International Journal of Food Microbiology 230 (2016) 45–57

Contents lists available at ScienceDirect

International Journal of Food Microbiology

journal homepage: www.elsevier.com/locate/ijfoodmicro

Predicting outgrowth and inactivation of Clostridium perfringens in meat

products during low temperature long time heat treatment
Zhi Duan a,b, Terese Holst Hansen b, Tina Beck Hansen a,⁎, Paw Dalgaard a, Susanne Knøchel b
a
Division of Microbiology and Production, National Food Institute (DTU Food), Technical University of Denmark, Kgs. Lyngby, Denmark
b
Department of Food Science, University of Copenhagen, Frederiksberg C, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: With low temperature long time (LTLT) cooking it can take hours for meat to reach a final core tempera-
Received 20 July 2015 ture above 53 °C and germination followed by growth of Clostridium perfringens is a concern. Available and
Received in revised form 23 December 2015 new growth data in meats including 154 lag times (tlag), 224 maximum specific growth rates (μmax) and 25
Accepted 20 March 2016
maximum population densities (Nmax) were used to developed a model to predict growth of C. perfringens
Available online 6 April 2016
during the coming-up time of LTLT cooking. New data were generate in 26 challenge tests with chicken
Keywords:
(pH 6.8) and pork (pH 5.6) at two different slowly increasing temperature (SIT) profiles (10 °C to 53 °C)
Slowly increasing temperature (SIT) followed by 53 °C in up to 30 h in total. Three inoculum types were studied including vegetative cells,
Low temperature cooking non-heated spores and heat activated (75 °C, 20 min) spores of C. perfringens strain 790-94. Concentra-
Fail-safe acceptable zone (FSAZ) approach tions of vegetative cells in chicken increased 2 to 3 log CFU/g during the SIT profiles. Similar results
were found for non-heated and heated spores in chicken, whereas in pork C. perfringens 790-94 increased
less than 1 log CFU/g. At 53 °C C. perfringens 790-94 was log-linearly inactivated. Observed and predicted
concentrations of C. perfringens, at the time when 53 °C (log(N53)) was reached, were used to evaluate the
new growth model and three available predictive models previously published for C. perfringens growth
during cooling rather than during SIT profiles. Model performance was evaluated by using mean deviation
(MD), mean absolute deviation (MAD) and the acceptable simulation zone (ASZ) approach with a zone of
± 0.5 log CFU/g. The new model showed best performance with MD = 0.27 log CFU/g, MAD =
0.66 log CFU/g and ASZ = 67%. The two growth models that performed best, were used together with a
log-linear inactivation model and D 53 -values from the present study to simulate the behaviour of
C. perfringens under the fast and slow SIT profiles investigated in the present study. Observed and predict-
ed concentrations were compared using a new fail-safe acceptable zone (FSAZ) method. FSAZ was defined
as the predicted concentration of C. perfringens plus 0.5 log CFU/g. If at least 85% of the observed log-
counts were below the FSAZ, the model was considered fail-safe. The two models showed similar perfor-
mance but none of them performed satisfactorily for all conditions. It is recommended to use the models
without a lag phase until more precise lag time models become available.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction (National Research Council, 1999; Voidarou et al., 2011). Vegetative

Clostridium perfringens is an anaerobic and endospore-forming bac-
terium able to produce toxin that causes a broad spectrum of human
and veterinary diseases. It occurs naturally in many environments in-
cluding soil, water and the intestinal tract of animals and humans
(McClane, 2001). C. perfringens contamination of raw meat is common
with high natural occurrence for various types of raw meat products
⁎ Corresponding author at: Division of Microbiology and Production, National Food
Institute (DTU Food), Technical University of Denmark, Mørkhøj Bygade 19, DK-2860
Søborg, Denmark.
E-mail address: tibha@food.dtu.dk (T.B. Hansen).
cells of C. perfringens are inactivated by pasteurization (D70-value of
0.32 ± 0.37 min) whereas spores can survive this heat treatment
(D 70 -value of 407 ± 454 min) (Van Asselt and Zwietering, 2006).
During subsequent cooling spores may germinate and are able to
grow rapidly at about 45 °C (Labbe and Huang, 1995) but fast
cooling from 54.4 °C to 26.6 °C within 1.5 h and from 26.6 °C to
4.4 °C within 5 h minimize germination and outgrowth (EFSA,
2005; Taormina and Dorsa, 2004). Martens (1999) recommended
a heat treatment of 90 °C for 10 min combined with chill storage
(b 10 °C) to manage non-proteolytic Clostridium botulinum. Howev-
er, guidelines related to control of C. perfringens during low temper-
ature long time (LTLT) cooking are lacking. A well-established
example of this type of heat treatment is sous-vide cooking where

http://dx.doi.org/10.1016/j.ijfoodmicro.2016.03.019
0168-1605/© 2016 Elsevier B.V. All rights reserved.
46 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57

ingredients are vacuum-packaged and heated at temperatures as
low as 60 to 65 °C for 2–8 h. Advantages of this technology include
higher water retention of meat, improved texture and tenderness as
well as higher flavour intensity of the final products (Church and
Parsons, 2000; Garcia-Linares et al., 2004). For LTLT cooking at below
60–65 °C these properties are further increased (Christensen et al.,
2011; Vaudagna et al., 2002). However, Willardsen et al. (1978) showed
that extensive growth of C. perfringens can occur during LTLT cooking
with slowly increasing temperature (SIT) profiles and they questioned
the safety of this type of heat treatment.
In relation to sensory properties of products, numerous SIT
profiles can be relevant for LTLT cooking of meat. Therefore, predic-
tive modelling of C. perfringens growth and inactivation seems an
interesting approach to evaluate the safety of these processes. A
few growth models (Jaloustre et al., 2011; Juneja et al., 2011; Le
Marc et al., 2008) and some inactivation models (Foegeding and
Busta, 1980; Jaloustre et al., 2012; Van Asselt and Zwietering,
2006) are available to predict responses of C. perfringens in meat.
However, the performance of available growth models has exclu-
sively been evaluated for cooling profiles (Jaloustre et al., 2011;
Juneja et al., 2011; Le Marc et al., 2008) and these models remain
to be evaluated for SIT profiles in relation to LTLT meat cooking.
The objective of the present study was to evaluate the performance
of C. perfringens growth and inactivation models under conditions
of relevance for safety evaluation of LTLT meat cooking. Specifically,
the effect of two different SIT profiles on growth and inactivation of
C. perfringens was studied and the performance of both a new and
three existing C. perfringens growth and subsequent inactivation
models were evaluated under these conditions.
2. Materials and methods
2.1. Experimental design
Growth curves for C. perfringens in various meats were collected
from 15 previously published studies (Table 1) and used for develop-
ment of a new growth model. Seven growth curves in pork at a con-
stant temperature of 45 °C were generated in the present study to
supplement those data. For evaluation of the developed and existing
models, growth and subsequent inactivation of C. perfringens in meat
products was studied in a series of challenge tests performed with
two different SIT profiles. Chicken or pork meat was inoculated
with vegetative cells, non-heated spores or heat activated (75 °C,
20 min) spores of C. perfringens strain 790-94 (see Section 2.2). For
the two different SIT profiles this resulted in 15 challenge tests
with chicken and 11 challenge tests with pork. Fourteen growth
and inactivation curves for SIT treatments of beef were collected
from the literature (Foegeding and Busta, 1980; Willardsen et al.,
1978; 1979) and included in the model evaluation.
2.2. Bacterial strain, reagents and pre-culture conditions
The studied C. perfringens isolate 790-94 originated from a human
food poisoning incidence and it is carrying a chromosomal cpe gene

Table 1
Literature growth data used for determination of lag time (tlag) and maximum specific growth rate (μmax) values.

