You are on page 1of 9

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

net/publication/45093721

Reconsideration of the derivation of Most Probable Numbers, their standard


deviations, confidence bounds and rarity values

Article  in  Journal of Applied Microbiology · November 2010


DOI: 10.1111/j.1365-2672.2010.04792.x · Source: PubMed

CITATIONS READS

129 5,332

3 authors:

Basil Jarvis Cordula Wilrich


University of Reading Bundesanstalt für Materialforschung und -prüfung
104 PUBLICATIONS   2,847 CITATIONS    68 PUBLICATIONS   237 CITATIONS   

SEE PROFILE SEE PROFILE

Peter Wilrich
Freie Universität Berlin
92 PUBLICATIONS   588 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The book "Statistical aspects of the microbiological examination of foods" 3rd Edn View project

Classification/testing strategy for physical hazards according to the UN-GHS View project

All content following this page was uploaded by Cordula Wilrich on 12 August 2021.

The user has requested enhancement of the downloaded file.


Journal of Applied Microbiology ISSN 1364-5072

ORIGINAL ARTICLE

Reconsideration of the derivation of Most Probable


Numbers, their standard deviations, confidence bounds
and rarity values
B. Jarvis1, C. Wilrich2 and P.-T. Wilrich3
1 Ross Biosciences Ltd., Upton Bishop, Ross-on-Wye, UK
2 Bundesanstalt für Materialforschung und –prüfung, Berlin, Germany
3 Institut für Statistik und Ökonometrie, Freie Universität Berlin, Berlin, Germany

Keywords Abstract
confidence bounds, microbiological analyses,
most probable number, MPN, rarity value. Aims: The purpose of this work was to derive a simple Excel spreadsheet and a
set of standard tables of most probable number (MPN) values that can be
Correspondence applied by users of International Standard Methods to obtain the same output
Basil Jarvis, Ross Biosciences Ltd, Upton values for MPN, SD of the MPN, 95% confidence limits and test validity. With
Bishop, Ross-on-Wye HR9 7UR, UK.
respect to the latter, it is considered that the Blodgett concept of ‘rarity’ is
E-mail: basil.jarvis@btconnect.com
more valuable than the frequently used approach of improbability (vide de
2010 ⁄ 0180: received 31 January 2010, Man).
revised 3 June 2010 and accepted 4 June Methods and Results: The paper describes the statistical procedures used in the
2010 work and the reasons for introducing a new set of conceptual and practical
approaches to the determination of MPNs and their parameters. Examples of
doi:10.1111/j.1365-2672.2010.04792.x MPNs derived using these procedures are provided. The Excel spreadsheet can
be downloaded from http://www.wiwiss.fu-berlin.de/institute/iso/mitarbeiter/
wilrich/index.html.
Conclusions: The application of the revised approach to the determination of
MPN parameters permits those who wish to use tabulated values, and those
who require access to a simple spreadsheet to determine values for nonstandard
test protocols, to obtain the same output values for any specific set of multiple
test results. The concept of ‘rarity’ is a more easily understood parameter to
describe test result combinations that are not statistically valid. Provision of
the SD of the log MPN value permits derivation of uncertainty parameters that
have not previously been possible.
Significance and Impact of the Study: A consistent approach for the derivation
of MPNs and their parameters is essential for coherence between International
Standard Methods. It is intended that future microbiology standard methods
will be based on the procedures described in this paper.

recording the number of tubes showing growth at each


Introduction
level of inoculum. The estimate of density is based on the
The most probable number (MPN) procedure is used application of the theory of probability based on certain
widely to estimate microbial densities in many matrices assumptions. The first primary assumption is that the
including foods and water. The procedure, derived from inoculum contains a random distribution of microbial
the original work of McCrady (1915), consists of adding cells. This implies that each dilution is thoroughly mixed
a volume of each of several serial dilutions of a sample to and clumps or aggregates of cells do not occur or, if they
a number of replicate tubes of culture medium and, do, that they are not disrupted during further dilution
following incubation under appropriate conditions, stages – an assumption that may not always be correct.

