INTERNATIONAL ISO

STANDARD 29201

First edition
2012-01-15

Water quality — The variability of

test results and the uncertainty of
measurement of microbiological
enumeration methods
Qualité de l’eau - Variabilité des résultats d’essais et incertitude de
mesure des méthodes d’énumération microbienne

iTeh STANDARD PREVIEW

(standards.iteh.ai)
ISO 29201:2012
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012

Reference number
ISO 29201:2012(E)

© ISO 2012
ISO 29201:2012(E)

iTeh STANDARD PREVIEW

(standards.iteh.ai)
ISO 29201:2012
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012

COPYRIGHT PROTECTED DOCUMENT

©  ISO 2012
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or ISO’s
member body in the country of the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail [email protected]
Web www.iso.org
Published in Switzerland

ii  © ISO 2012 – All rights reserved

 ISO 29201:2012(E)

Contents Page

Foreword.............................................................................................................................................................................. v
Introduction........................................................................................................................................................................ vi
1 Scope....................................................................................................................................................................... 1
2 Key concepts......................................................................................................................................................... 1
2.1 Uncertainty of measurement............................................................................................................................. 1
2.2 Estimation of the uncertainty of measurement............................................................................................ 1
2.3 Intralaboratory reproducibility.......................................................................................................................... 2
2.4 Combined standard uncertainty....................................................................................................................... 2
2.5 Relative standard uncertainty........................................................................................................................... 2
2.6 Relative variance.................................................................................................................................................. 3
2.7 Expanded uncertainty and expanded relative uncertainty....................................................................... 3
3 Microbiological methods.................................................................................................................................... 4
3.1 Common basis...................................................................................................................................................... 4
3.2 Quantitative instruments.................................................................................................................................... 4
3.3 Uncertainty structure.......................................................................................................................................... 4
3.4 Expression of combined uncertainty.............................................................................................................. 4
4 Choices of approach........................................................................................................................................... 5
4.1 General.................................................................................................................................................................... 5
4.2 Choices of evaluation approach...................................................................................................................... 6
4.3 Choices of expression and use of measurement uncertainty................................................................. 7
iTeh STANDARD PREVIEW
5 The component approach to the evaluation of operational uncertainty............................................... 7
(standards.iteh.ai)
5.1 General.................................................................................................................................................................... 7
5.2 Identification of the components of uncertainty.......................................................................................... 7
5.3 Evaluation...............................................................................................................................................................
ISO 29201:2012 7
6 https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
The global approach to the determination of the operational uncertainty........................................... 8
0b55923db79c/iso-29201-2012
6.1 General.................................................................................................................................................................... 8
6.2 Evaluation............................................................................................................................................................... 9
7 Combined uncertainty of the test result...................................................................................................... 10
7.1 Basic principle.................................................................................................................................................... 10
7.2 Operational variability....................................................................................................................................... 10
7.3 Intrinsic variability.............................................................................................................................................. 10
7.4 Combined uncertainty....................................................................................................................................... 10
7.5 Borderline cases................................................................................................................................................. 10
Annex A (informative) Symbols and definitions........................................................................................................ 11
Annex B (normative) General principles for combining components of uncertainty..................................... 13
Annex C (normative) Intrinsic variability — Relative distribution uncertainty of colony counts................ 18
Annex D (normative) Intrinsic variability of most probable number estimates................................................ 20
Annex E (normative) Intrinsic variability (standard uncertainty) of confirmed counts.................................. 23
Annex F (normative) Global approach for determining the operational and combined uncertainties...... 26
Annex G (normative) Component approach to evaluation of the combined relative uncertainty under
intralaboratory reproducibility conditions.................................................................................................. 31
Annex H (normative) Experimental evaluation of subsampling variance.......................................................... 35
Annex I (normative) Relative repeatability and intralaboratory reproducibility of
volume measurements...................................................................................................................................... 38
Annex J (normative) Relative uncertainty of a sum of test portions................................................................... 40
Annex K (normative) Relative uncertainty of dilution factor F ............................................................................. 44

© ISO 2012 – All rights reserved  iii

ISO 29201:2012(E)

Annex L (normative) Repeatability and intralaboratory reproducibility of counting...................................... 46

Annex M (normative) Incubation effects — Uncertainty due to position and time.......................................... 50
Annex N (informative) Expression and use of measurement uncertainty......................................................... 55
Bibliography...................................................................................................................................................................... 61

iTeh STANDARD PREVIEW

(standards.iteh.ai)
ISO 29201:2012
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012

iv  © ISO 2012 – All rights reserved

 ISO 29201:2012(E)

Foreword

ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the International
Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.