Type of meat pH NaNO2 (ppm) aw (w/w% NaCl) No. of strains No. of tlag-values No. of μmax-values Reference
inoculated
Reported Used Reported Used

Poultry (autoclaved) 6.3 0 0.999 (0)a 2 10 8b 10 8b Andersen et al. (2004)

0
Poultry (80 °C cooked) 6.0 (adjusted) 0 0.999 (0)a 3 2 1b 2 1b Juneja and Marmer (1996)
0.993 (1)a 2 2 2 2
0.985 (2)a 2 2 2 2
0.977 (3)a 2 0b 2 0b
Poultry (75 °C, 20 min) 6.2c 0 0.999 (0)a 3 32 28b,d,e 32 29b,d Juneja et al. (2009)
Poultry (60 °C cooked) 6.2c 120 0.974 (3.5)b 3 20 16b,f 20 16b,f Juneja and Marks (2002)
Pork (75 °C, 20 min) 5.8c 0 0.999 (0)a 3 32 24d–f 32 28d,f Juneja et al. (2010)
Pork (fresh) 5.6 0 0.999 (0)a 1 3 3 3 3 Present study
Pork (75 °C, 20 min) 5.9 (adjusted) 1 1 1 1
Pork (autoclaved) 5.8 1 1 1 1
5.6 (adjusted) 1 1 1 1
5.9 1 1 1 1
Pork (heat-shocked) 5.8c 156 0.990 (1.25)a 3 4 4 4 4 Amezquita et al. (2005)
Beef (autoclaved) 5.9 0 0.999 (0)a 4 none none 18 18g Labbe and Huang (1995)
Beef (80 °C cooked) 7.0 (adjusted) 0 0.999 (0)a 3 4 1b,d 4 1b,d Juneja and Majka (1995)
0.975 (3)a 3 1b,d 3 1b,d
Beef (autoclaved) 6.3 0 0.999 (0)a 1 8 6b 8 6b Juneja et al. (1994a)
Beef (autoclaved) 6.2 0 0.999 (0)a 3 none none 7 7g Juneja et al. (1994b)
Beef (autoclaved) 6.2c 0 0.999 (0)a 1 none none 9 9 Foegeding and Busta (1980)
Beef (fresh) 5.8c 0 0.999 (0)a 8 4 4 4 4 Willardsen et al. (1979)
Beef (autoclaved) 6.2c 2 2 2 2
Beef (autoclaved) 6.2c 0 0.999 (0)a 8 none none 25 25 Willardsen et al. (1978)
Beef (steamed) 5.8 0 0.999 (0)a 1 none none 4 4 Schroder and Busta (1971)
Beef (60 °C cooked) 6.0c 120 0.999 (0)a 3 26 20b,d,f 26 20b,d,f Juneja et al. (2001b)
Beef (75 °C, 20 min) 6.1a 0 0.999(0) 3 32 28d,e 32 30d Juneja et al. (2008)
Total number 192 154 255 224
a
The estimated aw values were calculated from concentration of NaCl and dry matter of meat by using “% water phase salt (%WPS) = %NaCl × 100 / (100% × dry matter + %NaCl)” and
“aw = 1–0.0052471 × %WPS – 0.00012206 × %WPS2” from Ross and Dalgaard (2004). Dry matter content was obtained from the Danish National Food Institute's Food Composition
Databank (www.foodcomp.dk), as 28.8% for turkey, 26.0% for chicken, 30.4% for pork and 34.6% for beef. Concentrations of NaCl were calculated from molar equivalent of sodium in
the meat from the same Databank.
b
Data from poorly fitted curves (R2 b 0.9) or data not consistent with model (including no growth) were removed.
c
Estimated values determined by measuring pH of similar types of meat.
d
Growth curves with less than three cell concentrations (log CFU/g) in the exponential phase and/or no cell concentrations in late lag phase were not used.
e
Reported lag times of 0 were removed.
f
Poorly fitted curves and/or curves without enough data points but at same the condition were pooled together.
g
Reported μmax-values from the Gompertz model were corrected by multiplying with a factor 0.8521 (see Section 2.6.1).
 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57
(Brynestad et al., 1997). A mixture of vegetative cells and spores, was
prepared in 10 ml of cooked meat medium (Difco 226730, BD Corpo-
rate, NJ, USA) as described by McDonel and McClane (1988) and main-
tained at −20 °C. All chemicals and culture reagents were purchased
from Sigma-Aldrich (St. Louis, MO, USA), unless otherwise specified.
2.2.1. Sporulation and spore purification
C. perfringens 790-94 was heated for 20 min at 75 °C in freshly
prepared fluid thioglycollate medium (FTG; Oxoid CM0173,
Basingstoke, UK) and incubated (18 h, 37 °C) in a 20-ml tube with
a screw cap at an atmosphere with 0% oxygen, 1.0–2.5% hydrogen,
and N97.5% nitrogen using an anaerobic chamber (Vinyl Anaerobic
Airlock Chamber, Coy Laboratory Products INC., MI, USA). Subse-
quently, 1.0 ml was transferred to 10 ml of freshly prepared FTG
and incubated (4 h, 37 °C) with the same atmosphere. Two ml
was then transferred to 200 ml of modified Duncan Strong (DS)
medium and incubated for 24 h at 37 °C. The original DS formula-
tion was modified by replacing starch with 0.4% raffinose to im-
prove sporulation (Juneja et al., 1993). The sporulation
percentage was determined after 8 h (Andersen et al., 2004) in
modified DS medium by using a phase contrast microscope, 400 ×
magnification and a digital camera (Olympus BH-2 and Dp11,
Olympus Denmark A/S, Denmark). The number of phase bright
spores out of approximately 1000 cells was determined
(Andersen et al., 2004). After 24 h at 37 °C, cells and spores were
harvested by centrifugation (8000 × g, 20 min, 4 °C) and re-
suspended in 50 ml of cold (5 °C) sterile deionized water. The sus-
pension was stored at 10 °C for 5–6 d to release spores from the
mother cells (Yang et al., 2009). Then sonication (4 × 1 min, 0 °C)
with an ultrasonic processor (Vibra Cell, model 72,434, Bioblock
Scientific, Illkirch, France) was performed at 50% of max power to
release spores from vegetative cells. To remove debris of vegetative
cells the spores were washed three times using sterile deionized
water (100 ml, 5 °C) and centrifugation (8000 × g, 20 min, 4 °C).
The spore pellet was re-suspended in sterile deionized water
(50 ml, 5 °C) and stored at 4 °C for up to 3 months. Phase contrast
microscopy was used to determine the ratio between free spores
and vegetative cells in the suspension.
2.2.2. Pre-cultures
For pre-culturing of vegetative cells, 0.1 ml of the spore suspen-
sion (see Section 2.2.1) was transferred to 10 ml of freshly pro-
duced FTG and then heated (75 °C, 20 min) and incubated (24 h,
37 °C) with the same anaerobic atmosphere described above. This
resulted in a stationary-phase pre-culture of vegetative cells
which was diluted in sterile water to obtain an inoculum with ca.
5 × 10 6 CFU/ml. Non-treated spores were prepared directly from
the spore suspension (see Section 2.2.1) by diluting 40-fold in ster-
ilized water to obtain about 5 × 106 CFU/ml. This solution was heat-
ed (75 °C, 20 min) to obtain heat-treated spores. The pre-culture of
heat-treated spores was cooled to room temperature prior to
inoculation.
2.3. Challenge tests
2.3.1. Meat preparation for challenge tests
To study the growth potential of C. perfringens in LTLT meat hav-
ing a wide range of pH values, pork (pH 5.6) and chicken (pH 6.8)
were selected for investigation. As several previous challenge tests
had studied growth in LTLT beef (Foegeding and Busta, 1980;
Willardsen et al., 1978; 1979), this kind of meat was not experimen-
tally tested in the present study.
Chicken legs purchased from a local supermarket were found to
have a high level of microorganisms that interfered with the detection
of C. perfringens on the selective TSC agar used for enumeration (see
Section 2.4). To overcome this problem, chicken breasts were used
47
and pH was adjusted from 6.2 to 6.8 by 1 M NaOH to mimic the high
pH of chicken legs.
All meat used for challenge tests was ground. Batches of 10 kg of
pork tenderloins (porcine M. psoas major) and chicken breasts,
respectively, were obtained from a local supermarket. Pork tender-
loins were surface sterilized by immersion in boiling rapeseed oil
(240 °C, 5 s). Inside a horizontal laminar airflow clean bench
(Holten Laminair HBB2460, Thermo Fisher Scientific Inc., MA,
USA), the fried surface was removed using a sterilized knife. This
procedure was not required for the chicken meat as the microbiota
was below 2 log CFU/g at the time of purchase. Both types of meat
were passed through a 3 mm holeplate by using a grinder (OBH
Nordica 6681, OBH Nordica Denmark A/S, Taastrup, Denmark),
which had been decontaminated using 70% ethanol. Three kg of
each type of ground meat was placed in sterile plastic bags, and
ground chicken was mixed with an appropriate amount of 1 M
NaOH to increase pH to 6.8 as described above. The ground meat
was manually tumbled for at least 5 min to make the batch as
homogenous as possible. Portions of 500 g were vacuum packaged
using sterile plastic bags and stored at − 20 °C until use. The
microbiota in ground pork and chicken was checked using TSC
agar (see Section 2.4) and exclusively batches with concentrations
below the detection limit (2 log CFU/g) was applied in the chal-
lenge tests.
2.3.2. Inoculation
The 500 g portions of meat were thawed overnight at 5 °C and
10 ± 0.1 g were aseptically transferred into sterile laminar-film-
bags (65 mm × 140 mm) with low-oxygen permeability (thickness
90 μm; oxygen transmission 50 cm 3 /m 2 d bar; Allfo
Vakuumverpackungen, Hans Bresele KG, Waltenhofen, Germany).
The 10 g meat samples were added 20 μl of one of the three pre-
cultures, resulting in 3–4 log CFU/g. The meat was manually mas-
saged for 3 min to evenly distribute the inoculum, and then
vacuum-packaged at − 683 mbar (Komet X200, Maschinenfabrik,
Stuttgart, Germany). To mimic chilled storage of meat prior to
LTLT cooking, the samples inoculated with vegetative cells or
non-heated spores were stored overnight at 5 °C before starting
the challenge tests. Heated spores were prepared immediately be-
fore starting challenge tests to reduce germination into vegetative
cells before inoculating the meat samples.
2.3.3. SIT treatments
Two SIT profiles were studied. The fastest SIT mimicked the tem-
perature profile used in a study of the sensory quality of LTLT meat
cubes (Line Christensen, University of Copenhagen, personal com-
munication). The 10-g-meat-samples were transferred from the
5 °C storage to a water bath at 25 °C and the temperature was then
increased following a predetermined SIT profile up to 53 °C
(Figs. 1a and 2a). A slow SIT profile was adopted from Smith et al.
(1980). In brief, the 10-g-meat-samples at 5 °C were transferred to
400 ml water at 10 °C. This water bath, with the samples, was then
moved to room temperature and left there until samples and water
together reached 22 °C which took approximately 45 min. Then sam-
ples were placed in a water bath at 25 °C and the temperature was in-
creased following a predetermined SIT profile up to 53 °C (Figs. 1b
and 2b). For both fast and slow SIT profiles, heating at 53 °C contin-
ued for up to 30 h including the coming-up time. Temperature pro-
files of 10-g-meat-samples were recorded by a thermocouple data
logger (EL-USB-TC-LCD with a K-type Probe, Lascar Electronics Ltd.,
Salisbury, UK).
2.4. Microbiological sampling and analysis
The 10-g-meat-samples were removed at regular time intervals
during the SIT treatments and cooled to 10 °C to stop growth of
48 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57

