International Journal of Food Microbiology 228 (2016) 58–66

Contents lists available at ScienceDirect

International Journal of Food Microbiology

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

Effect of slaughterhouse and day of sample on the probability of a pig

carcass being Salmonella-positive according to the Enterobacteriaceae
count in the largest Brazilian pork production region
Luís Gustavo Corbellini a,d,⁎, Alfredo Bianco Júnior b, Eduardo de Freitas Costa a, Ana Sofia Ribeiro Duarte d,
Elenita Ruttscheidt Albuquerque b, Jalusa Deon Kich c, Marisa Cardoso a, Maarten Nauta d
a
Departamento de Medicina Veterinária Preventiva, Faculdade de Veterinária, Universidade Federal do Rio Grande do Sul, Brazil
b
Ministério da Agricultura, Pecuária e Abastecimento, Florianópolis, SC, Brazil
c
Embrapa Suínos e Aves, Concórdia, SC, Brazil
d
Division of Epidemiology and Microbial Genomics, National Food Institute, Technical University of Denmark, Denmark

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

Article history: Sources of contamination of carcasses during slaughter include infected pigs as well as environmentally related
Received 18 April 2014 sources. There are many microbial indicators that can be used in the processing of food to assess food hygiene
Received in revised form 22 January 2016 and the safety of food processing. The presence of some microbial indicators can be viewed as a result of direct
Accepted 26 March 2016
or indirect contamination of a food with fecal material. The presence of Enterobacteriaceae is often used as a hy-
Available online 5 April 2016
giene indicator, as they are found both in the environment and in the intestine of warm-blooded animals. An as-
Keywords:
sociation between Salmonella isolation and Enterobacteriaceae count (EC) on pre-chill carcasses has been
Slaughterhouse size described, however the impact of slaughterhouse and the day of sampling on the occurrence of Salmonella has
Swine not been previously investigated. To this end, mixed logistic regressions (MLRs) with random effects and fixed
Food safety slopes were performed to assess the change in EC and its correlation with Salmonella occurrence using two
Indicator data sets. The first describes the EC and Salmonella isolation in 60 pork carcasses in one slaughterhouse sampled
Bacteriology at 11 different slaughter steps, including the carcass as a random effect. The second describes the EC and
Mixed models Salmonella isolation on 1150 pre-chill carcasses sampled in 13 slaughterhouses over 230 sampling days, and
the model combined two random intercepts, slaughterhouse and date of sampling nested with slaughterhouse
(day/slaughterhouse). Statistically significant associations (p b 0.0001) between the log of the EC and Salmonella
occurrence were found in all models. Nevertheless, although a strong association was found between Enterobac-
teriaceae and Salmonella contamination in pork carcasses, this association was not constant, given that there was
a high variation in the probability of a carcass being positive for Salmonella according to the EC mainly between
days of samples. The effect of the day of sampling on Salmonella prevalence was so large that the predictive value
of the EC count for Salmonella isolation on a daily basis was compromised. It is possible that on some days batches
with a high prevalence of Salmonella carriers shedding a high number of Salmonella were slaughtered. On these
days, the potential for contamination/cross-contamination of carcasses will be so large that even hygienic
slaughter, confirmed by the low EC on carcasses, will not be able to prevent the presence of Salmonella on
some carcasses. The results of this study demonstrate that, despite the statistically significant association
found, it may be difficult to predict when hygiene failure measured via EC actually indicates Salmonella contam-
ination, and neither the inverse.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction
Salmonella is one of the main pathogens associated with food-borne
disease worldwide, and is the most common cause of disease outbreaks
⁎ Corresponding author at: EPILAB, Departamento de Medicina Veterinária Preventiva,
Faculdade de Veterinária, Universidade Federal do Rio Grande do Sul (UFRGS), Av. Bento
Gonçalves 9090, CEP 91540-000 Porto Alegre, RS, Brazil.
E-mail addresses: luis.corbellini@ufrgs.br, lgcorbellini@hotmail.com (L.G. Corbellini).
reported in Brazil (Gomes et al., 2013). Pork is considered the third most
common foodborne source of human salmonellosis (Arguello et al.,
2013). The main source of contamination of carcasses is pigs introduced
into the slaughterhouse carrying Salmonella in the intestinal tract (EFSA,
2008). This is because infected swine may carry a high number of
Salmonella per gram of feces (Berends et al., 1997; Botteldoorn et al.,
2003), and contaminated feces can leak out throughout the slaughter
process. However, environmentally related sources of contamination
such as contact surfaces of equipment and handling by workers have
also been described (Berends et al., 1997; Botteldoorn et al., 2003;