ª 2010 The Authors


1660 Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology
B. Jarvis et al. Derivation of MPNs and their parameters

Secondly, that each volume of inoculum containing at described originally by Haldane (1939), whereas the tables
least one viable organism will exhibit growth when incu- are based on a modification of the method devised by de
bated in the culture medium. If these assumptions are Man (1983). For a more detailed consideration of the dif-
not met, then the MPN procedure may underestimate ferent approaches that have been made to determination
true microbial cell density. of MPN confidence intervals, see Garthright and Blodgett
Over time, different approaches have been made to (2003).
derive MPN values. Seminal papers include those by Another important aspect of deriving MPN estimates is
Halvorson and Ziegler (1933), Barkworth and Irwin that of improbable outcomes. Such outcomes may occur,
(1938), Haldane (1939), Finney (1947), Cochran (1950) for instance, as a consequence of laboratory errors. Sup-
and de Man (1975, 1977, 1983). Although much early pose three sets of results based on five tubes at each of
work was concerned with testing replicate tubes at a three successive tenfold serial dilutions give outcomes of
single dilution level, the concept of repeating the tests 0-0-5. The calculated estimate of this MPN is 9Æ05 per ml
with replicate tubes at multiple dilution levels soon if the inocula are 0Æ1, 0Æ01 and 0Æ001 ml of original sam-
became normal practice. Haldane (1939) and Cochran ple. But the statistical likelihood for the outcome is only
(1950) describe the statistical theory behind single and 0Æ000025. In de Man’s (1983) MPN tables, and also in
multiple dilution tests and explain some of the problems those of the US FDA (2006), an improbability index pro-
associated with the calculations of MPN values for vides guidance to practitioners as to the acceptability or
multiple tube tests. Cochran (1950) also stressed the need otherwise of a set of results. However, Blodgett (2002)
to ensure that the level of inoculum should lie within suggested an alternative approach to the assessment of
certain limits in order properly to cover the expected level improbability that he describes as the ‘rarity’ value
of organisms in the original sample and provide a means (Blodgett 2008). Improbability is the sum of the probabil-
of assessing the standard deviation of the log10 of the cell ities of all possible outcomes as likely, or less likely, than
density for different combinations of dilution ratios and the actual outcome. In contrast, the ‘rarity’ value mea-
replicate tests at each level. sures the probability of the actual outcome divided by the
Since that time, several authors have published tables probability of the most likely outcome. In this work, we
of MPN values for various combinations of replicate have developed a rarity score that is used to indicate
tubes and dilution levels. Pre-eminent are those of de whether a particular outcome is likely or not.
Man (1983) that provide confidence bounds for the MPN With the plethora of publications on MPN methods, it
estimates together with a measure of the improbability of is pertinent to ask why we have considered it necessary to
the estimate. In addition, various workers have developed ‘reinvent the wheel’ with yet another publication. In
computer software to enable the calculation of MPN esti- 2008, a number of mistakes were identified in part of a
mates – these include Hurley and Roscoe (1983), Klee revised International Standard (Anon 2007) dealing inter
(1993), Curiale (2000), Garthright and Blodgett (2003), alia with MPN techniques. The statistics working group
US FDA (2006) and La Budde (2008). (SWG) of ISO TC34 ⁄ SC9 was asked to recommend
Garthright and Blodgett (2003) describe the FDA’s pre- amendments to this standard. In so doing, we reviewed
ferred MPN methods for standard tests (e.g. five tubes at all international standards on food, dairy and water
three dilution levels) and also for large and unusual com- microbiology and determined that of 15 published stan-
binations of tests. Such test systems may use 25 or more dards, only five standards (Anon 2003, 2004, 2005,
replicate tests at each dilution level (Moruzzi et al. 2000) 2006a,b) cross-refer to Anon (2007), or its predecessor,
or may use well trays with automatic pipetting for serial whilst others use different sets of MPN tables and ⁄ or
dilution assays (Irwin et al. 2000; Walser 2000). One of refer the user to one of several different software systems
the benefits of the US FDA (2006) spreadsheet system is for estimation of MPN values. The SWG recommended
that it can accommodate situations where one or more that all relevant ISO microbiological standards should be
tube in a series has been lost or where overgrowth makes revised to include reference to a revised edition of Anon
a tube reading unreliable. However, use of the US FDA (2007), which would include both tables of MPNs for
(2006) spreadsheet system results in determination of standard combinations of tests and reference to a specific,
confidence bounds that differ from those cited in de generally available software that could be used for any
Man’s and other standard tables, including the FDA combination of test systems. We evaluated the ‘free’ soft-
tables, thus risking confusion amongst practitioners. The ware available and concluded that new approaches would
reason for these differences is that different approxima- be desirable. We recommended that for an international
tions have been used to derive the confidence bounds. standard, it was essential that both the tabulated parame-
Blodgett (personal communication) says that the FDA ters and those determined by use of a spreadsheet should
spreadsheet approach uses the normal approximation give the same results. We concluded, also, that it would