Fig. 1. Observed growth and subsequent inactivation of Clostridium perfringens 790-94 in chicken heated under slowly increasing temperature (SIT) profiles. Measured pH (⋅⋅⋅⋅⋅⋅⋅⋅) and
temperatures (——) are shown for fast (a) and slow (b) SIT. Vegetative cells during fast (c) or slow (d) SIT, non-heated spores during fast (e) and slow (f) SIT and heated spores during fast
(g) and slow (h) SIT. Observed data from trial 1 (♦, ▲), trial 2 (□, ×) and trial 3 ( ) with ▲ and × indicating values below the limit of detection. Vertical dotted lines show the time when
53 °C was reached.

C. perfringens. Meat from each sample was diluted 10-fold in physio- UK). Appropriate ten-fold serial dilutions in PSP were spread plated
logical saline (PSP, Difco 211,820, added 0.1% peptone) and homog- (0.1 ml) or drop plated (3 × 15 μl) (Herigstad et al., 2001) on
enized for 2 min with a Stomacher 400 (Seward Limited, London, Tryptose Sulfite Cycloserine (TSC) agar (Perfringens agar base
 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57 49

Fig. 2. Observed growth and subsequent inactivation of Clostridium perfringens 790-94 in pork heated under slowly increasing temperature (SIT) profiles. Measured pH (⋅⋅⋅⋅⋅⋅⋅⋅) and
temperatures (——) are shown for fast (a) and slow (b) SIT. Vegetative cells during fast (c) or slow (d) SIT, non-heated spores during fast (e) and slow (f) SIT and heated spores during
fast (g) and slow (h) SIT. Observed data from trial 1 (♦, ▲), trial 2 (□, ×) and trial 3 ( ) with ▲ and × indicating values below the limit of detection. Vertical dotted lines show the
time when 53 °C was reached.

(Oxoid CM0587) with TSC supplement (Oxoid SR0088)). These 2.5. pH and aw of meat samples
plates were dried, overlaid with 10 ml TSC at 48 °C and incubated
in an anaerobic chamber (24 h, 37 °C). Black colonies were counted pH was measured using a Knick Struers Portamess 751 calimatic
as C. perfringens. pH metre (Knick, Berlin, Germany) in 10 g of ground meat
50 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57
homogenized with 10 ml of sterile deionized water. 4.0 g of ground
meat was used for measuring the water activity (AquaLab 3TE, Deca-
gon Devices, Inc., WA, USA).
2.6. Available C. perfringens growth data
Data describing growth of C. perfringens in meats, either growth
curves, generation times, maximum specific growth rates (μmax) and/
or lag times (tlag), were collected from 15 previously published studies
(Table 1). As different primary growth models had been used for fitting
of the growth parameters in these studies, refinement and standardiza-
tion of the growth data was required.
2.6.1. Refinement of isothermal growth data
Four different methods of refinement were used. 1) As the modi-
fied Gompertz model overestimates μmax-values by 10 to 20% com-
pared to the Logistic and to the Baranyi and Roberts models
(Baranyi et al., 1993; Dalgaard et al., 1994; McMeekin et al., 1993),
growth curves originally fitted with the modified Gompertz model
to estimate μmax and t lag-values were re-fitted using the Baranyi
and Roberts model (Baranyi and Roberts, 1994). 2) When growth
curve data were reported in studies having used the modified
Gompertz model, a factor of 0.852 was used to correct the reported
μmax-values. This correction factor corresponds to the 10 to 20% dif-
ference previously determined from a large number of growth
curves for different microorganisms. It was estimated here based
on four growth curves for C. perfringens (Andersen et al., 2004;
Foegeding and Busta, 1980; Labbe and Huang, 1995) which were
fitted with both the Gompertz and Baranyi and Roberts models and
the average value of the relation between the two μmax-values was
used as the correction factor. 3) Data from poorly fitted growth curves,
i.e. R2 b 0.9, and data from conditions showing no growth were omitted.
4) Growth curves with less than three cell counts (log CFU/g) observed
in exponential phase and/or no cell counts observed in late lag phase
were not used directly. For these growth curves, tlag and μmax-values
were not used if only one growth curve was available. However, when
several growth curves were determined for the same specific condition,
then all replicates were pooled prior to fitting of tlag and μmax-values by
using the Baranyi and Roberts model.
2.6.2. Estimation of pH and aw values
When pH was not reported in the studies from literature, the same
kind of meat was bought at a local supermarket in Denmark and pH
was measured (see Section 2.5 and Table 1). With aw-values missing
in studies from literature, these were calculated from the concentra-
tions of NaCl and dry matter of similar meat cuts, collected from the
Danish Food Composition Databank (www.foodcomp.dk) at DTU
Food, using the method suggested by Ross and Dalgaard (2004)
(Table 1).
2.7. New model for growth of C. perfringens
To take into account the effect of both processing temperature and
product characteristics (water activity, pH and sodium nitrite) a new
model for growth of C. perfringens was developed.
2.7.1. Primary growth model
The logistic model with delay (Rosso et al., 1996) was used as pri-
mary model to describe growth of C. perfringens in meat (Eq. (1)).
8
> ðN0 Þ
Log0 1 t b t lag
>
>
>
< fraction approach was used to calculate the effect of changing tempera-
LogðNt Þ ¼ B N max C ð1Þ
> B
 C t ≥ t lag
> Log@
> N −μ max ðt−t lag Þ
A
>
: 1þ
max
−1  e
N0
where t is time, Nt is cell concentration at time t (CFU/g), N0 is the ini-
tial cell concentration (CFU/g), Nmax is maximum population density
(CFU/g), μmax is the maximum specific growth rate (1/h), and tlag the
lag time (h). This simple primary model was chosen as it directly and
accurately estimates the kinetic parameters tlag, μmax and Nmax
(Rosso, 1995).
2.7.2. Secondary growth models
The cardinal parameter model of Le Marc et al. (2008) was modified
to include a pH term as suggested by Presser et al. (1997) and a simple
term to take into account the effect of sodium nitrite (NIT) (Eq. (2)).

 
logðμ max Þ ¼ log bðT  T min Þ2  1  ecðTT max Þ  ðaw  aw min Þ
 
NIT 
max  NIT
ð2  aw  aw min Þ∙ 1  10ðpH min pHÞ 
NIT max
ð2Þ
where μmax is maximum specific growth rate (1/h), b and c are con-
stants, T is temperature (°C), Tmin is the theoretical minimum tem-
perature of growth (°C), T max is the theoretical maximum
temperature of growth (°C), aw is water activity, awmin is the theoret-
ical minimum water activity for growth, pHmin is theoretical mini-
mum pH for growth and NIT max is theoretical maximum sodium
nitrate concentration for growth. With T b Tmin, T N Tmax, aw b awmin,
pH b pHmin or NIT N NITmax, μmax was defined as zero.
A simple secondary lag time model Eq. (3) relying on a constant rel-
ative lag time (RLT)-values was used (Mellefont and Ross, 2003).

