, but represents the first step or the first level of quality assurance within a
testing and measurement analytical laboratory. Validation is realized according to national
and international standards (ISO 9000, ISO 17025), but also according to BPL-OECD
norms regarding good laboratory practice within testing and measurement in analytical
laboratory (GLP) (HG 63/2002 modified with HG 266/2006 and HG 448/2007).
For chloride anion determination from waters are used, with good results, both gravimetric
and volumetric analysis methods, as well as instrumental analysis methods. Classical
methods presume gravimetric determination by insoluble chloride precipitation, or
volumetric titration [1÷4]. From instrumental methods can be reminded UV-VIS molecular
absorption spectrometry (mercury tiocyanate method introduced by Japanese scientists
[5÷7] in 1952, mercury chloroanylate [8,9]), atomic absorption spectrometry (indirect
method) [10,11] and direct and indirect potentiometry [12÷18].
Analele Universit !ii din Bucure"ti – Chimie, Anul XVI (serie nou ), vol. II, pag. 25 – 32
Copyright © 2007 Analele Universit !ii din Bucure"ti
26 I.G. T#NASE! D.-E. POPA! M. BULEANDR#
Experimental
All the used reagents had analytical purity quality. A 10.000 ppm chloride stock solution
(WTW Germany, NIST (National Institute of Standards and Technology) traceable) was
used to obtain standard solutions. KNO3 was used to obtain the inert electrolyte solution
(1 mol.L-1). For standard and working solutions preparation adequate volumes of stock
standard solution were measured and then they were put in volumetric flasks. Tri-distilled
water was used for all solutions preparation and for vessel cleaning.
The necessary measurements for quantitative determination of chloride ion from drinking
and surface waters have been carried out using a pH/mV-metre WTW Inolab 740, with 6
point calibration; chloride-selective electrode as working electrode; R503/D-WTW
reference electrode; 50, 100, 500, 1000 mL volumetric flasks (A class); 100 mL Berzelius
glasses; 1, 2, 5, 10 mL pipettes (A class).
In order to prepare the standard solutions, in 50 mL volumetric flasks were put: fixed
volumes (5 mL) of 1 M KNO3 solution, increasing volumes of standard chloride solution
(100 ppm) and tri-distilled water. Drinking or surface water samples were prepared as
follows: in a 50 mL volumetric flask was added 25 mL from the water sample, 5 mL inert
electrolyte solution (1 mol L-1 KNO3) and tri-distilled water. The samples were decanted
into the potentiometric cell and emf was measured.
detection limit, quantification limit, robustness. From these, in this first part will be
discussed the working range and the linearity.
Before establishing the working range, the chloride-selective electrode response is followed
up on a wide concentration range (1.10-6÷1.10–1 mol.L–1), in aqueous medium of 0.1 mol.L–1
KNO3. The experimental results are graphically represented in Fig. 1. They emphasize that
the response curve of chloride-selective electrode presents a sigmoid form; there is a large
enough concentration range for which the dependence is linear; this can be used for
quantitative determination of chloride ion.
0,35
0,3
0,25
E(V)
0,2
0,15
0,1
0,05
-7 -6 -5 -4 -3 -2 -1 0
lg c
If the working range is established by calibration [23], then is worth mentioning that every
calibration experiment begins with selection of a preliminary working range. The
informative values are presented in Table 1 (for the lowest concentration were made 10
replicates and for the highest concentration were also made 10 replicates). Informative
concentration values were equidistantly distributed on the entire established preliminary
working range.
Table 1 Calibration data for establishing the working range used for quantitative determination
of chloride from drinking and surface waters by direct potentiometry with chloride-selective
electrode (preliminary range 1.10–5 – 2,5.10–3 mol L–1)
j xi
yi,1 yi,2 yi,3 yi,4 yi,5 yi,6 yi,7 yi,8 yi,9 yi,10
i (mol L–1)
1 1.0 10–5 0.3025 0.2975 0.2973 0.2964 0.2967 0.2969 0.2970 0.2971 0.2966 0.2966
2 2.5 10–5 0.2868
3 5.0 10–5 0.2650
4 7.5 10–5 0.2557
–4
5 1.0 10 0.2509
6 2.5 10–4 0.2269
7 5.0 10–4 0.2092
–4
8 7.5 10 0.1989
9 1.0 10–3 0.1910
10 2.5 10–3 0.1676 0.1658 0.1679 0.1665 0.1685 0.1662 0.1689 0.1652 0.1661 0.1669
For i = 1 and i = 10 there were obtained: y1 = 0.28746; s1 = 0.00180207; s12 = 3.25156 10–6;
2 = 1.49 10–6.
y10 = 0.166696; s10 = 0.00122; s10
ANALYTICAL DETN. OF CHLORIDE FROM DRINKING AND SURFACE WATER 29
Obtained dispersions were tested using the F test to examine the significant differences at
the working concentration range limits. The F test requires the testing value, PG,
determination starting with the formula:
s12
PG # 2
for s12 % s10
2
, (2)
s10
that is PG # 2,17554 .
