This article was downloaded by: [171.67.34.

205]
On: 28 June 2013, At: 15:00
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Critical Reviews in Food Science and Nutrition

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/bfsn20

Reinterpretation of Microbial Survival Curves

a b
Micha Peleg & Martin B. Cole
a
Department of Food Science, Chenoweth Laboratory, University of Massachusetts, Amherst,
MA 01003
b
Nabisco, Inc., 200 DeForest Avenue, East Hanover, NJ 07936
Published online: 03 Jun 2010.

To cite this article: Micha Peleg & Martin B. Cole (1998): Reinterpretation of Microbial Survival Curves, Critical Reviews in
Food Science and Nutrition, 38:5, 353-380

To link to this article: http://dx.doi.org/10.1080/10408699891274246

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
 Critical Reviews in Food Science, 38(5):353–380 (1998)

Reinterpretation of Microbial Survival Curves

Micha Peleg1,* and Martin B. Cole 2
1
Department of Food Science, Chenoweth Laboratory, University of Massachusetts, Amherst,
MA 01003; 2Nabisco, Inc., 200 DeForest Avenue, East Hanover, NJ 07936

* Corresponding author.

ABSTRACT: The heat inactivation of microbial spores and the mortality of vegetative cells
exposed to heat or a hostile environment have been traditionally assumed to be governed by first-
order reaction kinetics. The concept of thermal death time and the standard methods of calculating
the safety of commercial heat preservation processes are also based on this assumption. On closer
scrutiny, however, at least some of the semilogarithmic survival curves, which have been
considered linear are in fact slightly curved. This curvature can have a significant effect on the
Downloaded by [171.67.34.205] at 15:00 28 June 2013

thermal death time, which is determined by extrapolation. The latter can be considerably smaller
or larger depending on whether the semilogarithmic survival curve has downward or an upward
concavity and how the experimenter chooses to calculate decimal reduction time. There are also
numerous reports of organisms whose semilogarithmic survival curves are clearly and character-
istically nonlinear, and it is unlikely that these observations are all due to a mixed population or
experimental artifacts, as the traditional explanation implies. An alternative explanation is that the
survival curve is the cumulative form of a temporal distribution of lethal events. According to this
concept each individual organism, or spore, dies, or is inactivated, at a specific time. Because
there is a spectrum of heat resistances in the population — some organism or spores are destroyed
sooner, or later, than others — the shape of the survival curve is determined by its distributions
properties. Thus, semilogarithmic survival curves whether linear or with an upward or a down-
ward concavity are only reflections of heat resistance distributions having a different, mode,
variance, and skewness, and not of mortality kinetics of different orders. The concept is demon-
strated with published data on the lethal effect of heat on pathogens and spores alone and in
combination with other factors such as pH or high pressure. Their different survival patterns are
all described in terms of different Weibull distributions of resistances as a first approximation,
although alternative distribution functions can also be used. Changes in growing or environmental
condition shift the resistances distribution’s mode and can also affect its spread and skewness.
The presented concept does not take into account the specific mechanisms that are the cause of
mortality or inactivation — it only describes their manifestation in a given microbial population.
However, it is consistent with the notion that the actual destruction of a critical system or target
is a probabilistic process that is due, at least in part, to the natural variability that exists in
microbial populations.

KEY WORDS: thermal processing, thermobacteriology, predictive microbiology, mortality

curves, population dynamics, kinetics.

I. INTRODUCTION modeling in food microbiology. It is now

possible to estimate the likely effects of
In the last decade, there have been sig- key intrinsic factors such a pH, aw, and tem-
nificant advances in the area of microbial perature on the growth of all of the food

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

353
 pathogens of major concern (McClure et al.,
1994).
Importantly, these empirical models al-
low, for the first time, the possibility of truly
quantitative risk assessments in the field of
food microbiology (Whiting and Buchanan,
1996). The use of these techniques is gain-
ing acceptance internationally as a means of
comparing the equivalency of different food
processes in ensuring safe food and for en-
suring the stringency of risk management
procedures such as HACCP (ICMSF, 1997).
Thermal processing is frequently cited
as an example of where mathematical mod-
eling has been used traditionally for many
years to predict the outcome of different
processing conditions on the survival of
Downloaded by [171.67.34.205] at 15:00 28 June 2013
microorganisms. Ironically though, as our
ability to quantify the effects of key param-
eters on microbial growth has improved, at
the same the underlying basis of traditional
thermal processing is being questioned.
The origins of modern thermobac-
teriology can be traced back as far as 1920
(Bigelow and Esty, 1920; reviewed by Ander-
son, 1997), and the principles of this still
form the basis of all thermal processes (aimed
at microbial inactivation) used currently by
the food industry (Ball and Olson, 1957;
Stumbo, 1973; Teixeira, 1992; Jay, 1996;
Holdsworth, 1997).
The approach is based on the assump-
tion that all the cells or spores in a popula-
tion have identical sensitivity to heat, and
that it is the chance of a quanta of heat
striking the cell or target within the cell that
will determine the death rate. According to
this theory the destruction rate of microbial
spores or cells at a given temperature can be
described by the differential equation
–dN/dt = kN
where N is the momentary number of spores
or cells in the medium, dN/dt is the momen-
tary rate of change in the number, and k a
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
354
rate constant (1/time units). The magnitude
of k depends on the particular organism, the
temperature, the nature of the medium (e.g.,
pH, the presence of salts and/or nutrients),
and sometimes the organism’s previous ther-
mal history (see below). Integration of Eq. 1
results in a single term exponential decay
function
N = No exp – (kt) (2)
where No is the initial number of the cells or
spores.
Rearrangement of Eq. 2 gives the famil-
iar relationship
loge [N/No ] = – kt (3)
or
log10 S = k′t (4)
where S is the survival ratio (S = N/No, 0 ≤
S ≤ 1) and k′ = k/loge 10.
A plot of log10 S vs. t, according to this
model, is a straight line originating at log10
S = 0 with a slope of –k′ (Figure 1).
A microbial ‘Thermal Death Time’
(TDT), or τ, is defined as the length of time
at different temperatures necessary to com-
pletely destroy a definite concentration of
cells or spores (Bigelow, 1921). In the case
of the “botulinum cook” (Stumbo, 1973) this
is defined as the time needed to reduce the
original population by 12 orders of magni-
tudes, or in terms of the model to reach
S = 10–12 i.e.,
τ = 12/k′ (5)
The relationship between τ and temperature
is also assumed to be semilogarithmic and
has been used traditionally to design and
(1) assess the safety of commercial sterilization
and pasteurization processes (Teixeira, 1992;
Jay, 1996). The impressive safety record of
the canning and other food industries has
been considered as compelling evidence that
 Downloaded by [171.67.34.205] at 15:00 28 June 2013

without the consent of the publisher is prohibited.