  RLT ln ð2Þ
log t lag ¼ log ð3Þ
μ max
where tlag (h) is lag time and RLT is the relative lag time.
The maximum population density (Nmax) was determined as the av-
erage value (8.27 log CFU/g) from five growth curves in pork at 45 °C
from the present study as well as from 20 previously published growth
curves at 30 to 50 °C, that clearly had entered the stationary phase
(Amezquita et al., 2005; Juneja et al., 1994a; 2008; 2009; 2010).
2.7.3. Combined growth and inactivation model to predict effects of SIT
profiles
The secondary growth models (Eq. (2), Eq. (3) and log(Nmax) were
combined with a log-linear thermal inactivation model (Eq. (4)) to pre-
dict the effect of SIT profiles on growth and subsequent inactivation of
C. perfringens. The simple log-linear model was chosen as it seemed suf-
ficient to describe the observed inactivation kinetics (Figs. 1 and 2).
logðN t Þ ¼ logðN53 Þ  ðt  t 53 Þ=D53 ð4Þ
where Nt is the cell concentration at time t, N53 is the cell concentration
at the time (t53) when the SIT profile reaches 53 °C and D53 is the deci-
mal reduction time in min at 53 °C. D53-values in the present study were
obtained by fitting the cell concentrations (log CFU/g) observed after
the SIT profile reached 53 °C to Eq. (4). Student's t-test was performed
to examine whether there were difference of D53-values among meat
types (chicken or pork), type of inoculums (vegetative cells, non-
heated or heated spores) or SIT profiles (fast or slow).
Numerical integration was used to predict the effect of SIT profiles
on growth before the temperature reached 53 °C and on the subsequent
thermal inactivation after 53 °C was reached. The Euler method was
used for numerical integration (Press et al., 2007) in combination with
a short time step of 0.167 min during growth and 30 min during thermal
inactivation to obtain accurate predictions (Press et al., 2007). The lag
tures during the lag phase of C. perfringens (Zwietering et al., 1994).
Microsoft Excel 2010 was used for numerical integration (Microsoft
Corp., Redmond, WA, USA).
 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57
2.8. Validation of models
The effect of SIT profiles on growth and inactivation of C. perfringens
in meat samples was evaluated in two ways. First, growth of
C. perfringens during SIT cooking until 53 °C was simulated using both
the developed model (see Sections 2.7.1 and 2.7.2) and three available
growth models (Jaloustre et al., 2011; Juneja et al., 2011; Le Marc
et al., 2008). From each of these simulated growth curves, the final con-
centrations of C. perfringens were predicted and compared to observed
values of log(N53) (log CFU/g). Next, combined growth and subsequent
inactivation of C. perfringens was simulated using the two growth
models, showing best performance when predicting log(N53). Growth
and subsequent inactivation of C. perfringens at 53 °C was predicted by
using the growth models in combination with Eq. (4). Observed log-
counts were compared to simulated log-counts using the growth
models with and without lag time.
2.8.1. Evaluation of available growth models for C. perfringens
The three available growth models were evaluated using all meat tri-
als, i.e. pork, chicken and beef (n = 39; Table 4). The growth model of Le
Marc et al. (2008), included the effect of temperature, pH and aw on μmax
and tlag in broth, whereas Nmax was assumed to be a constant of
9.0 log CFU/g. The models of Juneja et al. (2011) included the effect of
temperature on μmax and tlag for growth of C. perfringens in meat either
from beef or other meats than beef, e.g. pork and chicken. For the models
of Juneja et al. (2011), Nmax for beef was assumed to be the average of
the maximum population densities (MPD) found by Juneja et al.
(2008), whereas Nmax for other meats was assumed to be the average
of MPDs found for pork in Juneja et al. (2009) and for chicken in
Juneja et al. (2010). The model of Jaloustre et al. (2011) included the ef-
fect of temperature on μmax and tlag in beef. Also in that model Nmax was
assumed to be a constant of 9.0 log CFU/g. When necessary the α0, q0
and h0 values from these models were converted into relative lag time
(RLT)-values (RLT = tlag ⋯μmax/ln (2)) by using Eq. (5).


1
ln 1 þ
 ln ðα 0 Þ q0 h0
RLT ¼ ¼ ¼
ln 2 ln2 ln2
where α0 is the physiological state (Le Marc et al., 2008), q0 is the phys-
iological state constant (Juneja et al., 2011) and h0 is a parameter char-
acterizing the “work-to-be-done” for cells to reach the exponential
phase (Jaloustre et al., 2011).
For evaluation of the performance of these models, observed and
predicted log(N53) values (see Section 2.7.3) were compared using
i) the mean deviation (MD; log CFU/g), ii) the mean absolute deviation
(MAD; log CFU/g) and iii) the percentage of observed values inside an
acceptable simulation zone (ASZ) of ±0.5 log CFU/g as previously ap-
plied in several studies including Møller et al. (2013).
2.8.2. Evaluation of model to predict growth and subsequent inactivation of
C. perfringens
The performance of models to predict growth and subsequent
inactivation of C. perfringens during SIT profiles was evaluated by a
fail-safe acceptable zone (FSAZ) method. FSAZ was defined as the
predicted concentration of C. perfringens (log CFU/g) plus
0.5 log CFU/g. If an observed concentration was below the FSAZ
then the prediction for that particular point was defined as safe and
the evaluated model was considered fail-safe when the percentage of
observed counts below the FSAZ was at least 85%. A percentage of 85
means, that no more than 100% − 85% = 15% of the observed concen-
trations can be under-predicted for the model to be evaluated as fail-
safe. Thus, the FSAZ method considers all over-predicted concentrations
as acceptable. This FSAZ method is a modification of the ASZ method
(see Section 2.8.1) where 70% has been used as acceptability level
51
(Møller et al., 2013). Besides the FSAZ%, also the MD and MAD measures
were calculated (see Section 2.8.1).
2.9. Statistical analyses and curve fitting
The Baranyi and Roberts model was fitted using DMFit (http://
www.combase.cc, UK Food Standards Agency). Parameter values in
Eqs. (2)–(4) were estimated with non-linear regression using
SigmaStat 3.5, from Systat Software Inc. (San Jose, CA, USA) (http://
www.sigmaplot.com). To stabilize variance of data, μmax (Eq. (2)),
tlag- (Eq. (3)) and Nt -values (Eq. (4)) were log-transformed.
Square-root transformation of μmax-data and of the right-hand side
in Eq. (2) was evaluated and resulted in similar parameter estimates.
The log-transformation of both μmax- and tlag-data was selected as it
resulted in close-to-constant variance of the data (Fig. 3). The root mean
squared error (RMSE) was used to evaluate the goodness-of-fit.
3. Results and discussion
3.1. pH and water activity of meat products during SIT profiles
Increasing pH-values were observed during the coming-up time of
SIT profiles (Figs. 1a, b, 2a, b) whereas water activity did not change
(P N 0.05). For chicken with initial pH of 6.8 the increase was from
0.04 to 0.15 pH-units and the observed changes were unlikely to signif-
icantly influence growth of C. perfringens. The opposite can be expected
for growth in pork where pH increased from 5.6 to 5.8 which is expected
to increase the growth rate significantly (Le Marc et al., 2008). These re-
sults indicated that it was important to include the observed dynamic
changes in pH when predicting growth of C. perfringens in pork during
SIT conditions.
3.2. Growth and subsequent inactivation of C. perfringens in SIT challenge
tests
There are marked differences in the ability of C. perfringens strains to
grow in the temperature interval from 12 to 50 °C (Andersen et al.,
ð5Þ 2004; Xiao et al., 2015). For the present study, we chose an isolate,
C. perfringens 790-94, with a high growth rate in the upper temperature
range as this is most relevant for LTLT cooking. Following inoculation
with vegetative cells or non-heated spores the C. perfringens 790-94
counts in chicken increased 2.3–2.9 log CFU/g which was more than
the 1.5–2.0 log CFU/g observed for heated spores (Fig. 1). For pork
with lower pH, the growth was in general less pronounced but more
variability of the growth response was observed for vegetative cells
(Fig. 2). These results suggest that low pH meat contaminated with veg-
etative cells represents a higher risk of growth than similar meat con-
taminated with spores. This is important for LTLT cooking where
contamination of the meat raw material with vegetative cells of
C. perfringens is likely (National Research Council, 1999; Voidarou
et al., 2011).
Lag times for C. perfringens 790-94 were short (b 1 h) and seemed in-
dependent of inoculum type (vegetative cells, non-heated spores or
heated spores), type of meat (chicken or pork) or SIT profiles (slow or
fast) (Figs. 1 and 2). Slightly longer lag times of approx. 1.5–2 h were ob-
served for vegetative C. perfringens cells in beef when exposed to slow
linear heating SIT profiles with 4.1–8.5 °C/h (Foegeding and Busta,
1980; Willardsen et al., 1978; 1979). However, these studies did not
apply overnight 5 °C-storage of inoculated samples prior to the LTLT
cooking which could explain the longer lag times.
Independent of SIT conditions and of the type of meat the highest
concentration of C. perfringens 790-94 cells was always observed at
the time when the product temperature approached 53 °C. The excep-
tion was the absence of growth in pork inoculated with heated spores
(Figs. 1 and 2). After 53 °C was reached, an apparently log-linear inacti-
vation of vegetative C. perfringens 790-94 cells was observed (Figs. 1 and
52 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57

Fig. 3. Previously published maximum specific growth rates (μmax, 1/h, n = 224) and lag times (tlag, h, n = 154) for Clostridium perfringens growing in different types of meat. μmax- (a) and
tlag-values (c) were shown for beef ( ), beef with NaNO2 added (♢), pork ( ), pork with NaNO2 added (□), poultry (▲) or poultry with NaNO2 added (△). Published μmax- and tlag-values
were compared with prediction (——) from Eq. (2) (b) and Eq. (3) (d) using parameter estimates shown in Table 2.