http://dx.doi.org/10.1016/j.ijfoodmicro.2016.03.030
0168-1605/© 2016 Elsevier B.V. All rights reserved.
 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66
Duggan et al., 2010; Giovannacci et al., 2001; Hald et al., 2003), and di-
rect transfer of house flora present in the slaughterhouse has been
found to be an important source of contamination for carcasses (Smid
et al., 2012; Van Hoek et al., 2012). Moreover, the extent of contamina-
tion may vary significantly depending on the slaughterhouse and sam-
pling day (Berends et al., 1997; Botteldoorn et al., 2003; Hald et al.,
2003; Silva et al., 2012).
There are many microbial indicators that can be used in the process-
ing of food to assess food hygiene and the safety of food processing.
Their meaningfulness depends on the type of food analyzed as well as
the species or group of bacteria employed as an indicator. As most of
the pathogenic microorganisms potentially found in pork are of enteric
origin (Ghafir et al., 2003; Ray, 2005), fecal indicators are preferred dur-
ing the slaughtering process, as they fulfill many of the criteria set for
ideal food safety indicators such as easy and inexpensive detection
and enumeration, sharing the same habitat as enteric pathogens and
being detected when the pathogen is present in food (Ghafir et al.,
2008). Their presence can be viewed as a result of direct or indirect con-
tamination of food with fecal material (Ray, 2005). In this context, the
specificity of Escherichia coli as a fecal contaminant is higher than
enterobacteria and coliforms (Ray, 2005). On the other hand, the enu-
meration of a broader group of bacteria, such as those belonging to
the family Enterobacteriaceae, may be a better indicator of sanitation
level, including possible fecal contamination and presence of enteric
pathogens (Ray, 2005). For this reason, Enterobacteriaceae are often
used as a hygiene indicator (Prendergast et al., 2008), since they are
found both in the environment and in the intestine of warm-blooded
animals.
The U.S. Food Safety and Inspection Service (FSIS) considers generic
E. coli as the process control indicator organism, and has adopted a
quantitative E. coli standard as a measure of process control with respect
to the prevention and reduction of fecal contamination in slaughter
plants (FSIS, 1996). On the other hand, aerobic plate counts (APCs)
and Enterobacteriaceae are employed as hygiene indicator criteria for
cattle, sheep, goat, horse and pork carcasses in the European Union
[Commission regulation (EC) No. 2073, 2005]. Moreover, the correlation
between the presence of Enterobacteriaceae and Salmonella in pig cuts
and carcasses has been previously reported (Ghafir et al., 2008;
Prendergast et al., 2008). Salmonella contamination in pork carcasses
can originate from either direct or indirect contact with feces from in-
fected swine, or from contact with house flora present in the slaughter-
house environment. Given this, it can be hypothesized that Salmonella
occurrence is associated with Enterobacteriaceae counts (EC) on the
carcass surface. However, the extent of this association requires
investigation, as many factors such as variation provided by slaugh-
terhouse and day of sampling can influence this relationship. Micro-
biological studies are necessary to assess the effectiveness of HACCP
programs and good manufacturing practices and to propose specific
intervention measures (Brown et al., 2000). In the absence of ongo-
ing monitoring programs, baseline studies can serve as a means of
benchmarking the current levels of hygiene in slaughterhouses
(Bohaychuk et al., 2011).
The overall aims of this study were to describe the occurrence of En-
terobacteriaceae and Salmonella sp. in pre-chill pork carcasses as well as
throughout the slaughter steps in the first baseline study undertaken in
large slaughterhouses in Brazil, and to assess the association between
EC and Salmonella on pork carcasses taking into account sources of var-
iation such as slaughterhouse and day of sampling.
2. Material and methods
2.1. Sampling
All slaughterhouses were located in the largest Brazilian pork pro-
duction region (ABIPECS, 2012) and processed on average between
600 and 5280 pigs/day (median number of 2500 pigs/day). Sample
59
were collected in two different studies and recorded in two data sets.
In study one (data set A), 60 carcasses (C1–60) were sampled at 11 se-
quential steps: 1) before scalding, 2) after scalding, 3) after dehairing,
4) after singeing, 5) after polishing, 6) before evisceration, 7) after evis-