ª 2010 The Authors


Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology 1661
Derivation of MPNs and their parameters B. Jarvis et al.

be sensible to replace the concept of improbability with We assume that the target micro-organisms are
the ‘rarity’ approach of Blodgett (2002, 2008) and to randomly distributed within the matrix, that they are
provide also an estimate of standard deviation that can separate, not clustered together, and that they do not
be used to derive the measurement uncertainty of MPN repel each other. The numbers of micro-organisms in
values. This paper describes the statistical approach used each quantum of inoculum are independent. We
and provides examples of the outputs. assume also that every tube whose inoculum contains
at least one viable target micro-organism will show the
presence of the micro-organism after incubation. Under
Methods
these assumptions, it is reasonable to assume a Poisson
MPN estimations are essentially ‘presence or absence’ distribution of the number of micro-organisms in the
tests performed using serial dilutions to estimate the tubes.
density l [as colony forming units (CFU) per g or ml] of Let k be the number of dilutions, di the relative
a target micro-organism in a test matrix. Samples drawn dilution level per tube (i.e. for a tenfold serial dilution
from the matrix are diluted to several different levels, and from 10)1 to 10)3, di = 0Æ1, 0Æ01 and 0Æ001, respec-
at each dilution level several tubes of culture medium are tively), wi the volume (or weight) of the inoculum at
inoculated with a portion (typically 1 ml) of that dilution. dilution level i, ni the number of tubes and xi the
After incubation under defined conditions, the number of number of positive tubes (tubes that show the presence
tubes that show the presence of the micro-organism at of the micro-organism) at dilution i; i ¼ 1; 2; :::; k.
each separate inoculum level is counted. These counts are Hence, a serial dilution test with k dilutions and its
the basis of the calculation of the MPN as an estimate of results can be described by the k quadruples (di, wi, -
the concentration l. ni, xi), as illustrated in Fig. 1.

Figure 1 Screenshot of Excel spreadsheet for determination of Most Probable Number estimates and their parameters.

ª 2010 The Authors


1662 Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology
B. Jarvis et al. Derivation of MPNs and their parameters

log (f(x)) of a function has its maximum at the same value


Determination of the MPN
as the function f(x), we use the Loglikelihood function
At each dilution, i the number Y of micro-organisms in
the tubes follows the Poisson distribution with expecta- ln L
tion diwil. The probability of y ¼ 0; 1; 2; ::: micro- Xk    
ni
organisms in a tube is given by ¼ ln þ xi lnð1  expðdi wi lÞÞ  ðni  xi Þdi wi l
i¼1 xi

ðdi wi lÞy and calculate the MPN as the value l ^ at which the first
PðY ¼ yÞ ¼ expðdi wi lÞ; y ¼ 0; 1; 2; :::
y! derivative of the Loglikelihood function with respect to l,

Particularly, the probability of no micro-organism in a k  


@ ln L X xi di wi expðdi wi lÞ
tube is ¼  ðni  xi Þdi wi
@l i¼1
1  expðdi wi lÞ
PðY ¼ 0Þ ¼ expðdi wi lÞ
is 0:
and the probability of at least one micro-organism in a
Xk  
tube is xi di wi expðdi wi l ^Þ xi di wi ð1expðdi wi l
^ÞÞ
þ ni di wi
i¼1
1expðdi wi l ^Þ ^Þ
1expðdi wi l
pi ¼ PðY>0Þ ¼ 1  PðY ¼ 0Þ ¼ 1  expðdi wi lÞ: X k  
xi di wi
¼ ni di wi ¼0:
i¼1
1expðd ^Þ
i wi l
The number Xi of positive tubes at dilution step i follows
the binomial distribution with parameters pi and ni. The
probability of xi ¼ 0; 1; :::; ni positive tubes is given by An estimate V^arð^ lÞ of the variance Varð^ lÞ of l
^ is
obtained from the second derivative of the Loglikelihood
  function,
ni xi
PðXi ¼ xi Þ ¼ p ð1  pi Þðni xi Þ " #
xi i @ 2 ln L Xk
xi ðdi wi Þ2 expðdi wi lÞ
  ¼ ;
ni @l2 ð1  expðdi wi lÞÞ2
¼ ð1  expðdi wi lÞÞxi ðexpðdi wi lÞÞðni xi Þ : i¼1
xi
as

