Eurachem - The Fitness for Purpose of Analytical Methods

8/21/2019 Eurachem - The Fitness for Purpose of Analytical Methods

1/70

The Fitness fAnalytica

A Laboratory Guide to Method

Second E

r Purpose ofl Methods

Validation and Related Topics

ition 2014


2/70


3/70

i

Eurachem Guide

The Fitness for Purpose of Analytical Methods

A Laboratory Guide to Method Validation and Related Topics

Second edition

AcknowledgementsThis document has been produced by members of the Eurachem Method Validation Working Group and othersco-opted for this task. Those who have contributed to this edition are listed below.

Project group

Vicki Barwick LGC (UK)Pedro P. Morillas Bravo Canal de Isabel II Gestin (ES)Stephen L. R. Ellison LGC (UK)Joakim Engman National Food Agency (SE)Elin L. F. Gjengedal Norwegian University of Life Sciences (NO)

Ulla Oxenbll Lund Eurofins Milj A/S (DK)Bertil Magnusson (editor) SP Technical Research Institute of Sweden (SE)Hans-Thomas Mller Mersin (TR)Marina Patriarca Istituto Superiore di Sanit (IT)Barbara Pohl Merck KGaA (DE)Piotr Robouch European Commission (EU)Lorens P. Sibbesen (chairman) Labquality International (DK)Elvar Theodorsson University Hospital in Linkping (SE)Florent Vanstapel University Hospital Leuven, Leuven (BE)Isabelle Vercruysse BELAB (BE)Aysun Yilmaz Cevre Food and Industrial Analysis Laboratory (TR)Perihan Yolci meroglu Okan University (TR)Ulf rnemark (editor) Emendo Dokumentgranskning (SE)

Copyright Copyright of this document is held by the contributing authors. All enquiries regarding reproduction in anymedium, including translation, should be directed to the Eurachem secretariat. The text may not be copied forresale.

Recommended citation

This publication should be cited* as: B. Magnusson and U. rnemark (eds.) Eurachem Guide: The Fitness forPurpose of Analytical Methods A Laboratory Guide to Method Validation and Related Topics, (2 nded. 2014).ISBN 978-91-87461-59-0. Available from www.eurachem.org.

*Subject to journal requirements


4/70

ii


5/70

The Fitness for Purpose of Analytical Methods Eurachem Guide

iii

Contents

Foreword to the second edition ................................................................................................. 1

Foreword to the first edition ..................................................................................................... 2

Abbreviations and symbols ........................................................................................................ 3

1 Introduction ........................................................................................................................ 5

1.1 Rationale and scope for this Guide ...................................................................................... 5

1.2 Notes on the use of this Guide .............................................................................................. 51.2.1 Terminology ...................................................................................................................................... 51.2.2 Quick References .............................................................................................................................. 6

2 What is method validation? ............................................................................................... 7

2.1 Definitions .............................................................................................................................. 7

2.2 What is the difference between validation and verification? ............................................ 7

3

Why is method validation necessary? ................................................................................ 9

3.1 Importance of analytical measurement ............................................................................... 9

3.2 The professional duty of the analytical chemist .................................................................. 9

3.3 Method development ............................................................................................................. 9

4 When should methods be validated or verified? ............................................................. 11

4.1 Method validation................................................................................................................ 11

4.2 Method verification ............................................................................................................. 11

5 How should methods be validated? ................................................................................. 13

5.1

Who carries out method validation? .................................................................................. 135.1.1 Approaches to method validation ............................................................ .........................................13

5.1.2 Interlaboratory approach ..................................................................................................................135.1.3 Single-laboratory approach ..............................................................................................................13

5.2 Extent of validation studies ................................................................................................ 13

5.3 Validation plan and report ................................................................................................. 14

5.4 Validation tools .................................................................................................................... 155.4.1 Blanks...............................................................................................................................................15 5.4.2 Routine test samples .......................................................... ...............................................................155.4.3 Spiked materials/solutions ............................................................ ....................................................155.4.4 Incurred materials .............................................................. ...............................................................15

5.4.5

Measurement standards ................................................................ ....................................................155.4.6 Statistics ...........................................................................................................................................16

5.5 Validation requirements ..................................................................................................... 16

5.6 Method validation process .................................................................................................. 16

6 Method performance characteristics ............................................................................... 19

6.1 Selectivity ............................................................................................................................. 196.1.1 Terms and definitions ........................................................ ...............................................................196.1.2 Effects of interferences ................................................................. ....................................................196.1.3 Assessment of selectivity .................................................................................................................19


6/70


iv

6.2 Limit of detection and limit of quantification ................................................................... 206.2.1 Terms and definitions ........................................................ ...............................................................206.2.2 Determination of the standard deviation at low levels .....................................................................216.2.3 Estimating LOD ...............................................................................................................................246.2.4 Estimating LOQ ...............................................................................................................................246.2.5 Alternative procedures .....................................................................................................................25

6.2.6

Capability of detection for qualitative analysis ................................................................................25

6.3 Working range ..................................................................................................................... 276.3.1 Definition .........................................................................................................................................276.3.2 Considerations for the validation study .............................................................. ..............................276.3.3 Method and instrument working range.............................................................................................276.3.4 Assessing instrument working range ....................................................... .........................................276.3.5 Assessing method working range ............................................................ .........................................28

6.4 Analytical sensitivity ........................................................................................................... 306.4.1 Definition .........................................................................................................................................306.4.2 Applications .....................................................................................................................................30

6.5 Trueness ............................................................................................................................... 30

6.5.1

Terminology to describe measurement quality ................................................................................30

6.5.2

Determination of bias .......................................................................................................................31

6.5.3 Interpreting bias measurements ............................................................... .........................................34

6.6 Precision ............................................................................................................................... 356.6.1 Replication .......................................................................................................................................356.6.2 Precision conditions .........................................................................................................................356.6.3 Precision limits ........................................................ ................................................................. ........366.6.4 Simultaneous determination of repeatability and intermediate precision .........................................36

6.7 Measurement uncertainty ................................................................................................... 38

6.8 Ruggedness ........................................................................................................................... 386.8.1 Definition .........................................................................................................................................386.8.2 Ruggedness test ................................................................. ...............................................................38

7

Using validated methods .................................................................................................. 41

8

Using validation data to design quality control .............................................................. 43

8.1 Introduction ......................................................................................................................... 43

8.2 Internal quality control ....................................................................................................... 43

8.3 External quality control ...................................................................................................... 44

9 Documentation of validated methods .............................................................................. 45

9.1 From draft to final version ................................................................................................. 45

9.2

Recommendations ............................................................................................................... 459.2.1 Checking the instructions .................................................................................................................45

9.2.2 Recommendations in standards ............................................................... .........................................459.2.3 Document control .............................................................. ...............................................................45

10 Implications of validation data for calculating and reporting results........................ 47

Annex A Method documentation protocol .......................................................................... 49

Annex B Statistical basis of limit of detection calculations ................................................ 53

Annex C Analysis of variance (ANOVA) ............................................................................ 54

Annex D Notes on qualitative analysis ................................................................................ 56

Bibliography ............................................................................................................................ 59


7/70


MV 2014 1

Foreword to the second edition

Since the first edition of this Guide in 1998, a number of important developments in analytical qualityhave taken place. Firstly, the ISO 9000 series of standards, which is widely used to provide a basis fora quality management system, has been revised. Its philosophy forms an integral part of internationalconformity assessment standards and guides, which underpins competence requirements forlaboratories, proficiency testing (PT) providers and reference material (RM) producers. Thesedocuments all stress the importance of using validated methods.

Secondly, several general or sector-specific guides on method validation have been revised ordeveloped. EU legislation contains mandatory requirements for analytical measurements in manysectors.