2). In line with these observations, log-linear inactivation was also ob-
served for C. perfringens in beef at similarly low temperatures
(Foegeding and Busta, 1980). The thermal inactivation at 53 °C resulted
in C. perfringens 790-94 concentration in chicken being lower than the
initial concentration after about 15 h (Fig. 1). For pork this overall reduc-
tion of C. perfringens 790-94 concentration was obtained after 1–15 h
depending on the inoculum and the SIT profile (Fig. 2).
The new secondary μmax-model (Eq. (2)) was based on data for all
types of meat and described data appropriately with statistically signif-
icant parameter estimates, relatively small standard errors, a RMSE-
value of 0.137 and a R2-value of 0.909 for log-transformed μmax-values
(Table 2 and Fig. 3b).
The estimated Tmin (10.3 °C) and Tmax (52.9 °C) values were both a
little lower than the values of 12.2 °C and 54.5 °C reported by Le Marc
et al. (2008). Importantly, the estimated Tmax-value of 52.9 °C in the

3.3. New growth model for C. perfringens Table 2

Estimated parameter values in secondary Clostridium perfringens growth models (Eqs. (2)
and (3)) determined by fitting literature data (Table 1).
154 out of 192 tlag-values and 224 out of 255 μmax-values from liter-
ature studies were selected for development of a new C. perfringens Response variable Parameter Estimate Std. error P value
growth model (Table 1). Added curing agents, like NaCl and NaNO2, re- μmax (n = 224) b 1.705 0.608 0.006
duced μmax-values and extended the tlag of C. perfringens in meat. Poultry Tmin (°C) 10.3 0.23 b0.001
appeared to result in slightly faster growth of C. perfringens (Fig. 3a), but c 0.166 0.032 b0.001
Tmax (°C) 52.9 0.42 b0.001
there was no obvious difference in the μmax-values and tlag determined
pHmin 4.7 0.39 b0.001
for different types of meat as also found in the study of Juneja et al. awmin 0.945 0.009 b0.001
(2011). The high variability of observed μmax-values at different storage NITmax (ppm NaNO2) 334 37 b0.001
temperatures (Fig. 3a), especially for beef, was most likely due to differ- tlag (n = 154) RLT 5.97 0.35 b0.001
ences in product pH and differences between the many strains studied Nmax (n = 25) Nmax (log CFU/g) 8.27 0.088 n.a.

in those experiments (Table 1). n.a.: Not applicable.

 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57
present study corresponded well with the observed inactivation of
C. perfringens 790-94 starting at 53 °C (Figs. 1 and 2). The estimated
awmin-value (0.945) was lower than the value of 0.976 reported by Le
Marc et al. (2008) and the new estimate is in agreement with growth
of C. perfringens observed for beef with aw of 0.974–0.975 (Juneja and
Marks, 2002). The estimated pHmin-value (4.7) was similar to the 4.76
reported by Le Marc et al. (2008) and both values corresponded well
with the pH growth boundary of 5.0 reported by Smith (1972). The es-
timated NITmax-value of 334 ppm sodium nitrite corresponded to stud-
ies observing growth in meat with 200 ppm (Juneja et al., 2013) but we
have found no cardinal parameter values for the effect sodium nitrite on
growth of C. perfringens in the literature.
Considerable variability was observed for the lag times and RLT-
values obtained from isothermal growth in meat products (Figs. 3c, d),
but no significant effect of temperature on RLT-values was observed
(P N 0.05). Short tlag-values below 1 h [log(tlag) b 0] were observed
mainly for growth in uncured meats (Fig. 3d). The fitted RLT-value
was 5.97 (Table 3), which corresponded to values reported by others,
e.g. 6.5 (chicken), 5.7 (pork) and 5.8 (beef) by Le Marc et al. (2008),
6.6 (chicken), 6.2 (pork) and 3.9 (beef) by Juneja et al. (2011), and 6.5
(beef) by Jaloustre et al. (2011).
The estimated Nmax-value of 8.27 log CFU/g in the developed model
(Table 2) was slightly lower than the level of 9 log CFU/g used in both Le
Marc et al. (2008) and Jaloustre et al. (2011). However, Nmax in the pres-
ent study was comparable to the levels 8.5 log CFU/g (beef) and
8.1 log CFU/g (pork and chicken) estimated from the studies of Juneja
et al. (2008; 2009; 2010) and also comparable to the levels found by
Roy et al. (1981) after growth at 37, 41, 45 and 49 °C in beef.
3.4. Thermal inactivation model for C. perfringens in meat after SIT
treatment
D53-values determined for inactivation of vegetative cells of
C. perfringens 790-94 at 53 °C (Fig. 1, Fig. 2) were significantly different
(P = 0.04) for the two SIT profiles but no significant difference was
found between the types of meat (Table 3). The slow SIT profile resulted
in higher heat resistance of C. perfringens 790-94 and in a 1.3-fold higher
D53-value compared to the fast SIT profile (Table 3). A similar effect of
SIT profiles has previously been observed for C. perfringens (Roy et al.,
1981), Salmonella (Stasiewicz et al., 2008) and Listeria monocytogenes
(Hassani et al., 2005; Stephens et al., 1994). The D53-values determined
in the present study were close to those determined by Foegeding and
Busta (1980) for C. perfringens in beef but markedly higher than D53-
values extrapolated from both Van Asselt and Zwietering (2006) and
Jaloustre et al. (2012) (Table 3). Those lower D53-values were extrapo-
lated from D- and z-values determined at higher temperatures. This un-
derlines the importance of validation when D-values are extrapolated.
Table 3
Decimal reduction times at 53 °C (D53-values) for vegetative cells of Clostridium perfringens.
Reference Treatment before heat inactivation
Present study Fast SIT
Slow SIT
Foegeding and Busta (1980) Growth at 45 °C followed by heating at constant temperature
of 53 °C
Van Asselt and Zwietering (2006) Average value from five studiesb including inactivation in FTG
broth and beef
Jaloustre et al. (2012) Average value from eight studiese including inactivation in FTG
broth and beef
a
Estimated values (see Table 2).
b
Studies included ICMSF (1996), Juneja and Marmer (1998), Juneja et al. (2001a), Novak et al. (2001) and Roy et al. (1981).
c
Converted from D70–value of 0.38 min (log10 D70 = −0.42) and z-value of 10.3 °C.
d
Converted from D60–value of 3 min and z-value of 5.21 °C.
e
Studies included Foegeding and Busta (1980), Heredia et al. (1997), Juneja and Marmer (1998), Juneja et al. (2001a), Novak et al. (2001), Roy et al. (1981), Sarker et al. (2000) and
Smith et al. (1981).
53
3.5. Model performance
3.5.1. Growth models
Based on the coming-up time and temperature profiles, characteriz-
ing the fast and the slow SIT conditions, the log-concentration of
C. perfringens obtained when 53 °C (log(N53)) was reached was predict-
ed using three previously reported growth models as well as the new
model developed in the present study. The same procedure was repeat-
ed for linear SIT profiles for beef collected from Foegeding and Busta
(1980) and Willardsen et al. (1978; 1979) (Fig. 4). Table 4 shows the
comparison of these predictions to the observations. The new growth
model developed in the present study showed overall the best perfor-
mance for predicting log(N53), with the lowest MD (0.27 log CFU/g)
and MAD (0.66 log CFU/g) and the highest total ASZ (67%). These values
show that, on average, the new model overestimated the observed
C. perfringens log(N53) counts (Fig. 4d). From a food safety perspective,
overestimation of log(N53) could be considered as fail-safe if it provides
the basis for calculation of the number of log-reductions required for the
LTLT meat product to be safe to eat. Figs. 4a, b in combination with the
positive MD-values (Table 4) for the models from Le Marc et al.
(2008) and Juneja et al. (2011) showed that these models also generally
overestimated log(N53), whereas the model from Jaloustre et al. (2011)
with a negative MD, on average, underestimated the log(N53) values
(Fig. 4c).
Le Marc et al. (2008) used what they described as robust strains to
build their model, i.e. strains with the cpe gene on their chromosome
resulting in a high heat tolerance of spores and strains showing good
growth potential after heating/cooling regimes. The model of Jaloustre
et al. (2011) was built from growth of many strains and relied on data
reported in 14 different studies with beef. The model of Juneja et al.
(2011) relied on growth of a cocktail of three strains but this model ex-
clusively included the effect of the storage temperature on growth. The
better performance of the Le Marc et al. (2008) model mainly resulted
from its ability to take into account the effect of the product's pH. Al-
though, it performs better than other available C. perfringens growth
models, the Le Marc et al. (2008) model has a limited ability to predict
the effect of reduced water activity. This results from an awmin-value
of 0.976 indicating that C. perfringens is predicted not to grow at aw-
values below 0.976. However, Juneja and Majka (1995) and Juneja
and Marks (2002) observed growth of C. perfringens in beef with aw of
0.974–0.975. Furthermore, the Le Marc et al. (2008) model has a theo-
retical maximum temperature of growth (Tmax) of 54.5 °C resulting in
growth being predicted at 53 °C where the present study observed
C. perfringens 790-94 to be inactivated (Figs. 1 and 2). Tmax of the new
model was 52.9 °C (Table 2) and, therefore, more realistic also resulting
in fewer observed log(N53) being overestimated compared to the Le
Marc et al. (2008) model (Fig. 4).
Tested medium pH aw D53-value
(min)
Pork or chicken 5.60/6.80 0.999a 256 ± 60
Pork or chicken 5.60/6.80 0.999a 320 ± 78
Autoclaved ground beef 6.20a 0.999a 278.3
17.0c
66.2d
54 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57