The probability of observing (x1, x2, ..., xk) positive tubes 1 1


lÞ ¼ 
V^arð^  ¼ " #:
2  Pk ðd Þ2

at the k dilutions is @ ln L  x i i i expðdi wi l
w
@l2  l ¼ l
^
^ÞÞ2
ð1  expðdi wi l
i¼1
PðX1 ¼ x1 ; :::; Xk ¼ xk Þ
Yk  
ni
¼ ð1  expðdi wi lÞÞxi ðexpðdi wi lÞÞðni xi Þ : The standard deviation of the estimate l
^ is given by
i¼1 xi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Varð^ lÞ:
Given the results (x1, x2, ..., xk) of the serial dilution test,
this is the likelihood function of l, If only negative test results have been observed, xi = 0 for
i = 1, ..., k, the MPN is l ^ ¼ 0; and if only positive test
L ¼ Lðl; di ; wi ; xi ; i ¼ 1; :::; kÞ
results have been observed, xi = ni for i = 1, ..., k, the
Yk  
ni MPN is infinity.
¼ ð1  expðdi wi lÞÞxi ðexpðdi wi lÞÞðni xi Þ :
i¼1 x i

A confidence interval for the concentration


The likelihood function L(l) gives, for each possible
concentration l, the probability of the result of the serial There are various ways to construct a confidence interval
dilution test that has been observed. As an estimated for the concentration l. The calculation of an ‘exact’
MPN of the concentration l, we use the value of l that 1 ) a = 95% confidence interval [lL, lU] for the concen-
maximizes the likelihood function. As the logarithm1 tration l is tedious. It is described in detail by de Man
(1983). Its limits additionally depend on the rule by
which the a = 5% improbable concentrations are divided
1
For convenience we use natural logarithms ln(x) = loge(x). into those below lL and those above lU and hence,

ª 2010 The Authors


Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology 1663
Derivation of MPNs and their parameters B. Jarvis et al.

different authors end up with different confidence inter- positive tubes we find MPN = l ^ ¼ 006 CFU ml)1. How-
vals. We propose to use an easy approximation. As a ever, the result x1 ¼ 0; x2 ¼ 5; x3 ¼ 10 strongly contra-
maximum likelihood estimator of ln l, the (natural) loga- dicts our expectation that, with increasing dilution levels,
rithm ln l
^ of l
^ follows an approximately normal distribu- the numbers of positive results should decrease. Hence,
tion with estimated variance this result violates our assumptions underlying the MPN
determination. As the MPN calculation works irrespective
^2ln l^ ¼ V^arðln l
r ^Þ ¼ V^arð^ l2 :
lÞ=^ of the unlikeliness of the result of the serial dilution test,
we need a measure of this unlikeliness to decide whether
Hence, the interval to trust the result of the test or not.
 The rarity index (r), introduced by Blodgett (2002,
^  2^
ln l ^ þ 2^
rln l^ ; ln l rln l^ 2008), is the ratio of two likelihood values:

