International Journal of Food Microbiology 104 (2005) 239 – 248

www.elsevier.com/locate/ijfoodmicro

Prediction of pathogen growth on iceberg lettuce under real

temperature history during distribution from farm to table
Shigenobu Koseki*, Seiichiro Isobe
Food Processing Laboratory, National Food Research Institute, 2-1-12, Kannondai, Tsukuba, Ibaraki 305-8642, Japan
Received 29 November 2004; received in revised form 1 February 2005; accepted 19 February 2005

Abstract

The growth of pathogenic bacteria Escherichia coli O157:H7, Salmonella spp., and Listeria monocytogenes on iceberg
lettuce under constant and fluctuating temperatures was modelled in order to estimate the microbial safety of this vegetable
during distribution from the farm to the table. Firstly, we examined pathogen growth on lettuce at constant temperatures,
ranging from 5 to 25 8C, and then we obtained the growth kinetic parameters (lag time, maximum growth rate (l max), and
maximum population density (MPD)) using the Baranyi primary growth model. The parameters were similar to those predicted
by the pathogen modelling program (PMP), with the exception of MPD. The MPD of each pathogen on lettuce was 2–4 log10
CFU/g lower than that predicted by PMP. Furthermore, the MPD of pathogens decreased with decreasing temperature. The
pffiffiffiffiffiffiffiffiffi
relationship between lmax and temperature was linear in accordance with Ratkowsky secondary model as was the relationship
between the MPD and temperature. Predictions of pathogen growth under fluctuating temperature used the Baranyi primary
microbial growth model along with the Ratkowsky secondary model and MPD equation. The fluctuating temperature profile
used in this study was the real temperature history measured during distribution from the field at harvesting to the retail store.
Overall predictions for each pathogen agreed well with observed viable counts in most cases. The bias and root mean square
error (RMSE) of the prediction were small. The prediction in which l max was based on PMP showed a trend of overestimation
relative to prediction based on lettuce. However, the prediction concerning E. coli O157:H7 and Salmonella spp. on lettuce
greatly overestimated growth in the case of a temperature history starting relatively high, such as 25 8C for 5 h. In contrast, the
overall prediction of L. monocytogenes under the same circumstances agreed with the observed data.
D 2005 Elsevier B.V. All rights reserved.

Keywords: Pathogen growth; Lettuce; Fluctuating temperature; Real temperature history; Predictive microbiology

