You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225356794

Estimation of measurement uncertainty in food microbiology: The ISO


approach

Article  in  Accreditation and Quality Assurance · April 2006


DOI: 10.1007/s00769-005-0085-5

CITATIONS READS

17 2,363

1 author:

Bertrand Lombard
Agence Nationale de Sécurité Sanitaire de l'Alimentation, de l'Environnement et du Travail
53 PUBLICATIONS   937 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Bertrand Lombard on 09 September 2014.

The user has requested enhancement of the downloaded file.


Accred Qual Assur (2006) 17: 94–100
DOI 10.1007/s00769-005-0085-5 LEGISLATION AND NORMS

Bertrand Lombard
Estimation of measurement uncertainty
in food microbiology: The ISO approach

Abstract Given the current interest envisaged for the estimation of the
Received: 13 May 2005
Accepted: 19 December 2005 in measurement uncertainty (MU) in reproducibility standard deviation, in
Published online: 14 February 2006 food microbiology, in particular for a decreasing order of preference: The

C Springer-Verlag 2006 laboratory accreditation purposes, intra-laboratory standard deviation,
and the need to have harmonized the inter-laboratory standard
Presented at AOAC Europe/Eurachem reference documents specifically in deviation derived from method
Symposium March 2005, Brussels, Belgium
that area at the international level, validation, and the inter-laboratory
B. Lombard () ISO is conducting works to meet this standard deviation derived from
AFSSA-LERQAP (Agence Française need. An ISO Technical Specification proficiency testing.The uncertainty of
de Sécurité Sanitaire de Aliments– (ISO/TS 19036) is being prepared on qualitative determinations is still
Laboratoire d’Etudes et de Recherches MU estimation for quantitative under investigation, and will be
sur la Qualité des Aliments et sur les
Procédés Agro-alimentaires),
determinations. A global approach covered by a separate ISO
23 avenue du Général De Gaulle, has been chosen, based on the publication.
94 706 Maisons-Alfort, France reproducibility standard deviation of
e-mail: b.lombard@afssa.fr the final result of the measurement Keywords Food . Microbiology .
Tel.: +33-1-49-77-26-96 process. Three possibilities are Measurement uncertainty . ISO
Fax: +33-1-43-68-97-62

Introduction the finalization of international standardization works that


are presented in this article. In addition to the accreditation
Microbiological laboratories, like the overall analytical requirements, the particular significance of microbiologi-
community, are more and more addressing the question cal analysis may be emphasized: It detects or quantifies in
of measurement uncertainty (MU) attached to the results many cases a direct hazard for consumers’ heath, making
they produce. The main reason for this trend is certainly it all the more critical to assess the level of confidence that
to be found in the frame of laboratory accreditation, where can be put in the outcome of the analysis. The variability
the Standard EN ISO 17025 [1], in its clause 5.4.6, re- associated with many reference methods used to determine
quires laboratories to estimate MU associated to their re- colony counts in food microbiology is large, typically in the
sults, and there is no reason that this general requirement 0.1–1.0 log10 cfu/g range (Table 7.6 in [3]). In particular
would not apply to microbiology. The same standard also for the interpretation of the results, the need to estimate and
states that if rigorous/statistically valid calculation of MU is control this large variability is thus reinforced. Finally, most
not possible, the MU components are to be identified, and analyses for food pathogens are currently performed on a
to be reasonably estimated. The requirements regarding presence/absence basis. The way to express the uncertainty
MU for the accreditation of microbiological laboratories of qualitative determinations needs to be investigated, no
have been clarified at European level by the European co- solution being currently available.
operation for Accreditation (EA) in the Guide EA-04/10 These reasons have led ISO/TC 34/SC 9 “Food
[2]. In France, the body in charge of laboratory accred- products—Microbiology” to address the question of MU
itation, COFRAC (Comité Français d’Accréditation) has in food microbiology at its Bangkok meeting in December
recognized that these requirements should apply to food 2002. ISO/TC 34/SC 9 is in charge of standardization in the
microbiology; it requests that laboratories address the ques- field of food microbiology, preparing the so-called “hori-
tion of MU, identify and control the main sources of un- zontal” ISO Standards, which are applicable to the analysis
certainty, but it has delayed the need to estimate MU up to of all foods and feeds, including the environment of food
production and food handling. At the 2002 meeting, it has
95