ceration, 8) after splitting (before veterinary inspection), 9) before final
rinse, 10) before chilling (after final rinse) and 11) after chilling. One
carcass per day was randomly sampled throughout the slaughter line
approximately 2 h after slaughter beginning. Swabs were taken over a
300 cm2 area on the shoulder (in the upper part when the sample was
collected before a given step, and in the lower part when the sample
was collected after a given step). The swabs were placed in sterile plastic
bags containing 100 mL peptone water 0.1%.
In study two (data set B), carcasses were sampled before chilling in
13 slaughterhouses (A–M) following the Commission Regulation (EC)
No. 2073/2005 as a part of a baseline study conducted by the Brazilian
Ministry of Livestock and Food Supply. The sampling plan to establish
the microbiological criteria in carcasses of pigs was comprised of 50
samples derived from 10 sampling sessions (7–10 day intervals on
average) in which five carcasses from different batches were
randomly sampled in each session. A total of two sampling plans
(n = 50; 100 carcasses) were performed per slaughterhouse (A–J).
In the remaining three slaughterhouses (K, L and M), one sampling
plan (n = 50) was performed. A total of 1150 carcasses were sampled
from August 2011 to September 2012. Carcasses were swabbed in
four parts (belly, hind leg, loin and jowl), each covering a 100 cm2
area, and then placed together in sterile plastic bags containing 40 mL
peptone water 0.1%.
2.2. Bacteriology
Bacterial samples taken from 60 pork carcasses at 11 different
slaughter steps in one slaughterhouse were analyzed for the pres-
ence of Salmonella using the immunoassay VIDAS®-SLM method
(bioMérieux, France) following the instructions of the supplier.
Quantitative assays for Enterobacteriaceae were performed using En-
terobacteriaceae Count Plate Petrifilm (3M Microbiology, St. Paul,
MN, USA) according to the instructions of the supplier. One milliliter
of the appropriate dilution was spread on Petrifilm and subsequently
incubated at 35 °C for 23–25 h. Petrifilm plates were then counted on
a standard colony counter. Samples taken from the 1150 pre-chill
carcasses were analyzed for the presence of Salmonella and quanti-
fied for Enterobacteriaceae according to standards from Normative
Instruction No. 62 from the Brazilian Ministry of Livestock and
Food Supply (MAPA, 2003). For Salmonella isolation, a 25 mL aliquot
from each homogenized sample was transferred to 225 mL buffered
peptone water (BPW) and incubated overnight at 37 °C for approxi-
mately 18 h. Aliquots of this buffered peptone water culture were
then transferred to both Rappaport–Vassiliadis broth and selenite
cystine broth. After incubation at 41 °C for 24 h, a 10 μL loop of
each culture was streaked onto brilliant green phenol red lactose su-
crose (BPLS) agar and xylose lysine deoxycholate (XLD) agar, both
incubated for 24 h at 37 °C. Suspect Salmonella colonies were con-
firmed by biochemical and serological tests. Detection and enumera-
tion of Enterobacteriaceae were performed using violet red bile
glucose (VRBG) agar.
Serial dilutions (100, 10−1, 10−2) from the original sample were pre-
pared and 1 mL of each dilution was transferred in duplicate to Petri
dishes followed by the addition of VRBG. After the incubation at 37 °C
for 24 h, plates containing up to 250 typical colonies were counted
manually using a standard colony counter. At least three presumptive
purple/pink colonies were confirmed by biochemical tests.
In both cases, the final EC estimate for each sample was obtained
from the mean of the confirmed colony counts and multiplied by the di-
lution where the counting was performed, and adjusted according to
the area sampled on the carcass. All counts were expressed as colony
forming units (CFU/cm2).
60 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66
2.3. Statistical analysis
2.3.1. Fitting probability distribution
For carcasses with Enterobacteriaceae concentrations below the
limit of detection (LOD) it was assumed that they were all contaminated
with low concentrations of this hygiene indicator. For the regression
analysis, in which Enterobacteriaceae concentrations are paired with
the observed presence or absence of Salmonella, (see Section 2.3.2) it
is necessary to have estimates for these low concentrations. Assuming
arbitrary numbers in non-detects such as half of the LOD was not
considered appropriate because it introduces bias in the data set