Fig. 4. Comparison of observed and predicted concentrations (log CFU/g) of Clostridium perfringens 790-94 when 53 °C was reached (log(N53)) in pork exposed to fast (□) or slow (■) SIT
profiles and in chicken exposed to fast (△) or slow (▲) SIT profiles. Predicted values are from models of Le Marc et al. (2008) (a), Juneja et al. (2011) (b), Jaloustre et al. (2011) (c) and the
present study (d).

In particular, the new model performed well for beef (ASZ = 93%)
and chicken (ASZ = 73%) but poorly for pork (ASZ = 20%) (Fig. 4d).
The Le Marc et al. (2008) model also performed well for beef trials
(ASZ = 71%) and relatively well for slow SIT chicken trials whereas
log(N53) for fast SIT chicken and most pork trials were overestimated
with more than 1 log-unit (Fig. 4a). With the exception of the model
of Jaloustre et al. (2011), the models performed better for heated spores
as compared to vegetative cells and non-heated spores (Table 4). As the
majority of the available growth data of C. perfringens at isothermal con-
ditions (Table 1) used heated spores as inocula, and the fact that the
model of Le Marc et al. (2008) also was developed using heated spores,
this indicated that different models might be needed for other types of
inocula. Other researchers have shown that spore germination can
take hours (Mafart, 1995) which will result in higher RLT values for bac-
terial spores compared to vegetative cells. This appeared not to be the
case in the present study as Figs. 1 and 2 do not indicate longer tlag for
spores than for vegetative cells of C. perfringens 790-94 under the non-
isothermal growth conditions. However, as shown in Fig. 2, there
seemed to be a difference in the time point where the decrease of the
population started for inocula consisting of vegetative cells, heated
and non-heated spores in pork. The decrease of C. perfringens 790-94,
when either heated or non-heated spores were used, started before
53 °C was reached and, therefore, earlier than observed for the vegeta-
tive cells. This could mean that Tmax in the μmax model was lower for
spore inocula of C. perfringens 790-94 in pork. As this was not observed
for chicken (Fig. 1), which had a much higher pH (6.8), an interaction

Table 4
Evaluation of available growth models for predicting the effect of slowly increasing temperature profiles on the number of Clostridium perfringens (log CFU/g) at the time when the tem-
perature reached 53 °C (log(N53)).

Evaluation measure Condition evaluated n Le Marc et al. (2008) Juneja et al. (2011) Jaloustre et al. (2011) New model from
model models model present study

MDa (log CFU/g) All data 39 0.40 0.96 −0.35 0.27

MADb (log CFU/g) All data 39 0.79 1.29 1.00 0.66
ASZc (%) All data 39 41 28 26 67
Beefd 14 71 7 57 93
Porke 10 30 0 20 20
Chickene 15 20 67 0 73
Fast SITe 13 15 38 8 54
Slow SITe 12 33 42 8 50
Vegetative cellse 9 22 22 22 56
Sporese 9 11 33 0 33
Heated sporese 7 43 71 0 71
a
Mean deviation.
b
Mean absolute deviation.
c
Percentage of observed values inside the acceptable simulation zone of ±0.5 log CFU/g.
d
Pooled data from Foegeding and Busta (1980) and Willardsen et al. (1978; 1979).
e
Observed data from the present study.
 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57 55

Table 5
Evaluation of growth models, combined with the thermal inactivation model from the present study, for predicting the effect of slowly increasing temperature profiles on the number of
Clostridium perfringens 790-94 (log CFU/g) using the mean deviation (MD), mean absolute deviation (MAD) and fail-safe acceptable zone (FSAZ) methods.

Type of meat Temperature profiles n Le Marc et al. (2008) model New model from present study

Fitted RLT RLT = 0 Fitted RLT RLT = 0

MD MAD FSAZ MD MAD FSAZ MD MAD FSAZ MD MAD FSAZ

Pork Fast SITa 54 0.64 0.80 98 2.03 2.04 100 0.91 0.97 100 2.39 2.40 100
Slow SITb 105 0.08 0.76 76 1.55 1.61 98 1.07 1.28 90 2.53 2.39 100
Chicken Fast SITc 129 0.83 0.96 95 2.27 2.27 100 −0.20 0.48 64 1.35 1.35 100
Slow SITd 93 −0.43 0.51 60 1.21 1.21 100 −0.33 0.45 65 1.10 1.10 100
All SIT challenge tests 381 0.29 0.77 82 1.78 1.80 99 0.27 0.76 77 1.76 1.72 100
a
Data in Fig. 2a, c, e. Both growth and thermal inactivation phases were included.
b
Data in Fig. 2b, d, f, h. Both growth and thermal inactivation phases were included.
c
Data in Fig. 1a, c, e, g. Both growth and thermal inactivation phases were included.
d
Data in Fig. 1b, d, f, h. Both growth and thermal inactivation phases were included.

between pH and temperature may also be involved. Such an interaction
could also explain part of the overestimations of log(N53) found for pork
with all four models (Fig. 4).
As the growth model of Le Marc et al. (2008) and the new model
showed better performance than the models of Juneja et al. (2011) or
Jaloustre et al. (2011), the former were selected for further performance
evaluation in combination with the thermal inactivation model deter-
mined in the present study.
3.5.2. Combined growth and thermal inactivation model for SIT profiles
None of the validated models was found acceptable at the overall
level as the FSAZ values were below 85% (Table 5). The new model
had an overall FSAZ value of 77% which was slightly lower than the
82% found for the model of Le Marc et al. (2008) when combined with
D53-values obtained in the present study (Table 5). Both models, how-
ever, performed to an acceptable extent for pork prepared at fast SIT
profiles which is illustrated in Fig. 5a. For pork, prepared from slow
SIT profiles (Fig. 5b), only the new model performed satisfactorily
with an FSAZ value of 90%, whereas only the Le Marc et al. (2008)
model performed satisfactorily for chicken prepared from fast SIT pro-
files having an FSAZ value of 95% (Fig. 5c; Table 5). As shown in
Fig. 5d, both models underestimated the numbers of C. perfringens
790-94 surviving during heating of chicken at 53 °C after a slow SIT pro-
file resulting in FSAZ values as low as 60–65% and negative MD values
(Table 5). For both models, this appeared to be caused by underestima-
tion of log(N53) which could be a result of the models predicting too
long tlag, in particular for the slow SIT trials (Figs. 5b, d). It is well
known that the temperature history of pre-cultures and temperature
shifts influence subsequent tlag (Ross and Dalgaard, 2004), and the
short tlag of C. perfringens 790-94 exposed to SIT profiles could be due
to both the overnight incubation of inoculated meat at 5 °C (see
Section 2.3.2) and the dynamic temperature cooking. However, as