International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.

The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.

Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.

ISO  29201 was prepared by Technical Committee ISO/TC  147, Water quality, Subcommittee SC  4,
Microbiological methods.

iTeh STANDARD PREVIEW

(standards.iteh.ai)
ISO 29201:2012
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012

© ISO 2012 – All rights reserved  v

ISO 29201:2012(E)

Introduction

Testing laboratories are required to apply procedures for estimating uncertainty of measurement (see
ISO/IEC  17025[5]). Without such an indication, measurement results cannot be compared, either among
themselves or with reference values (see ISO/IEC Guide 98‑3:2008[7]).

General guidelines for the evaluation and expression of uncertainty in measurement have been elaborated by
experts in physical and chemical metrology, and published by ISO and IEC in ISO/IEC  Guide  98‑3:2008.[7]
However, ISO/IEC Guide 98‑3:2008[7] does not address measurements in which the observed values are counts.

The emphasis in ISO/IEC Guide 98‑3:2008[7] is on the “law of propagation of uncertainty” principle, whereby
combined estimates of the uncertainty of the final result are built up from separate components evaluated by
whatever means are practical. This principle is referred to as the “component approach” in this International
Standard. It is also known as the “bottom‑up” or “step‑by‑step” approach.

It has been suggested that the factors that influence the uncertainty of microbiological enumerations are not
well enough understood for the application of the component approach (see ISO/TS  19036:2006[6]). It is
possible that this approach underestimates the uncertainty because some significant uncertainty contributions
are missed. Reference [19] shows, however, that the concepts of ISO/IEC Guide 98‑3:2008[7] are adaptable
and applicable to count data as well.

Another principle, a “black‑box” approach known as the “top‑down” or “global” approach, is based on statistical
analysis of series of repeated observations of the final result (see ISO/TS 19036:2006[6]). In the global approach
it is not necessary to quantify or even know exactly what the causes of uncertainty in the black box are.

iTeh STANDARD PREVIEW

According to the global philosophy, once evaluated for a given method applied in a particular laboratory,
the uncertainty estimate may be reliably applied to subsequent results obtained by the method in the same
(standards.iteh.ai)
laboratory, provided that this is justified by the relevant quality control data (EURACHEM/CITAC CG 4[10]). Every
analytical result produced by a given method thus should have the same predictable uncertainty. This statement
is understandable against its background of chemical analysis. In chemical analyses the uncertainty of the
ISO 29201:2012
analytical procedure and the https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
uncertainty of the final result of analysis are usually the same. The global principle
dismisses the possibility that there might be something unique about the uncertainty of a particular analysis.
0b55923db79c/iso-29201-2012
The uncontrollable “variation without a cause” that always accompanies counts alters the situation for
microbiological enumerations. The full uncertainty of a test result can be estimated only after the final result
has been secured. This applies to both the global and the component approaches.

The unpredictable variation that accompanies counts increases rapidly when counts get low. The original
global design is therefore not suitable for low counts, and therefore also not applicable to most probable
number (MPN) methods and other low‑count applications, such as confirmed counts.

It is often necessary, and always useful, to distinguish between two precision parameters: the uncertainty of the
technical measuring procedure (operational variability), which is more or less predictable, and the unpredictable
variation that is due to the distribution of particles. A modification of the global principle that takes into account
these two sources of uncertainty is free from the low‑count restriction. This is the global model detailed in this
International Standard.

In theory, the two quantitative approaches to uncertainty should give the same result. A choice of two approaches
is presented in this International Standard. Offering two approaches is appropriate not only because some
parties might prefer one approach to the other. Depending on circumstances one approach may be more
efficient or more practical than the other.

Neither of the main strategies is, however, able to produce unequivocal estimates of uncertainty. Something
always has to be taken for granted without the possibility of checking its validity in a given situation. The estimate
of uncertainty is based on prior empirical results (experimental standard uncertainties) and/or reasonable
general assumptions.

vi  © ISO 2012 – All rights reserved

INTERNATIONAL STANDARD ISO 29201:2012(E)

Water quality — The variability of test results and the uncertainty

of measurement of microbiological enumeration methods

1 Scope
This International Standard gives guidelines for the evaluation of uncertainty in quantitative microbiological
analyses based on enumeration of microbial particles by culture. It covers all variants of colony count methods
and most probable number estimates.