been decided to prepare at first the ISO guidelines on MU food microbiologists and where the approach selected has
estimation in quantitative microbiology. MU for qualitative not yet been widely used. Thus, it justifies the choice of
determinations would be dealt with in a second step, since a status of reference document lower than an International
this second topic would require more investigation before Standard, being still “tentative” for a limited period of time
an approach could be agreed upon. and needing a review at short-term. Finally, this status is
meant to encourage the users of the Technical Specification
to report their experience on the implementation of the ap-
Measurement uncertainty in quantitative microbiology proach described. After the 3-year period, ISO/TS 19036
would then be reviewed by ISO/TC 34/SC 9 in the light of
Presentation the experience gained. After improvements and sufficient
support by users, it is intended to become an International
The member countries and international organisations in Standard. In particular, certain aspects not dealt with in the
liaison of ISO/TC 34/SC 9 have agreed on the need to first edition would need to be addressed, such as MU at low
develop guidelines specific to food microbiology, even if levels of contamination and MU of Most Probable Number
a large set of reference documents was already available techniques. In addition, a choice between the three options
to address and estimate MU of quantitative determinations left open to estimate the reproducibility standard deviation
from a more general point of view, such as the basic refer- may have to be made (see later).
ence document on MU published by ISO, GUM [4], as well The principle of the approach was easily agreed upon by
as ISO/TS 21748 [5] or the EURACHEM/CITAC Guide ISO/TC 34/SC 9 at its 2002 meeting, with a broad consen-
[6]. ISO/TC 34/SC 9 considered that it was necessary to sus of countries and observing international organizations,
address the specificities of microbiology, the analyte being based on a first series of trials performed earlier in the year
a living microorganism and some main reagents used in that had aimed at testing the feasibility of the approach
enumeration techniques (such as culture media, antibod- proposed.
ies for confirmation) being of biological nature, resulting ISO/TS 19036 applies not only to colony-count tech-
in the large variability of the results recalled earlier in the niques, the reference methods in quantitative food micro-
introduction. In addition, it can be noted that the existing biology, but also to other quantitative techniques regarded
ISO/CEN Standards describing reference methods in food as alternative methods, which can be of instrumental nature
microbiology provide at best, in a limited number of cases, (such as impedancemetry, ATP-metry, flow cytometry,. . .)
values of precision data derived from inter-laboratory trials and whose signal is converted into number of colonies, the
(see later), but no limits to the values of the major sources unit of the reference methods.
of MU, which could have enabled laboratories to avoid
to estimate MU themselves, according to EN ISO 17025
(Note 2 in clause 5.4.6.2). The approach selected
Moreover, the need to have a reference document harmo-
nized at the international level was recognized by ISO/TC Two main approaches can be distinguished to estimate MU
34/SC 9. The variety of existing documents introducing dif- for any type of measurement. The first one consists in esti-
ferent approaches was regarded as a reason to produce an mating the individual sources of variability (variances) that
harmonized reference document in food microbiology. The contribute to the uncertainty in the measurement process,
intention was to avoid that each country, accreditation body and MU is derived using formal principles of uncertainty
or even each laboratory would adopt its own approach, that propagation by combination of variances. This “bottom
could lead to marked inconsistency in the approaches to up” or “step-by-step” approach is a widely adopted stan-
MU and to the difficulty of comparing MU values from dard approach, described in the GUM as well as in several
different laboratories for the same determination (differ- other publications, particularly in a more practical way by
ences could arise from differences in the approach to MU, the EURACHEM/CITAC Guide for analytical chemistry.
and not from differences in the analytical method chosen It has been applied also to quantitative microbiology by the
nor in the proficiency of the laboratories). Finnish Centre for Metrology and Accreditation (MIKES)
At its Bangkok meeting, ISO/TC 34/SC 9 decided to es- [9]. This approach can be rather easily implemented in
tablish an ISO Technical Specification, ISO/TS 19036 [7], water microbiology, where the matrix is relatively simple
and not an International Standard. It should be noted that and homogeneous. However, this approach has not been
a Technical Specification cannot be regarded as having the recommended by ISO/TC 34/SC 9 for food microbiology,
same status as an International Standard, that it cannot have in accordance with other reference documents and authors
the same degree of reference and that competing Techni- (such as ISO/TS 21748, Horwitz [10]). The following rea-
cal Specifications on the same subject are permitted [8]. sons can be given to justify this choice:
One reason for this choice is that such a status enables a
1. The bottom up approach, to be implemented routinely,
quicker publication for a topic that was considered as one of
requires the use of standard or assumed variances for
the first priorities of the committee. In addition, the status
the individual steps that align the combined estimate to
of Technical Specification requires a 3-year revision. This
the basic method rather than the analytical result.
relatively short life time was found appropriate for a topic
that has only recently raised interest in the community of
96