FIGURE 1. Simulated microbial survival curves generated with the cumulative Weibull distribution as a model with n = 1. Note the similarity to a first-order
destruction kinetic.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material

355
 the theory is sound and there has been little
incentive to question its foundations. Thus,
most of the advances made in the math-
ematical aspects of thermal processing of
foods have been in the application of the
theory to different thermal histories in order
to facilitate the calculation of optimal pro-
cessing conditions.
Although generally accepted in the food
community and by regulating agencies, the
theory has had its critics. Their main points
have been (e.g., Casolari, 1988, 1994; Cole,
1997):
1. The experimental survival curve from
which the thermal death time is deter-
mined rarely covers more than 5 or 6
Downloaded by [171.67.34.205] at 15:00 28 June 2013
decades. Thus, the assumption that log
S vs. t remains linear (or even approxi-
mately linear) in the remaining 6 to 7
decades is not supported by experimen-
tal evidence (see below).
2. There is a growing number of published
observations that show that log S vs. t
is clearly a nonlinear relationship and
hence that the mortality of at least some
microorganisms, and spores, is certainly
not a process governed by a first-order
kinetics (see below).
3. The first-order kinetic does not account
for the various possible mechanisms
that are responsible for a vegetative
cell demise or a spore thermal inactiva-
tion.
Obviously, the model expressed in Eqs.
1 to 4 is inadequate for organisms whose
survival curve has a pronounced upward or
downward concavity when plotted in semi-
logarithmic coordinated. Consequently,
such survival curves have been described by
a variety of alternative mathematical and
kinetic models (e.g., Whiting, 1995; Hills
and Mackey, 1995; Linton et al., 1995;
Holdsworth, 1997). They have also been ex-
plained as representing a mixture of heat
resistant and sensitive subpopulations
(Stumbo, 1973; Bunning et al., 1990; LeJean
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
356
et al., 1994; Linton et al., 1995), each obey-
ing a first-order mortality kinetics with a
characteristic rate constant k (Eq. 4).
The alternative approach is to consider
the different shapes of thermal inactivation
curves as different cumulative forms of dis-
tributions of heat resistances or sensitivities.
The origin of this approach has a long his-
tory. For example, Withell (1942) showed
that arithmetic survival data having skewed
distributions conformed to a log normal dis-
tribution.
More recently, Cole et al. (1993) used a
logistic equation to describe an accumula-
tive distribution of heat sensitivities of List-
eria monocytogenes within a narrow tem-
perature range but under different conditions
of pH and salt concentration. When the data
were transformed to S vs. log time relation-
ships, a single distribution type accounted
for the inactivation kinetics over the condi-
tions studied.
A log-logistic equation has been used to
describe the effect of heating rate on the
inactivation of L. monocytogenes (Stephens
et al., 1994), as well as the effect of tempera-
ture on the inactivation of Salmonella
typhimurium (Ellison et al., 1994), the sur-
vival of Yersinia enterocolylica at sub-opti-
mal pH (Little et al., 1994) and temperature,
and the thermal inactivation of Clostri-dium
botulinum (Anderson et al., 1996).
The approach has been applied recently
to the characterization of the effect of heat
on the survival of Listeria, whose log S vs.
t relationship has a very strong upward con-
cavity (Augustin et al., 1998). The authors
considered L. monocytogenes as having a
spectrum of resistances with a log normal
distribution and hence the long “tail” to the
right. (The actual mathematical model that
they used was cfu(t) = cfu(0)/(1 + exp[(x–
m)/s2, where x is log10 t and m and s2 con-
stants representing the distribution (logarith-
mic) center and spread. Each survival curve
was converted into a frequency (density) form
of the distribution by derivation.) They were
able to show that different pretreatments, or
 growing conditions, produced shifts of
different magnitudes in the resistances
distribution’s mode, and in some cases in its
spread as well (see below). The concept can
also be related to the probabilistic nature of
certain inactivation mechanisms (Casolari,
1988). Whether and how early a critical tar-
get or targets are hit beyond repair, thus
causing mortality or inactivation, is a matter
of chance and hence, according to this theory,
the resulting mortality events must have a
temporal distribution. The reader will notice
that the specific cause of the inactivation is
unimportant in this context, although differ-
ent destructive mechanisms, and their com-
binations, may produce distributions with
different characteristics.
Downloaded by [171.67.34.205] at 15:00 28 June 2013
A similar approach has been developed
independently (Peleg, 1996, 1997; Peleg et
al., 1997) to account for the different shapes
of microbial dose response curves. When
expressed as log S vs. dose relationships,
they can be linear, or curved, with upward or
downward concavity, and with or without a
“lag”. The dose response curves have been
treated as being a reflection of the pop-
ulation’s spectrum of sensitivities, or resis-
tances, to the lethal agent. Or in other words,