Thirdly, much effort has been invested by the analytical community in implementing the uncertaintyconcept. For example, in its Harmonized guidelines for single-laboratory validation of methods ofanalysis (2002) IUPAC predicted that, ...with an increasing reliance on measurement uncertainty as akey indicator of both fitness for purpose and reliability of results, analytical chemists will increasinglyundertake measurement validation to support uncertainty estimation.... In the following years,accreditation bodies issued policies and guidance documents clearly recognising the use of methodvalidation data in the measurement uncertainty estimation process.

Furthermore, the International vocabulary of metrology Basic and general concepts and associatedterms (VIM) has been substantially revised, taking into account chemical and biologicalmeasurements. Although terminology related to method validation is far from harmonised, thesituation has improved. VIM is also a normative document for laboratories accredited to, e.g. ISO/IEC17025 and ISO 15189.

The second edition of this Guide aims to reflect changes in international standards and guidancedocuments and puts less emphasis on terms and definitions. Instead the Guide refers to the VIM andother readily available sources. As a consequence, the list of terms and definitions has been omitted

from the Annex. Literature cited in this edition of this Guide are listed in the Bibliography at the end.Additional sources and literature related to method development and validation is available as aReading list under the menu item Publications on the Eurachem website at www.eurachem.org.Annex A is revised as a consequence of changes to ISO 78-2. This edition has also been extended toinclude information on the statistical basis of limit of detection calculations (Annex B), analysis ofvariance (Annex C) and qualitative analysis (Annex D).

It is becoming increasingly common among routine laboratories, especially in the clinical sector, touse commercially available measuring systems. This means that the responsibility for validationmainly lies with the manufacturer. The laboratorys work will focus on verifying the manufacturers

published performance data and demonstrate that the method works on the end-users premises.

However, looking back to the foreword to the first edition, we conclude that the six principles stated

there are still relevant, and are consistent with the requirements of international standards such asISO/IEC 17025.


8/70


MV 2014 2

Foreword to the first edition*

An initiative in the UK to promote good practice in analytical measurement has identified sixprinciples of analytical practice which, taken together, are considered to constitute best practice. Thesix principles which are described in more detail in a separate guideare:

1. Analytical measurements should be made to satisfy an agreed requirement. (i.e. to a definedobjective).

2. Analytical measurements should be made using methods and equipment which have beentested to ensure they are fit for purpose.

3. Staff making analytical measurements should be both qualified and competent to undertakethe task. (and demonstrate that they can perform the analysis properly).

4. There should be a regular independent assessment of the technical performance of alaboratory.

5. Analytical measurements made in one location should be consistent with those made

elsewhere.6. Organisations making analytical measurements should have well defined quality control and

quality assurance procedures.

These principles are equally relevant to laboratories whether they are working in isolation orproducing results which need to be compared with those from other laboratories.

This document is principally intended to assist laboratories in implementing Principle 2, by givingguidance on the evaluation of testing methods to show that they are fit for purpose.

*The first edition (1998) of this Guide was developed by a Eurachem Working Group from a draft originallyproduced by LGC. The following persons were members of the Eurachem group at that time:D. Holcombe, P. De Bivre, D. Bttger, C. Eastwood, J. Hlavay, M. Holmgren, W. Horwitz, M. Lauwaars, B.Lundgren, L. Massart, J. Miller, J. Morkowski, B. te Nijenhuis, B. Nyeland, R. Philipp, P. Radvila, J. Smeyers-Verbeke, R. Stephany, M. Suchanek, C. Vandervoorst, H. Verplaetse, H. Wallien, M. Walsh, W. Wegscheider,D. Westwood, H. J. van de Wiel.

The managers guide to VAM, UK Department of Trade and Industry, Valid Analytical MeasurementProgramme. Published as VAM Principles M. Sargent. Anal. Proc., 1995, 32, 201-202.


9/70


MV 2014 3

Abbreviations and symbols

The following abbreviations, acronyms and symbols occur in this Guide.

AMC Analytical Methods CommitteeANOVA Analysis of variance

AOAC International a globally recognized standards developing organization

ASTM International a globally recognized standards developing organization

BIPM International Bureau of Weights and Measures

CCQM Consultative Committee for Amount of Substance Metrology in Chemistry

CEN European Committee for Standardization

CITAC Cooperation on International Traceability in Analytical Chemistry

CLSI Clinical and Laboratory Standards Institute

CRM certified reference material

EA European co-operation for Accreditation

EC European Commission

EPA Environmental Protection Agency

EQA external quality assessment

EU European Union

GUM Evaluation of measurement data Guide to the expression of uncertainty inmeasurement

ICH International Conference on Harmonisation of Technical Requirements for

Registration of Pharmaceuticals for Human Use

IEC International Electrotechnical Commission

ISO International Organization for Standardization

IUPAC International Union of Pure and Applied Chemistry

JCGM Joint Committee for Guides in Metrology

LOD limit of detection

LOQ limit of quantification

NATA National Association of Testing Authorities

QA quality assurance

QC quality control

RSC Royal Society of Chemistry

SANCO European Commissions Directorate-General for Health and Consumers

SOP standard operating procedure

PT proficiency testing

RM reference material

RSD relative standard deviation

UV/VIS ultraviolet/visible

VIM International vocabulary of metrology Basic and general concepts and associatedterms


10/70


MV 2014 4

b absolute bias

b(%) relative bias in %

kQ multiplier used in calculating limit of quantification

m number of measurements

n number of replicate observations averaged when reporting results

nb number of blank observations averaged when calculating the blank correction

r repeatability limit

R reproducibility limit

(%) relative recovery (apparent recovery) in per cent(%) relative spike recovery in per cent

s standard deviation

s0 estimated standard deviation of single results at or near zero concentration

standard deviation used for calculating an LOD or LOQsI intermediate precision standard deviation

sr repeatability standard deviation

sR reproducibility standard deviation

u standard uncertainty

x mean value (arithmetic average)

refx reference value

mean value of measurements with an alternative method, e.g. a reference method

mean value of spiked sample in a recovery experiment added concentration in a recovery experiment


11/70


MV 2014 5

1 Introduction

1.1 Rationale and scope for thisGuide

Method validation is an important requirement inthe practice of chemical analysis. Most analyticalchemists are aware of its importance, but why itshould be done and when, and exactly whatneeds to be done, is not always clear to them.Some analysts used to see method validation assomething that can only be done in collaborationwith other laboratories and therefore refrainedfrom it. Requirements in standards such asISO/IEC 17025 [1], ISO 15189 [2] and ISO15195 [3] have helped in clarifying this. Forexample, the need to demonstrate that methods

are fit for purpose is stressed in Clause 5.4.2 ofISO/IEC 17025:

The laboratory shall use test and/or calibrationmethods, including methods for sampling, whichmeet the needs of the customer and which are

appropriate for the tests and/or calibrations itundertakes... and further: When the customerdoes not specify the method to be used, thelaboratory shall select appropriate methods....

The purpose of this Guide is to discuss theissues related to method validation and increase

readers understanding of what is involved, whyit is important, and give some idea of how it can

be accomplished.

The Guide is expected to be of most use to a)laboratory managers responsible for ensuringthat the methods under their supervision areadequately validated and b) analysts responsiblefor planning and carrying out studies on methodsfor validation purposes. Other staff may find theguidance of use as a source of backgroundinformation senior staff from a management

point of view and junior staff from a technical oreducational point of view.

The Guide focuses on single-laboratoryvalidation. It aims to direct the reader towardsestablished protocols where these exist andwhere they do not, give a simple introduction tothe processes involved in validation and providesome basic ideas to enable the reader to designtheir own validation strategies. It includesreferences to further material on particulartechnical aspects of validation.