Fig. 5. Observed and predicted concentrations (log CFU/g) of vegetative cells of Clostridium perfringens 790-94 in pork exposed to fast (a) or slow (b) SIT profiles and in chicken exposed to
fast (c) or slow (d) SIT profiles. Observed data from trial 1 (♦) and trial 2 (□, ×) with × indicating values below the limit of detection. Concentrations were predicted using primary growth,
Eq. (1) and inactivation, Eq. (4) models in combination either with the secondary growth models, Eqs. (3) and (4) from the present study (——) or with the secondary models from Le Marc
et al. (2008) (⋅⋅⋅⋅⋅⋅⋅⋅).
56 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57
Table 6
Evaluation of growth models, combined with the thermal inactivation model from the present study, for predicting the processing time needed to reduce the increment of Clostridium
perfringens 790-94 back to the initial level.
Type of meat Temperature profiles Predicted values
Le Marc et al. (2008)
model
Fitted RLT RLT = 0
Pork Fast SITa 8h 15 h
Slow SITb 8h 17 h
Chicken Fast SITc 18 h 24 h
Slow SITd 14 h 24 h
a
Data in Fig. 2a, c, e, g. Both growth and thermal inactivation phases were included.
b
Data in Fig. 2b, d, f, h. Both growth and thermal inactivation phases were included.
c
Data in Fig. 1a, c, e, g. Both growth and thermal inactivation phases were included.
d
Data in Fig. 1b, d, f, h. Both growth and thermal inactivation phases were included.
mentioned in section 3.2, longer lag times have been observed for
C. perfringens growth at SIT profiles in beef but without pre-incubation
at 5 °C. Further studies are needed to determine the quantitative contri-
butions of temperature shifts on tlag of C. perfringens. Importantly for the
safety assessment of LTLT cooked meat, Smith et al. (1980) found very
short lag times for C. perfringens in naturally contaminated beef kept
overnight at 4 °C before exposure to SIT profiles. This illustrates the rel-
evance of predicting C. perfringens growth with short or no lag time for
LTLT cooking.
Using a worst case approach and excluding tlag, by setting RLT equal
to 0, from both models resulted in overall FSAZ values of 99 and 100% for
the Le Marc et al. (2008) model and the new model, respectively
(Table 5). As indicated by the positive MD values and the high MAD
values in Table 5, the consequence of using such a worst case approach
was over-prediction of C. perfringens 790-94 concentrations in all cases,
being most pronounced for the pork trials and the trials using the fast
SIT profile.
One way to use the new combined growth and inactivation model
for evaluation of LTLT cooking could be to predict heat-treatments suf-
ficient to reduce the increment of C. perfringens back to its starting
level. As shown in Table 6, this approach resulted in safe predictions
for pork with the new model without excluding tlag, whereas excluding
tlag was required for chicken. Another way to use the results from the
present study could be to predict heat-treatments sufficient to result
in a 3 log-reduction of concentrations of C. perfringens. As the maximal
concentration of C. perfringens rarely exceeds 8 log CFU/g in LTLT cooked
meats (Roy et al., 1981), a 3 log-reduction will reduce the concentration
below 5 log CFU/g which is considered safe according to USDA (1999). If
a log-reduction higher than 3 is applied, LTLT treatments may become
so long that unacceptable growth of thermophilic Bacillus cereus may
be supported. For the slow SIT profiles in the present study this resulted
in the need for 16 h of heating time at 53 °C and a total process time of
20 h. For extrapolation to other temperatures in the range from 53 °C to
61 °C, we suggest using a z-value of 5.2 °C as published by Foegeding
and Busta (1980) and Roy et al. (1981) for these relatively low heating
temperatures.
4. Conclusions
C. perfringens 790-94 had shorter lag times and higher D53-values at
SIT conditions compared to isothermal conditions. A new model to pre-
dict growth and inactivation of C. perfringens during the studied SIT and
other LTLT treatments of meat was developed. Application of the new
growth or the Le Marc et al. (2008) model in combination with the
new inactivation data and without lag time resulted in fail-safe predic-
tions. Until we have a better understanding of the effect of SIT condi-
tions on the lag time, it is recommended to use the models without
lag time to obtain fail-safe predictions of concentrations C. perfringens
during commercial LTLT cooking of meats.
Observed values
New model from present Vegetative cells Non-heated spores Heated spores
study
Fitted RLT RLT = 0
9h 17 h 10 h 2½ h No growth
15 h 24 h 12–14 h 4–5 h 4–5 h
13 h 19 h 15–17 h 15–16 h 13–14 h
14 h 23 h 18–19 h 16–17 h 16–17 h
Acknowledgements
We thank Sigrid Brynestad and Per Einar Granum from the Norwe-
gian School of Veterinary Science, Oslo, Norway for providing the Clos-
tridium perfringens isolate 790-94. The work was supported by a grant
3304-FVFP-07-044 from the Danish Agriculture and Food Council and
by DTU Food.
References
Amezquita, A., Weller, C.L., Wang, L., Thippareddi, H., Burson, D.E., 2005. Development of
an integrated model for heat transfer and dynamic growth of Clostridium perfringens
during the cooling of cooked, boneless ham. Int. J. Food Microbiol. 101, 123–144.
Andersen, K.G., Hansen, T.B., Knøchel, S., 2004. Growth of heat-treated enterotoxin-
positive Clostridium perfringens and the implications for safe cooling rates. J. Food
Prot. 67, 83–89.
Baranyi, J., Roberts, T.A., 1994. A dynamic approach to predicting bacterial-growth in food.
Int. J. Food Microbiol. 23, 277–294.
Baranyi, J., Roberts, T.A., McClure, P., 1993. A non-autonomous differiential equation to
model bacterial growth. Food Microbiol. 10, 43–59.
Brynestad, S., Synstad, B., Granum, P.E., 1997. The Clostridium perfringens enterotoxin gene
is on a transposable element in type A human food poisoning strains. Microbiology
UK 143, 2109–2115.
Christensen, L., Ertbjerg, P., Aaslyng, M.D., Christensen, M., 2011. Effect of prolonged heat
treatment from 48 degrees C to 63 degrees C on toughness, cooking loss and color of
pork. Meat Sci. 88, 280–285.
Church, I.J., Parsons, A.L., 2000. The sensory quality of chicken and potato products pre-
pared using cook-chill and sous vide methods. Int. J. Food Sci. Technol. 35, 155–162.
Dalgaard, P., Ross, T., Kamperman, L., Neumeyer, K., McMeekin, T.A., 1994. Estimation of
bacterial-growth rates from turbidimetric and viable count data. Int. J. Food
Microbiol. 23, 391–404.
EFSA, 2005. Opinion of the scientific panel on biological hazards on the request from the
commission related to Clostridium spp. in foodstuffs. EFSA Journal 199, 1–65.
Foegeding, P.M., Busta, F.F., 1980. Clostridium perfringens cells and phospholipase-c activ-
ity at constant and linearly rising temperatures. J. Food Sci. 45, 918–924.
Garcia-Linares, M.C., Gonzalez-Fandos, E., Garcia-Fernandez, M.C., Garcia-Arias, M.T.,
2004. Microbiological and nutritional quality of sous vide or traditionally processed
fish: influence of fat content. J. Food Qual. 27, 371–387.
Hassani, M., Manas, P., Raso, J., Condon, S., Pagan, R., 2005. Predicting heat inactivation of
Listeria monocytogenes under nonisothermal treatments. J. Food Prot. 68, 736–743.
Heredia, N.L., García, G.A., Luévanos, R., Labbé, R.G., García-Alvarado, J.S., 1997. Elevation
of the heat resistance of vegetative cells and spores of Clostridium perfringens type a
by sublethal heat shock. pp. 998–1000.
Herigstad, B., Heersink, J., Hamilton, M., 2001. How to optimize the drop plate method for
enumerating bacteria. J. Microbiol. Methods 44, 121–129.
ICMSF, 1996. Microorganisms in Foods. 513. Blackie Academic and Professional, London.
Jaloustre, S., Cornu, M., Morelli, E., Noel, V., Delignette-Muller, M.L., 2011. Bayesian model-
ing of Clostridium perfringens growth in beef-in-sauce products. Food Microbiol. 28,
311–320.
Jaloustre, S., Guillier, L., Morelli, E., Noel, V., Delignette-Muller, M., 2012. Modeling of Clos-
tridium perfringens vegetative cell inactivation in beef-in-sauce products: a meta-
analysis using mixed linear models. Int. J. Food Microbiol. 154, 44–51.
Juneja, V.K., Call, J.E., Miller, A.J., 1993. Evaluation of methylxanthines and related com-
pounds to enhance Clostridium perfringens sporulation using a modified duncan and
strong medium. J. Rapid Methods Autom. Microbiol. 2, 203–218.
Juneja, V.K., Majka, W.M., 1995. Outgrowth of Clostridium perfringens spores in cook-in-
bag beef products. J. Food Saf. 15, 21–34.
Juneja, V.K., Marks, H.M., 2002. Predictive model for growth of Clostridium perfringens dur-
ing cooling of cooked cured chickens. Food Microbiol. 19, 313–327.