the different shapes of the mortality curves
examined could all be explained in terms of
the statistical properties of different under-
lying distributions of resistances or sensi-
tivities, rather than being a manifestation of
lethality kinetics of different orders.
The objectives of this work are to rein-
troduce and discuss this unifying approach
in more detail and to explore its potential
implications in the safety evaluation of ther-
mal and other processes of food preserva-
tion.
II. THE MODEL
Let us assume that after exposure of a
single microorganism to a high-temperature
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
T for a duration t it is either alive or dead, or,
in the case of a single spore, the latter is
either viable or not. (The possibilities of in-
jury and recovery, scenarios where com-
pounds released from dead organisms can
protect the survivors and enhance their resis-
tance or when the process is slow enough to
allow for adaptation and a resistance build
up, are not considered.) If there are only two
states, life and death, the survival function,
Si(t), of an individual organism, or a single
spore, i, is given by:
For t < tci Si(t) = 1 (alive) (6)
t ≥ tci Si(t) = 0 (dead)
where tci marks the time at which the organ-
ism dies or the spore loses its viability. (From
now on the term ‘organism’ will include
spores unless otherwise stated.) It is obvious
that a resistant organism will have a rela-
tively high tc, while a sensitive one a lower
one. Hence, tc is a measure of the organism’s
heat resistance or sensitivity.
For any discrete assembly of such or-
ganisms the survival curve will be:
S(t) = ΣSi(t) (7)
If the assembly is very large, as is the case in
most microbial populations, and if its mem-
bers’ heat resistance expressed in terms of
their tci can be considered as having a con-
tinuous distribution, then the assembly’s
survival curve S(t) can be written as:
∫ f [t, t (φ)] dφ
1
S( t ) = c (8)
0
where f[t, tc(φ)] is a function of the exposure
time, t, and the fraction of organisms φ shar-
ing any given tc.
If we assume that Eq. 8 holds and that tc
has a Weibull distribution, i.e.,
(
dφ dt c = bnt cn –1 exp – bt cn ) (9)
357
 where b and n are constants then Eq. 8 be-
comes:
S(t) = exp (–btn) (10)
the cumulative form of the Weibull distribu-
tions.
Or, if presented as a semilogarithmic
relationship
log S(t) = –btn (11)
Examples of simulated survival curves
using Eq. 10, or Eq. 11, as a model are given
in Figures 1 to 3. They show that if n = 1 the
survival curve will appear linear in semi-
logarithmic coordinates and hence the ap-
pearance of a first-order kinetics. However,
Downloaded by [171.67.34.205] at 15:00 28 June 2013
because according to the presented concept
the exact nature of the underlying distribu-
tion (Eq. 9) is usually unknown and cannot
be assumed a priori, the existence of cases
where n ≠ 1 cannot be excluded. Thus, when
n < 1, that is, when the distribution has a
strong right skew (see below), the semiloga-
rithmic survival curve has a noticeable up-
ward concavity. When n > 1, that is, the
semilogarithmic survival curve has a pro-
nounced downward concavity, the underly-
ing distribution can have a left skew depend-
ing on the value of b (Peleg et al., 1997).
These qualitative relationships between the
shapes of the survival curves and those of
their underlying distributions of resistances
are not limited to the Weibull distribution.
They are observed with other skewed
unimodel distributions, such as the asym-
metric beta distributions (see below). (Un-
like the Weibull distribution, which always
has a 0–∞ range, the beta distribution has a
finite upper limit and it can have a lower
limit ≠0 as well.) If an experimental survival
curve is considered as a cumulative form of
an underlying frequency distribution of sen-
sitivities, or resistances, then the statistical
properties of the frequency distribution can
be determined directly from the survival data.
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
358
Thus, if the Weibull distribution is consid-
ered as an appropriate representative math-
ematical model of the population, its param-
eters b and n can be estimated by fitting Eq.
10 to the survival data. Once calculated, these
values of b and n can be used to plot the
sensitivities/resistances frequency curve us-
ing Eq. 9, and to calculate the distribution’s
mode, tcm, mean, -tc, variance, σ2tc and coeffi-
cient of skewness, ν1, from the following
(Patel et al., 1976):
t cm = [(n – 1) nb]
1/ n
(12)
{ }
t c = Γ[(n – 1) n] b1/ n (13)
{ )}
(
σ 2tc = Γ[(n + 2) n] – Γ[(n + 1) n]
2
b 2 / n (14)
and
ν1 = µ 3 µ 32 / 2 (15)
where Γ is the gamma function, µ3 = Γ(1+2/
n)/b3/2 and µ2 = Γ(1+2/n)/b2/n.
Similar expressions are available for the
beta and other common unimodel distribu-
tion functions (Patel et al., 1976).
A. Linear vs. Semilogarithmic
Coordinates
If the Weibull distribution is selected as
a model the question that arises is whether
the estimation of its constants b and n should
be done by fitting the “raw” S(t) vs. t data
using Eq. 10 as a model, or after the data
conversion into a log S(t) vs. t file using
Eq. 11 as a model. If there is no data scatter
and the distribution is indeed of the Weibull
type, the calculated values of b and n will
be the same regardless of how it is deter-
mined. However, if the Weibull distribu-
 Downloaded by [171.67.34.205] at 15:00 28 June 2013