This Guide is aimed at the validation ofquantitative methods. However, some of the

principles described here are also relevant for

qualitative methods for determining the presenceof one or more analytes, e.g. the concepts ofselectivity and limit of detection (LOD).

The Guide avoids emphasis on the use ofstatistics although undoubtedly those with aworking knowledge of elementary statistics willfind the method validation process easier tounderstand and implement. Several referencesare made to publications on basic statistics forchemists [4, 5, 6].

The analysts understanding of methodvalidation is inhibited by the fact that many ofthe metrological and technical terms used todescribe processes for evaluating methods varyin different sectors of analytical measurement,

both in their meaning and the way they aredetermined. This Guide cannot say where a termis used correctly or incorrectly although it isintended to provide some clarification. The bestadvice when using a term that may bemisunderstood, is to state the source and whichconvention has been used.

It is implicit in the method validation processthat the studies to determine method performancecharacteristics* are carried out using equipment

that is within specification, working correctly,and adequately calibrated. Therefore, this Guidedoes not cover specifically the concepts ofequipment qualification or instrumentqualification. Likewise the analyst carrying outthe studies must be competent in the field ofwork under study, and have sufficient knowledgerelated to the work to be able to makeappropriate decisions from the observationsmade as the study progresses.

1.2 Notes on the use of this Guide

1.2.1 Terminology

In the revision of this Guide the main focus hasbeen on updating the terminology and literaturereferences to reflect developments since theGuide was first published fifteen years ago. Withregards to terminology we have, where possible,followed the 3rd edition of the VIM first

published in 2007 [7, 8]. This has beensupplemented, where necessary, with

*

Commonly used synonyms for method performancecharacteristics are method performance parameters,metrological characteristics and performance

properties.


12/70


MV 2014 6

terminology used in ISO/IEC 17025:2005 [1],other ISO documents [9, 10, 11] and the IUPACHarmonized Guidelines for Single-LaboratoryValidation from 2002 [12] to reflect termscommonly used in analytical laboratories.

In some cases it may be difficult to decide whichterm to use when several similar terms are in use.For clarity it has been considered important touse a term consistently throughout the Guide.One example is the term used to describe thedocument that gives a detailed description of themethod to be validated using personnel andequipment in a particular laboratory. Forquantitative analysis VIM refers to themeasurement procedure, in ISO/IEC 17025 thisis the method, in ISO 15189 [2] it is theexamination procedure and many laboratoriesrefer to their standard operating procedure(SOP). The working group has decided to adhereto ISO/IEC 17025 and use the generic termmethod. Consequently, this Guide uses thecommonly recognised term method validationalthough procedure validation would be morecorrect.

The terms ruggedness and selectivity arepreferred to robustness and specificity [13]since the former are used by IUPAC [12].

Various terms, e.g. calibration, measurement,testing, analysis and examination are usedto describe laboratory work. This Guide usesanalysis in a general sense and specifies, wherenecessary, the circumstances. Similarly, this

Guide often refers to a measured concentrationalthough several other quantities are regularlyinvestigated in the chemistry laboratory [14].

In the processes of sampling, sample preparationand analysis terms such as sampling target,

primary sample, increment, compositesample, subsample, laboratory sample, testsample, test portion and test solution may beused [15, 16]. In this Guide we normally use thegeneral term sample or test sample [17].*Themost important terms used in the Guide aredefined in the text. Definitions in VIM, ISO 9000[9] and IUPAC [17, 18] have been providedwherever possible. The terms in VIM related toanalytical chemistry are further explained in theEurachem Guide Terminology in analyticalmeasurement [8]. Users should note that there isstill no universal agreement on the definition ofsome of the terms used in method validation.

1.2.2 Quick References

In Section 6, the shaded boxes provide QuickReference advice related to the specificperformance characteristic of a method.However, it is recognised that in many caseslaboratories will not have the time and resourcesto carry out experiments in the detail describedhere. Carrying out the operations described in the

boxes, using less replication than suggested, willstill yield useful information and is certainly

better than no work at all. However, theinformation provided will be less reliable than iffull replication had been utilised.

*Test sample: Sample, prepared from the laboratorysample, from which test portions are removed fortesting or for analysis [17].


13/70


MV 2014 7

2 What is method validation?

2.1 Definitions

Definitions of validationfrom three international

documents are given in Table 1. Methodvalidationis basically the process of defining ananalytical requirement, and confirming that themethod under consideration has capabilitiesconsistent with what the application requires.Inherent in this is the need to evaluate themethods performance. The judgement ofmethod suitability is important; in the pastmethod validation tended to concentrate only onevaluating the performance characteristics.

Method validation is usually considered to bevery closely tied to method development. Manyof the method performance characteristics (Table2) that are associated with method validation areusually evaluated, at least approximately, as partof method development. However, it is importantto remember that formal validation of the finalversion of the method (the documented

procedure) should be carried out.

Some sectors use the concepts of primaryvalidation and secondary validation, the latterin the sense of verification [19]. The conceptsqualification and metrological confirmation

[20] also seem to cover verification (Table 1).

2.2 What is the difference betweenvalidation and verification?

ISO 9000 [9] defines verification asconfirmation, through provision of objectiveevidence, that specified requirements have beenfulfilled. This is very similar to the definition ofvalidation in Table 1. The VIM [7] states thatverification is provision of objective evidencethat a given item fulfils specified requirementsand that validation is a verification, where thespecified requirements are adequate for anintended use.

A laboratory may adopt a validated procedurewhich, e.g. has been published as a standard, or

buy a complete measuring system to be used fora specific application from a commercialmanufacturer. In both these cases, basicvalidation work has already been carried out butthe laboratory will still need to confirm its abilityto apply the method. This is verification. Itmeans that some experimental work must bedone to demonstrate that the method works in theend-users laboratory. However, the workload islikely to be considerably less compared tovalidation of a method that has been developed

in-house.The terms validation and verification are furtherdiscussed in the Eurachem Guide on terminologyin analytical measurement [8].

Table 1 Definitions of the concept validation in ISO 9000, ISO/IEC 17025 and VIM

Definition Reference

confirmation, through the provision of objective evidence, that the requirementsfor a specific intended use or application have been fulfilled

ISO 9000 [9]a

confirmation by examination and provision of objective evidence that theparticular requirements for a specific intended use are fulfilled

ISO/IEC 17025 [1]

verification, where the specified requirements are adequate for an intended use VIM [7]

aISO 9000 defines qualification process as process to demonstrate the ability to fulfil specifiedrequirements.

bVIM defines verification as provision of objective evidence that a given item fulfils specifiedrequirements


14/70


MV 2014 8

Table 2 Overview of performance characteristics commonly

evaluated during method validation

Selectivity

Limit of detection (LOD) and limit of quantification (LOQ)Working range

Analytical sensitivity

Truenessbias, recovery

Precision repeatability, intermediate precision and reproducibility

Measurement uncertaintya

Ruggedness (robustness)

a

Strictly, measurement uncertainty is not a performance characteristic of aparticular measurement procedure but a property of the results obtainedusing that measurement procedure.


15/70


MV 2014 9

3 Why is method validation necessary?

3.1 Importance of analyticalmeasurement

Millions of tests, measurements andexaminations are made every day in thousands oflaboratories around the world. There areinnumerable reasons underpinning them, forexample: as a way of valuing goods for trade

purposes; supporting healthcare; checking thequality of drinking water, food and feed;analysing the elemental composition of an alloyto confirm its suitability for use in aircraftconstruction; forensic analysis of body fluids incriminal investigations. Virtually every aspect ofsociety is supported in some way by analytical

work.The cost of carrying out these measurements ishigh and additional costs may arise fromdecisions made on the basis of the results. Forexample, tests showing food to be unfit forconsumption may result in compensation claims;tests confirming the presence of banned drugscould result in fines, imprisonment or even, insome countries, execution. Clearly it is importantto make a correct measurement and be able toshow that the result is correct.