1. Introduction

* Corresponding author. Tel.: +81 29 838 8029; fax: +81 29 838

8122.
E-mail address: koseki@nfri.affrc.go.jp (S. Koseki).
0168-1605/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijfoodmicro.2005.02.012
Fruit and vegetable contamination can occur during
the growing stage. Fruit and vegetables can be contam-
inated during growth from many sources, such as soil,
240 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248
water, wild animals, birds, and insects. Following pro-
duction, processes involving harvesting, washing, cut-
ting, packaging, and shipping could create additional
contamination. When fruit or vegetables are consumed
raw, as is the case with salads, harmful microorganisms
may be present and ingested. Traditionally, eating raw
fresh fruits and vegetables from the field was consid-
ered safe; however, bacterial pathogens are currently
being found in or on fruits and vegetables (Ackers et al.,
1998; Beuchat, 1996; De Roever, 1998; Mead et al.,
1999). Pathogenic bacteria of concern include Escher-
ichia coli O157:H7, Salmonella spp., and Listeria
monocytogenes. Although various sanitizers have
been examined for their effectiveness, most treatments
have been shown to have minimal effects, resulting in a
less than 2 log10 CFU/g reduction of bacterial numbers.
The bacteria remaining in or on the contaminated pro-
duce after treatment will inevitably grow during distri-
bution. Since sanitizer treatment alone is not enough to
keep the produce safe, low temperature management
and a presentation of the information concerning bac-
terial growth predictions in the food chain will play an
important role.
Modelling the growth and survival of pathogenic
and spoilage microorganisms is a basic tool for the
prediction of food safety and the microbial deteriora-
tion of food products in the food chain (McMeekin et
al., 1993). Although numerous bacterial growth models
have been published, few predictive models have been
constructed with respect to fresh produce (Riva et al.,
2001; Rodriguez et al., 2000; Viswanathan and Kaur,
2001). These models, however, provide predictions of
bacterial growth during constant environmental condi-
tions. A dynamic model has been developed by Bar-
anyi and Roberts (1994) which can successfully
provide predictions of bacterial growth in fluctuating
temperature conditions (Baranyi et al., 1995). Bovill et
al. (2000, 2001) reported that in general, the Baranyi
model provided accurate predictions of L. monocyto-
genes and Salmonella growth in a broth, as well as in
meat and milk products. Shorten et al. (2004) applied
the Baranyi model to Erwinia carotovora in vegetable
juice under conditions of fluctuating temperature.
However, at present, there are no reports concerning
bacterial growth predictions of salad vegetables under
fluctuating temperature.
In the present study, we modelled the growth of
pathogenic bacteria E. coli O157:H7, Salmonella
spp., and L. monocytogenes on iceberg lettuce under
fluctuating temperature during distribution of the let-
tuce from the field to the retail store. We examined
pathogen growth on lettuce at constant temperatures,
ranging from 5 to 25 8C, and we then calculated the
growth kinetic parameters and secondary models. The
derived parameters and secondary models were then
applied to predict pathogen growth under fluctuating
temperature. The temperature profile of the lettuce was
measured from the field at harvesting to the retail store
in order for our experimental protocol to reflect actual
distribution conditions. Predictions of pathogenic
growth under real temperature history during distribu-
tion from the farm to the retail store were validated on
lettuce kept in a programmable incubator. Our findings
will serve as a basis for the strengthening of tempera-
ture management strategies, and will provide valuable
safety information to consumers.
2. Materials and methods
2.1. Bacteria strains
Six strain suspensions of E. coli O157:H7 were
used: ATCC 35150, ATCC 43889, ATCC 43895,
ATCC 51657, ATCC 700378, and ATCC BAA-460.
The Salmonella spp. used represented a mixture of
Salmonella enteritidis and Salmonella typhyimurium.
Two strain suspensions of S. enteritidis were used:
ATCC BAA-708 and ATCC 4931. Three strain sus-
pensions of S. typhyimurium were used: ATCC
29057, ATCC 29629, and 29630. Six strain suspen-
sions of L. monocytogenes were used: ATCC 19111,
ATCC 19117, ATCC 19118, ATCC 13932, ATCC
15313, and ATCC 35152. All strains were maintained
at  80 8C in tryptic soy broth (TSB, Merck, Darm-
stadt, Germany) containing 10% glycerol. On the day
before the experiment, a sterile disposable plastic
stick was used to transfer the frozen bacterial cultures
to 10 ml of TSB for E. coli O157:H7 or Salmonella
spp. in a glass tube. Brain heat infusion broth (BHI,
Merck, Darmstadt, Germany) was used to transfer the
frozen L. monocytogenes cultures. The cultures were
incubated without agitation at 37 8C for 4 h, then
transferred to another tube of fresh TSB or BHI broth
and incubated at 37 8C for 18 h without agitation. A
mixture containing equal numbers of cells from each
 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248
strain of each kind of bacterium was used as the
inoculum.
2.2. Inoculation and incubation of lettuce
Iceberg lettuce was purchased from a local super-
market. The outer three or four leaves and the core
were removed from the lettuce head and discarded.
Intact leaves were cut into 3  3 cm pieces using a
sterile surgical knife.
The 100-g leaf samples were placed into separate
plastic bags (25  20 cm) to facilitate bacterial inoc-
ulation. Each bag was inoculated with 1 ml of 0.1%
peptone water (pH 7.1; Merck) containing each path-
ogen cocktail at a final concentration of 3 to 4 log10
CFU/g, and then shaken gently 30 times to ensure an
even distribution of bacterial cells on the lettuce. The
inoculated leaves were air-dried under a safety cabi-
net at room temperature for 1 h before use in the
experiment.
The inoculated lettuce samples (10 g) were asepti-
cally divided into 400-ml stomacher bags (Gunze,
Tokyo, Japan) at each sampling interval, and stored
at constant or fluctuating temperature.
2.3. Growth under constant temperature
The procedure followed is as outlined in a previous
paper concerning L. monocytogenes (Koseki and
Isobe, 2005). Lettuce samples inoculated with E.
coli O157:H7 or Salmonella spp. (10 g) were each
aseptically divided into 400-ml stomacher bags
(Gunze, Tokyo, Japan) at each sampling interval,
and stored at 5, 10, 15, 20, and 25 8C until they
reached the stationary phase. Sampling was generally
carried out at 1-h intervals for the first 20 h of incu-
bation at 15, 20, and 25 8C, although longer sampling
intervals were used after this time. However, longer
intervals were selected for the experiment at 5 and 10
8C experiments. Each experiment consisted of three
trials with two replicates per trial and two plates per
replicate at each time interval.
The growth curves were fitted and growth para-
meters (maximum growth rate (l max), lag time, and
maximum population density (MPD)) were deter-
mined using the Baranyi model (Baranyi and Roberts,
1994) with DMFit Excel Add-In software (kindly
provided by J. Baranyi, Institute of Food Research,
241
Norwich, UK). As a comparison, the US Department
of Agriculture (USDA) Agricultural Research Servi-
ce’s Pathogen Modelling Program (PMP, version 7.0)
was used to predict the growth parameters of each
pathogen in the broth.
2.4. Growth under real temperature history during
distribution
Temperature fluctuations experienced by the let-
tuce during distribution from the farm to the retail
store were measured by a compact data logger with a
built-in temperature sensor (Cool Memory, f 17
mm  5.8 mm, SEC-CD16TB, SANYO, Osaka,
Japan). The logger was inserted into the centre of
the lettuce when harvested in the field in Numata,
Gunma, Japan. Following transportation, the logger
was removed when the lettuce arrived at the super
market in Tsukuba. The data were then stored on
computer disk in the form of an ExcelR file.
Lettuce samples inoculated with E. coli O157:H7,
Salmonella spp., or L. monocytogenes (10 g) were
aseptically divided into 400-ml stomacher bags, and
then put in the programmable incubator (SH-221,
ESPEC, Tokyo, Japan) which set a temperature regime
that reflected real temperature fluctuations collected
during lettuce distribution by the method described
above. Sampling was generally carried out at 6-h inter-
vals up to 120 h. Each experiment consisted of three
trials with two replicates per trial and two plates per
replicate at each time interval.
2.5. Enumeration of bacterial population
Each 10-g sample of lettuce was combined with
90 ml of 0.1% peptone water in a 400-ml stomacher
bag and pummelled for 2 min in a blender (TC82;
CDC, Italy). The sample was then serially diluted in
0.1% peptone water and plated in duplicate (0.1 ml)
on each appropriate selective agar plates. E. coli
O157:H7 was enumerated on sorbitol MacConkey
agar (SMAC, Merck, Germany) supplemented with
CT selective supplement (Merck, Germany), and in-
cubated at 37 8C for 24 h. The presence of E. coli
O157:H7 was confirmed using the latex agglutination
test (E. coli single-path, Merck, Germany). Salmo-
nella spp. was enumerated on bismuth sulfate agar
(BSA, Merck, Germany) incubated at 37 8C for 24 h.
242 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248