without the consent of the publisher is prohibited.

FIGURE 2. Simulated microbial survival curves generated with the cumulative Weibull distribution as a model with n < 1.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material

359
 Downloaded by [171.67.34.205] at 15:00 28 June 2013

360
without the consent of the publisher is prohibited.
FIGURE 3. Simulated microbial survival curves generated with the cumulative Weibull distribution as a model with n > 1

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
 tion function is only an approximation, and
if there is a considerable data scatter as is
usually the case, then the discrepancy be-
tween the calculated values of b and n can
be substantial. In such a case the corre-
sponding distributions calculated by the two
methods can be very different. The discrep-
ancy stems from the different relative weight
that is assigned to the deviations from the
fitted values when the data are expressed as
absolute values or their logarithm. Conse-
quently, the meaning of a distribution cal-
culated from experimental log S(t) vs. t
data is unclear (see below). Nevertheless,
when microbial populations are concerned
the difference between, survival ratios of
let’s say 10–3 and 10–5 can be meaningful
Downloaded by [171.67.34.205] at 15:00 28 June 2013
even though both are numerically negligible
compared with, for example, 0.5 or 0.1.
Consequently, the presentation of the sur-
vival curve in linear coordinates may not be
sensitive to the detailed shape of the
distribution’s tail especially when the distri-
bution is strongly skewed to the right. Thus,
if the sole purpose of the fit is to accurately
describe the survival of the most resistant
members of the populations, then Eq. 11 or
a similar semilogarithmic relationship can
and should be used as the model. If, how-
ever, the b and n values so determined are
used to generate a distribution of resistances,
the result may be an exaggerated fraction of
the most resistant survivors.
B. Mixed Populations
A concave log S(t) vs. t curve can be
interpreted as indicative of a mixed popula-
tion whose subpopulations all have a first-
order mortality kinetics with different rate
constants (Stumbo, 1973; Rank and Pflug,
1977). It can equally be interpreted as in-
dicative of the existence of a mixture whose
subpopulations have a symmetric or asym-
metric heat sensitivity/resistance distribu-
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
tions (Peleg et al., 1997; Peleg, 1997), or,
in terms of the Weibull distribution’s pa-
rameters, with n = 1 and/or n ≠ 1. To test
which is the true scenario one must grow
the survivors after repeated partial heat treat-
ments and to determine if their logarithmic
survival curve becomes linear. (Such a test
may still be inconclusive, as the effects of
natural selection and/or adaptation cannot
be completely ruled out.) If this indeed
happens, then the conventional first-order
model and the proposed distribution model
with n = 1 become equally viable and
equivalent alternatives. However, if no such
shift can be observed then the hypothesis of
a mixture ought to be dropped and the as-
sumption of a first-order mortality kinetics
dismissed. In such a case it is much more
likely that the population has undergone, or
that the process induces, a spectrum of le-
thal events whose temporal distribution can
be symmetric or skewed in either direction.
In a number of microorganisms where the
survivor’s survival curve had been deter-
mined, no evidence was found of a dra-
matic change in the curve’s shape. Thus,
one must conclude that at least these organ-
isms have a mortality pattern that is gov-
erned by a distribution of lethal events rather
than by a first-order kinetics. (Obviously, if
the process is repeated for a sufficient num-
ber of times a more heat-resistant popula-
tion may eventually emerge as a result of
“natural selection”. But as has already been
stated this scenario, as well as that where a
slow heating process which allows for physi-
ological adaptation, should not concern us
here.)
C. Effect of the Medium
It is well established that the efficacy of
microorganism’s destruction by heat (and
other lethal agents) can be affected not only
by the temperature but also by the mediums’
361
 pH, water activity, the presence of chemical
preservatives, etc. In terms of the first-order
kinetics model, these affect the magnitude
of the rate constant k. In at least some micro-
organisms, however, a change in the tem-
perature or pH can also change the shape of
the semilogarithmic survival curve and en-
hance (e.g., Ama et al., 1994; Ballestra et al.,
1996) or even invert its concavity (e.g., Rank
and Pflug, 1977). In such cases it seems
more plausible to describe the effect in terms
of changes in the distribution of the
population’s sensitivities or resistances than
to contemplate the ad hoc creation of new
mixtures just in order to save the first-order
kinetics model (see below).
Downloaded by [171.67.34.205] at 15:00 28 June 2013
III. ANALYSIS OF PUBLISHED
SURVIVAL CURVES
A survival curve of C. botulinum spores
plotted on semilogarithmic coordinates is
shown in Figure 4. Traditionally, this kind
FIGURE 4. Semilogarithmic survival curve of C. botulinum spores fitted with Eq. 11 as a model. Note the
slight but noticeable downward concavity. (Data from Gillespy [1962] as appears in Jay [1996]).
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
362
of curve has been considered linear and
was actually used to calculate the D values
of spores. On closer inspection one can
detect a slight curvature in the curve with a
noticeable downward concavity. As can be
seen from the figure, the fit of Eq. 11 is
almost perfect and it yields an n value, which
is substantially different from one. When
the same data are plotted on linear coordi-
nates, Figure 5, the typical sigmoid shape
of a cumulative distribution clearly appears.
This survival curve can be described in terms
of a cumulative form of the Weibull distri-
bution (Eq. 10), from which the frequency
distribution of the heat resistances can be
generated using Eq. 9. The calculated val-
ues of b and n can also be used to calculate
the distribution parameters, namely, the
mode, mean, variance, and coefficient of
skewness using Eqs. 12–15 (see Table 1). It
ought to be mentioned that the selection of
the Weibull distribution was based on con-
venience only, and that other unimodal dis-
tribution functions can have a comparable
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 5. The survival curves of the C. botulinum spores fitted with Eq. 10 as a model and its underlying
distribution of resistances.

fit. This is demonstrated in Figure 6, where function. As expected, the resulting fre-
the same published data are fitted with the quency distribution is almost exactly the
cumulative form of the beta distribution same as that calculated from the Weibull

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

363
 TABLE 1
Heat Resistance Distribution Parameters of the Organisms Discussed in This Work

Weibull distribution parameters

Coeff. of
Mode Mean skewness Variance
Organism Treatment (min) (min) (–) (min2) Data source

C. botulinum 240°F 4.9 6.1 1.3 10.3 Jay (1996)

(spores)
B. stearothermophilus 123°C 1.5 1.6 1.2 5.3 Le Jean et al.
(spores) (1994)
S. typhimurium 56°C — 5.1s 5.5 111s2 Ellison et al.
(1994)
L. monocylogenes 55°C — 0.6 6.6 2.0 Augustin et
al. (1998)
L. monocytogenes
grown at 19°C 60°C 2.3 2.2 1.0 0.07 Bhaduri et
al. (1991)
grown at 37°C 60°C 3.1 4.8 1.4 31.1 Bhaduri et
Downloaded by [171.67.34.205] at 15:00 28 June 2013

al. (1991)
E. coli 5 MPa/25°C 20.4 20.6 1.2 51.6 Ballestra et
al. (1996)
5 MPa/35°C 5.6 9.6 1.5 39.3 Ballestra et
al. (1996)
5 MPa/45°C 2.5 5.8 1.6 18.2 Ballestra et
al. (1996)
1.2 MPa/45°C 20.1 26.0 1.3 196 Ballestra et
al. (1996)
2.5 MPa/45°C 5.6 9.6 1.5 39.3 Ballestra et
al. (1996)
5.0 MPa/45°C 2.5 5.8 1.6 18.2 Ballestra et
al. (1996)
V. vulnificus 50°C/pH 6.2 3.7 7.9 1.6 94.7 Ama et al.
(1994)
50°C/pH 4.6 0.6 0.6 1.2 0.06 Ama et al.
(1994)