(Busschaert et al., 2010). Therefore, for each slaughterhouse and each
slaughter step, the log transformed Enterobacteriaceae concentrations
within the batch were assumed to follow a normal distribution. A meth-
od for distribution fitting with censored data was implemented in MS
Excel 2007, using the inverse of the cumulative normal distribution
function (the function NORMINV (probability, mean, standard_dev))
and Excel Solver. Assuming a data set of k EC samples with kn non-
detects (negative results) and kp EC count data (positive results), so
that kn + kp = k, the following procedure was applied: 1) using the
Excel function RANDBETWEEN(), kn different rank numbers i = 1…kn
were randomly assigned to carcass samples that gave the negative
results. Alternatively, the positive results ECi (expressed in log CFU per
cm²) were sorted from low to high, and got rank numbers i =
kn + 1…k; 2) a mean M and standard deviation SD of the log CFU per
cm2 were defined, based on a first guess of their values; 3) for each of
the kp positive results (ECi for i = kn + 1…k), the difference ECi −
NORMINV((i − 0.5) / k,M,SD) was calculated. These differences were
squared and summed to a sum of squared differences (SSQ); 4) Excel
Solver was used to find the values of the variables M and SD that gave
the minimum value of SSQ. With this method, M and SD estimates
were obtained, based on the available positive results and the number
of negative results. These negative results don't influence the SSQ by
an observed value ECi, but by the rank numbers of the first and subse-
quent positive values i = kn + 1…k and the value of k = kn + kp, as
used in the NORMINV function; and 5) fitted values for ECi concentra-
tions were calculated using the function NORMINV((i − 0.5) / k,M,SD)
for all samples (negative and positive results, rank numbers i = 1…k)
using the estimates for M and SD found by the Excel Solver in step 4.
This procedure was applied separately for each slaughter step (data
set A) and for each of the 13 slaughterhouses (data set B). Values of the
fitted distribution were tested for normality using both PROC UNIVARI-
ATE in SAS and by plotting the fitted data in the graph along with the EC
data. The fitted distribution was included in the mixed regression anal-
ysis used to assess the relationship between Salmonella and EC.
2.3.2. Mixed logistic regressions
Statistical models were developed to assess the change in EC and its
correlation with Salmonella occurrence using two data sets with paired
data (data sets A and B). Statistical analyses were carried out on the
fitted log-transformed values for EC (log CFU/cm2) obtained by the pro-
cedure described above (Section 2.3.1). The assignment of random rank
numbers to negative results was done once for each data set. To check if
the randomization influences the results, regression analyses were per-
formed using data obtained in four different randomizations and the re-
sults were all similar as expected (since only the rank of the first positive
point has influence on the minimization, data not shown). Mixed logis-
tic regressions with random effects (intercepts) and fixed slopes were
performed.
In data set A, a logistic regression with mixed effects was performed
(termed MLR 1). The model dimension accounted for the structure of
the data, wherein 60 carcasses were sampled over the slaughter line
at 11 points (660 data points) and then the variance matrix was blocked
by carcasses (subject). Because one carcass was sampled per day and re-
peatedly measured, random intercepts representing each carcass was
included in the model and the fitted log of the EC was included as a
fixed effect. Therefore, a constant variance was assumed between the
slaughter steps.
In data set B, a logistic regression model with mixed effects (MLR
2) combining two random intercepts, slaughterhouse and date of sam-
pling nested with slaughterhouse (slaughterhouse/day), was per-
formed using data from the 13 slaughterhouses. Fitted log of the EC
was included as a fixed effect. Using two random intercepts, the propor-
tion of variance accounted due to group levels slaughterhouse and date
of sample nested with slaughterhouse, can be obtained.
To check whether or not the number of swine slaughtered per
slaughterhouse had an effect on the probability of a carcass being
Salmonella-positive, slaughterhouse size (A ≤ 663; B N 663 ≤ 2000;
C N 2000 ≤ 2500; and D N 2500 pigs per day) was also tested as a fixed
effect along with fitted log of the EC (MLR 3). The models can be repre-
sented by the following equation:


logit Y ij ¼ β0 þ β1 xi j þ u j ð1Þ
where β0 and β1 are the intercept and fixed effects estimates respective-
ly, and u is the random intercept with uj ~ N(0,σ2u), while Yij is the binary
variable indicating Salmonella status of carcass i in group j. In MLR 1, xij is
the log of the EC estimated at observation i along the slaughter process
in carcass j, and uj is the random intercept for each carcass j (60 sub-
jects). In MLR 2, xij is the log of the EC in carcass i at slaughterhouse j,
and uj is the random intercept for each slaughterhouse j (13 subjects);
in this model, for each subject (i.e. slaughterhouse) the Z matrix con-
tains 20 intercept columns that represent 20 sampling days nested
within slaughterhouse (Appendix A provides the SAS code and model
dimensions). MLR 3 has the same dimension as MLR 2 but with two
fixed effects: fitted log of the EC and slaughterhouse size. The estimated
variances (σ2u) in logit scale measures the degree of heterogeneity in the
probability of being Salmonella-positive between subjects groups
(i.e., carcasses in MLR 1 and slaughterhouses and slaughterhouse/day
in MLRs 2 and 3). Predicted probabilities conditional on the random in-
tercepts were obtained for each observation i in the data sets using the
PRED statement in PROC GLIMMIX in SAS (Appendix A).
A Box–Tidwell test was done to verify the null hypothesis that the
predictor variable fitted log of the EC is of a linear term using the soft-
ware STATA. Results show a p-value N 0.05 (p = 0.19 and p = 0.38 in
data sets A and B respectively), suggesting that EC is suitable for using
as a continuous variable in both data sets.
To check the effect of fitting a distribution to the data (see
Section 2.3.1), all models were also tested including the non-fitted log
of EC as a fixed effect, adding 0.001 CFU/cm2 to all data. This gives a
value of −3.00 to zero counts on the log scale and makes no difference
to the log values of non-zero counts (Prendergast et al., 2008). This was
performed to check whether there was a substantial difference in the
fitted data on the results.
Mixed logistic regressions were performed in SAS (version 9.2, Cary,
NC, USA) using the PROC GLIMMIX. Appendix A provides the SAS code
explanation and model dimensions.
3. Results
3.1. Descriptive statistics
3.1.1. Data set A
Salmonella and Enterobacteriaceae were isolated from 76 (11.5%) and
523 (79.2%) of the 660 data points respectively. Enterobacteriaceae were
recovered from 71 (93%) of 76 Salmonella-contaminated carcasses.
Salmonella occurrence followed the trend of Enterobacteriaceae
concentration during the slaughter steps (Fig. 1). The prevalence of
Salmonella and mean fitted number of the EC before scalding were
43% (26/60) and 3.5 log CFU/cm2 respectively. Mean levels of Enterobac-
teriaceae contamination were reduced drastically after scalding (3.6 log
reductions) and markedly increased after dehairing (3.1 log increase).
 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66 61