Two approaches, the component (also known as bottom‑up or step‑by‑step) and a modified global (top‑down)
approach are included.

The aim is to specify how values of intralaboratory operational variability and combined uncertainty for final
test results can be obtained.

The procedures are not applicable to methods other than enumeration methods.

NOTE 1 Most annexes are normative. However, only the annexes relevant to each case are to be applied. If the choice
is the global approach, then all normative annexes that belong to the component approach can be skipped and vice versa.

NOTE 2 Pre-analytical sampling variance at the source is outside the scope of this International Standard, but needs
to be addressed in sampling designs and monitoring programmes.
iTeh STANDARD PREVIEW
NOTE 3 The doubt or uncertainty of decisions based on the use of analytical results whose uncertainty has been
(standards.iteh.ai)
estimated is outside the scope of this International Standard.

NOTE 4 The extra‑analytical variations observed in proficiency tests and intercalibration schemes are also not
ISO 29201:2012
detailed in this International Standard, but it is necessary to take them into consideration in analytical control. The use
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
of intercalibration data in uncertainty estimation offers the possibility for the bias between laboratories to be included
0b55923db79c/iso-29201-2012
(Nordtest Report TR 537[12]).

2 Key concepts

2.1 Uncertainty of measurement

Uncertainty of measurement according to ISO/IEC Guide 98‑3:2008[7] is defined as a “parameter, associated
with the result of measurement, that characterizes the dispersion of the values that could reasonably be attributed
to the measurand”. It is a measure of imprecision. The parameter is expressed as a standard uncertainty or
relative standard uncertainty.

2.2 Estimation of the uncertainty of measurement

According to ISO/IEC Guide 98‑3:2008,[7] the parameter can be evaluated by statistical analysis of series of
observations. This is termed type A estimation of uncertainty.

Any other type of procedure is called type B estimation of uncertainty. The most common type B estimates in
microbiological analysis are those based on assumed statistical distributions in the component approach.

Types A and B may refer to the uncertainty of individual components of uncertainty as well as to the combined
uncertainty of the final result.

Type A evaluations of standard uncertainty are not necessarily more reliable than type B evaluations. In many
practical measurement situations where the number of observations is limited, the components obtained from
type B evaluations can be better known than the components obtained from type A evaluations (ISO/IEC Guide
98‑3:2008[7]).

© ISO 2012 – All rights reserved  1

ISO 29201:2012(E)

2.3 Intralaboratory reproducibility

A somewhat abstract expression of uncertainty, intralaboratory reproducibility, is frequently considered
the most appropriate parameter of the uncertainty of measurement, see ISO/TS  19036:2006.[6] It is also
known as intermediate reproducibility or intermediate precision, e.g. [time + equipment + operator]-different
intermediate precision standard uncertainty as defined by ISO 5725‑3.[2] The idea is to evaluate how much the
analytical result might have varied if the analysis had been made by another person in the same laboratory
using different equipment and batches of material and different analytical and incubation conditions than those
actually employed. The value of intermediate precision estimated never belongs to any actual analytical result,
but is assumed to give a general estimate of reasonable uncertainty for the application of a method in one
particular laboratory.

Intralaboratory reproducibility is estimated either by combining separate components of uncertainty determined

under intralaboratory reproducibility conditions (component approach) or by special experiments in which the
analytical conditions are varied by design (global approach).

2.4 Combined standard uncertainty

2.4.1 General

The final test results of microbiological analyses are calculated from intermediate observed values. The
main intermediate observation is the count. Most of the other observed values are connected with volume
measurements.

Combined standard uncertainty, as defined in ISO/IEC Guide 98‑3:2008,[7] is the “standard uncertainty of

iTeh STANDARD PREVIEW
the result of a measurement when that result is obtained from the values of a number of other quantities, equal
to the positive square root of a sum of terms, the terms being variances or covariances of these other quantities
(standards.iteh.ai)
weighted according to how the measurement result varies with changes in these quantities”.

NOTE 1 Observation of covariances is only necessary if significant correlations occur between components of
ISO 29201:2012
uncertainty. Otherwise a simplehttps://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
root sum of variances is sufficient (see 2.4.2 and 2.5).
0b55923db79c/iso-29201-2012
NOTE 2 In cases of microbiological enumeration, it can be assumed that all components of uncertainty are independent,
i.e. statistically uncorrelated. In such instances, the combined standard uncertainty is the positive square root of the sum
of component variances, i.e. the root sum of squares (Annex B). (ISO/IEC Guide 98‑3:2008.[7])

2.4.2 Significant property of combined uncertainties

According to EURACHEM/CITAC CG 4[10], “Unless there is a large number of them, components (standard
uncertainties) that are less than one-third of the largest need not be evaluated in detail”. This statement implies
that in borderline cases, even a single component might provide an adequate estimate of the combined
uncertainty. To decide when a component is unimportant, its approximate size should be known in relation to
other components. Generally at least two, usually more, components are significant and should be included.