is an approximate 95% confidence interval for ln l and Lð^lÞ


r¼ ;
 L0 ð^

l rln l^ Þ; l
^  expð2^ ^  expð2^
rln l^ Þ
In the numerator, we have the likelihood
is an approximate 95% confidence interval for the X k   
ni
concentration l. lÞ ¼ exp
Lð^ ln þ xi lnð1  expðdi wi l
^ÞÞ
If only negative test results have been observed, xi = 0 for i¼1 xi
i = 1, ..., k, the lower confidence limit lL is 0, and the !
upper confidence limit l U is the value l for which the ðni  xi Þdi wi l
^
likelihood function Lðl U Þ ¼ 0  025; where 0Æ025 =
(1 ) 0Æ95) ⁄ 2 : of the result (x1, x2, ..., xk) of the serial dilution test, i.e.
Y
k the value that we find if we insert l ^ into the likelihood
expðdi wi l U Þni ¼ 0  025: function L(l).
i¼1 In the denominator, we have the maximum of the likeli-
hood Lð^lÞwith respect to (x1, ..., xk),
( !)
Xk   
max max ni
L0 ð^
lÞ ¼ lÞg ¼
fLð^ exp ln þ xi lnð1  expðdi wi l
^ÞÞ  ðni  xi Þdi wi l
^ ;
ðx1 ; :::; xk Þ ðx1 ; :::; xk Þ i¼1 xi

This gives i.e. the value of the likelihood if the result of the serial
dilution test was most likely under a concentration l
ln 0025 ln 40 equal to the estimate l ^ of the concentration; this maxi-
lU ¼  ¼ k :
Pk P mum is achieved if
di wi ni di wi ni
i¼1 i¼1

If only positive test results have been observed, xi = ni for xi ¼ ½ðni þ 1Þð1  expðdi wi l
^Þ; i ¼ 1; :::; k
i = 1, ..., k, the upper confidence limit lU is infinity, and
the lower confidence limit lL is the value for which the where [x] denotes the largest integer not larger than x.
likelihood function L(lL) = 0Æ025: The rarity index is a value between 0 and 1. It is 1 if
the result of the serial dilution test is most likely a
Y
k
ð1  expðdi wi lL ÞÞni ¼ 0025: concentration equal to the estimated MPN. If it is in the
i¼1 neighbourhood of 0, the result of the serial dilution test
is very unlikely for a concentration equal to the estimated
This is a nonlinear equation for lL. MPN. Following the approach of de Man (1975, 1983),
we use three categories of rarity:
Category 1: The MPN value would be very likely to
The rarity index
occur if its rarity value falls within the range 0Æ05–1Æ00
If we perform a serial dilution test with k = 3 dilutions, (0Æ05 £ r £ 1Æ00).
dilution factors d1 ¼ 1; d2 ¼ 01; d3 ¼ 001, inocula Category 2: The MPN value would be expected to occur
volumes w1 = w2 = w3 and numbers of tubes only rarely if its rarity value falls within the range 0Æ01–
n1 = n2 = n3 = 10 and observe x1 ¼ 0; x2 ¼ 5; x3 ¼ 10 0Æ05 (0Æ01 £ r < 0Æ05).

ª 2010 The Authors


1664 Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology
B. Jarvis et al. Derivation of MPNs and their parameters

Category 3: The MPN value would be expected to occur Columns 1–3 show the numbers of positive test results
extremely rarely if its rarity value falls within the range 0– for inoculum quanta of 1Æ0, 0Æ1 and 0Æ01 g or ml of sam-
0Æ01 (0 < r < 0Æ01). ple, respectively.
If only negative results or only positive results have Column 4 shows the derived MPN estimates referenced
been observed, the rarity index is r = 1 and hence, the to the primary level of inoculum (i.e. 1Æ0 g or ml)
category is also 1. rounded to two significant figures. Respectively, MPN
values for larger or smaller quantities of primary inocula
should be divided or multiplied by the additional factor
Software
involved, e.g. if the series contains 10, 1 and 0Æ1 g inocu-
An Excel spreadsheet for estimation of the MPN and its lum, the listed MPN value per g should be divided by 10;
parameters has been developed that is freely available if the series contains 0Æ01, 0Æ001 and 0Æ0001 g inoculum,
from http://www.wiwiss.fu-berlin.de/institute/iso/mitarbe the MPN value per g should be multiplied by 100.
iter/wilrich/index.html. Columns 5 and 6 show the log MPN and the standard
deviation, respectively, of the log MPN. Thus, it is possi-
ble to derive an estimate of the expanded microbiological
Results
uncertainty for the calculated MPN value and combine it
Table 1 illustrates the MPN parameters for a three-tube with uncertainties stemming from other sources.
assay with tenfold dilutions. The MPN values, derived Columns 7 and 8 provide approximate bounds of the
using the procedures described by Arndt et al. (1981), are 95% confidence interval for each MPN value.
essentially identical to those found in the tables of de Column 9 lists the calculated ‘rarity value’ for the MPN
Man (1983) and on the BAM website (Garthright and result, based on the procedure of Blodgett (2008) that is
Blodgett 2003; US FDA, 2006). The data columns in the used to determine the acceptability category of potential
table are as follows: MPN results (Column 10).