3.2 The professional duty of theanalytical chemist

If the result of an analysis cannot be trusted thenit has little value and the analysis might as wellhave not been carried out. When customerscommission analytical work from a laboratory, itis assumed that the laboratory has a degree ofexpert knowledge that the customers do not havethemselves. The customer expects to be able totrust results reported and usually only challengesthem when a dispute arises. Thus the laboratory

and its staff have an obvious responsibility tojustify the customers trust by providing the rightanswer to the analytical part of the problem, inother words results that have demonstrablefitness for purpose. Implicit in this is that thetests carried out are appropriate for the analytical

part of the problem that the customer wishessolved, and that the final report presents theanalytical data in such a way that the customercan readily understand it and draw appropriateconclusions. Method validation enables chemiststo demonstrate that a method is fit for purpose.

For an analytical result to be fit for its intendeduse it must be sufficiently reliable that any

decision based on it can be taken withconfidence. Thus the method performance must

be validated and the uncertainty on the result, ata given level of confidence, estimated.Uncertainty should be evaluated and quoted in away that is widely recognised, internallyconsistent and easy to interpret [21]. Most of theinformation required to evaluate uncertainty can

be obtained during validation of the method. Thistopic is dealt with briefly in Section 6.7 and inmore detail in the Eurachem/CITAC GuideQuantifying Uncertainty in AnalyticalMeasurement [22].

Regardless of how good a method is and how

skilfully it is used, an analytical problem can besolved by the analysis of samples only if thosesamples are appropriate to the problem. Takingappropriate samples is a skilled job, requiring anunderstanding of the problem and its relatedchemistry. A laboratory should, wherever

possible, offer advice to the customer on thetaking of samples as part of its customer care.Clearly there will be occasions when thelaboratory cannot themselves take or influencethe taking of the samples. On these occasionsresults of analysis will need to be reported on the

basis of the samples as received, and the reportshould make this distinction clear.

We have mostly (and rightly) focused on theoverall objective of performing methodvalidation, i.e. demonstrating that methods arefit for purpose. However, it should berecognised that a method validation study givesadditional benefits to the laboratory undertakingthe validation. It provides a solid knowledge andexperience of the practical details of performingthe method, including awareness of any critical

steps in the process. Validation gives thelaboratory and its employees a greaterconfidence in their own results.

3.3 Method development

The validation work is preceded by adevelopment phase which may involve differentstaff and which can take a number of forms.

At one extreme, it may involve adapting anexisting method by making minor changes sothat it is suitable for a new application. For

example, a method required to determine toluenein water might be adapted from an establishedmethod for benzene in water. The matrix is thesame, and the two analytes have broadly similar


16/70


MV 2014 10

properties. It is likely that the same principles ofisolation, identification, and quantification thatare applied to benzene can also be applied totoluene. If, on the other hand, a method isrequired to determine benzene in soil, adaptation

of the benzene in water method may not be thebest option. Adaptation of some other method fordetermining organics in soil may be a betterstarting point.

At the other extreme, the analytical chemist maystart out with a few sketchy ideas and applyexpertise and experience to devise a suitablemethod. This clearly involves a great deal morework and a degree of doubt as to whether the

final method will be successful. It is not unusualfor method development to involve work on anumber of different ideas simultaneously beforeeventually choosing one winner.

Regardless of how much effort has been invested

during method development, there is noguarantee the method will perform adequatelyduring validation (or under routine conditions ina particular laboratory). When different staff areinvolved in the development and validation

phase this offers the possibility of checking thatthe instructions (the measurement procedure) can

be understood and implemented.


17/70


MV 2014 11

4 When should methods be validated or verified?

4.1 Method validation

A method should be validated when it is

necessary to demonstrate that its performancecharacteristics are adequate for use for aparticular purpose. For example, it is stated inClause 5.4.5.2 of ISO/IEC 17025 [1] that thelaboratory shall validate:

non-standard methods;

laboratory-designed/developed methods;

standard methods used outside their intendedscope;

amplifications and modifications of standard

methods.Validation must be as extensive as necessary tomeet the requirements in connection with thegiven use or the given application [23]. Theextent (scale, scope) of validation will dependon the application, the nature of the changesmade, and the circumstances in which themethod is going to be used.

Validation is also required when it is necessaryto demonstrate the equivalence of resultsobtained by two methods, e.g. a newly developed

method and an existing standard/regulatorymethod.

4.2 Method verification

For standard(ised) methods, such as those

published by, e.g. ISO or ASTM, validation bythe laboratory using the method is not necessary.However, the laboratory needs to verify the

performance of the method as detailed inISO/IEC 17025 Clause 5.4.2:

The laboratory shall confirm that it canproperly operate standard methods beforeintroducing the tests or calibrations.

Verification is also required when there is animportant change such as a new but similarinstrument, relocation of equipment etc.

In laboratory medicine a majority ofmeasurements and tests are performed withcommercial procedures which have already beenvalidated by the manufacturer, but which need to

be verified by the end-user [24]. ISO 15189 [2]stresses that examination procedures usedwithout modification shall be subject toindependent verification by the laboratory beforebeing introduced into routine use. This couldalso include when an instrument is updated withnew software, or when quality control indicates

that the performance of an established method ischanging with time.


18/70


MV 2014 12


19/70


MV 2014 13

5 How should methods be validated?

5.1 Who carries out methodvalidation?

5.1.1 Approaches to methodvalidation

Once the initial method development is finished,the laboratory should document the measurement

procedure in detail (see Annex A). It is thisdocumented procedure that is taken forward forthe formal validation.

There are two main approaches to methodvalidation; the interlaboratory comparisonapproach and the single-laboratory approach.Regardless of the approach, it is the laboratory

using a method which is responsible for ensuringthat it is fit for the intended use and, if necessary,for carrying out further work to supplementexisting validation data.

5.1.2 Interlaboratory approach

Much has been published in the literatureconcerning method validation by dedicatedinterlaboratory comparisons often referred to ascollaborative studies or cooperative studies.There are a number of protocols relating to thistype of validation [25, 26, 27, 28], as well as the

ISO 5725 standards [29] which can be regardedas the most generally applicable. If a method isbeing developed which will have wide-ranginguse, perhaps as a published standardised

procedure, then a collaborative study involving agroup of laboratories is probably the preferredway of carrying out the validation. A publishedmethod validated in this way is demonstrated to

be robust. Published information normallycontains precision (repeatability, reproducibilityand/or corresponding precision limits) and,sometimes, bias estimates. Where a method has

been validated by a standards approvingorganisation, such as ISO, CEN or AOACInternational, the user will normally need only toverify published performance data and/orestablish performance data for their own use ofthe method. This approach, therefore, reduces theworkload for the laboratory using the method.

5.1.3 Single-laboratory approach

Laboratories will from time to time find that amethod is needed but not available as a publishedstandard. If the method is developed for use in

one laboratory, for example because there is nogeneral interest in the method or because other

laboratories are competitors, the single-laboratory approach is appropriate [12].

Whether or not methods validated in a singlelaboratory will be acceptable for regulatory

purposes depends on any guidelines covering thearea of measurement concerned. It shouldnormally be possible to get a clear policystatement from the appropriate regulatory body.

5.2 Extent of validation studies

The laboratory has to decide which performancecharacteristics (see Table 2 and Section 6) needto be investigated in order to validate the methodand, in some cases, how detailed theinvestigation of a single performancecharacteristic should be. The IUPAC protocol[12] lists a number of situations, which takes intoaccount, among other things, the status of themethod and the competence of the laboratory.

Where the scope of the analytical work is welldefined and applications are similar over time, itmay be possible for an organisation or sector toissue general guidelines for the extent ofvalidation studies. An example from the

pharmaceutical sector is shown in Table 3.