EXAMPLE The combined uncertainty of two components, one three times the other, is calculated as
2 2
uc ( y) = 3 + 1 = 10 = 3,16 .

Without the smaller component, the estimate would be 3,00. Ignoring the smaller component underestimates the
combined uncertainty in this case by about 5 %. For the sake of caution, setting a four‑fold difference as the limit might
be recommended.

2.5 Relative standard uncertainty

2.5.1 General

The formula for the final results of microbiological analyses involves only multiplication and division. Under
such conditions, the combined standard uncertainty should be calculated from components expressed as
relative standard uncertainties (ISO/IEC Guide 98‑3:2008[7])(see Annex B).

2  © ISO 2012 – All rights reserved

 ISO 29201:2012(E)

With both type A and type B estimates, the symbol chosen to represent the relative standard uncertainty is urel.

NOTE 1 Relative standard uncertainty is often expressed as a percentage. The term commonly used for this expression
is coefficient of variation (CV), CV.

NOTE 2 When it is important to stress that the standard uncertainty has been calculated by the type A process, the
symbol used is s.

NOTE 3 Any systematic or random variation that takes place in the process before the final suspension, such as
subsampling, matrix, and dilution effects, influence the target concentration in the final suspension proportionally. Relative
variances of these components are therefore additive. Such effects after inoculation as incubation, and reading, can be
more complicated statistically and are not well enough known. Proportionality can still be the best simple approximation.
Systematic errors in these influences are usually treated as if they were random effects.

2.5.2 Logarithms and relative standard uncertainty

“Global” estimates of experimental standard uncertainty are traditionally made by calculation with common
logarithms. When using such estimates in further calculations together with other estimates, it is necessary
to express all components of uncertainty on the same scale of measurement, either by converting relative
standard uncertainties into logarithms or logarithms into relative standard uncertainties.

In most cases, absolute standard uncertainty calculated in natural logarithmic scale and the relative standard
uncertainty in interval scale can be assumed to be numerically equal. Values calculated in common logarithms
can be converted to natural logarithms and vice versa by use of appropriate coefficients. The mathematical
relationships between relative standard uncertainty and standard uncertainty on different logarithmic scales
are shown in B.9.
iTeh STANDARD PREVIEW
2.6 Relative variance
(standards.iteh.ai)
The square of the relative standard uncertainty is called the relative variance (ISO/IEC Guide 98‑3:2008).[7]
ISO 29201:2012
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
2.7 Expanded uncertainty and expanded relative uncertainty
0b55923db79c/iso-29201-2012
Especially when the test result is used for assessing limits concerned with public health or safety, it is pertinent
to give an uncertainty value that encompasses a large fraction of the expected range of the observed values.
The parameter is termed the expanded uncertainty, for which the symbol is U.

The value of U is obtained by multiplying the combined uncertainty with a coverage factor k:

U = ku c ( y )

The value of k is typically in the range 2 to 3. On the relative scale

U rel = ku c,rel ( y )

For normal distributions, about 95 % of the results are covered by the expanded uncertainty interval m ± U , where
m is the mean, when the coverage factor k = 2 is chosen. When k = 3, coverage corresponds to about 99 %.

Microbiological test results almost never fit a normal distribution perfectly. Distributions are often markedly
asymmetrical (skewed). When there are sufficient reasons for assuming distributions to be other than normal
(e.g. Poisson or negative binomial or log–normal distributions) and plausible estimates of the relevant
parameters are available, upper and lower 95 % boundaries can be based on these distributions. Annex N
gives more details about estimation and use of expanded uncertainty.

© ISO 2012 – All rights reserved  3

ISO 29201:2012(E)

3 Microbiological methods

3.1 Common basis

Microbiological enumeration methods based on culture are technical variants of the same basic principle.
The analysis often begins with the mixing of a measured portion of the laboratory sample into a suitable liquid
medium to produce a homogenate called the initial suspension. It may have to be diluted further to produce
a final suspension of appropriate density for detection and enumeration of the target microorganism. In water
analysis, the water sample is the initial suspension and, when dilution is unnecessary, also directly serves as
the final suspension.