Table 1 Most Probable Number (MPN) table for a 3 · 3 design (i.e. three sequential tenfold dilution levels and a reference quantum of 1Æ0 g) for
outcomes with rarity index category 1 and 2

Number of positive results for 95% confidence


inoculum volume (ml or g) MPN SD of limits
per ml or log10 log10 Rarity
1Æ00 0Æ10 0Æ01 per g MPN MPN Lower Upper index Category

0 0 0 0 NA* NA 0 1Æ1 1Æ000 1


0 1 0 0Æ30 )0Æ52 0Æ43 0Æ041 2Æ3 0Æ087 1
1 0 0 0Æ36 )0Æ45 0Æ44 0Æ048 2Æ7 1Æ000 1
1 0 1 0Æ72 )0Æ14 0Æ31 0Æ17 3Æ0 0Æ021 2
1 1 0 0Æ74 )0Æ13 0Æ31 0Æ18 3Æ1 0Æ211 1
1 2 0 1Æ1 0Æ056 0Æ26 0Æ35 3Æ7 0Æ021 2
2 0 0 0Æ92 )0Æ037 0Æ32 0Æ21 4Æ0 1Æ000 1
2 0 1 1Æ4 0Æ16 0Æ26 0Æ42 4Æ8 0Æ041 2
2 1 0 1Æ5 0Æ17 0Æ27 0Æ43 5Æ0 0Æ426 1
2 1 1 2Æ0 0Æ31 0Æ23 0Æ69 6Æ0 0Æ019 2
2 2 0 2Æ1 0Æ32 0Æ24 0Æ71 6Æ2 0Æ069 1
3 0 0 2Æ3 0Æ36 0Æ31 0Æ55 9Æ7 1Æ000 1
3 0 1 3Æ8 0Æ59 0Æ31 0Æ93 16 0Æ084 1
3 1 0 4Æ3 0Æ63 0Æ33 0,95 19 1Æ000 1
3 1 1 7Æ5 0Æ87 0Æ30 1Æ9 30 0Æ209 1
3 1 2 12 1Æ1 0Æ26 3Æ6 37 0Æ021 2
3 2 0 9Æ3 0Æ97 0Æ32 2Æ2 40 1Æ000 1
3 2 1 15 1Æ2 0Æ27 4Æ4 51 0Æ420 1
3 2 2 21 1Æ3 0Æ24 7Æ2 64 0Æ068 1
3 3 0 24 1Æ4 0Æ32 5Æ6 100 1Æ000 1
3 3 1 46 1Æ7 0Æ34 9Æ6 220 1Æ000 1
3 3 2 110 2Æ0 0Æ32 25 480 1Æ000 1
3 3 3 ¥ NA NA 36 ¥ 1Æ000 1

*NA, not available.

ª 2010 The Authors


Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology 1665
Derivation of MPNs and their parameters B. Jarvis et al.

Table 2 Examples of Most Probable Number (MPN) estimates for large and unusual combinations of tests

95%
confidence
limits
Weight of MPN per log10 log10 SD Rarity Rarity
Inocula (ml) sample* (g) No. tubes No. positives ml or per g MPN MPN Lower Upper index category