Fig. 1. Observed numbers of Salmonella-contaminated carcasses and mean log of the Enterobacteriaceae count (log CFU/cm2) in 60 pork carcasses collected along the slaughter line.
BSCAL = before scalding; ASCAL = after scalding; ADEH = after dehairing; ASING = after singeing; APOLIS = after polishing; BEVISC = before evisceration; AEVISC = after
evisceration; ASPLIT = after splitting; BRINSE = before rinse; BCHILL = before chilling; ACHILL = after chilling.

Singeing decreased this number again by around 2.5 log units. During isolated in 80 of 100 Salmonella-positive carcasses corresponding to an
polishing, levels of EC increased around 1.3 log units, resulting in a average sensitivity of 80% (Table 3). In contrast, Enterobacteriaceae
mean EC of 1.7 log CFU/cm2. Subsequently, the log of the EC decreased were not recovered in 661 of 1050 Salmonella-negative carcasses corre-
gradually during the slaughter steps down to 1 log CFU/cm2 at the sponding to an average specificity of 63% (Table 3).
post-chill stage (Fig. 1).
3.2. MLR 1 (data set A)
3.1.2. Data set B
Descriptive statistics of the baseline study were performed using the Results of the models are described in Table 4. There was a statisti-
original data set without fitting (Tables 1 and 2), i.e., with the log trans- cally significant association between the log of the EC and the number
formed data from the enumeration of Enterobacteriaceae colony count of Salmonella-contaminated carcasses during slaughter (t value = 7.3;
technique. Salmonella and Enterobacteriaceae were isolated from 100 p b 0.0001). The number of observations used in the model was 652
(8.7%) and 469 (40.8%) pre-chill carcasses respectively out of 1150 sam- (8 missing), of which 76 were positive for Salmonella. The ratio of the
pled. The proportion of Salmonella-positive carcasses ranged from 2% to chi-square to the degree of freedom was 0.69, which means that there
20% between slaughterhouses, while the proportion of carcasses testing is no residual over-dispersion. The model predicts that for a 1-unit in-
positive for Enterobacteriaceae varied between 8% and 73% (Table 1). crease in the log of the EC, the expected (mean) probability of Salmonel-
The median log CFU/cm² of the EC found in the 13 slaughterhouses la detection increased by a factor of exp (0.87) = 2.39 [95% CI
varied from 0.30 to 1.52 among samples where Enterobacteriaceae (confidence interval): 1.89–3.03]. There was a high variability in the
were detected. The distribution of Salmonella-positive samples accord- probability of being Salmonella-positive between carcasses, as the
ing to the log of the EC is shown in Table 2. Enterobacteriaceae were mean covariance parameter estimate (σ2u) was 0.97 (Table 3). As an ex-
ample, in a carcass with a highly negative intercept value (μ = −0.76,
Table 1 carcass C8) in which 3.8 log CFU/cm² of Enterobacteriaceae were found
Descriptive statistics of the Enterobacteriaceae counts and Salmonella isolation from swab at the sampling i before scalding, the predicted probability of being
samples collected from 1150 carcasses (100 per slaughterhousea) in 13 pig slaughter- Salmonella-positive was 19.6% (logit of −1.41). In contrast, in a carcass
houses in Brazil. with a highly positive intercept value (μ = 1.56, carcass C9), in which
Slaughterhouse Salmonella Enterobacteriaceae count (log CFU/cm²)b 3.0 log CFU/cm² of Enterobacteriaceae were found at the same slaughter
(%) step, the predicted probability of being Salmonella-positive was 56.7%
Positive Median 75th 90th 95th Max
(%)c (logit of 0.27).
A 8 28 1.37 1.95 2.43 2.90 3.97
B 2 26 0.72 1.29 4.22 4.25 4.41 3.3. MLR 2 and MLR 3 (data set B)
C 15 51 0.49 0.88 1.43 1.54 2.73
D 4 26 0.42 0.96 1.56 2.09 3.40 The median (Q1, Q3) of the estimated log of the EC recovered from
E 9 37 0.54 0.70 1.04 1.28 2.45 Salmonella positive and negative carcasses was 0.56 (0.05, 1.16) and
F 5 21 0.44 0.74 0.95 1.10 1.63
−0.35 (−1.22, 0.40) log CFU/cm2 respectively. In MLR 2, a statistically
G 20 55 1.52 2.17 3.29 3.58 3.93
H 4 73 1.00 1.36 1.60 1.72 3.92 significant association was found between the log of the EC and the iso-
I 11 58 0.32 0.52 0.83 0.99 2.04 lation of Salmonella (t value = 5.5; p b 0.0001). On average, for a one
J 9 39 0.30 0.83 1.31 1.40 2.08 unit increase in the log of the EC, the probability of Salmonella isolation
K 20 40 1.34 1.89 2.22 2.44 4.26
increased by a factor of exp (0.52) = 1.69 (95% CI: 1.40–2.05). Estimat-
L 2 8 0.60 0.73 0.76 0.77 0.78
M 4 62 0.52 0.98 1.28 1.55 3.95 ed variances (σ2u) for the slaughterhouse and the slaughterhouse/day
Total 8.7 40.8 were 0.20 and 0.72 respectively (Table 3). The proportion of variance
a
50 samples were collected in K, L and M.
due to slaughterhouse was 22% (0.20/[0.20 + 0.72]) and for date of
b
Count distribution of positive samples (limit of detection of −1 log CFU/cm²). sample nested with slaughterhouse was 78% (0.72 / [0.20 + 0.72]), in-
c
Prevalence of Enterobacteriaceae. dicating that most of the heterogeneity in the probability of a carcass
62 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66