3.2 Quantitative instruments

Measured portions of the final suspension are transferred into a detection instrument for quantitative evaluation.

The detection instruments in microbiological analyses vary from a single Petri dish to systems of many parallel
plates in different dilutions and to most probable number (MPN) systems of diverse complexity.

3.3 Uncertainty structure

A complete microbiological analytical procedure consists of five or six successive steps:

a) subsampling and mixing;

b) dilution;
iTeh STANDARD PREVIEW
c) delivery of test portions(s) into the detection system of nutrient media;
(standards.iteh.ai)
d) development during incubation;
ISO 29201:2012
e) counting and possibly confirming the (presumptive) target organisms.
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012
The operational variability consists of the effects of these technical steps. They are individually estimated for
use in the component approach. When estimating the uncertainty of the final result, the uncertainty due to
random distribution of particles in suspension is additionally taken into account (5.2). In the traditional global
approach all operational components and the random distribution of particles are estimated together.

3.4 Expression of combined uncertainty

3.4.1 Two‑component model

For many practical and illustrative purposes it is sufficient to consider the uncertainty of microbiological test
results to consist of two groups of components.

Combined uncertainty of measurement is obtained by combining the operational variability and the intrinsic
variability (distribution uncertainty).

In microbiological contexts both variances are to be expressed as relative (or logarithmic) variances. The
symbols used in this connection in this International Standard are:

2 2
u c,rel ( y ) = u o,rel + u d,rel (1)

where

4  © ISO 2012 – All rights reserved

 ISO 29201:2012(E)

uc,rel(y) is the combined relative standard uncertainty;

uo,rel is the relative operational variability (experimental relative standard uncertainty);

ud,rel is the relative intrinsic variability (relative distribution uncertainty).

Equation (1) is applied in both the modified global and the component approaches to construct the combined
relative uncertainty of measurement of the final result.

NOTE Subscripts can be used to indicate the experimental conditions or level of uncertainty (r for repeatability, Rʹ for
intermediate or intralaboratory repeatability and R for interlaboratory repeatability).

3.4.2 Operational variability (technical uncertainty)

Operational variability is the combination of all the uncertainties associated with the technical steps of the
analytical procedure. It includes the variability of the subsampling, mixing, and dilution of the laboratory sample
to prepare the final test suspension. It also includes the possible effects of incubation and the uncertainty of
reading the result. Bias components are involved but form parts of random variation.

3.4.3 Intrinsic variability (distributional uncertainty)

Intrinsic variability is the unavoidable variation without a cause that is associated with the distribution of particles
in the final suspension and in the detection instrument. In microbiological suspensions it is usually believed to
follow the Poisson distribution. When partial confirmation is practised or the MPN principle is used, the intrinsic
variation increases considerably and no longer follows the Poisson distribution (Annexes D and E).
iTeh STANDARD PREVIEW
NOTE  The intrinsic variability can be decreased by using replicate plates and for MPN estimates by increasing the
number of parallel tubes. (standards.iteh.ai)
ISO 29201:2012
4 Choices of approach
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012
4.1 General
The tradition of evaluation, presentation and use of measurement uncertainty is short in microbiology.
Different parties still have different interpretations and understanding of the meaning and use of measurement
uncertainty. Because of this fluid state, there is no unique right way of determining, expressing and using the
uncertainty of measurement.

This International Standard is primarily intended to provide guidelines for laboratories on how to get started
with establishing the practices of evaluating the uncertainty of measurement. Basic global and component
approaches are described. While the recommendations presented do give valid approaches to the evaluation
of measurement uncertainty for many purposes, there exist other uncertainty evaluation systems, both wider
and narrower in scope than the present protocol. They can provide solutions to specific demands or different
quality control situations. Some of them are briefly characterized in the remainder of this subclause.

In addition to the two basic approaches in this International Standard, there exist other approaches to the
analysis and expression of the uncertainty of measurement. They have gained favour particularly in those
parts of the world where they have been developed. Five examples are given below. Their common feature
is that they are mainly or completely based on data generated in connection with internal and external quality
assurance activity. They address the technical aspects of method validation and analytical competence of
laboratories, and the associated uncertainties, rather than the measurement uncertainty of test results. Also
the statistics may differ. For instance, robust statistics instead of standard statistics may be employed.