1Æ0, 0Æ1, 0Æ01 1Æ0 20, 20, 20 20, 14, 3 13 1Æ1 0Æ11 7Æ6 21 0Æ794 1
1Æ0, 0Æ1, 0Æ01 1Æ0 50, 50, 50 50, 35, 7 13 1Æ1 0Æ071 9Æ0 17 0Æ806 1
1Æ0, 0Æ1, 0Æ01 2Æ0 50, 50, 50 50, 35, 7 6Æ3 0Æ80 0Æ071 4Æ5 8Æ7 0Æ806 1
1Æ0, 0Æ1, 0Æ01 2Æ0 50, 49, 49 50, 34, 7 6Æ2 0Æ79 0Æ071 4Æ5 8Æ6 0Æ746 1
10Æ0, 1Æ0, 0Æ1, 0Æ01 1Æ0 1, 10, 10, 10 1, 9, 4, 1 3Æ3 0Æ51 0Æ15 1Æ7 6Æ4 0Æ089 1
10Æ0, 1Æ0, 0Æ1, 0Æ01 1Æ0 1, 10, 10, 10 1, 4, 2, 1 0Æ80 )0Æ096 0Æ17 0Æ37 1Æ7 0Æ017 2
10Æ0, 1Æ0, 0Æ1, 0Æ01 1Æ0 1, 10, 10, 10 0, 5, 1, 0 0Æ33 )0Æ49 0Æ18 0Æ14 0Æ74 0Æ004 3
5Æ0, 1Æ0, 0Æ5, 0Æ1, 0Æ05 1Æ0 1, 5, 5, 5, 5 1, 5, 3, 1, 1 2Æ7 0Æ42 0Æ16 1Æ3 5Æ5 0Æ512 1

*Quantity of sample in 1 ml of the initial homogenate (wi).

In Table 1, and also in the output of the computer Acknowledgements


programme, we show that if results at all test levels are
negative, then the MPN is 0; this value is supplemented The authors are grateful to their colleagues on the SWG
by the 95% confidence interval (with the lower confi- of ISO TC34 ⁄ SC9 and the anonymous reviewers for help-
dence limit 0). Most published MPN tables give a value ful comments and discussion.
<x where x is the MPN for the next set of results with
the same number of tubes and only one positive result. References
For instance, if in the design (3 · 1Æ0, 3 · 0Æ1,
3 · 0Æ01 ml) all results are negative, we report Anon (2003) Microbiology of Food and Animal Feeding Stuffs –
MPN = 0 with a 95% confidence interval [0, 1Æ1], Horizontal Method for the Enumeration of Coagulase-
whereas for this design some tables (e.g. de Man 1983) positive Staphylococci (Staphylococcus aureus and Other
Species) – Part 3: Detection and MPN Technique for Low
and computer programmes (e.g. Curiale 2000) show
Numbers. ISO 68883:2003. Geneva: International
MPN < 0Æ30, with no confidence limits. We believe that
Organisation for Standardisation.
our statement is much more informative: i.e. 0 is the
Anon (2004) Microbiology of Food and Animal Feeding Stuffs –
most probable concentration, but a concentration up to
Horizontal Methods for the Detection and Enumeration of
1Æ1 is possible with 95% confidence. Similarly, if all test
Enterobacteriaceae – Part 1: Detection and Enumeration by
results are positive the MPN is infinity (¥) with 95% MPN Technique with Pre-enrichment. ISO 21528-1:2004.
confidence interval [36, ¥], whereas de Man (1983), Geneva: International Organisation for Standardisation.
Curiale (2000) and others give MPN > 110 with no Anon (2005) Microbiology of Food and Animal Feeding Stuffs –
confidence limits. Horizontal Method for the Enumeration of Beta-glucuroni-
Table 2 illustrates results for some different combina- dase-positive Escherichia coli – Part 3: Most Probable Num-
tions of dilution factor, inoculum level and number of ber Technique Using 5-bromo-4-chloro-3-indolyl-beta-D-
replicate tests undertaken. These values were derived glucuronide. ISO 16649-3:2005. Geneva: International
using the Excel spreadsheet version of the programme, Organisation for Standardisation.
for which a partial screenshot is presented as Fig. 1. Anon (2006a) Microbiology of Food and Animal Feeding Stuffs
– Horizontal Method for the Detection and Enumeration of
Coliforms – Most Probable Number Technique. ISO
Discussion
4831:2006. Geneva: International Organisation for
The procedure described here provides a means of Standardisation.
obtaining MPN estimates, and their parameters, for both Anon (2006b) Microbiology of Food and Animal Feeding Stuffs
standard and nonstandard assay combinations using – Horizontal Method for the Determination of Low Numbers
either derived tables of values or a freely available spread- of Presumptive Bacillus cereus – Most Probable Number
sheet, both of which provide identical outputs for the Technique and Detection Method. ISO 21871:2006. Geneva:
same inputs. The system is to be included in the revised International Organisation for Standardisation.
Anon (2007) Microbiology of Food and Animal Feeding
ISO 7218 Standard to which other international standards
Stuffs – General Requirements and Guidance for
will cross-refer.