The presence of Salmonella spp. was confirmed by 3. Results

testing reactions on triple sugar iron (Merck) slants.
L. monocytogenes was enumerated on Oxford Lis- 3.1. Growth modelling of E. coli O157:H7 and Sal-
teria-selective agar supplemented with Oxford Lis- monella spp. on lettuce under constant temperature
teria-selective supplement (Merck), which was
incubated at 37 8C for 24 h. The presence of L. Representative data on the growth of E. coli
monocytogenes was confirmed using a Listeria diag- O157:H7 and Salmonella on lettuce at different tem-
nosis kit (SinglepathR Listeria; Merck). peratures are shown in Fig. 1 with fitted growth
curves produced using the Baranyi model; these pro-
2.6. Mathematical model vided a good statistical fit to the observed data. The
curves obtained from DMFit at all temperatures ex-
The model applied to growth data obtained from cept 5 8C showed a high correlation coefficient
the real temperature history experiment was the cou- (R 2 N 0.95).
pled set of differential equations of Baranyi et al.
(1995):
E. coli O157:H7
7
dq 25°C
¼ lmax q; qð 0Þ ¼ q0 E.coli O157:H7 (Log10 CFU/g)
20°C
dt
  15°C
6 10°C
dx q x
¼ lmax 1 x; x ð 0Þ ¼ x 0 ð1Þ
dt 1þq xmax
where x(t) denotes the natural logarithm of the cell 5
5°C
concentration, q(t) is a dimensionless quantity related
to the physiological state of the cells, l max is the
4
maximum specific growth rate, x max represents the
maximum population density (MPD), and x 0 is the
initial population size. The lag time variable q evolves 3
according to logistic growth and q 0 determines the 0 20 40 60 80 100 120
duration of the lag time in microbial growth. It is Time (h)
assumed that the maximum specific growth rate,
l max, is determined by the external environment. In 7
Salmonella
the present study, the external environmental is defined
25°C
only by temperature. The effect of temperature on the
Salmonella (Log10 CFU/g)

maximum growth rate (l max) of a microbial population 6 20°C

can be described using the simple square root model of

15°C
Ratkowsky et al. (1982), 10°C

pffiffiffiffiffiffiffiffiffi 5
lmax ¼ bðT  Tmin Þ ð2Þ
5°C
where b is a constant, T is the temperature, and T min is
4
the conceptual minimum temperature for microbial
growth.
The model for l max was substituted into the 3
above differential equation, and the temperature 0 20 40 60 80 100 120
allowed to be dependent on time. The system was Time (h)
solved numerically by the forth-order Runge–Kutta
Fig. 1. The observed growth of E. coli O157:H7 and Salmonella
method as a means of obtaining predictions of spp. on lettuce at constant temperature ranging from 5 to 25 8C. The
bacterial concentration during time-dependent tem- growth curves were obtained using the Baranyi model. Data are
perature fluctuations. mean values of triplicate trials F standard error.
 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248 243