Table 2
Frequency distribution of Enterobacteriaceae counts (EC) and Salmonella in 1150 swabs of carcasses collected in 13 slaughterhouses in Brazil.

Percentile Enterobacteriaceae Salmonella (values per category of EC)

Category Samples Cumulative frequency (%) Positive Frequency (%)

(log CFU/cm²)

0 BLODa 681 59.2 20 2.9

b25th −1 to 0.29 116 69.3 18 15.5
≥25th b 50th 0.30–0.67 115 79.3 21 18.2
≥50th b 75th 0.68–1.24 120 89.7 18 15.0
≥75th b 90th 1.25–1.85 71 95.9 11 15.5
≥90th ≥1.86 47 100.0 12 25.5
Total 1150 – 100 8.7
a
BLOD = below limit of detection (log − 1 log CFU/cm²)

being Salmonella-positive occurred between days of sampling. Predicted
probabilities estimates can be seen in Table 5.
No statistically significant association (p = 0.71) was found when
the size of the slaughterhouse was included as a fixed effect along
with EC (MLR 3, Table 3). The EC estimates [β = 0.54, exp (β) =
1.71)] did not change notably compared with the EC estimates in MLR
2 that did not include the size of the slaughterhouse. Adding size of
the slaughterhouse in the model did not change substantially the esti-
mated variances. No residual over-dispersion was found in MLRs 2
and 3 (Table 4).
Based on fixed effects only, predicted values in the final linearized
model obtained in MLR 1 and MLR 2 evidenced a linear relationship be-
tween Enterobacteriaceae count (EC) and Salmonella in pork carcasses.
Including random effects in the model predictions, a high variability
could be observed (Fig. 2).
Results of the models (MLR 1, MLR 2 and MLR 3) using the non-fitted
EC were all similar to those obtained with the fitted distributions. In
general, slightly smaller β estimates and standard errors, and similar
variances (σ2u) were found in models using non-fitted EC.
4. Discussion
Our findings showed that the probability of being Salmonella-
positive is associated with increasing EC on a carcass, although there
was a high heterogeneity in the probability of Salmonella isolation that
cannot be explained by EC. Ghafir et al. (2008) reported a good correla-
tion between EC and Salmonella in pork, and concluded that Enterobac-
teriaceae would be a better food safety indicator in comparison to E. coli.
It has been postulated that a direct relationship between the level of an
indicator and the probability of an enteric pathogen presence in
food should exist (Ray, 2005). The results of this study demonstrate
that, despite the association found between Enterobacteriaceae and
Salmonella contamination in pork carcasses, it may be difficult to predict
when hygiene failure measured via EC actually indicates Salmonella
contamination.
In general, the dynamics of Salmonella contamination and EC along
the slaughter line found in data set A were similar to previous published
studies (Berends et al., 1997; Bolton et al., 2002; Pearce et al., 2004). The
Table 3
Cross table showing isolation of Salmonella for qualitative results of Enterobacteriaceae
from 1150 swabs of carcasses collected in 13 slaughterhouses in Brazil.
Enterobacteriaceae Salmonella isolation
Positive Negative
Positive 80 (80% ) a
389
Negative 20 661 (63%b)
Total 100 1050
a
Sensitivity of Enterobacteriaceae microbiology test.
b
Specificity of Enterobacteriaceae microbiology test.
findings of these studies demonstrated similar trends in the dynamics of
microbial levels (Enterobacteriaceae, total viable counts and/or coli-
forms) and Salmonella contamination. These dynamics were character-
ized by high levels of both indicators and Salmonella before scalding, a
marked decrease after scalding and singeing, and an increase again
after dehairing, polishing and evisceration. In this study, no consider-
able increase in EC and Salmonella prevalence was observed after evis-
ceration and after splitting. Berends et al. (1997) reported that after
singeing, it is mainly evisceration that leads to the contamination of car-
casses with Enterobacteriaceae and also Salmonella (Delhalle et al., 2008;
Pearce et al., 2004). The presence of Salmonella in the house flora
associated with the splitting saw was also considered critical for carcass
contamination during splitting (Smid et al., 2012). Similar findings were
found in a survey that included some slaughterhouses enrolled in
this study (Silva et al., 2012), in which no distinct increase in either
Salmonella number or prevalence was found after evisceration. Bolton
et al. (2002) also observed no increase in total bacterial count after evis-
ceration, suggesting good practices in the slaughterhouse concerning
training of operators and hygiene. One possible explanation is the use
of a plastic bag to seal off the rectum before bung loosening, as this
prevents the dissemination of pathogens via the feces (Hald et al.,
2003), as well as efficient cleaning and disinfection procedures.
However, variation in carcass contamination along the slaughter
line is a complex function of many factors. These are related to
batch prevalence levels, environmental contamination and slaugh-
terhouse practices such as hygiene, cleaning and disinfection, educa-
tion and training (Arguello et al., 2013; De Busser et al., 2013;
Giovannacci et al., 2001). Before scalding and up until the post-chill
stage, a significant decrease (89%) in the number of Salmonella-
contaminated carcass (from 27 carcasses to 3) and reduction in EC
levels (from 3.5 log CFU/cm2 to 1 log CFU/cm2) suggests that the process
was effective in reducing the level of contamination, and is indicative of
good hygiene practices.
The isolation of Salmonella in only one carcass after polishing was
unexpected. Dehairing and polishing can cause cross-contamination as
fecal matter can leak out from the carcass due to pressure (Borch
et al., 1996; Smid et al., 2012). Bacteria that survive the singeing process
can also be further spread over the carcass by the polishing machine
(Borch et al., 1996; Pearce et al., 2004). Persistent contamination and
isolation of Salmonella from the polishing equipment has also been de-
scribed (Giovannacci et al., 2001; Hald et al., 2003), since the equipment
is difficult to clean and disinfect (Borch et al., 1996). Nevertheless, some
authors reported that polishing is not an important point of recontami-
nation (Arguello et al., 2013; Pearce et al., 2004). It might be hypothe-
sized that the reduction in Salmonella concentration after singeing
Total brings it to a level below the detection limit, and that further spread
469 over the carcass precludes its isolation, in contrast with the higher num-
681 ber of Enterobacteriaceae still remaining.
1150 The mean prevalence (8.7%) of Salmonella found in pre-chill car-
casses from large slaughterhouses in Brazil (data set B) was similar to
the overall prevalence of external contamination with Salmonella
 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66 63