distribution — compare the bottom part of
Figures 5 and 6. Although not visible on
linear coordinates, the major difference
between the two functions, as already men-
tioned, is in the detailed shape of the
distribution’s tail.
Another example of a semilogarithmic
survival curve of bacterial spores is shown
in Figure 7. Again, although this curve of B.
stearothermophilus previously has been con-
sidered linear in order to calculate the spores’
thermal death time, it has a noticeable down-
ward concavity. This curve can also be de-
scribed by Eq. 11 as a model with n ≠ 1, as
shown in the figure. When plotted on linear
coordinates the data have the characteristic
sigmoid shape of a cumulative Weibull dis-
tribution function (Figure 8). As before, the
latter could be used to generate the spores
frequency distribution of heat resistances and
to calculate its statistical parameters
(Table 1).
Examples of clearly concave upward
semilogarithmic survival curves, of Salmo-
nella and Listeria are given in Figures 9 and
11. Both could be fitted with Eq. 11 as a
model with n < 1 as shown in these figures.
In these cases, however, the frequency
distributions do not appear to have a peak, as
seen in Figures 10 and 12. In fact, the great

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

364
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 6. The survival curve of the C. botulinum spores fitted with a cumulative beta distribution function
and its corresponding frequency distribution. Note that the latter is practically the same as that calculated on
the basis of the Weibull distribution shown in Figure 5.

majority of both organisms were destroyed mental data relate to the comparatively very
within a very short time after being exposed small number of survivors that were still
to heat. Thus, much of the reported experi- viable after a considerably long time.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

365
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 7. Semilogarithmic survival curve of B. stearothermophilus spores fitted with Eq. 11 as a model.
Note the slight but noticeable downward concavity. (Data from Le Jean et al. [1994]).

Whether this kind of a distribution, that is,
with an extreme skewness to the right, is
characteristic of all or most heat sensitive
pathogens is uncertain.
A. Expressing the Effects of
External Conditions in Terms of
Shifts in the Resistance
Distributions
Semilogarithmic survival curves of L.
monocytogenes grown at 19 and 37°C are
shown in Figure 13. Although the data scat-
ter data were considerable the curves could
still be fitted by Eq. 11 as a model with a
reasonable degree of fit. Similarly, the sur-
vival curves plotted on linear coordinates
could be fitted with Eq. 10, from which the
frequency distributions could be generated,
as shown in Figure 14. This figure clearly
shows that when the organism was grown at
19°C, it was not only more heat sensitive on
average but that it has a very narrow distri-
bution of heat resistances. When grown at
37°C, there was only a relatively small shift
in the distribution’s mode, but a very sub-
stantial increase in its overall spread and its
skewness to the right. (This means that a
substantial fraction of the Listeria popula-
tion grown at 37°C can survive long after the
majority has been destroyed.) Although be-
cause of the scatter the values of the
distribution’s statistical parameters that are
reported in Table 1 are only estimates, the
qualitative effect of the growing tempera-
ture is unmistakable. In this form of presen-
tation the effect is also expressed in a man-
ner and terms that are both intuitively
meaningful and visually clear.
Another example is the combined effect
of high carbon dioxide pressure and tem-
perature on the survival of E. coli. As shown
in Figures 15 and 16, an elevated tempera-

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

366
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 8. The survival curves of the B. stearothermophilus spores fitted with Eq. 10 as a model and its
underlying distribution of resistances.

ture increased the efficacy of the high-pres-
sure treatment. This was manifested in a
noticeable shift in the distribution’s mode
toward a lower value. The same thing hap-
pened when the temperature was held con-
stant and the pressure was increased, as
shown in Figures 17 and 18 — another dem-
onstration of the synergistic effect of high
temperature and high pressure on microbial
lethality.
The effect of pH on the efficacy of a heat
treatment can be described in a similar man-
ner (Figures 19 and 20). In this case lower-
ing the pH shifted the mode of the Vibrio’s

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

367
Downloaded by [171.67.34.205] at 15:00 28 June 2013
FIGURE 9. Semilogarithmic survival curve of S. typhimurium fitted with Eq. 11 as a model. Note the
noticeable upward concavity. (Data from Ellison et al. [1994]).
distribution of resistances to a lower level
and rendered the distribution considerably
narrower.
The reader will notice that in all the
examples shown here the log S vs. t relation-
ship had a noticeable curvature, with up-
ward or downward concavity. The analysis,
however, would be exactly the same had
these relationships been all linear. In terms
of the proposed concept, a linear semiloga-
rithmic survival curve only means that the
spectrum of the heat resistances of the or-
ganism in question happens to have a Weibull
distribution with n = 1.
IV. IMPLICATIONS IN FOOD
PRESERVATION
Given the large number of papers report-
ing deviations from first-order microbial in-
activation kinetics (see above), it seems in-
credulous that it was ever assumed that a
population of bacteria or spores are identical
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
368
in their resistance to a lethal stress factor.
Indeed, it is difficult to think of another ex-
ample from the field of biology where the
members of a population would be assumed
to have identical resistance to a lethal agent.
However, it is worth remembering that at the
time when the original ideas of the “botuli-
num cook” were derived (Bigelow and Esty,
1920) not only were microbiology experi-
ments more rudimentary but also the degree
of process control possible in the emerging
canning industry was very limited (Thorne,
1986). This meant that anything beyond a
simple robust control target would have been
inappropriate. For example, at the time when
Stumbo first proposed the 12 decades reduc-
tion of Clostridium botulinum spores as the
target for food sterility, the canning industry
was using manually controlled retorts with
mercury thermometers and stop clocks to
monitor the autoclavization process (Thorne,
1986). By the time Stumbo published
Thermobacteriology in Food Process-
ing (Stumbo, 1973), the canning indus-
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 10. The survival curves of the S. typhimurium fitted with Eq. 10 as a model and its underlying
distribution of resistances.

try had introduced semiautomatically oper-
ated steam valves but was still operating
mainly large vertical batch retorts for large
cans. In an attempt to assure that all cans in
a retort experience greater than the mini-
mum required heat treatment, retorts are
equipped with many thermocouples and the
process set on the coolest part of the product
(usually the geometric center) in the coolest
can. The consequence of this is that most
sterilized products receive an average lethal
treatment far in excess of the target of 3 min
at 121°C. Thus, it is not surprising that the
concept of minimum “botulinum cook” has