Table 1
Estimated maximum growth rate, lag time and maximum population density (MPD) of E. coli O157:H7 or Salmonella inoculated on the lettuce
and calculated from pathogen modelling program (PMP)
Temperature E. coli O157:H7a Salmonellaa
(8C) Growth rate Lag time MPD Growth rate Lag time MPD (log10 CFU/g)
(log10 CFU/h) (h) (log10 CFU/g) (log10 CFU/h) (h)
Lettuce PMP Lettuce PMP Lettuce PMP Lettuce PMP Lettuce PMP Lettuce PMP
b
5 ND ND ND ND ND ND ND ND ND ND ND ND
10 0.03 0.06 32.58 47.12 5.89 9.02 0.02 0.02 35.06 39.87 5.04 9.00
15 0.12 0.15 7.40 13.40 6.02 9.05 0.13 0.10 6.45 13.41 5.39 8.74
20 0.26 0.29 5.17 4.54 6.33 9.10 0.26 0.25 5.23 4.74 6.01 8.99
25 0.44 0.53 4.44 2.21 6.80 9.01 0.41 0.48 2.78 3.20 6.51 8.93
a
Mean value of triplicate trials.
b
Not determined.

The growth rates of each bacterium on lettuce were shown in Fig. 2. The transportation distance was
similar to those predicted by the PMP at all tempera- about 400 km and the total time from farm to retail
tures except 5 8C (Table 1). Both of these pathogenic store was 72 h. The temperature of the lettuce just
bacteria on lettuce showed a shorter lag time than that after being harvested was 15–17 8C, and 3 h later the
predicted by the PMP at 10 and 15 8C. In contrast, a lettuce was pre-cooled by 5 8C in a vacuum refriger-
longer lag time was shown in both pathogens on
lettuce at 20 and 25 8C relative to that calculated by Table 2
the PMP. A 2–3 log10 CFU/g lower MPD on lettuce The relationship between the square root of maximum growth rate
pffiffiffiffiffiffiffiffiffi
was shown for both pathogens relative to that pre- ( lmax Þ and temperature in each pathogen inoculated on lettuce
dicted by the PMP. As the temperature rose, the MPD and calculated from pathogen modelling program (PMP)
of both pathogens on lettuce gradually increased. Bacteria Square root equationa R2
However, the MPD measured on lettuce was smaller E. coli O157:H7b Lettuce 0.99
than that predicted by PMP. The MPD predicted by pffiffiffiffiffiffiffiffiffi
lmax ¼ 0:033ðT  4:54Þ
PMP was almost the same value (9 log10 CFU/ml),
regardless of the incubation temperature. PMP 0.99
Square root analysis was carried out to determine pffiffiffiffiffiffiffiffiffi
lmax ¼ 0:032ðT  2:67Þ
the relationship between l max and the temperature of
each bacterium on lettuce and the PMP-predicted data. Salmonella b
pffiffiffiffiffiffiffiffiffi Lettuce 0.98
The relationship between lmax and temperature was pffiffiffiffiffiffiffiffiffi
lmax ¼ 0:033ðT  4:96Þ
linear with a high correlation coefficient of linearity
(R 2 N 0.98, Table 2). The square root model adequately
pffiffiffiffiffiffiffiffiffi PMP 0.99
described this temperature dependence. The lmax of pffiffiffiffiffiffiffiffiffi
lettuce and PMP-predicted data showed almost the lmax ¼ 0:037ðT  6:27Þ
same slope for E. coli O157:H7 and Salmonella
pffiffiffiffiffiffiffiffiffi L. monocytogenes c Lettuce 0.99
spp. In contrast, the lmax of PMP-predicted data in
pffiffiffiffiffiffiffiffiffi
L. monocytogenes showed a larger slope than that lmax ¼ 0:016ðT þ 4:26Þ
observed for lettuce.
PMP 0.99
pffiffiffiffiffiffiffiffiffi
3.2. Temperature history during distribution from lmax ¼ 0:027ðT þ 0:44Þ
farm to table
a
T is the temperature (8C).
The real temperature history of lettuce collected b
Data from this study.
c
from the farm to the retail store in early October is Data from literature (Koseki and Isobe, 2005).
244 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248