2. ISO/TC 34/SC 9 considered that such a process might Measurement methods and results,” which has prepared
underestimate the true MU value in the case of the micro- ISO/TS 21748, a publication that creates a bridge between
biological analysis of food where it is difficult to build GUM and the ISO 5725 series dedicated to the estimation
a really comprehensive model of the measurement pro- of trueness and precision of analytical methods, and that
cess. The measurement process of colony-count tech- justifies the consistency of the global approach with GUM.
niques is complex and includes several steps, the im- The global approach can be considered as a special case of
portant ones being: Sub-sampling of the test portion, GUM’s type A experimental estimation. Finally, the Inter-
preparation of the initial suspension and of the serial national Laboratory Accreditation Cooperation (ILAC) in
decimal dilutions, plating on isolation agars, incubation, its general guide on MU in the context of ISO/IEC 17025
counting of colonies, confirmation step in some cases. A [12], encourages the use of existing experimental data from
significant source of variation may thus be overlooked. quality control charts or inter-laboratory trials, which are
3. It appears difficult to quantify accurately the contribu- at the basis of the global approach.
tion of each individual step of the analytical process
in food microbiology, where (a) the analyte is a liv-
ing microorganism, whose physiological state can be The main steps
largely variable, (b) the analytical target includes dif-
ferent strains, or different species, or different gen- In May 2003, the first draft of ISO/TS 19036 was prepared
era, and (c) the enumeration techniques use reagents on the basis of the first series of trials of 2002 and the
of biological nature (such as culture media, antibod- decisions taken in Bangkok, the author acting as Project
ies for serological confirmation) that may largely vary Leader for ISO/TC 34/SC 9.
in an unpredictable manner. In other words, it is diffi- From December 2003 to March 2004, a second series of
cult to build a metrologically rigorous and statistically ISO trials has been conducted. Their purpose was to quan-
valid estimation of MU for the microbiological anal- tify the MU component linked to (i) the sub-sampling of the
yses of foods, as it has been recognized by the EA test portion, and (ii) the preparation of the initial suspension
Guide 04/10. (see details later). The burden was shared by 79 laboratories
4. Some error factors can be unpredictable, such as drop- from 10 countries, a level of participation not yet encoun-
ping the thermometer or missing a decimal point, and tered in trials launched by ISO/TC 34/SC 9, denoting the
are not subject to a predictive statistical description that interest of the analytical community for that topic!
is needed in the decomposition approach (Horwitz [10]). Based on the comments received from on member coun-
5. A purely modelling approach would not take account of tries and organizations in liaison, and the outcome of the tri-
the variability due to difference in sample types and to als, a Project Group recommended in a meeting during 22–
the sub-sampling of the test portion. A restricted view 23 March 2004, modifications to the first draft in order to
to MU may exclude the sub-sampling of the test por- prepare the final draft of ISO/TS 19036, and these were
tion from the estimation of MU that would be limited endorsed by ISO/TC 34/SC 9 at its meeting in Parma,
to the further analytical steps. A larger view includes in April 2004.
MU all steps performed by the laboratory, and in food
microbiology, the heterogeneous distribution of the mi-
croorganisms in most food types (solid ones) can be the The Technical Specification ISO/TS 19036
source of a major part of the total uncertainty, at the
sub-sampling step. The final draft of ISO/TS 19036 has been submitted to
6. Given the complexity of the measurement process in the official voting period until 11 May, and it will soon be
food microbiology, the step-by-step approach is rather published after being modified according to the comments
heavy to implement in this analytical field. received during the vote.
Right from the title of ISO/TS 19036, the intention to
ISO/TC 34/SC 9 has chosen the second approach, based
define a “guide” or guidelines is expressed. It means that
on the reproducibility standard deviation (sR ) of the final
ISO/TC 34/SC 9 was conscious that other publications or
result of the entire measurement process, here the number
reference documents are available on the same topic, and
of colonies per gram or millilitre of sample. This experi-
that the approach of ISO/TS 19036 is not the unique nor the
mental approach makes no attempt to set generic estimates
mandatory approach to be followed in this field is unique,
of uncertainty for specific test methods and rightly aligns
this approach being nevertheless the recommended one in
the MU estimate with a specific analysis, or set of anal-
food microbiology. The choice of a Technical Specifica-
yses. It is likely that most of the potential variability fac-
tion status had a similar objective (see earlier). Moreover,
tors that are likely to be encountered in practice, including
the guidelines mention in the introduction the step-by-step
the sub-sampling of the test portion and the unpredictable
approach as well as its application to microbiology.
tricks, will be introduced (Horwitz, [10]). This “global” or
As in any ISO reference document, one of the first sec-
“top down” approach has been endorsed by different refer-
tions provides definitions, including MU itself. They are all
ence structures, including Codex Alimentarius (the Codex
extracted from GUM.
Committee on Methods of Analysis and Sampling [11]),
The Technical Specification then describes the global
and ISO/TC 69/SC 6 “Application of statistical methods—
approach, which can be conveniently illustrated by a
97