The methods in the Nordtest Report TR 537[12] and NMKL Procedure No. 8[11] are based on stable reference
samples which permit some control of the bias components within and between laboratories. The connection
and applicability to microbial populations of real natural samples necessarily remain somewhat obscure.

NMKL Procedure No. 8[11] for the uncertainty in quantitative microbiological examinations is widely accepted
among food analysts in the Nordic countries. It uses internal control data as well as results of validation

© ISO 2012 – All rights reserved  5

ISO 29201:2012(E)

data from collaborative studies for estimating the measurement uncertainty at the participating laboratories.
Nordtest Report TR 537[12] deals for example with intralaboratory bias.

The most comprehensive of the systems is, at the time of publication, under preparation by AFNOR (see
Reference  [9]). It is reported to address the evaluation of different levels of uncertainty (repeatability,
intermediate and interlaboratory reproducibility) from internal and external quality control data and to employ
Bayesian statistics in confidence interval (CI) estimation.

BS 8496[8] is designed to detect the presence of overdispersion (termed “uncertainty of measurement”) between
duplicate counts of natural drinking water samples. A value for the uncertainty of measurement is not determined.

A system in use in New Zealand (Reference [16]) is based on special experiments with natural water samples.
The design is an extension of the basic global design. Data by three technicians analysing several water
samples in quintuplicate are used to estimate general measurement uncertainty values. Operational and
intrinsic components are not separated. As a consequence, low and “normal” counts (limit set at 20 colonies)
require separate assessment.

4.2 Choices of evaluation approach

The uncertainty of measurement established under intralaboratory reproducibility conditions (the intermediate
precision) is the focus of this International Standard. Under these conditions, the components of uncertainty
can be identified and both the global and component approach basically apply. Experiments based on natural
samples are considered important.

Both main approaches to uncertainty of measurement described in this International Standard should, in
principle, give the same results. There are few objective reasons for choosing one approach rather than the
iTeh STANDARD PREVIEW
other. Subjective preferences or requests by a customer or an accreditation authority may be equally valid
(standards.iteh.ai)
reasons. Neither of these approaches might be the one to choose, if one of the approaches outlined in 4.1 is
more fitting to the quality control system and quality control data possessed by the laboratory.
ISO 29201:2012
If a laboratory already has a good quality control system for monitoring details of the analytical procedure, it probably
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
has most of the necessary data available to calculate a component uncertainty estimate. If not, then the global
0b55923db79c/iso-29201-2012
approach would seem to provide the fastest way to get started with estimation of the uncertainty of measurement.

According to recent observations, two components of operational uncertainty are expected to be larger than
others. They are the subsampling variance (matrix effect) with solid materials, and the incubation effect with
many methods. Subsampling variance often exceeds the particle distribution effect in solid samples. With
difficult microbial populations and poor selective methods, the incubation effect can become as important as
the particle distribution effect, whereas with good selective methods, simple microbial populations, and easily
interpreted colony morphology, the incubation effect is insignificant.

The incubation effect is evaluated by observing the possible overdispersion of parallel counts of final
suspensions. Such tests belong to the quality control arsenal of all laboratories irrespective of which evaluation
approach they prefer. Evaluation of the incubation‑effects component is therefore usually possible without any
special arrangements.

With water samples, the subsampling variance is not expected to exceed the Poisson distribution variance
significantly. Other liquid samples and finely powdered materials might be in the same category. In laboratories
where the quality control is based on details of procedure, the estimate of measurement uncertainty can be
constructed from the normally available quality control data. A global approach in such cases would be superfluous.

With solid samples the situation is different. Significant overdispersion between subsamples is the rule. In
these cases, either the global approach should be chosen or the subsampling variance component should
be evaluated by a dedicated experiment. The dedicated experiment for subsampling variance (Annex  H) is
a statistically somewhat more complex design than the entire global experiment (Annex F). The question is
which is considered a more useful parameter to evaluate, the global operational uncertainty or the subsampling
variance. The choice depends on subjective preference.

Evaluation of the operational variance by the global approach is based on subtraction. The smaller the
operational component is in comparison with the distribution uncertainty, the greater its relative imprecision.
The global approach is less efficient with low counts than the component approach. This is a typical situation

6  © ISO 2012 – All rights reserved

 ISO 29201:2012(E)

with water samples. With increasing heterogeneity of samples, the efficiency of the global approach improves
progressively. As soon as the operational variance is expected to become larger than the distribution variance,
the global approach is a reasonable choice. This is a likely situation with solid samples.