30

25

20
Temperature (°C)
Harvest
Pre-cooling

15

10

Field Storage Transportation Display

0
0 10 20 30 40 50 60 70

Time (h)

Fig. 2. A temperature history of lettuce from the field at harvesting to the retail store.

ator. The pre-cooled lettuces were stored at 5 8C in a where T was temperature. These equations were also
storehouse for around 5 h until they were shipped. substituted into Eq. (1) to take varying MPD into
Trucks without any form of temperature control trans- consideration. For the initial value of q(0) for Eq.
ported the lettuces. Lettuce temperature fluctuated in (1), the procedure of Baranyi et al. (1995) was fol-
the range from 3 to 15 8C during transportation. The lowed; a geometric mean value for the physiological
lettuces were displayed on a showcase set at 7 8C after state parameter a 0 = q(0) / ( 1+ q(0)) was estimated
arrival in the retail store. from the constant temperature experimental data
obtained from this study, our previous study, and
3.3. Pathogen growth modelling under real tempera- PMP predicted data (Table 3). It should be noted
ture history during distribution that the relationship between lag time and a 0 can be
shown as follows (Baranyi and Roberts, 1994):
pffiffiffiffiffiffiffiffiffi
Equations of lmax on lettuce and of the PMP-
predicted data for each pathogen were substituted into  lnða0 Þ
Lag time ¼ :
Eq. (1) for a comparison of growth under real tem- lmax
perature history. The maximum population density
(MPD) of E. coli O157:H7 and Salmonella spp. on The viable counts of E. coli O157:H7, Salmonella
lettuce under constant temperature varied depending spp. and L. monocytogenes on lettuce under real tem-
on the temperature. In contrast, the MPD of L. mono- perature conditions are shown in Fig. 3 with growth
cytogenes was almost constant, regardless of the in-
cubation temperature. The relationship between the Table 3
MPD of each pathogen and temperature was described Estimates of the parameter (a 0) of the physiological initial state of
by a linear equation as follows: the cell from constant temperature experiment
Bacteria Base model
MPD(x max, ln CFU/g) = 0.218T + 10.320 (R 2 = 0.97) Lettuce PMP
(E. coli O157:H7) E. coli O157:H7a 0.056 0.033
MPD(x max, ln CFU/g) = 0.257T + 8.641 (R 2 = 0.98) Salmonella spp.a 0.097 0.056
(Salmonella spp.) L. monocytogenes b 0.072 0.048
MPD(x max, ln CFU/g) = 0.037T + 12.434 (R 2 = 0.88) a
Data from this study.
(L. monocytogenes) b
Data from literature (Koseki and Isobe, 2005).
 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248 245

curves predicted using the Baranyi–Ratkowsky model

6 20 in combination with the relationship between MPD and
E. coliO157:H7 temperature. Overall predictions for each pathogen
Predicted (Lettuce based)
Predicted (PMP based)
agreed well with observed viable counts. The predic-
E. coli O157:H7 (log10 CFU/g)

Observed
5 15 tion in which l max was based on PMP tended to be
overestimated. The bias and root mean squared error

Temperature (°C)
(RMSE) of the prediction were small (Table 4, temper-
4 10
ature history dLowT). A prediction was also obtained
under another temperature history that involved an
initial temperature of around 25 8C for 5 h (Fig. 4).
3 5
This situation simulated the harvesting of lettuce on a
Temperature (°C)
hot summer day. We actually measured the temperature
2 0
of lettuce at harvest in a field on a hot summer day; it
0 10 20 30 40 50 60 70 80 was around 25 8C for 3–5 h prior to the lettuce being
Time (h) placed in pre-cooling facilities. The prediction for E.
6 20 coli O157:H7 and Salmonella spp. was greatly over-
Salmonella
estimated, especially for the initial 5 h. A large bias and
Predicted (Lettuce based)
Predicted (PMP based) RMSE of the prediction for E. coli O157:H7 and
Salmonella (Log10 CFU/g)

5 Observed 15
Salmonella spp. resulted from a high temperature his-
Temperature (°C )

tory (Table 4 temperature history dHighT). On the other

hand, the overall prediction for L. monocytogenes was
4 10
in agreement with the observed data.
The large errors associated with the prediction are
3 5 due mainly to inappropriate adjustments of the initial
Temperature (°C )
parameter of a 0. We chose appropriate a 0 values to fit
the curves by a process of trial and error (Fig. 5). The
2 0 curves described by the solid line were obtained by
0 10 20 30 40 50 60 70 80
substituting the appropriate values of a 0 in the differ-
Time (h) ential equation Eq. (1). These lines were the fitted
20
L. monocytogenes curves, while the other dashed-line curves were pre-
6
Predicted (Lettuce based)
Predicted (PMPbased) dictions from independent static data.
L. monocytogenes (log10 CFU/g)