Fig. 1 Diagram of the main Excluded (out of


sources of uncertainty in food the analytical
microbiology, and the process)
“black-box” approach to
measurement uncertainty Serial Counting/
Sub-sampling/ dilutions/ confirmation
Sampling Primary dilution plating of colonies

Laboratory
sample
Black box Result

Matrix Residual Operator/ Equipment/ Bias


random time culture
errors media/
reagents Total bias not covered
(empirical nature of
microbial counts)
Low levels excluded

“black box” diagram as in Fig. 1. This diagram mentions ISO/TS 19036 does not deal with MU estimation at low
the main uncertainty sources that are covered, partly levels (<10 colonies counted on a plate) due to a lack of
covered or not covered by the approach chosen, such as a simple agreed approach, but it is expected that the next
sampling, excluded from the measurement process senso edition of these guidelines would cover it.
stricto. Regarding the bias, part of it can be estimated
in the experimental conditions of the intra-laboratory
reproducibility and to a larger extent in the conditions of Intra-laboratory sR
the inter-laboratory reproducibility (see later). However,
the empirical nature of microbial counts, where the analyte The intra-laboratory sR is a particular case of “intermediate
is defined by the method used, justifies that in the general precision” as defined in the Standard ISO 5725-3 [13]; the
case the total bias cannot be included in MU estimation, reproducibility conditions are varied within a laboratory:
the true value being unknown. Analysis of the same sample in conditions differing as much
ISO/TS 19036 insists that the global approach is to be as possible within the same laboratory (see details later).
implemented in a laboratory under quality control, as to This is the preferred option, since it provides an MU value
ensure the validity of MU estimation in the routine ana- that is linked to the laboratory, thus nearer to the basic MU
lytical activity, and not only at a given time. It means that definition which relates MU even to one individual test
the critical factors associated with the method or the lab- result.
oratory that are likely to affect the measurement result, An experimental protocol is described in ISO/TS 19036,
should be identified and demonstrated to be under control. and is represented in Fig. 2. From the same food sample,
A re-assessment of MU estimation is required following two test portions are taken by two different operators (the
changes of any of these factors. In other words, the global “operator” may be a team of operators—technicians, each
approach does not preclude that the laboratory identifies of whom performing a given part of the procedure). From
the main components of uncertainty, in agreement with the each test portion, one initial suspension is prepared, and
requirements of ISO/IEC 17025, but the quantification of analysed once, as in routine testing. Conditions A and B
each component is not required. should be as different as possible and should ideally include
The following principles are then defined: as many variations as may be encountered from one day
(i) Expanded MU (U)=2 sR , the usual value of 2 being to another within the laboratory, in terms of technicians,
chosen for the coverage factor (95% confidence level); batches of culture media and reagents, vortex mixer, pH
(ii) sR is to be estimated per microorganism (or consistent meter, incubators, time of analysis, etc. If the contamina-
group of), per matrix type (or consistent group of), for tion of the food sample is demonstrated to be sufficiently
a given physiological state of the organism and for a stable (which is rarely the case in food microbiology), the
given method that the laboratory uses for producing in conditions A and B should relate also to different days of
routine its results. The term “consistent” means that analysis. The purpose is to cover the extremes of method-
the group would give equivalent MU values. ological variability likely to occur in a given laboratory
(iii) Three options are recommended for estimating sR in order to derive a maximum estimate of MU that would
reflect the range of individual results likely to be obtained
• Intra-laboratory sR ; in the same laboratory. The laboratory can then safely as-
• Inter-laboratory sR derived from method validation; sociate this MU value with its results without the risk of
• Inter-laboratory sR derived from proficiency testing. underestimating MU.
98