Starting with a carefully considered analyticalspecification given in the scope of thedocumented procedure (see A.5 in Annex A)

provides a good base on which to plan thevalidation process, but it is recognised that in

practice this is not always possible. Theassessment of method performance may beconstrained. This is acknowledged in ISO/IEC17025, clause 5.4.5.3 as Validation is always abalance between costs, risks and technicalpossibilities. The laboratory should do its bestwithin the constraints imposed, taking into

account customer and regulatory requirements,existing experience of the method, availabletools (Section 5.4), and the need for metrologicalcompatibility [7] with other similar methodsalready in use within the laboratory or used byother laboratories. Some performancecharacteristics may have been determinedapproximately during the method developmentor method implementation stage. Often a

particular set of experiments will yieldinformation on several performancecharacteristics, so with careful planning the

effort required to get the necessary informationcan be minimised.


20/70


MV 2014 14

Table 3 Extent of validation work for four types of analytical applications. Example from the

pharmaceutical sector [13]. x signifies a performance characteristic which is normally validated.

Performance characteristic

Type of analytical application

Identificationtest

Quantitative testfor impurity

Limit testfor impurity

Quantification ofmain component

Selectivity x x x x

Limit of detection x

Limit of quantification x

Working range includinglinearity

x x

Trueness (bias) x x

Precision (repeatability andintermediate precision)

x x

NOTE The table is simplified and has been adapted to the structure and terminology used in this Guide.

The implications of the constraints discussedabove are particularly critical where the methodis not going to be used on a routine basis. The

process of validating methods which are going tobe used on a routine basis is comparatively well-defined. Clearly the same principles apply for adhoc analysis as for routine testing. It is necessaryto have an adequate level of confidence in theresults produced. Establishing the balance

between time and cost constraints and the need tovalidate the method is difficult. In somecircumstances it may be more appropriate tosubcontract the analyses to another laboratorywhere they can be performed on a routine basis.

5.3 Validation plan and report

The validation work shall be performed, and theresults reported, according to a documented

procedure.

The outline of a validation plan (validationprotocol) and validation report may be stated in

sectoral guidelines (see Section 5.5). Nationalaccreditation bodies may point to minimumrequirements for this documentation [23].However, a simple template for a combinedvalidation plan and validation report could, e.g.consist of the following sections.

Title: This section should identify the methodand when and who is performing the work.Brief information about the method scope anda short description of the method should begiven, as well as details of the status of the

method (e.g. an international standard, a

method developed in-house etc.), the analyte,measurand, measurement unit, types ofsample and the intended use. Sampling andsubsampling can be part of the measurement

procedure and must, in those cases, bevalidated. Even if these steps are performedelsewhere, it is useful to include informationabout them in the validation plan/report.

Planning: This section should outline thepurpose, e.g. full validation of a new method,verification of performance of a standardisedmethod, extension to method scope, etc. Theextent of the validation work should beindicated, i.e. the performance characteristicswhich will be investigated and any associatedrequirements.

Performance characteristics: This sectionshould give a brief explanation of the

performance characteristic, repeat anyspecific requirements, outline the experimentswhich will be done and how the results are to

be evaluated. Results and conclusions fromthe experiments should be stated. Separatesections are used for each performancecharacteristic.

Summary: The last section should summarisethe validation work and the results.Implications concerning routine use, andinternal and external quality control, can begiven. Most importantly, a concludingstatement as to whether the method is fit for

purpose shall be given. Note that this is arequirement in ISO/IEC 17025 [1].


21/70


MV 2014 15

5.4 Validation tools

5.4.1 Blanks

Use of various types of blanks enablesassessment of how much of the measured signalis attributable to the analyte and how much toother causes. Various types of blank are availableto the analyst:

Reagent blanks*: Reagents used during the

analytical process (including solvents used forextraction or dissolution) are analysed inorder to determine whether they contribute tothe measurement signal.

Sample blanks. These are essentially samplematrices with no analyte present, e.g. a humanurine sample without a specific drug of abuse,or a sample of meat without hormoneresidues. Sample blanks may be difficult toobtain but such materials are necessary togive a realistic estimate of interferences thatwould be encountered in the analysis of testsamples.

5.4.2 Routine test samples

Routine test samples are useful because of theinformation they provide on precision,interferences etc. which could be realisticallyencountered in day-to-day work. If the analytecontent of a test material is accurately known, itcan be used to assess measurement bias. Anaccurate assessment of analyte content can beobtained using a reference method, althoughsuch methods are not always available.

5.4.3 Spiked materials/solutions

These are materials or solutions to which theanalyte(s) of interest have been deliberatelyadded. These materials or solutions may alreadycontain the analyte of interest so care is neededto ensure the spiking does not lead to analytelevels outside of the working range of themethod. Spiking with a known amount of analyteenables the increase in response to the analyte to

be measured and calculated in terms of theamount added, even though the absolute amountsof analyte present before and after addition of thespike are not known. Note that most methods ofspiking add the analyte in such a way that it willnot be as closely bound to the sample matrix as itwould be if it was present naturally. Therefore,

bias estimates obtained by spiking can beexpected to be over-optimistic.

*A reagent blank taken through the entire analyticalprocedure is sometimes called a procedural blank.

Spiking does not necessarily have to be restrictedto the analyte of interest. It could includeanything added to the sample in order to gaugethe effect of the addition. For example, thesample could be spiked with varying amounts of

a particular interference in order to judge theconcentration of the interferent at whichdetermination of the analyte is adverselyaffected. The nature of the spike obviously needsto be identified.

5.4.4 Incurred materials

These are materials in which the analyte ofinterest may be essentially alien, but has beenintroduced to the bulk at some point prior to thematerial being sampled. The analyte is thus moreclosely bound in the matrix than it would be had

it been added by spiking. The analyte value willdepend on the amounts of analyte in contact withthe material, the rate of take-up and loss by thematrix and any other losses through metabolism,spontaneous disintegration or other chemical or

physical processes. The usefulness of incurredsamples for validation purposes depends on howwell the analyte value can be characterised. Thefollowing are examples of incurred materials:

1. Herbicides in flour from cereal sprayed withherbicides during its growth;

2. Active ingredients in pharmaceuticalformulations added at the formulation stage.

3. Egg-white powder (known protein content)added to a cookie dough before baking wheninvestigating allergens.

5.4.5 Measurement standards

Care must be taken when referring to standardsas the term also applies to written documents,such as ISO standards. Where the term refers tosubstances used for calibration or identification

purposes it is convenient to refer to them asmeasurement standards or calibrants/calibrators[7]. These are traditionally thought of assolutions of single substances but in practice can

be anything in which a particular parameter orproperty has been characterised to the extent itcan serve as a metrological reference.

It is important to distinguish between referencematerials (RMs) and certified reference materials(CRMs) [7, 30] because of the significantdifference in how they can be used in the methodvalidation process (6.5.2). RMs can be virtuallyany material used as a basis for reference, and

could include laboratory reagents of knownpurity, industrial chemicals, or other artefacts.The property or analyte of interest needs to be


22/70


MV 2014 16

stable and homogenous but the material does notneed to have the high degree of characterisation,metrological traceability, uncertainty anddocumentation associated with CRMs.

The characterisation of the parameter of interest

in a CRM is generally more strictly controlledthan for an RM, and in addition the characterisedvalue is certified with a documented metrologicaltraceability and uncertainty. Characterisation isnormally done using several different methods,or a single primary measurement procedure, sothat as far as possible, any bias in thecharacterisation is reduced or even eliminated.

Assessment of bias requires a reliable referencepoint, preferably, a CRM with the same matrixand analyte concentrations as the test samples.