Observed
15
Temperature (°C)

5
4. Discussion
10
An important issue in predictive food microbiol-
ogy is the extent to which results obtained under
4
5 controlled laboratory conditions can be applied to the
Temperature (°C) less controlled environment of a produce distribution
process or an industrial process. The objective of the
3 0 present study was to examine whether the growth
0 10 20 30 40 50 60 70 80
curve under fluctuating temperature conditions, such
Time (h)
as those experienced during lettuce distribution,
Fig. 3. The observed growth of E. coli O157:H7, Salmonella spp. could be predicted from the measured kinetic para-
and L. monocytogenes on lettuce under real temperature history meters of bacterial growth under constant tempera-
involving a relatively low initial temperature, such as 15 8C. The
growth curves were predicted using the Baranyi–Ratkowsky model
ture conditions.
based on the lettuce and PMP model separately. Observed data are Firstly, a series of 5 growth curves for each path-
mean values of triplicate trials F standard error. ogen on lettuce was produced at temperatures of 5, 10,
246 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248
Table 4
Evaluation of the prediction of pathogen growth on the lettuce under real temperature history during distribution calculated by Baranyi–
Ratkowsky model
Bacteria Temperature history Growth rate obtained from
E. coli O157:H7 Low Lettuce
PMP
High Lettuce
PMP
Salmonella spp. Low Lettuce
PMP
High Lettuce
PMP
L. monocytogenes Low Lettuce
PMP
High Lettuce
PMP
a
Root mean squared error.
15, 20, and 25 8C. PMP is probably the most well-
known predictive software package in the world. As
anybody can freely download and use the program
(http://www.arserrc.gov/mfs/pathogen.htm), we chose
the growth model of PMP for our comparative study.
Our results demonstrated that the growth rate pre-
dicted from PMP for E. coli O157:H7 and Salmonella
spp. almost agreed with that observed on lettuce.
However, the PMP predicted growth rate of L. mono-
cytogenes was greater than that observed on lettuce
(Koseki and Isobe, 2005). This variance would be due
to differences in bacterial characteristics.
The MPD of each pathogen on lettuce was mark-
edly limited in comparison to that derived from PMP
predictions. This will be due to competitive native
bacterial flora. Similar findings have been reported
for L. monocytogenes growing on endive (Carlin et
al., 1996) and alfalfa sprouts (Palmai and Buchanan,
2002), Listeria innocua growing on lettuce and cole-
slaw (Francis and O’Beirne, 1998, 2002), and E. coli
O157:H7 on ground beef (Tamplin, 2002; Vold et al.,
2000). In the present study, moreover, the MPD of E.
coli O157:H7 and Salmonella spp. decreased with
decreasing temperature. These limitations of MPD
and its temperature dependence would be due to the
growth of native bacterial flora on lettuce. The domi-
nant component of the bacterial flora on lettuce is
represented by Pseudomonas spp. and they are
regarded as psychrotrophic bacteria (Nguyenthe and
Carlin, 1994). Since the native bacterial flora would
grow faster than the inoculated E. coli O157:H7 or
Bias (log10 CFU/g) RMSEa (log10 CFU/g)
0.03 0.15
0.25 0.41
1.23 1.47
0.96 1.24
0.04 0.16
0.13 0.20
1.21 1.39
1.41 1.57
0.03 0.11
0.10 0.20
0.08 0.15
0.24 0.33
Salmonella spp. at low temperature, the MPD of E.
coli O157:H7 and Salmonella spp. would be inhibited.
The model applied in the present study accurately
predicted the growth of pathogens when the temper-
ature history started at a relatively low temperature
(15 8C) in the first 5 h. Since we applied equations
incorporating the relationship between MPD and tem-
perature in the present model, predictions for the later
stage of growth were also in agreement with observed
data. A model that did not consider the MPD variation
showed large deviations from the observed data dur-
ing the later stages of growth (data not shown). Indi-
cations showed that a modification of the model with
respect to MPD would yield better predictions on later
stages of growth (Baranyi et al., 1995). However, in
the case of the temperature history involving a higher
initial temperature (25 8C) in the first 5 h, the predic-
tions for E. coli O157:H7 and Salmonella spp. devi-
ated markedly from the observed data. This deviation
is caused from an inappropriate initial setting for the
parameter a 0 of the model. The model was made to fit
the observed data by adjusting a 0 (Fig. 5). However,
this procedure is not prediction but only fitting. Al-
though the parameter a 0 used in the model was cal-
culated from the growth parameter obtained on
lettuce, the model deviated greatly from the observed
data. Bovill et al. (2000) mentioned that this is a
consequence of the fact that lag time depends on the
physiological state (and therefore the previous envi-
ronment) of the bacterial cells which cannot be pre-
dicted from other independent data. Baranyi (2002)
 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248 247