Food sample 2. The precision attained by the measurements within the


laboratory is compatible with that expected from re-
peatability and reproducibility estimates derived from
1st operator (conditions A) 2nd operator (conditions B) the inter-laboratory study;
3. The inter-laboratory study has correctly covered all the
Initial suspension Artificial contamination Initial suspension sources of uncertainty (especially sample preparation
(if needed) and homogenisation).
Analysis Analysis
These pre-requisites are justified by the fact that the stan-
dard deviation derived from an inter-laboratory study is
linked to the method, and not to a given laboratory which
is to report MU attached to its results. The procedure to
check that these conditions are met, and how to form a
Different conditions combined uncertainty estimate with the possible additional
Fig. 2 Experimental protocol for the intra-laboratory standard de- factors not covered by the inter-laboratory study, are not
viation of reproducibility described in detail in ISO/TS 19036 which refers for these
aspects to ISO/TS 21748.
This protocol is to be repeated for at least 10 samples for This approach has the advantages to use precision data
a given combination (target microorganism, matrix type); already generated and to enable a laboratory that has taken
a consistent group of each of them can also be considered. part in an inter-laboratory study to assess its laboratory
The number of matrix types to be tested depends on the bias, part of the bias component of MU. This aspect is not
diversity of the matrices analysed in routine by the labo- detailed in ISO/TS 19036.
ratory. The selected matrices should be representative, in However, in food microbiology, there are several limita-
terms of MU values, of the matrix types analysed by the tions to this approach, justifying its consideration only as a
laboratory and should also be relevant to the microorgan- second option:
isms for which the test is to be done. This selection is – In addition to the need to check the pre-requisites listed
certainly the most challenging aspect in this approach, and before, only a limited number (5) of reproducibility pa-
an inappropriate choice would lead to an unrealistic MU rameters have been derived from inter-laboratory studies
estimation. Moreover, it makes the protocol less practicable for standardized reference methods in food microbiol-
when the laboratories analyse a large variety of matrices, ogy.
each requiring a separate MU assessment (such as in the – It may be difficult to generalise values obtained in an
case of official food control laboratories). inter-laboratory study to the routine samples analysed
This protocol incorporates the effect of the sub-sampling by the laboratory. Precision values of an inter-laboratory
of the test portion in the evaluation of the total uncertainty. study will be obtained under a limited and precisely de-
Since it is well known in food microbiology that the nat- fined combination of matrices, strains and physiological
ural contamination of certain food products can be highly state of microorganism, contaminations level, etc. and
heterogeneous, this MU component may have a major con- for a given background microflora (if present).
tribution to the total MU and cannot be avoided, especially – Given the homogeneity requirements for the samples
when a judgement on conformity of the sample analysed used for inter-laboratory studies, the need to send to the
against limits (such as microbiological criteria) is to be laboratories homogenised and stabilised samples induces
made. a reduction of the “natural” variation in sample con-
ISO/TS 19036 then provides the different calculations tamination that may be found in practice, which causes
up to MU, not detailed here. The data are initially log- underestimation of the uncertainty.
transformed, to stabilize the reproducibility variance over
the contamination levels, given that low levels are not con-
sidered here. Inter-laboratory sR derived from proficiency testing