5.4.6 Statistics

Statistical methods are essential for summarisingdata and for making objective judgements ondifferences between sets of data (significancetesting). Analysts should familiarise themselveswith at least the more basic elements of statisticaltheory particularly as an aid to evaluation of

precision, bias, linear range, LOD, LOQ andmeasurement uncertainty. A number of useful

books introducing statistics for analyticalchemistry are referenced [5, 6, 31, 32, 33, 34].

5.5 Validation requirements

Requirements for how to carry out methodvalidation may be specified in guidelines withina particular sector relevant to the method [13, 25,35 for example]. Where such requirements exist,it is recommended they are followed. This willensure that particular validation terminology,together with the statistics used, is interpreted ina manner consistent within the relevant sector.Official recognition of a method may requirecharacterisation using a collaborative study.

5.6 Method validation process

Faced with a particular customer problem, thelaboratory must first set the analyticalrequirement which defines the performancecharacteristics that a method must have to solvethat problem (Figure 1).

In response to these requirements, the laboratoryneeds to identify a suitable existing method, or ifnecessary develop/modify a method. Note that

certain regulations may require a particularmethod to be followed. Table 4 shows the type ofquestions which might be posed in formalisingan analytical requirement (column 1) and thecorresponding performance characteristics of the

method which may need to be evaluated (column2). The laboratory will then identify and evaluaterelevant performance characteristics and checkthem against the analytical requirement. Thevalidation process ends with a conclusion andstatement of whether or not the analyticalrequirement is met. If the analytical requirementis not met, further method development isnecessary. This process of development andevaluation continues until the method is deemedcapable of meeting the requirement.

In reality an analytical requirement is rarelyagreed with the customer beforehand in such aformal way. Customers usually define theirrequirements in terms of cost and/or time andrarely know how well methods need to perform,although performance requirements for methodsmay be specified where the methods support aregulatory requirement or compliance with aspecification. For example, the European Union(EU) have published requirements, e.g. for theanalysis of drinking water [36], for analyses

performed within the water framework directive

[37], for the determination of the levels ofveterinary drug residues in food of animal origin[38] and of pesticide residues in food and feed[39].

However, it will usually be left to the analystsdiscretion to decide what performance isrequired. Very often this will mean setting ananalytical requirement in line with the methodsknown capability (e.g. as published instandardised methods, as observed in proficiencytesting (PT) schemes or estimated frommathematical models, such as the Horwitzfunction [40]).

Financial constraints may dictate thatdevelopment of a method that satisfies a

particular analytical requirement is noteconomically feasible, in which case the decisionmust be taken whether to relax the requirementto a more achievable level or rethink the

justification for the analysis.


23/70


MV 2014 17

Figure 1 The method validation process: from the customer problem to the laboratory decision on

whether or not the customer request can be carried out with an identified method. Note: method

validation consists of a stage where performance characteristics are evaluated and then compared with

analytical requirements. Regardless of what existing performance data may be available for the method,

fitness for purpose will be determined by how the method performs when used by the designated analyst

with the available equipment/facilities.

Identify/modifyexisting method or

develop new method

Develop methodfurther

Relaxanalytical

requirement?

Furtherdevelopment

feasible?

Issue validationreport

Is methodfit for purpose?

Evaluatemethod

performance

YES

YES

NO NO

NOYES

Customer problemto be solved.Set analyticalrequirement

Restate analyticalrequirements

Use method Unable touse method


24/70


MV 2014 18

Table 4 Questions which might be posed in formalising an analytical requirement, and related

performance characteristics with references to the appropriate sections in this Guide

Question Performance characteristic Section Note

Do resource constraints apply and how people,time, money, equipment and reagents, laboratory

facilities?

- - a)

Is sampling and subsampling required (and will thisbe done within the laboratory)?Are there any restrictions on samplesize/availability?What is the chemical, biological and physical natureof the matrix?Is the analyte dispersed or localised?

Is a qualitative or quantitative answer required? SelectivityLOD and LOQ

6.16.2

What are the analytes of interest and the likely levelspresent (%, g/g, ng/g, etc.....)? Are the analytes

present in more than one chemical form (e.g.oxidation states, stereoisomers), and is it necessary to

be able to distinguish between different forms?

SelectivityLOD and LOQ

Working and linear ranges

6.16.2

6.3

What quantity is intended to be measured (themeasurand)? Is it the total concentration of theanalyte present that is of interest, or the amountextracted under specified conditions?

Recovery 6.5

What trueness and precision are required? What isthe target uncertainty and how is it to be expressed?

Trueness and recovery 6.5

b)Repeatability, intermediate

precision, reproducibility6.6

Uncertainty 6.7What are the likely interferences to the analyte(s)? Selectivity 6.1

Have tolerance limits been established for allparameters, critical for performing the analysis (e.g.time of extraction, incubation temperature)?

Ruggedness 6.8 c)

Do results need to be compared with results fromother laboratories?

Uncertainty 6.7b)

Do results need to be compared with externalspecifications?

Uncertainty 6.7b)

a) Not all of the elements of the analytical requirement link directly to method validation requirements butdictate more generally as to whether particular techniques are applicable. For example, different techniqueswill be applicable according to whether the analyte is dispersed through the sample or isolated on thesurface.

b) One essential element of the analytical requirement is that it should be possible to judge whether or not amethod is suitable for its intended purpose and thus must include the required uncertainty expressed either as

a standard uncertainty or an expanded uncertainty.c) Published standardised procedures have normally been shown to be rugged within the scope of the

procedure, i.e. matrix types and working range. Therefore single-laboratory verification for implementationof a published standardised procedure need not normally include ruggedness.


25/70


MV 2014 19

6 Method performance characteristics

6.1 Selectivity

6.1.1 Terms and definitions

Analytical selectivity relates to the extent towhich the method can be used to determineparticular analytes in mixtures or matriceswithout interferences from other components ofsimilar behaviour[41].

Definitions in various documents [7, 18, 42]more or less agree with this interpretation. WhileIUPAC recommends the term selectivity, someareas, e.g. the pharmaceutical sector [13], usespecificity or analytical specificity. The latteris recommended to avoid confusion with

diagnostic specificity as used in laboratorymedicine [43].

6.1.2 Effects of interferences

In general, analytical methods can be said toconsist of a measurement stage which may ormay not be preceded by an isolation stage. In themeasurement stage, the concentration of ananalyte is normally not measured directly.Instead a specific property (e.g. intensity of light)is quantified. It is, therefore, crucial to establishthat the measured property is only due to the

analyte and not to something chemically orphysically similar, or arising as a coincidencethus causing a bias in the measurement result.The measurement stage may need to be preceded

by an isolation stage in order to improve theselectivity of the measuring system.

Interferences may cause a bias by increasing ordecreasing the signal attributed to the measurand.The size of the effect for a given matrix isusually proportional to the signal and is thereforesometimes called a proportional effect. It

changes the slope of the calibration function, butnot its intercept. This effect is also calledrotational [44].

A translational or fixed effect arises from asignal produced by interferences present in thetest solution. It is therefore independent of theconcentration of the analyte. It is often referredto as a background or baseline interference. Itaffects the intercept of a calibration function, butnot its slope.

It is not unusual for both proportional and

translational effects to be present simultaneously.The method of standard additions can onlycorrect for proportional effects.

6.1.3 Assessment of selectivity

The selectivity of a procedure must be

established for in-house developed methods,methods adapted from the scientific literatureand methods published by standardisation bodiesused outside the scope specified in the standardmethod. When methods published bystandardisation bodies are used within theirscope, selectivity will usually have been studiedas part of the standardisation process.

The selectivity of a method is usuallyinvestigated by studying its ability to measurethe analyte of interest in samples to whichspecific interferences have been deliberatelyintroduced (those thought likely to be present insamples). Where it is unclear whether or notinterferences are already present, the selectivityof the method can be investigated by studying itsability to measure the analyte compared to otherindependent methods. Example 1 and Example 2

below and Quick Reference 1 illustrate thepractical considerations regarding selectivity.