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

369
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 11. Semilogarithmic survival curve of L. monocytogenes at 55°C fitted with Eq. 11 as a model. Note
the noticeable upward concavity. (Data from Augustin et al. [1998]).

never really been challenged. During the
1980s and 1990s, there have been a number
of advances in sterilization equipment and
process control. They included the applica-
tion of overpressure to small cans and semi-
rigid containers, the use of rotary retorts for
faster heat transfer and, more recently, the
installation of computerized controls based
on advanced algorithms.
Progress in technologies such as ultra
high temperature treatment and microwave
and ohmic heating opens the possibility to
adhere more closely to the target of minimal
process requirements. These technologies and
other nonthermal preservation methods such
as the exposure to ultra-high hydrostatic pres-
sure or pulsed high-voltage electric field
(Barbosa-Canovas et al., 1998) have arisen
because of the demand for shelf-stable foods
that are less heavily processed. However,
while the demand for fresher and more “natu-
ral” food products increased, so did the con-
cern for their safety. The above trends mean
that now more than ever before there is a
need for a unified approach to the evaluation
of the efficacy of microbial inactivation pro-
cesses. It will have to be suitable not only for
different conditions but also for different
technologies and their combinations. Some
of the possible implications of such an ap-
proach to food preservation are given below.
A. Over- and Underprocessing
In the traditional method of thermal pro-
cess calculation based on first-order mortal-
ity kinetics, and a survival ratio of 10–12
serving as the lethality criterion, the thermal
death time is determined by extrapolation of
the linear log10 S(t) vs. t relationship to –12
(Eq. 5). If the relationship is not linear, that
is n ≠ 1, a discrepancy will arise between the
calculated thermal death time calculated by
the two methods. As the objective of this
paper is not to challenge the validity of the

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

370
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 12. The survival curve of the L. monocylogenes at 55°C fitted with Eq. 10 as a model and its
underlying distribution of resistances.

S(t) = 10–12 criterion, although it too may
need a revision, we will proceed as if the
reduction of a microbial population size by
12 orders of magnitude is an acceptable
measure of microbial lethality. The question
of how valid is any value calculated by ex-
trapolation over a range of 5 to 7 orders of
magnitudes of missing data will also be ig-
nored, and it will be assumed that any math-
ematical relationship describing the S(t) or
log S(t) vs. t relationship indeed accounts for
survival in the whole pertinent time range.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

371
Downloaded by [171.67.34.205] at 15:00 28 June 2013
FIGURE 13. Semilogarithmic survival curves of L. monocytogenes at 60°C grown at 19 and 37°C fitted with
Eq. 11 as a model. (Data from Bhaduri et al. [1991]).
It is obvious that if the semilogarithmic
survival curve has n > 1, that is, the log S(t)
vs. t has a downward concavity, then the
extrapolated time to reach lethality can be
considerably shorter than that based on first-
order kinetics. The difference will depend
on the actual curvature of the log S(t) vs. t
relationship. In such cases the conventional
calculation of a process efficacy, based on
the assumption of a first-order kinetics, will
result in a theoretical overkill and therefore
it has no safety implications. (It may affect
other quality attributes, but these should not
concern us here.) The situation is very dif-
ferent, however, if n < 1, that is, when the
log S(t) vs. t has an upward concavity. In
such a case the calculated thermal death time
based on first-order kinetics, at least theo-
retically, will be shorter than that truly
needed. For example, in the work of Ander-
son et al. (1996), the log-logistic model used
suggested that processes below 111°C that
are currently deemed to be equivalent to a
process of 121°C by traditional log-linear
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
372
kinetic calculations may indeed be deliver-
ing less at a reduction in C. bolutinum spores
than expected. Obviously, if the value S(t) =
10–12 is itself an unrealistically low value,
then organisms having n < 1 can still be
effectively destroyed, although not by the
assumed kinetics. Mathematically, the dis-
crepancy between the thermal death time, τ,
calculated on the basis of the two interpreta-
tion methods is given by:
τ( b, n ) = ( kτ1st order b)
1/ n
(16)
where τ(b,n) is the thermal death times cal-
culated from the nonlinear log S(t) vs. t re-
lationship, and τ1storder the time calculated on
the assumption of a first-order kinetics with
a rate constant k. A similar kind of discrep-
ancy will also arise if any other unimodal
distribution is used to characterize the sur-
vival curve. (It can be shown that different
distributions can be used interchangeably if
the data have sufficient scatter. However,
and as has already been mentioned, the shape
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 14. The survival curves at 60°C of L. monocytogenes grown at 19 and 37°C fitted with Eq. 10 as
a model and their underlying distributions of resistances. Note that growing the organism at a higher
temperature not only shifted its distribution but also made it broader.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

373
Downloaded by [171.67.34.205] at 15:00 28 June 2013
FIGURE 15. Semilogarithmic survival curves of E. coli exposed to high hydrostatic pressure at different
temperatures fitted with Eq. 11. (Data from Ballestra et al. [1996].)
of the distribution’s “tail” can be highly sen-
sitive to the mathematical formulation of the
selected distribution function).
B. Establishing Equivalent
Preservation Regimes
One of the components of the GATT
“Sanitary and Phytosanitary” agreement that
will have far-reaching effects on interna-
tional trade in food and food products is the
requirement for countries to provide risk
assessments as part of the process of resolv-
ing disputes that involve food sanitary issues
(ICMSF, 1997).
Despite the extensive use of quantitative
risk assessment for the evaluation of risks
associated with chemicals, pharmaceuticals,
environmental impact of technologies and
economic investment strategies, the use of
quantitative risk assessment for evaluation of
microbiological concerns related to food safety
is still only an emerging field (ICMSF, 1997).
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
374
One important aspect of quantitative
microbial risk assessment is the ability to
reliably estimate the effect of different pro-
cessing parameters on the numbers of de-
stroyed, or surviving, microorganisms and
to compare the equivalency of different re-
gimes.
The comparison of the efficacy of differ-
ent inactivation regimes has undoubtedly
been hampered by a prior assumption that
the underlying kinetics must be of first or-
der. The decimal reduction time is usually
used for comparison between organisms and
to calculate the time needed for the required
destruction of the microbial population. Dis-
putes may arise because of the different ways
D values are calculated by different labora-
tories if the latter follow a different number
of log cycle reductions. If the underlying
kinetic is first order, then the death rate is
assumed to be constant and hence the esti-
mation of D value would be little affected
whether a 2 to 3 or 5 to 6 decades reduction
is followed. If, however, the underlying dis-
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 16. The survival curves of E. coli exposed to high hydrostatic pressure at different temperatures
and the underlying distributions of resistances to the treatment. Note the shift in the distributions mode and
the changes in their span.