When the estimate of uncertainty is to include interlaboratory biases, the evaluation is based on intercalibration
data using the same reference samples for all laboratories. In such cases, the analysis can only be based on
the global approach. The components of uncertainty cannot even be identified. Such evaluations are not within
the scope of this International Standard. Those interested in the approach are advised to consult relevant
protocols (e.g. Nordtest Report TR 537[12]).

4.3 Choices of expression and use of measurement uncertainty

Customers, accreditors and the laboratory may have different expectations and uses of the measurement
uncertainty information. Observation of these requirements determines whether the uncertainty should
be given as operational uncertainty, combined uncertainty, expanded uncertainty or an interval based on
expanded uncertainty of measurement, and in which specific form or scale of measurement. Both the use and
the expressions relevant to various uses are presented in Annex N.

5 The component approach to the evaluation of operational uncertainty

5.1 General
In the component estimation, individual contributions to the uncertainty of measurement (subsampling, dilution,
inoculation, incubation, and reading) evaluated separately are mathematically combined using the law of
iTeh STANDARD PREVIEW
propagation of uncertainty (ISO/IEC Guide 98‑3:2008[7]). Computationally, it means forming the root sum of
squares of the component uncertainties. The combined estimate produced can be called the intralaboratory
(standards.iteh.ai)
reproducibility when the components are determined under reproducibility conditions within one laboratory.
ISO 29201:2012
5.2 Identificationhttps://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
of the components of uncertainty
0b55923db79c/iso-29201-2012
Statistically thinking the uncertainty structure in microbiological enumerations consists of three layers: a) before;
2) within; and 3) after the final suspension. For a more detailed list, see 3.3.

Uncertainty before the final suspension consists of the subsampling and matrix variation, as well as dilution.
Influence quantities before the final suspension affect the combined uncertainty proportionally to the mean
concentration. Whatever additional variation occurs in subsampling or during dilution is transported to the
mean of the final suspension proportionally.

Uncertainty within the final suspension consists of the random distribution of particles in suspension. Together
with the distribution of colonies on the plate, and the possible contribution of the uncertainty of partial confirmation,
they constitute the intrinsic variation. Intrinsic variation does not contribute to the operational uncertainty.

Variation after the final suspension includes the uncertainties connected with the reading of the results and
influences of the incubation environment and time on the apparent observed result. The uncertainties may
include both additive (e.g. contamination) and proportional elements (e.g. uncertainty of counting).

Experiments and examples for the quantitative estimation of the components of uncertainty are detailed in the
annexes. The variance components for subsampling and incubation effects require special experiments. The
other three operational components are available from quality control procedures.

5.3 Evaluation
When components are independent (statistically uncorrelated) and the influence quantities are multiplicative, the
combined relative operational uncertainty is calculated as the positive square root of the sum of relative variances.

© ISO 2012 – All rights reserved  7

ISO 29201:2012(E)

The combined operational relative variance is obtained as the sum of the relative variances of the components.
2 2 2 2 2 2
u o,rel = u rel,M + u rel, F + u rel,V + u rel,I + u rel,L (2)

The meaning of the symbols is given in Table  1. Detailed instructions for the practical evaluation of the
components are found in the annexes listed in Table 1.

Table 1 — Meaning of symbols

Component Symbol Determination

Matrix and subsampling urel,M Annex H
Dilution factor urel,F Annex K
Test portion urel,V, urel,SV Annexes I, J
Incubation urel,I Annex M
Counting urel,L Annex L

The uncertainty of the first three analytical steps is largely independent of the methods and the operator, but
may depend on equipment and material. Once determined, the values can be used repeatedly and for several
microbiological methods.

In the fourth step, the uncertainty depends on the incubation conditions (temperature, atmosphere, and time) and
iTeh STANDARD PREVIEW
the target organism, but is independent of the operator. It can depend on the nutrient medium. It is method specific.

In the fifth step, the uncertainty depends on the operator or equipment used for detection or enumeration.
(standards.iteh.ai)
6 The global approach to the determination of the operational uncertainty
ISO 29201:2012
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012
6.1 General
If all of the quantities on which the result of a measurement depends are varied, its uncertainty can be evaluated
by statistical means (ISO/IEC Guide 98‑3:2008[7]). This is the theoretical basis of the global approach. Planning
a global experiment requires not only a good vision of the important influence quantities, but also a plan to vary
them in a plausible and realistic way.