indicated that a need exists for the development of a

30 stochastic model for the dynamics of individual cells
7
Predicted (Lettuce based)
Predicted (PMP based)
E. coli O157:H7 to predict bacterial lag time and growth more accu-
Observed
25 rately. In a future study, we will further improve the
E. coli O157:H7 (log10 CFU/g)

6 present model by taking into consideration the statis-

tical technique as reported in Shorten et al. (2004).

Temperature (°C)
20

5
In conclusion, our study revealed that the Baranyi–
15 Ratkowsky model in combination with MPD variation
4
is able to predict the growth of pathogens on lettuce
10 under real temperature history during distribution
from the farm to the retail store in most cases. Pre-
3 5
dicting the growth and behaviour of pathogenic bac-
Temperature (°C)
2 0
0 10 20 30 40 50 60 70 80 30
Time (h) E. coli O157:H7
7
α0=0.056
30 25
Predicted (Lettuce based) Salmonella E. coli O157:H7 (log10 CFU/g)
7 Predicted (PMP based) 6
Observed
25

Temperature (°C)
20
Observed
6
Salmonella (Log10 CFU/g)

5 α0=0.002
20
Temperature (°C)

15
5
15 4
10

4
10 3 5
Temperature (°C)
3 5
2 0
Temperature (°C) 0 10 20 30 40 50 60 70 80
2 0 Time (h)
0 10 20 30 40 50 60 70 80
30
Time (h)
7 Salmonella
30
Predicted (Lettuce based) L. monocytogenes α0=0.097 25
6 Predicted (PMP based)
Salmonella (Log10 CFU/g)

Observed
25 6
L. monocytogenes (log10 CFU/g)

20
Temperature (°C)
5 20 Observed
Temperature (°C)

5 α0=0.002
15
15
4 4
10
10
3 5
3
5 Temperature (°C)