The obtained PG value was compared to the table values for the F distribution. Consulting
the table containing values of F (dispersion homogeneity test for &1 = &2 = n – 1 = 9
freedom degree for dispersions s12 and s10
2 ), it was obtained F
9;9;0.99 = 5,35. In this case, PG
= 2,17554 and F9;9;0,99 = 5,35, so PG ' F9;9;0,99; the deviation between dispersions s12 and s10 2
. -5 . -3 . -1
wasn’t significant and the working range was correctly selected: 1 10 –2,5 10 mol L (0,5
– 50 mg .L-1) Cl-.
Comparison of calculated value PG, with the table value didn’t showed significant
difference. Dispersions are homogenous and are allowing the simple (linear) regression
analysis.
In the statistic linearity test, calibration data are used to calculate a linear calibration
function and also a non-linear calibration function, both of them showing a standard
residual deviation s y1 and s y2 . Dispersions difference DS2 can be calculated using the
formula:
DS 2 # N " 2 !s 2y " N " 3!s 2y , (3)
1 2
Coming back to calculation, for the linearity test was obtained: standard deviation of linear
regression s y = 0.00147358 and s 2 = 2.171.10-6; standard deviation of a nonlinear
1 y1
regression s y = 0.000738786 and s 2 = 5.4580.10-7.
2 y 2
2. Linearity
Linearity of a quantitative analytical method represents its ability to obtain results
proportional with analyte concentration from the sample. Linearity can’t be expressed, it
must be demonstrated directly on analyte or spiked samples, using at least five
concentrations on working range.
Besides the visual evaluation of the analytical signal as a concentration function, it is
recommended to do the corresponding statistical calculations, like linear regression.
Reporting statistical parameters like the slope of the calibration curve, b and its intersection
to the origin, a, residual values squares amount and correlation coefficient is a mandatory
condition.
Linear calibration function must result from obtained data, starting from a working range
from x1 to x10, as it was established starting from uncorrelated measurements for blank
samples. Usually, mustn’t be included any blank sample value in calibration experiments
and in least squares method.
Linear calibration function is described by the equation:
y # a ( bx (5)
The linear model is, no doubt, the most important one for processing dimensional data.
Over sizing the equations system represents the regression analysis focussing point,
meaning the determination of more coordinate pairs then the obtained ones (x1, y1) and (x2,
y2), requested to calculate a and b using the classic algebra. The least squares model was
selected, leading to a straight line subject (submitted) to the restriction:
1 1.0 10-5 - 5.000 0.3025 2.50E+01 -1.25E+02 6.25E+02 9.15E-02 -1.51E+00 7.56E+00 -2.00E-03
2 2.5 10-5 -4.602 0.2818 2.12E+01 -9.75E+01 4.49E+02 7.94E-02 -1.30E+00 5.97E+00 -2.53E-04
3 5.0 10-5 -4.301 0.2650 1.85E+01 -7.96E+01 3.42E+02 7.02E-02 -1.14E+00 4.90E+00 -7.64E-05
-5
4 7.5 10 -4.124 0.2557 1.70E+01 -7.01E+01 2.89E+02 6.54E-02 -1.05E+00 4.35E+00 6.06E-04
5 1.0 10-4 -4.000 0.2509 1.60E+01 -6.40E+01 2.56E+02 6.30E-02 -1.00E+00 4.01E+00 2.80E-03
6 2.5 10-4 -3.602 0.2269 1.30E+01 -4.67E+01 1.68E+02 5.15E-02 -8.17E-01 2.94E+00 1.25E-03
7 5.0 10-4 -3.301 0.2092 1.09E+01 -3.60E+01 1.19E+02 4.38E-02 -6.91E-01 2.28E+00 5.24E-04
-4
8 7.5 10 -3.124 0.1989 9.76E+00 -3.05E+01 9.52E+01 3.96E-02 -6.21E-01 1.94E+00 2.06E-04
ANALYTICAL DETN. OF CHLORIDE FROM DRINKING AND SURFACE WATER 31
9 1.0 10-3 -3.000 0.1910 9.00E+00 -2.70E+01 8.10E+01 3.65E-02 -5.73E-01 1.72E+00 -7.00E-04
-3
N=10 2.5 10 -2.602 0.1676 6.77E+00 -1.76E+01 4.58E+01 2.81E-02 -4.36E-01 1.13E+00 -1.65E-03
N
$ -37.656 2.3495 1.47E+02 -5.94E+02 2.47E+03 5.69E-01 -9.15E+00 3.68E+01 7.02E-04
i #1
Linear function y = a + bx
Equation y = 0,0564x + 0,0225
N N
Slope – sensibility, b !