tribution of sensitivities has n ≠ 1, then the ing distribution of resistances as closely as

“D value” will progressively change. In such
a case if an “average” or “overall” D is
calculated its magnitude will depend on the
number of log reductions that are followed.
For quantitative risk assessment it would
clearly be preferable to estimate the underly-
possible and this would require following a
large reduction in the population numbers,
that is, of as many decades as necessary to
accomplish stability. This would eliminate
the need to extrapolate and to assume that the
model is valid beyond the experimental data.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

375
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 17. Semilogarithmic survival curves of E. coli exposed to different levels of hydrostatic pressure
at 45°C fitted with Eq. 11 as a model. (Data from Ballestra et al. [1996]).

If the semilogarithmic survival curve
clearly shows a deviation from linearity, the
experimenter attempting to determine D val-
ues has the problem of which is the most
appropriate portion of the data to use in the
calculation. The experimenter usually takes
one of three options: (1) the D value is cal-
culated from the steepest (usually the
straightest) part of the curve in which case,
the resulting extrapolation could be frail-
dangerous, or (2) the D value is calculated
from the shallowest part of the curve in which
case the resulting process may be excessive,
or (3) all of the data points are used and the
calculated D value will be influenced by
both the shape of the curve and the experi-
mental time scale as already mentioned.
Another area where the a priori assump-
tion of first-order inactivation kinetics ought
to be reassessed is in the evaluation of novel
preservation methods such as ultra-high-pres-
sure and pulsed-field treatments (Barbosa-
Canovas et al., 1997). This is an area where
it will be important to use a consistent ex-
perimental protocol to derive data for the
purpose of establishing process safety crite-
ria (ICMSF, 1997).
C. Mechanistic Aspects
In addition to the practical implications
of the above approach, an appreciation of
the underlying population resistance distri-
bution and how its properties change under
different conditions can also be helpful in
understanding the mechanism of microbial
resistance. A good example here is the cor-
relation of the denaturation of the 30’s ribo-
some with the inactivation of L. mono-
cytogenes (Anderson et al., 1991). The
thermogram of the 30’s ribosome indicated
that its degree of deactivation with heat also
had a similar corresponding distribution.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

376
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 18. The survival curves of E. coli exposed to different levels of hydrostatic pressure at 45°C fitted
with Eq. 10 as a model and their underlying distribution of resistances to the treatment. Note the shifts on
the distributions mode and the changes on their span.

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

377
Downloaded by [171.67.34.205] at 15:00 28 June 2013

FIGURE 19. Semilogarithmic survival curves of

V. vulnificus exposed at 50°C at two pH levels fitted
with Eq. 11 as a model. (Data from Ama et al.
[1994]).

The counterargument to an underlying

population biovirability is the notion that
there is a distribution around the chance of a
quanta of heat striking a particular target or
targets within the cell (McKee and Gould,
1988). While it is true that the approach
described in this article does not distinguish
between these two possibilities, it is sug-
gested that the existence of a distribution FIGURE 20. Survival curves of V. vulnificus at 50°C
must be at least partly be due to the natural at two pH levels and their underlying distributions of
resistances. Note that lowering the pH had little
biovariability. One fact that is quite often effect on the distribution’s mode but a considerable
overlooked is that inactivation kinetics de- influence on its spread.
rived from observations of the ability of or
inability of spores, or cells, to germinate,
and/or grow, following the lethal treatment. produces the observed distributions. They
Therefore, it is interesting to note that spore may also assist in the development of mecha-
germination time itself shows a characteris- nistic inactivation models that will explain
tic skewed distribution (Coote et al., 1995). these distributions in terms of properties and
The recent development of single cell events at the cellular and molecular levels.
techniques such as flow cytometry (Nebe-
von-Cavon and Balley, 1996) and image
analysis (Coote et al., 1995) now allow direct ACKNOWLEDGMENT
measurement of microbial population distri-
butions. Therefore, they may offer a way for This is a contribution of the Massachu-
direct confirmation of the biovariability that setts Agricultural Experiment Station at

Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.