Juneja, V.K., Marks, H., Huang, L.H., Thippareddi, H., 2011. Predictive model for growth of
Clostridium perfringens during cooling of cooked uncured meat and poultry. Food
Microbiol. 28, 791–795.
 Z. Duan et al. / International Journal of Food Microbiology 230 (2016) 45–57
Juneja, V.K., Marks, H., Mohr, T., Thippareddi, H.H., 2013. Predictive model for growth of
Clostridium perfringens during cooling of cooked beef supplemented with NaCl, sodi-
um nitrite and sodium pyrophosphate. J. Food Process. Tech. 4, 275. http://dx.doi.org/
10.4172/2157-7110.1000275.
Juneja, V.K., Marks, H., Thippareddi, H., 2008. Predictive model for growth of Clostridium
perfringens during cooling of cooked uncured beef. Food Microbiol. 25, 42–55.
Juneja, V.K., Marks, H., Thippareddi, H., 2009. Predictive model for growth of Clostridium
perfringens during cooling of cooked ground chicken. Innovative Food Sci. Emerg.
Technol. 10, 260–266.
Juneja, V.K., Marks, H., Thippareddi, H.H., 2010. Predictive model for growth of Clostridium
perfringens during cooling of cooked ground pork. Innovative Food Sci. Emerg.
Technol. 11, 146–154.
Juneja, V.K., Marmer, B.S., 1996. Growth of Clostridium perfringens from spore inocula in
sous-vide turkey products. Int. J. Food Microbiol. 32, 115–123.
Juneja, V.K., Marmer, B.S., 1998. Thermal inactivation of Clostridium perfringens vegetative
cells in ground beef and turkey as affected by sodium pyrophosphate. Food Microbiol.
15, 281–287.
Juneja, V.K., Marmer, B.S., Miller, A.J., 1994a. Growth and sporulation potential of Clostridium
perfringens in aerobic and vacuum-packaged cooked beef. J. Food Prot. 57, 393–398.
Juneja, V.K., Novak, J.S., Eblen, B.S., McClane, B.A., 2001a. Heat resistance of Clostridium
perfringens vegetative cells as affected by prior heat shock. J. Food Saf. 21, 127–139.
Juneja, V.K., Novak, J.S., Marks, H.M., Gombas, D.E., 2001b. Growth of Clostridium
perfringens from spore inocula in cooked cured beef: development of a predictive
model. Innovative Food Sci. Emerg. Technol. 2, 289–301.
Juneja, V.K., Snyder, O.P., Cygnarowiczprovost, M., 1994b. Influence of cooling rate on out-
growth of Clostridium perfringens spores in cooked ground reef. J. Food Prot. 57,
1063–1067.
Labbe, R.G., Huang, T.H., 1995. Generation times and modeling of enterotoxin-positive
and enterotoxin-negative strains of Clostridium perfringens in laboratory media and
ground beef. J. Food Prot. 58, 1303–1306.
Le Marc, Y., Plowman, J., Aldus, C.F., Munoz-Cuevas, M., Baranyi, J., Peck, M.W., 2008.
Modelling the growth of Clostridium perfringens during the cooling of bulk meat.
Int. J. Food Microbiol. 128, 41–50.
Mafart, P., 1995. Modelling germination kinetics of spores of Clostridium tyrobutyricum: a
tool for predictive microbiology. J. Appl. Bacteriol. 78, 477–480.
Martens, T., 1999. Harmonization of safety criteria for minimally processed foods. FAIR
Concerted Action. FAIR CT96–1020. European Commission DE XII, p. 79.
McClane, B.A., 2001. The complex interactions between Clostridium perfringens enterotox-
in and epithelial tight junctions. Toxicon 39, 1781–1791.
McDonel, J.L., McClane, B.A., 1988. Production, purification, and assay of Clostridium
perfringens enterotoxin. In: Sidney, H. (Ed.), Methods in Enzymology Microbial
Toxins: Tools in Enzymology. Academic Press, pp. 94–103.
McMeekin, T.A., Olley, J., Ratkowsky, D.A., Ross, T., 1993. Predictive microbiology: theory
and application. John Wiley & Sons, Taunton, UK.
Mellefont, L.A., Ross, T., 2003. The effect of abrupt shifts in temperature on the lag phase
duration of Escherichia coli and Klebsiella oxytoca. Int. J. Food Microbiol. 83, 295–305.
Møller, C.O.A., Ilg, Y., Aabo, S., Christensen, B.B., Dalgaard, P., Hansen, T.B., 2013. Effect of
natural microbiota on growth of Salmonella spp. in fresh pork — a predictive microbi-
ology approach. Food Microbiol. 34, 284–295.
National Research Council, 1999. Committee on Drug Use in Food Animals, Panel on An-
imal Health, Food Safety, and Public Health. The use of drugs in food animals: benefits
and risks. National Academy Press, Washington DC, p. 128.
Novak, J.S., Tunick, M.H., Juneja, V.K., 2001. Heat treatment adaptations in Clostridium
perfringens vegetative cells. J. Food Prot. 64, 1527–1534.
Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 2007. Numerical Recipes 3rd
Edition: The art of scientific computing. Cambridge university press, Cambridge, UK.
Presser, K.A., Ratkowsky, D.A., Ross, T., 1997. Modelling the growth rate of Escherichia coli
as a function of pH and lactic acid concentration. Appl. Environ. Microbiol. 63,
2355–2360.
57
Ross, T., Dalgaard, P., 2004. Secondary models. In: In McKeller, R.C., Lu, X. (Eds.), Modeling
Microbial Responses in Foods CRC Press. Boca Raton, USA, pp. 63–150.
Rosso, L., 1995. Modélisation et microbiologie prévissionnelle: Élaboration d'un nouvel
outil pour l'agro-alimentaire (Ph.D. theses) Université Claude Bernard, Lyon, France.
Rosso, L., Bajard, S., Flandrois, J.P., Lahellec, C., Fournaud, J., Veit, P., 1996. Differential
growth of Listeria monocytogenes at 4 and 8 degrees C: consequences for the shelf
life of chilled products. J. Food Prot. 59, 944–949.
Roy, R.J., Busta, F.F., Thompson, D.R., 1981. Thermal inactivation of Clostridium perfringens
after growth at several constant and linearly rising temperatures. J. Food Sci. 46,
1586–1591.
Sarker, M.R., Shivers, R.P., Sparks, S.G., Juneja, V.K., McClane, B.A., 2000. Comparative ex-
periments to examine the effects of heating on vegetative cells and spores of Clostrid-
ium perfringens isolates carrying plasmid genes versus chromosomal enterotoxin
genes. Appl. Environ. Microbiol. 66, 3234–3240.
Schroder, D.J., Busta, F.F., 1971. Growth of Clostridium perfringens in meat loaf with and
without added soybean protein. J. Milk Food Technol. 34, 215–217.
Smith, L., 1972. Factors involved in the isolation of Clostridium perfringens. J. Milk Food
Technol. 35, 71–76.
Smith, L.B., Busta, F.F., Allen, C.E., 1980. Effect of rising temperatures on growth and sur-
vival of Clostridium perfringens indigenous to raw beef. J. Food Prot. 43, 520–524.
Smith, A.M., Evans, D.A., Buck, E.M., 1981. Growth and survival of Clostridium perfringens
in rare beef prepared in a water bath. J. Food Prot. 44, 9–14.
Stasiewicz, M.J., Marks, B.P., Orta-Ramirez, A., Smith, D.M., 2008. Modeling the effect of
prior sublethal thermal history on the thermal inactivation rate of Salmonella in
ground turkey. J. Food Prot. 71, 279–285.
Stephens, P.J., Cole, M.B., Jones, M.V., 1994. Effect of heating rate on the thermal inactiva-
tion of Listeria monocytogenes. J. Appl. Bacteriol. 77, 702–708.
Taormina, P.J., Dorsa, W.J., 2004. Growth potential of Clostridium perfringens during
cooling of cooked meats. J. Food Prot. 67, 1537–1547.
USDA (U.S. Department of Agriculture), Food safety and inspection service, 1999. Perfor-
mance standards for the production of certain meat and poultry products. Fed. Regist.
64, 732–749.
Van Asselt, E.D., Zwietering, M.H., 2006. A systematic approach to determine global ther-
mal inactivation parameters for various food pathogens. Int. J. Food Microbiol. 107,
73–82.
Vaudagna, S.R., Sanchez, G., Neira, M.S., Insani, E.M., Picallo, A.B., Gallinger, M.M., Lasta,
J.A., 2002. Sous vide cooked beef muscles: effects of low temperature-long time
(LT-LT) treatments on their quality characteristics and storage stability. Int. J. Food
Sci. Technol. 37, 425–441.
Voidarou, C., Alexopoulos, A., Plessas, S., Stavropoulou, E., Fotou, K., Tzora, A., Skoufos, I.,
Bezirtzoglou, E., 2011. Hygienic quality and antibiotic resistance profile of sliced
butchery. Anaerobe 17, 344–350.
Willardsen, R.R., Busta, F.F., Allen, C.E., 1979. Growth of Clostridium perfringens in 3 differ-
ent beef media and fluid thioglycollate medium at static and constantly rising tem-
peratures. J. Food Prot. 42, 144–148.
Willardsen, R.R., Busta, F.F., Allen, C.E., Smith, L.B., 1978. Growth and survival of Clostridi-
um perfringens during constantly rising temperatures. J. Food Sci. 43, 470–475.
Xiao, Y., Wagendorp, A., Abee, T., Wells-Bennik, M.H.J., 2015. Differential outgrowth po-
tential of Clostridium perfringens food-borne isolates with various cpe-genotypes in
vacuum-packed ground beef during storage at 12 °C. Int. J. Food Microbiol. 194,
40–45.
Yang, W.W., Crow-Willard, E.N., Ponce, A., 2009. Production and characterization of pure
Clostridium spore suspensions. J. Appl. Microbiol. 106, 27–33.
Zwietering, M.H., de Wit, J.C., Cuppers, H.G.A.M., van 't Riet, K., 1994. Modeling of
bacterial-growth with shifts in temperature. Appl. Environ. Microbiol. 60, 204–213.