Confirmatory techniques can be useful as ameans of verifying identities. The more evidenceone can gather, the better. Inevitably there is a

trade-off between costs and time taken foranalyte identification, and the confidence withwhich one can decide if the identification has

been made correctly.

Whereas evaluation of repeatability requires themeasurement to be repeated several times by onetechnique, confirmation of analyte identityrequires the measurement to be performed byseveral, preferably independent, techniques.Confirmation increases confidence in thetechnique under examination and is especially

useful when the confirmatory techniques operateon significantly different principles. In someapplications, for example, the analysis ofunknown organics by gas chromatography, theuse of confirmatory techniques is essential.When the measurement method being evaluatedis highly selective, the use of other confirmatorytechniques may not be necessary.

An important aspect of selectivity which must beconsidered is where an analyte may exist in thesample in more than one form such as: bound or

unbound; inorganic or organometallic; ordifferent oxidation states. The definition of themeasurand is hence critical to avoid confusion.


26/70


MV 2014 20

Example 1 Chromatography. A peak in achromatographic trace may be identified as being dueto the analyte of interest on the basis that an RMcontaining the analyte generates a signal at the same

point on the chromatogram. But, is the signal due tothe analyte or to something else which coincidentallyco-elutes, i.e. a fixed effect? It could be either or

both. Identification of the analyte, by this meansonly, is unreliable and some form of supportingevidence is necessary. For example, thechromatography could be repeated using a column ofdifferent polarity, employing a different separation

principle to establish whether the signal and thesignal generated by the RM still appear at the sametime. Where a peak is due to more than onecompound, a different polarity column may be agood way of separating the compounds. In many

cases modern mass spectrometric instruments canoffer a high selectivity, e.g. gas or liquidchromatography with mass spectrometric detection.

Example 2 Spectroscopy. In infraredspectroscopy, identification of unknowncompounds may be made by matching absorbancesignals (i.e. peaks) in the analyte spectrum with

those of reference spectra stored in a spectrallibrary. Once it is believed the correctidentification has been made, a spectrum of an RMof the analyte should be recorded under exactly thesame conditions as for the test portion. The largerthe number of peaks which match between analyteand RM, the better the confidence that can be

placed on the identification being correct. It wouldalso be worthwhile examining how dependant theshape of the spectrum was with respect to how theanalyte was isolated and prepared for infraredanalysis. For example, if the spectrum wasrecorded as a salt disc, the particle size distribution

of the test portion in the disc might influence theshape of the spectrum.

Quick Reference 1 Selectivity

What to doHow many

times

What to calculate/determine from the

dataComments

Analyse testsamples, and RMs

by candidate andother independentmethods.

1 Use the results from the confirmatorytechniques to assess the ability of themethod to confirm analyte identity andits ability to measure the analyte inisolation from other interferences.

Decide how much supportingevidence is reasonablyrequired to give sufficientreliability.

Analyse test samplescontaining varioussuspectedinterferences in the

presence of theanalytes of interest.

1 Examine effect of interferences. Doesthe presence of the interferent inhibitdetection or quantification of theanalytes?

If detection or quantificationis inhibited by theinterferences, further methoddevelopment will berequired.

6.2 Limit of detection and limit of

quantification6.2.1 Terms and definitions

Where measurements are made at lowconcentrations, there are three general conceptsto consider. First, it may be necessary toestablish a value of the result which is consideredto indicate an analyte level that is significantlydifferent from zero. Often some action isrequired at this level, such as declaring a materialcontaminated. This level is known as the criticalvalue, decision limit or, in EU directives, CC

[38].Second, it is important to know the lowestconcentration of the analyte that can be detected

by the method at a specified level of confidence.That is, at what true concentration will weconfidently exceed the critical value describedabove? Terms such as limit of detection (LOD),minimum detectable value, detection limit, or,in EU directives, CC [38] are used for thisconcept.

Third, it is also important to establish the lowestlevel at which the performance is acceptable fora typical application. This third concept isusually referred to as the limit of quantification(LOQ)*.

* Synonyms used include 'quantification limit,quantitation limit, limit of quantitation, limit of


27/70


MV 2014 21

Terminology relating to all these concepts is verydiverse and varies between sectors. For example,the terms limit of detection (LOD) or detectionlimit (DL) were previously not generallyaccepted, although used in some sectoral

documents [13, 38]. However, they are nowincorporated into the VIM [7] and IUPAC GoldBook [17]. ISO uses as a general term minimumdetectable value of the net state variable whichfor chemistry translates as minimum detectablenet concentration [45, 46, 47, 48]. In this Guidethe terms critical value, limit of detection(LOD) and limit of quantification (LOQ) areused for the three concepts above. In methodvalidation, it is the LOD and LOQ that are mostcommonly determined.

It is also necessary to distinguish between theinstrument detection limit and the methoddetection limit. The instrument detection limitcan be based on the analysis of a sample, often areagent blank, presented directly to theinstrument (i.e. omitting any sample preparationsteps), or on the signal-to-noise ratio in, e.g. achromatogram. To obtain a method detectionlimit, the LOD must be based on the analysis ofsamples that have been taken through the wholemeasurement procedure using results calculatedwith the same equation as for the test samples. It

is the method detection limit that is most usefulfor method validation and is therefore the focusof this Guide.

The following paragraphs describe theexperimental estimation of LOD and LOQ. Thestatistical basis for the calculation of the LOD isgiven in Annex B. Because the LOD and LOQ

both depend on the precision at or near zero,Section 6.2.2 first describes the experimentalestimation of the standard deviation of resultsnear zero.

6.2.2 Determination of the standarddeviation at low levels

Both LOD and LOQ are normally calculated bymultiplying a standard deviation ( ) by asuitable factor. It is important that this standarddeviation is representative of the precisionobtained for typical test samples, and thatsufficient replicate measurements are made togive a reliable estimate. In this section, thestandard deviation is based on a standarddeviations0for single results near zero, adjustedfor any averaging or blank correction used in

determination, reporting limit, limit of reportingand application limit.

practice (see below). Alternative approaches arediscussed in Section 6.2.5

The following issues should be considered indetermining LOD and LOQ from an experimentusing simple replication.

Suitable samples for estimating LOD and

LOQ: The samples used should preferably beeither a) blank samples, i.e. matrices containingno detectable analyte, or b) test samples withconcentrations of analyte close to or below theexpected LOD. Blank samples work well formethods where a measurable signal is obtainedfor a blank, such as spectrophotometry andatomic spectroscopy. However for techniquessuch as chromatography, which rely on detectinga peak above the noise, samples with

concentration levels close to or above the LODare required. These can be prepared by, forexample, spiking a blank sample (see Section5.4).

When blank samples or test samples at lowconcentrations are not available, reagent blanks*can often be used. When these reagent blanks donot go through the whole measurement

procedure, and are presented directly to theinstrument, the calculation based on thesemeasurements will give the instrument

LOQ/LOD.Covering the scope of the method: Formethods with a scope covering very differentmatrices it may be necessary to determine thestandard deviation for each matrix separately.

Ensuring representative replication: Thestandard deviation should be representative ofthe performance of the method as used in thelaboratory, i.e. the standard deviation is to becalculated based on test results where analysesare performed exactly according to the whole

documented measurement procedure, includingany sample preparation steps. The values usedfor calculating the standard deviation should

be in the measurement units specified in theprocedure.

Conditions of measurement: The standarddeviation is normally obtained underrepeatability conditions and this is the proceduredescribed in this section. However, a morereliable estimate can be obtained from the use of

* There is confusion regarding the terminologyrelating to blanks for further discussion see Section5.4.1.