b # $ xi " x yi " y ! $ xi " x ! # 0,0564
i #1 i #1
N
Residual standard deviation :
s y1 # $ ) yi " a ( bx !*2 N " 2 s y1 # 0,00147358
i #1
Intercept a:
a# $ yi " b$ xi " c$ xi2 ! N a # "0,009681719
Although the correlation coefficient (R) and the determination coefficient (R2) obtained
during the regression analysis, are often used to prove the linearity (R>0.997) [24], their
values are not able to do that; they can only prove if statistical processed values are
correlated (R=|1.000|) or uncorrelated (R=0.000).
An alternative approach for using the correlation coefficient to linearity establishment
consists in dividing the sample signal to the corresponding concentration (– lgcCl" ) and the
32 I.G. T#NASE! D.-E. POPA! M. BULEANDR#
graphical display of these “relative responses versus lg(– lgcCl" ). It must be obtained a
horizontal straight line with positive deviation at low concentrations and negative deviation
at large concentrations.
Conclusion
In this first part of the validation study it was proved that the method is linear and presents
an appropriate linear working range.
REFERENCES
1. Williams, W.J. (1979) Handbook of Anion Determinations, Butterworths, London, 291-330.
2. P troescu, C. and G nescu, I. (1980), Analiza apelor, Editura Scrisul Românesc, Craiova, 87-91.
3. M nescu, S., Cucu, M. and Diaconescu, I.L. (1978), Chimia sanitar a mediului, Editura Medical ,
Bucure"ti, 59-61
4. *** SR ISO 9297 (2001) Calitatea apei. Determinarea con inutului de cloruri. Titrare cu azotat de argint
utilizând cromatul ca indicator (Metoda Mohr), ASRO, Bucure"ti
5. Iwaski, I., Utsumi, S. and Ozawa, T. (1952) Bull. Chem. Soc. Japan 25, 226
6. Iwaski, I., Utsumi, S., Mageno, K. and Ozawa, T. (1956) Bull. Chem. Soc. Japan 29, 860
7. Zoll, D.M., Fischer, D. and Garner, M.Q. (1956) Anal. Chem. 28, 1665
8. Barney, J.E. and Bertolacini, R.J. (1957) Anal. Chem. 29, 1187
9. Bertolacini, R.J. and Barney, J.E. (1958) Anal. Chem. 30, 202
10. Reichel, W. and Acs, L. (1969) Anal. Chem. 41, 1886
11. Fuinuma, H., Kasama, K., Takeuchi, K. and Hirono, S. (1970) Japan Analyst 19, 1487
12. Prokopov, T.S. (1970) Anal. Chem. 42, 1096
13. Bladel, W.J., Lewis, W.B. and Thomas, J.W. (1952) Anal. Chem. 24, 509
14. Malmstad, H.W. and Winefronder, J.D. (1959) Anal. Chim. Acta 20, 283
15. Bishop, E. (1956) Mikrochim. Acta, 619
16. Bishop, E. (1958) Analyst 83, 212
17. Bishop, E. and Dhaneshar, R.G. (1962) Analyst 87, 845
18. Florence, T.M. (1971) J. Electroanal. Chem. 31, 77
19. T nase, I.G. (2000) Tehnici !i metode electrometrice de analiz , Ars Docendi, Bucure"ti, 128
20. T nase, I.G. (2007) Analiz instrumental . Partea I. Tehnici !i metode electrometrice de analiz , Editura
Universit !ii din Bucure"ti, Bucure"ti, 131
21. *** ICH Q2A 1994 Validation of Analytical Methods (Definitions and Terminology)
22. *** ICH Q2B 1996 Analytical Validation – Methodology
23. *** SR ISO 8468-1 (1999) Calitatea apei. Etalonarea !i evaluarea metodelor de analiz" !i estimarea
caracteristicilor de performan ". Partea 1. Evaluarea statistic" a func iei liniare de etalonare, IRS, Bucure"ti
24. Chan C.C. (2004) Potency Method validation in Analytical Method validation and instrument
Performance Verification, Chan C.C., Lam H., Lee Y.C. and Zhang X.M. (Eds.), Wiley Interscience,
Hobokein, New Jersey, p. 1-10