If the laboratory has taken part in an inter-laboratory pro-


Inter-laboratory sR derived from method validation ficiency testing (PT), it may use the standard deviation of
reproducibility from this PT to deduce its measurement
If the method used routinely by the laboratory has been uncertainty, under the following conditions:
submitted to an inter-laboratory study for validation, the – During the inter-laboratory study, the laboratory used the
laboratory may use the reproducibility standard deviation same method that it uses in routine analyses;
of the method for deriving its own MU under the following – The samples that were used in the study are comparable,
conditions: in terms of matrix and contamination level, to the ones
1. The laboratory bias is compatible with that expected on analysed routinely;
the basis of the repeatability and reproducibility esti- – The laboratories that participated in the study did not use
mates derived from the inter-laboratory study; different empirical methods, or a sufficient number of
99

participants used the same method, as to allow a correct Table 1 Values of standard deviation linked to sub-sampling of
estimation of the reproducibility standard deviation. test portion/preparation of initial suspension (sIS ), in log10 cfu/g, per
food category (2003–2004 trials)
This option presents the two same advantages than Food Minimum Median Maximum Mean (sIS )
the former option. However, most of the limitations category (sIS ) (sIS ) (sIS )
encountered for the former option apply here too, except
the limited availability of data, given the current large I 0.02 0.08 0.17 0.08
number of PT schemes organized in food microbiology. II 0.01 0.09 0.17 0.09
III 0.08 0.17 0.41 0.19
IV 0.05 0.20 0.74 0.25
Expression of measurement uncertainty in test reports Note. I: liquids and powders; II: well-mixed solids; III: small (or very
small) solids; and IV: other solids
ISO/TS 19036 proposes several possibilities to express
MU, including the following:
ANOVA (Kruskall–Wallis by ranks), the factor category
– y±2sR , where y=log10 x is the test result, sR expressed was shown to have a significant impact (p=5.10−7 ) on
in log units, the values of the MU component value linked to sub-
– after back transformation, x[10 y−2sR , 10 y+2sR ]. sampling/initial suspension.
Table 1 gives the values of the standard deviation linked
to sub-sampling of test portion/preparation of initial sus-
Results of trials pension for each of the food categories defined above.
For the first two categories, the matrix was responsible for
An informative annex of ISO/TS 19036 provides the results about 0.1 log units of the overall standard deviation, regard-
of the trials organized in 2003/2004 for ISO/TC 34/SC less of the laboratory or the microorganism. For the last two
9, which aimed at estimating the uncertainty component categories, it was not possible to assess the order of mag-
linked to the sub-sampling of the test portion and to the nitude of the effect independent of the microorganism and
preparation of the initial suspension. The protocol and the the laboratory. On the basis of these trials, ISO/TC 34/SC
results of these trials are detailed into a report prepared by 9 concluded that it was not possible to decompose total
the trials’ organizers [14] and are intended to be published MU between the component linked to sub-sampling/initial
in a separate paper. suspension (a typical value per type of matrix) and the
Each of the 79 participating laboratories elected to enu- component linked to the following analytical steps (to be
merate one or more target microorganisms in one or more determined by each laboratory).
matrix types naturally contaminated with the target mi-
croorganisms of interest. The participants followed a com-
mon experimental protocol laid down for the trials, corre- Analytical uncertainty in qualitative microbiology
sponding to the protocol of ISO/TS 19036 for the intra-
laboratory sR (see Fig. 2) except that for each condition Strictly speaking, a qualitative test is not a measurement,
A or B, two duplicates were tested under repeatability but it is still possible to characterize the uncertainty in the
conditions. The protocol was also to be repeated for 10 result of a qualitative determination, the term “analytical
samples of the same combination (matrix type, target mi- uncertainty” may be more appropriate in that context. The
croorganism). Ninety-six combinations (laboratory, target need to characterize the uncertainty of detection methods
microorganism, matrix type) meeting the acceptability cri- (presence/absence tests) is particularly important in food
teria defined for the trials were used to calculate, for each microbiology, as highlighted before, since most of the con-
combination, the intra-laboratory sR , as well as, with the trols on food-borne pathogens are currently conducted on
theory of the variance components, the three standard devi- a qualitative basis.
ations linked to (i) sub-sampling of test portion/preparation This section is not expanded, since preliminary works
of initial suspension, (ii) the intra-laboratory conditions are still under way, conducted by Working Group 2 “Statis-
(operator/time/reagents/equipment) as defined by ISO/TS tics” of ISO/TC 34/SC 9. For detection methods, it is en-
19036 in the experimental protocol for the intra-laboratory visaged to follow the proposal from one of the group’s
sR , and (iii) the residual errors (random errors under re- members, Hitchins (non-published), deriving the analyti-
peatability conditions). cal uncertainty from the 50% limit of detection (LOD50% ).
A classification of matrices related to physical criteria This is the value of the contamination level where a posi-
was tested: (i) liquids and powders (e.g. milk, coconut milk, tive result is obtained in 50% of the cases. This point cor-
dried milk); (ii) well-mixed solids (e.g. minced meat, me- responds to the inflexion point of the sigmoid curve (dose,
chanically separated meat, sausage meat, crushed meat, response) well-investigated in toxicological studies, and for
whipped cream, dairy ice cream, soya cream); (iii) small which several models have been developed. In particular,
(or very small) solids (e.g. dehydrated parsley/mushrooms, WG 2 of ISO/TC 34/SC 9 is currently investigating the
grated carrots/celeriac, salad, shrimps, cereals, feeding model which would be best suited to microbiological de-
stuffs, chopped hazel nuts); and (iv) other solids (e.g. non- tection methods, on the basis of experimental data already
minced meat, cheeses, pastry). Using a non-parametric generated.
100