378
 Amherst. The opinions expressed in this
paper are those of the authors only and need
not represent their respective institutions.
REFERENCES
Ama, A. A., Hamdy, M. K., and Toledo, R. T.
Effect of heating, pH and thermo radiation on
inactivation of V. vulnificus. Food Micobiol.,
11, 215, 1994.
Anderson, W. F., Hedges, H. D., Jones, M. V., and
Cole, M. B. Thermal inactivation of Listeria
monocytogenes studied by differential scan-
ning calorimetry, J. Gen. Microbiol., 137, 1419-
1424, 1991.
Downloaded by [171.67.34.205] at 15:00 28 June 2013
Anderson, W. F., McClure, P. J., Baird-Parker,
A. C., and Cole, M. B. The application of a
log-logistic model to describe the thermal inac-
tivation of C. botulinum 213B at temperatures
below 121.1°C. J Appl. Bacteriol., 80, 283,
1996.
Anderson, W. A. Non-Log-Linear Thermal Inacti-
vation of Clostridium botulinum Spores, Ph.D.
thesis, University of Reading, 1997.
Augustin, J. C., Carlier, V., and Roziier, J. Math-
ematical modeling of the heat resistance of
L. monocytogenes. Appl. Microbiology, 84, 185,
1998.
Bahduri, S., Smith, P. W., Palumbo, S. A., Turner-
Jones, C. O., Smith, J. L., Marmer, B. S.,
Buchanan, R. L., Zaika, L. L., and Williams,
A. A. Thermal destruction of L. monocylogenes
in liver sausage slurry. Food Microbiol., 8, 75,
1991.
Ball, C. O. and Olson, F. C. W. Sterilization in
Food Technology. McGraw-Hill, New York,
1957.
Ballestra, P., Abreu, daSilva, A., and Cuq, J. L.
Inactivation of E. coli by carbon dioxide under
pressure. J. Food Sci., 61, 829, 1996.
Barbosa-Canovas, G. V., Pothakamury, U. R., Palou,
E., and Swanson, B. G., Eds., Non-Thermal
Preservation of Foods, Marcel Dekker, Inc.,
New York, 1998.
Bigelow, W. D. The logarithmic nature of thermal
death time curves. J. Infect. Dis., 19, 528, 1921.
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
Biglow, W. D. and Esty, J. R. Thermal death point
in relation to time of typical thermophilic or-
ganisms. J. Infect. Dis., 17, 602, 1920.
Bunning, V. K., Crawford, R. G., Temey, J. T., and
Peeler, J. T. Thermoresistance of L. monocyto-
genes and S. typhimurium after sub lethal heat
shock. Appl. Environ. Mirobiol., 56, 3216, 1990.
Casolari, A. Microbial Death. In: Bazin, M. J. and
Prosser, J. I., Eds., Physiological Models in
Microbiology, Vol. II. CRC Press, Boca Raton,
FL, 1988, 1–44.
Casolari, A. About basic parameters of food steril-
ization technology. Food Microbiol. 11, 75,
1994.
Cole, M. B. The outlook for novel preservation
technologies: A food industry prospective. Pa-
per presented at the COFE 97, AICHE annual
meeting at Los Angeles, November 1997.
Cole, M. B., Davies, K. W., Munro, G., Holyoak,
C. D., and Kilsby, D.C. A vitalistic model to
describe thermal inactivation of L. mono-
cytogenes. J. Ind. Microbial., 12, 232, 1993.
Ellison, A., Anderson, W., Cole, M. B., and Stewart,
G. S. A. B. Modeling the thermal inactivation
of Solmonella typhimurium using biolumines-
cence data. Int. J. Food Microbiol., 23, 467,
1994.
Hills, B. P. and Mackey, B. M. Multi-compartment
kinetic models for injury, resuscitation induced
lag and growth in bacterial cell populations.
Food Microbiol., 12, 335, 1995.
Holdsworth, S. D. Thermal Processing of Pack-
aged Foods, Blackie Academic Professional,
London, 1997.
ICMSF. Principles for the establishment of control
measures to meet food safety objectives. Draft
document presented at the 30th Annual meet-
ing of the ICMSF, France, 1997.
Jay, J. M. Modern Food Microbiology, Chapman
Hall, New York, 1996.
LeJean, G., Abraham G., Debray, E., Candau, Y.,
and Piar, G. Kinetics of thermal destruction
of B. stearothermophilus spores using a two
reactions model. Food Microbiol., 11, 229,
1994.
Linton, R. H., Carter, W. H., Pierson, M. D., and
Hackney, C. R. Use of a modified Compertz
379
 equation to model nonlinear survival curves
for L. monocytogenes Scott, A. J. Food Pro-
tect., 58, 946, 1995.
Little, C. L., Adams, M. R., Anderson, W. A., and
Cole, M. B. Application of a log-logistic
model to describe the survival of Yersinis
entero colitica at sub- optimal pH and
temperature. Int. J. Food Microbiol., 22, 63,
1994.
McClure, P. J., Blackburn, C.de W., Cole, M. B.,
Curtis, P. S., Jones, J. E., Legan, J. D., Ogden,
I. D., Peck, M. W., Roberts, T. A., Sutherland,
J. P., and Walker, S. J. Modeling the growth,
survival and death of microorganisms in foods:
the U.K. food micromodel approach. Int. J.
Food Microbiol., 23, 265, 1994.
McKee, S. and Gould, G. W. A simple mathemati-
cal model of the thermal death of microorgan-
Downloaded by [171.67.34.205] at 15:00 28 June 2013
isms. Bull. Math. Biol., 50, 5, 493–501.
Nebe-von-Caron, G. and Balley, A. Bacterial char-
acterization by flow cytometry, In: Flow
Cytometry Applications in Cell Culture, Al-
Rubeai, M. and Emery, A. N., Eds., Marcel
Dekker, New York, 1996.
Patel, J. K., Kapadie, D. H., and Owen, D. B. Hand-
book of Statistical Distributions, Marcel-
Dekker, New York, 1976.
Peleg, M. Evaluation of the Fermi equation as a
model of doze response curves. Appl. Microbiol.
Biotechnol., 46, 303, 1996.
Peleg, M. Modeling microbial populations with the
original and modified version of the continu-
Copyright© 1998, CRC Press LLC — Files may be downloaded for personal use only. Reproduction of this material
without the consent of the publisher is prohibited.
380
ous and discrete logistic equations. Crit. Rev.
Food Sci. Nutrit., 37, 471, 1997.
Peleg, M., Normand, M. D., and Damrau, E. Math-
ematical interpretation of dose- response curves.
Bull. Math. Biol., 59, 747, 1997.
Rank, E. and Pflug, I. G. Dry heat destruction of
spores in metal surfaces and on potato during
baking. J. Food Protect., 40, 608, 1977.
Stephens, P. J., Cole, M. B., and Jones, M. V. Ef-
fect of heating rate on the thermal inactivation
of Listeria monocytogenes. J. Appl. Bacteriol.,
77, 702, 1994.
Stumbo, C. R. Thermobacteriology in Food Pro-
cessing, 2nd ed., Academic Press, New York,
1973.
Teixeira, A. Thermal processing calculations, in
Handbook of Food Engineering, Heldman,
D. R. and Lund, L. B., Eds., Marcel Dekker,
New York, 1992.
Thome, S. The History of Food Preservation,
Parthenon Publishing Group Ltd., Casterton
Hall, Lancs, UK, 1986.
Whiting, R. C. Microbial modeling in foods, Crit.
Rev. Food Sci. Nutrit., 35, 467, 1995.
Whiting, R. C. and Buchanan, R. L. Development
of quantitative model for Salmonella enteriti-
dis in pasteurized liquid egg. Int. J. Food
Microbiol., 36, 111, 1997.
Withell, E. R. The significance of the variation on
shape of time-survivor curves. J. Hygiene, 42,
2, 124, 1942.