ª 2010 The Authors


1666 Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology
B. Jarvis et al. Derivation of MPNs and their parameters

Microbiological Examinations. ISO 7218:2007. Geneva: Irwin, P., Tu, S., Damert, W. and Phillips, J. (2000) A
International Organisation for Standardisation. modified Gauss–Newton algorithm and ninety-six well
Arndt, G., Weiss, H. and Hampe, F. (1981) On the investiga- micro-technique for calculating MPN using Excel
tion of some statistical properties of the Most Probable spreadsheets. J Rapid Methods Autom Microbiol 8,
Number (MPN)-procedure for estimating the density of 171–191.
microorganisms by use of computer simulations. In Com- Klee, A.J. (1993) A computer program for the determination
putational Statistics ed. Büning, H. and Naeve, P. pp. 1–17. of most probable number and its confidence limits.
Berlin, NY: de Gruyter. J Microbiol Methods 18, 91–98.
Barkworth, H. and Irwin, J.O. (1938) Distribution of coliform La Budde, R. (2008) A simple MPN calculator. http://
organisms in milk and the accuracy of the presumptive www.lcfltd.com/customer/LCFMPNCalculator.exe (accessed
coliform test. J Hyg, Camb 38, 440–457. 23 June 2010).
Blodgett, R.J. (2002) Measuring improbability of outcomes de Man, J.C. (1975) The probability of most probable num-
from a serial dilution test, Communications in Statistics. bers. Eur J Appl Microbiol 1, 67–78.
Theory Methods 31, 2209–2223. de Man, J.C. (1977) MPN tables for more than one test. Eur
Blodgett, R.J. (2008) Judging a model of an experiment. A J Appl Microbiol 4, 307–316.
Paper Presented to the Joint Statistical Meetings of the de Man, J.C. (1983) MPN Tables corrected. Eur J Appl
American Statistical Association, Denver, 14 August 2008. Microbiol 17, 301–305.
Cochran, W.G. (1950) Estimation of bacterial densities by means McCrady, M.H. (1915) The numerical interpretation of
of the ‘‘Most Probable Number’’. Biometrics 6, 105–116. fermentation-tube results. J Infect Dis 17, 183–212.
Curiale, M. (2000) MPN Calculator for Excel. http:// Moruzzi, G., Garthright, W.E. and Floros, J.D. (2000) Aseptic
www.i2workout.com/mcuriale/mpn/index.html (accessed packaging machine pre-sterilization and package steriliza-
23 June 2010). tion: statistical aspects of microbiological validation. Food
Finney, D.J. (1947) The principles of biological assay. J Roy Control 11, 57–66.
Stat Soc, Ser B 9, 46–91. US FDA (2006) Bacteriological Analytical Manual Online
Garthright, W.E. and Blodgett, R.J. (2003) FDA’s preferred Appendix 2: Most Probable Number from Serial Dilutions.
MPN methods for standard, large or unusual tests, with a http://www.fda.gov/Food/ScienceResearch/Laboratory
spreadsheet. Food Microbiol 20, 439–445. Methods/BacteriologicalAnalyticalManualBAM/acm10956.
Haldane, J.B.S. (1939) Sampling errors in the determination of htm (accessed 23 June 2010).
bacterial or virus density by the dilution method. J Hyg, Walser, P.E. (2000) Using conventional microtiter plate
Camb 39, 289–293. technology for the automation of microbiological testing
Halvorson, H.O. and Ziegler, N.R. (1933) Application of statis- of drinking water. J Rapid Methods Autom Microbiol 8,
tics to problems in bacteriology. J Bacteriol 25, 101–121. 193–207.
Hurley, M.A. and Roscoe, M.E. (1983) Automated statistical
analysis of microbial enumeration by dilution series. J Appl
Bacteriol 55, 159–164.

ª 2010 The Authors


Journal of Applied Microbiology 109, 1660–1667 ª 2010 The Society for Applied Microbiology 1667

View publication stats

You might also like