Conclusion for other analytical fields where qualitative determinations


are performed.
The global approach retained at ISO level to estimate MU It is hoped that the publication of ISO/TS 19036 will
for quantitative determinations appears to be pragmatic, encourage the generalization of MU estimation in food
adapted to the complexity of the analytical process in food microbiology, thus leading the analytical field of food mi-
analysis, especially in food microbiology. It is in agreement crobiology from an empirical state to a more “scientific”
with some international reference documents and rules. one!
It is foreseen that ISO will soon agree on a way to express Acknowledgements The author wishes to thank Dr Marie Cornu
the uncertainty of presence/absence determinations, that and Professor Basil Jarvis for their help in improving the text of the
would be of interest not only for food microbiology, but also publication.

References
1. ISO (1999) General requirements for 6. Ellison SLR, Rosslein M, Williams A 11. Codex Alimentarius (2004) Draft
the competence of testing and (2000) EURACHEM/CITAC guide: guidelines on measurement uncertainty,
calibration laboratories, EN ISO 17025. Quantifying uncertainty in analytical Appendix IV to ALINORM 04/27/23
International Organization for measurement, 2nd edn. (Report of the 25th Session of Codex
Standardization, Geneva, Switzerland EURACHEM/CITAC, Committee on Methods of Analysis and
2. EA (2002) Accreditation for QUAM:2000.P1 Sampling, Budapest, 8–12 March,
microbiological laboratories, European 7. ISO (2003 1st version, 2005 2nd 2004). FAO/WHO Codex Secretariat,
co-operation for Accreditation, Guide version) Microbiology of foods and Roma, Italy
EA-04/10 rev.02 animal feeding stuffs—Guide on 12. ILAC (2002) Introducing the concept
3. Jarvis B (1989) Statistical aspects of estimation of measurement uncertainty of uncertainty of measurement in
the microbiological analysis of the for quantitative determinations, Draft testing in association with the
foods (coll. Progress in industrial ISO TS 19036. International application of the standard ISO/IEC
microbiology, vol. 21). Elsevier, Organization for Standardization, 17025, Guide ILAC-G17. International
Amsterdam, 179 p Geneva, Switzerland Laboratory Accreditation Cooperation
4. ISO (1995) GUM, Guide to the 8. ISO (2004) ISO/IEC Directives Part 2: 13. ISO (1994) Accuracy (trueness and
expression of uncertainty in Rules for the structure and drafting of precision) of measurement methods and
measurement, BIPM/IEC/IFCC International Standards. International results—Part 3: Intermediate measures
/ISO/IUPAC/OIML. International Organization for Standardization, of the precision of a standard
Organization for Standardization, Geneva, Switzerland measurement method, ISO 5725-3.
Geneva, Switzerland 9. Niemelä SI (2002) Uncertainty of International Organization for
5. ISO (2004) Guide for the use of the quantitative determinations derived by Standardization, Geneva, Switzerland
estimation of repeatability, cultivation of microorganisms, 14. Ah Soon C, Cornu M (2004) Report of
reproducibility and trueness in the Publication J3/2002. Centre for 2003/2004 ISO trials about uncertainty
estimation of measurement uncertainty, Metrology and Accreditation (MIKES), of measurement, June 2004. AFSSA
ISO/TS 21 748. International Helsinki, Finland Maisons-Alfort, France
Organization for Standardization, 10. Horwitz W (2003) The certainty of
Geneva, Switzerland uncertainty. J AOAC Int 86(1):
109–111

View publication stats

You might also like