The global approach presented in this International Standard is based on experiments identical in design
with the model given in ISO/TS  19036:2006.[6] The idea is to duplicate the whole analytical process from
preparation of the initial suspension to the final count. Natural samples shall be studied whenever possible. The
modification in this International Standard concerns only calculations.

The precision of the average estimate depends on the number of samples examined. According to the original
source, 10 samples might provide a sufficient database. In this International Standard at least 30, but preferably
more, samples are recommended.

The global estimate of uncertainty should be evaluated separately for every procedure, type of matrix, and every
target microorganism. To make the estimate realistically representative of the intralaboratory reproducibility, the
factors that are expected to be important should be varied during the experiment. The list presented in 3.3 is helpful.

It is implied in ISO/TS 19036:2006[6] that the generally validated estimate of uncertainty is determined under
conditions in which the distribution (intrinsic) uncertainty is negligible. These conditions are not met when
the counts are low, or even with high counts when the subsampling variation is low. For this reason, the
original global approach is not well suited for water, MPN methods, and low count applications in general. A
modification to the original approach is proposed.

The suggested modification in the calculations is to estimate the more stable part, the operational variance,
by subtraction (6.2). This should correct the low count restriction. Low counts and partial confirmation should
nevertheless be avoided, if possible, in the global experiment, because they increase the imprecision of the

8  © ISO 2012 – All rights reserved

 ISO 29201:2012(E)

estimate of operational uncertainty. If this cannot be avoided, the effect should be compensated by increasing
the number of samples studied.

6.2 Evaluation
The parameter initially determined is the combined standard uncertainty. It can be called the intralaboratory
reproducibility standard uncertainty s ′ (common logarithmic scale) when the experiments are carried out
R
under intralaboratory reproducibility conditions. Planning and analysing such experiments is detailed in Annex F.

After conversion of s to a natural logarithmic scale by u = 2, 303 s an estimate of the relative operational
R′ R ′ ,rel R′
2 2
variance, u o,rel is obtained by subtracting the intrinsic variance (distribution uncertainty), u d,rel , from the
intralaboratory reproducibility variance, u R2 ',rel (B.9).
2
u o,rel = u R2 ' ,rel − u d,rel
2
(3)

where

ud,rel is an estimate of the relative distribution standard uncertainty (intrinsic variability);

u R ',rel is the relative intralaboratory reproducibility standard uncertainty.

With colony counts it is usually assumed that the Poisson distribution is a plausible model for the intrinsic
iTeh STANDARD PREVIEW
variation of the final count.

NOTE 1 It also has been suggested (standards.iteh.ai)

that reproducibility standard uncertainties generated in interlaboratory method
performance trials and proficiency tests could serve a similar function to s (see ISO/TS 19036:2006 ). The parameter
R
[6]
evaluated is the interlaboratory reproducibility standard uncertainty, sR, which is not linked with any given analytical result
in one laboratory. Those interested in the use of ISO 29201:2012trial data in uncertainty evaluation are advised to consult,
interlaboratory
for instance, Nordtest Report TR 537. [12]
https://standards.iteh.ai/catalog/standards/sist/f5ea805a-545a-4ac0-8920-
0b55923db79c/iso-29201-2012
In keeping with ISO/TS 19036:2006,[6] the intralaboratory reproducibility is initially calculated using common
logarithms. An estimate denoted s R2 ' is computed. Conversion to natural logarithms enables the calculations
as above: u R2 ' ,rel = 5, 301 9 × s R2 ' . See also Annex B.

Calculations without taking logarithms are also possible (see Annex I).

In principle, the global approach is valid for any type of microbiological enumeration. However, it is not always
clear what the best estimate of distribution uncertainty is. It is therefore advisable to avoid situations in which
the distribution uncertainty might form the main source of uncertainty in global determinations. Low counts in
general, MPN counts, and especially partial confirmation are such situations. Application of the subtraction
approach is detailed in Annex F.

NOTE 2 If a considerable number of replicate series of analyses covering the whole practical counting range of colony
numbers within one dilution are available a regression approach can be applied (see ISO/TR 13843 [4]).

An estimate of the relative operational variance is obtained as the slope of the regression line fitted to a plot of the variance
to mean ratio, K, vs. the mean number of colonies, nC

2
K = a + u o,rel nC (4)

2
where u o,rel is the estimate of the relative operational variance.

The advantage of the regression approach is that it is not upset if the distribution of the final numbers of
2
colonies does not fit the Poisson distribution after incubation. The estimate of u o,rel is possibly more reliable
than that obtained by the subtraction approach.

© ISO 2012 – All rights reserved  9