Temperature (°C) 2 0
2 0 0 10 20 30 40 50 60 70 80
0 10 20 30 40 50 60 70 80 Time (h)
Time (h)
Fig. 5. The comparison of the predicted curve and fitted curve of E.
Fig. 4. The observed growth of E. coli O157:H7, Salmonella spp. coli O157:H7 and Salmonella spp. on lettuce under a real temper-
and L. monocytogenes on lettuce under real temperature history ature history involving an initial temperature of 25 8C. The dashed
involving a relatively high temperature, such as 25 8C. The growth line is the curve predicted using a value of a 0 determined by the
curves were predicted using the Baranyi–Ratkowsky model based parameter obtained from the independent constant temperature ex-
on the lettuce and PMP model separately. Observed data are mean periment. The lower solid line is a fitted curve obtained by trial and
values of triplicate trials F standard error. error with a different value of a 0.
248 S. Koseki, S. Isobe / International Journal of Food Microbiology 104 (2005) 239–248
teria in or on lettuce will help to reduce the microbial
risks associated with the consumption of salad vege-
tables such as lettuce, as well as provide valuable
information concerning the shelf life of products to
consumers. Since the prediction of pathogenic growth
during distribution will serve as proof of the impor-
tance of low-temperature management, it is useful to
thoroughly investigate all aspects of temperature man-
agement for those concerned with the distribution of
such products.
Acknowledgements
This work was supported by Research fellowships
of the Japan Society for the Promotion of Science
(JSPS) for Young Scientists. The authors wish to
thank JA Tone-Numata and the SANYO Electric
Co., Ltd. for their kind cooperation with the measure-
ment of lettuce temperature during distribution.
References
Ackers, M.L., Mahon, B.E., Leahy, E., Goode, B., Damrow, T.,
Hayes, P.S., Bibb, W.F., Rice, D.H., Barrett, T.J., Hutwagner, L.,
Griffin, P.M., Slutsker, L., 1998. An outbreak of Escherichia
coli O157:H7 infections associated with leaf lettuce consump-
tion. J. Infect. Dis. 177, 1588 – 1593.
Baranyi, J., 2002. Stochastic modelling of bacterial lag phase. Int. J.
Food Microbiol. 73, 203 – 206.
Baranyi, J., Roberts, T.A., 1994. A dynamic approach to predicting
bacterial-growth in food. Int. J. Food Microbiol. 23, 277 – 294.
Baranyi, J., Robinson, T.P., Kaloti, A., Mackey, B.M., 1995. Pre-
dicting growth of Brochothrix thermosphacta at changing tem-
perature. Int. J. Food Microbiol. 27, 61 – 75.
Beuchat, L.R., 1996. Listeria monocytogenes: incidence on vege-
tables. Food Control 7, 223 – 228.
Bovill, R., Bew, J., Cook, N., D’Agostino, M., Wilkinson, N.,
Baranyi, J., 2000. Predictions of growth for Listeria monocyto-
genes and Salmonella during fluctuating temperature. Int. J.
Food Microbiol. 59, 157 – 165.
Bovill, R.A., Bew, J., Baranyi, J., 2001. Measurements and predic-
tions of growth for Listeria monocytogenes and Salmonella
during fluctuating temperature II. Rapidly changing tempera-
tures. Int. J. Food Microbiol. 67, 131 – 137.
Carlin, F., NguyenThe, C., Morris, C.E., 1996. Influence of back-
ground microflora on Listeria monocytogenes on minimally
processed fresh broad-leaved endive (Cichorium endivia var
latifolia). J. Food Prot. 59, 698 – 703.
De Roever, C., 1998. Microbiological safety evaluations and recom-
mendations on fresh produce. Food Control 9, 321 – 347.
Francis, G.A., O’Beirne, D., 1998. Effects of the indigenous mi-
croflora of minimally processed lettuce on the survival and
growth of Listeria innocua. Int. J. Food Sci. Technol. 33,
477 – 488.
Francis, G.A., O’Beirne, D., 2002. Effects of vegetable type and
antimicrobial dipping on survival and growth of Listeria inno-
cua and E. coli. Int. J. Food Sci. Technol. 37, 711 – 718.
Koseki, S., Isobe, S., 2005. Growth of Listeria monocytogenes on
iceberg lettuce and on solid media. Int. J. Food Microbiol. 101
(2), 217 – 225.
McMeekin, T.A., Olley, J., Ross, T., Ratkowsky, D.A., 1993. Pre-
dictive Microbiology: Theory and Application. Research Stud-
ies Press, Taunton.
Mead, P.S., Slutsker, L., Dietz, V., McCaig, L.F., Bresee, J.S., Sha-
piro, C., Griffin, P.M., Tauxe, R.V., 1999. Food-related illness and
death in the United States. Emerg. Infect. Dis. 5, 607 – 625.
Nguyenthe, C., Carlin, F., 1994. The microbiology of minimally
processed fresh fruits and vegetables. Crit. Rev. Food Sci. Nutr.
34, 371 – 401.
Palmai, M., Buchanan, R.L., 2002. Growth of Listeria monocyto-
genes during germination of alfalfa sprouts. Food Microbiol. 19,
195 – 200.
Ratkowsky, D.A., Olley, J., McMeekin, T.A., Ball, A., 1982. Rela-
tionship between temperature and growth rate of bacterial cul-
tures. J. Bacteriol. 149, 1 – 5.
Riva, M., Franzetti, L., Galli, A., 2001. Microbiological quality and
shelf life modeling of ready-to-eat cicorino. J. Food Prot. 64,
228 – 234.
Rodriguez, A.M.C., Alcala, E.B., Gimeno, R.M.G., Cosano, G.Z.,
2000. Growth modelling of Listeria monocytogenes in packaged
fresh green asparagus. Food Microbiol. 17, 421 – 427.
Shorten, P.R., Membré, J.M., Pleasants, A.B., Kubaczka, M., Sobo-
leva, T.K., 2004. Partitioning of the variance in the growth
parameters of Erwinia carotovora on vegetable products. Int.
J. Food Microbiol. 93, 195 – 208.
Tamplin, M.L., 2002. Growth of Escherichia coli O157:H7 in raw
ground beef stored at 10 degrees C and the influence of com-
petitive bacterial flora, strain variation, and fat level. J. Food
Prot. 65, 1535 – 1540.
Viswanathan, P., Kaur, R., 2001. Prevalence and growth of patho-
gens on salad vegetables, fruits and sprouts. Int. J. Hyg. Envi-
ron. Health 203, 205 – 213.
Vold, L., Holck, A., Wasteson, Y., Nissen, H., 2000. High levels of
background flora inhibits growth of Escherichia coli O157:H7
in ground beef. Int. J. Food Microbiol. 56, 219 – 225.