28/70


MV 2014 22

intermediate precision conditions. This approachis discussed further in Section 6.2.5.

Number of observations: The number ofreplicates (m) should be sufficient to obtain anadequate estimate of the standard deviation.

Typically between 6 and 15 replicates areconsidered necessary; 10 replicates are oftenrecommended in validation procedures/protocols(see Section 6.2.5.1).

Allowing for averaging: In many measurementprocedures the mean of replicates is reported inroutine use of the method, where each replicateis obtained by following the entire measurement

procedure. In this case the standard deviation ofsingle results s0should be corrected by dividingwith the square root of n, where nis the number

of replicates averaged in routine use.Allowing for the effect of blank corrections: If

blank corrections are specified in themeasurement procedure, care needs to be takenwhen determining the standard deviation used tocalculate the LOD or LOQ. If the resultsobtained during the validation study were allcorrected by the same blank value the approachrecommended here for simplicity the standarddeviation of the results will be smaller than thatseen in practice when results are corrected by

different blank values obtained in different runs.In this cases0should be corrected by multiplying

by + where n is the number of replicateobservations averaged when reporting resultswhere each replicate is obtained following theentire measurement procedure, and nb is thenumber of blank observations used to calculatethe blank correction.

Note that under intermediate precision conditionsresults will be corrected by different blank values

so no correction of the standard deviation isnecessary (see Section 6.2.5).

Example 3 illustrates these calculations and theflow chart in Figure 2 summarises thecorrections required for averaging and blank

correction.

Example 3 A validation exercise is based on theanalysis of a sample blank. Ten (m) independentmeasurements of the sample blank are made underrepeatability conditions. The results have a meanvalue of 2 mg/kg and a standard deviation s0 of 1mg/kg.

Case 1 The measurement procedure states thattest samples should be measured once (n=1) andthe results corrected by the result for a singlesample blank sample (nb=1). In a series of

measurements each run consists of singlereplicates of routine samples and one (nb) blanksample. The standard deviation for calculatingLOD/LOQ is then, according to Figure 2 equalto:

= + = 1 + = 12 =1.4 mg/kg

Case 2 The measurement procedure states thattest samples should be analysed in duplicate(n=2) and also that the blank sample should beanalysed in duplicate. In a series of

measurements each run consists of duplicates(n=2) of routine samples and two (nb) blanksamples. The concentration obtained for routinesamples is corrected by subtracting the meanvalue of the two blank samples. The standarddeviation for calculating LOD/LOQ is then,according to Figure 2 equal to:

= + = 1 + =1 mg/kg


29/70


MV 2014 23

0 is the estimated standard deviation of msingle results at or near zero concentration. is the standard deviation used for calculating LOD and LOQ.n is the number of replicate observations averaged when reporting results where each replicate is obtained

following the entire measurement procedure.

is the number of blank observations averaged when calculating the blank correction according to themeasurement procedure.

Figure 2 Calculation of the standard deviation, to be used for estimation of LOD and LOQ. The flowchart starts with an experimental standard deviation, s0 calculated from the results of replicate

measurements under repeatability conditions on a sample near zero concentration, either without blank

correction or with a blank correction applied to all results as specified by the method. This blank

correction may be based on a single blank observation or on a mean of several blank observations.

NO

YES

From results ofm replicate

measurements during validation

calculate the standard deviation,

Use the calculated standard deviation,

, for calculating the LOD and LOQ

Will results be blank

corrected during

routine use of the

method?


30/70


MV 2014 24

6.2.3 Estimating LOD

For validation purposes it is normally sufficientto provide an approximate value for the LOD,i.e. the level at which detection of the analyte

becomes problematic. For this purpose the 3s

approach shown in Quick Reference 2 willusually suffice.

Where the work is in support of regulatory orspecification compliance, a more exact approachis required, in particular taking into account thedegrees of freedom associated with s0. This isdescribed in detail by IUPAC [49] and others[50, 51]. Where the critical value and/or LOD areused for making decisions, the precision should

be monitored and the limits may need to berecalculated from time to time. Different sectors

and/or regulations may use different approachesto LOD estimation. It is recommended that theconvention used is stated when quoting adetection limit. In the absence of any sectoralguidance on LOD estimation, the

approaches given in the Quick Reference 2 canbe used as a general guidance.

6.2.4 Estimating LOQ

The LOQ is the lowest level of analyte that canbe determined with acceptable performance.(Acceptable performance is variouslyconsidered by different guidelines to include

precision, precision and trueness, ormeasurement uncertainty [52]. In practice,however, LOQ is calculated by most conventionsto be the analyte concentration corresponding tothe obtained standard deviation ( ) at low levelsmultiplied by a factor, kQ. The IUPAC defaultvalue for kQ is 10 [49] and if the standarddeviation is approximately constant at lowconcentrations this multiplier corresponds to a

relative standard deviation (RSD) of 10 %.Multipliers of 5 and 6 have also sometimes beenused which corresponds to RSD values of 20 %and 17 % respectively [53, 54]. See furtherReference [8] and Quick Reference 3.

Quick Reference 2 Limit of detection (LOD)

What to do

How

many

times

What to calculate from the

dataComments

a) Replicate measurements ofblank samples,i.e. matricescontaining no detectableanalyte.

or

Replicate measurements oftest samples with lowconcentrations of analyte.

10 Calculate the standard

deviation, of the results.Calculate from following the flow chart inFigure 2.

Calculate LOD asLOD = 3 .

b) Replicate measurements ofreagent blanks.

or

Replicate measurements ofreagent blanks spiked withlow concentrations ofanalyte.

10 Calculate the standarddeviation,s0of the results.

Calculate

froms0

following the flow chart inFigure 2.

Calculate LOD asLOD = 3 .

Approach b) is acceptable, when itis not possible to obtain blanksamples or test samples at low

concentrations.When these reagent blanks are nottaken through the wholemeasurement procedure, and are

presented directly to the instrument,the calculation will give theinstrument LOD.

NOTES

1) For some analytical techniques, e.g. chromatography, a test sample containing too low a concentration ora reagent blank might need to be spiked in order to get a non-zero standard deviation.

2) The entire measurement procedure should be repeated for each determination.3) The standard deviation is expressed in concentration units. When the standard deviation is expressed in

signal domain the LOD is the concentration corresponding to the blank signal + 3 . A shortexample of LOD calculations in the signal domain is given also in Reference [5].


31/70


MV 2014 25

Quick Reference 3 Limit of quantification (LOQ)

What to do

How

many

times

What to calculate from the

dataComments

a) Replicate measurements

of blank samples, i.e.matrices containing nodetectable analyte.

or

Replicate measurementsof test samples with lowconcentrations of analyte.

10 Calculate the standard

deviation,s0of the results.Calculate froms0followingthe flow chart in Figure 2.

Calculate LOQ asLOQ = kQ .

The value for the multiplier kQis

usually 10, but other values such as 5or 6 are commonly used (based onfitness for purpose criteria).

b) Replicatemeasurements of reagent

blanks.

or

Replicate measurementsof reagent blanks spikedwith low concentrations ofanalyte.

10 Calculate the standarddeviation,s0of the results.

Calculate froms0followingthe flow chart in Figure 2.

Calculate LOQ asLOQ = kQ .

Approach b) is acceptable, when it isnot possible to obtain blank samplesor test samples at low concentrations.

When these reagent blanks are nottaken through the whole measurement

procedure and are presented directlyto the instrument the calculation willgive the instrument LOQ.

NOTES

1) For some analytical techniques, e.g. chromatography, a test sample containing too low a concentrationor a reagent blank might need to be spiked in order to get a non-zero standard deviation.

2) The entire measurement procedure should be repeated for each determination.3) The standard devia

Documents

Eurachem - The Fitness for Purpose of Analytical Methods