Table 4
Results of the mixed logistic regression models (MLRs 1, 2, and 3) with random intercepts and fixed slopes that included the presence of Salmonella in pork carcasses as response variable,
and fitted log of the Enterobacteriaceae count (EC) or size of the slaughterhouses along with EC as fixed effects.

Model Random effects Fixed effects Fita

Parameter Subject (No.) Estimate (SE)b Variable β (SE) p-Value

MLR 1 Intercept (σ2u) Carcass (60) 0.97 (0.38) EC 0.87 (0.12) b0.0001 0.69
MLR 2 Intercept (σ2u) Slaughterhouse (13) 0.20 (0.17) EC 0.52 (0.09) b0.0001 0.63
Intercept (σ2u) Day/slaughterhouse 0.72 (0.23)
(20c/slaughterhouse)
MLR 3 Intercept (σ2u) Slaughterhouse (13) 0.26 (0.24) EC 0.54 (0.09) b0.0001 0.64
Sized
Intercept (σ2u) Day/slaughterhouse 0.73 (0.27) A −0.02 (0.63)
(20/slaughterhouse) B −0.34 (0.52) 0.70
C −0.55 (0.50)
D 0
a
Chi-square/DF value.
b
Logit scale.
c
Minimum 10 and maximum 20 sampling days per slaughterhouse.
d
A ≤ 663; B N 663 ≤ 2000; C N 2000 ≤ 2500; and D ≥ 2500 pigs slaughtered per day.

reported in the EU, which was 8.3% (EFSA, 2008). In southern Brazil,
previous studies reported 24% (Kich et al., 2011) and 14.7% (Silva
et al., 2012) of Salmonella positive pre-chill carcasses. However, these
studies included only three slaughterhouses and used a distinct sam-
pling scheme, which may have led to frequencies that reflected the
slaughter day rather than the overall prevalence. On the contrary, the
results of the present study are obtained in a more comprehensive
sampling plan, and represent the first baseline study on Salmonella in
pork conducted in southern Brazil. Studies conducted in the United
States and Canada have reported a mean prevalence of 6.9% and 1.6%
in post-chill carcasses respectively (Bohaychuk et al., 2011; USDA,
1998).
A strong association between Salmonella isolation and Enterobacteri-
aceae levels on pre-chill carcass surfaces was found. Results are in agree-
ment with those of previous studies made on pre-chill carcasses
(Duggan et al., 2010; Ghafir et al., 2008) and in pork cuts (Prendergast
et al., 2008). However, the variation occurring between slaughterhouses
and day of sample may have been overlooked in previous studies. A
wide range in the percentages of Salmonella-positive carcasses between
slaughterhouses (between 2% and 20%, Table 1) as well as between days
of sampling was observed in our study. A high variation in the associa-
tion between Salmonella occurrence and EC was found mainly between
days of sampling (Table 4). MLR 1 corroborates the findings of MLR 2, as
a high variation between carcasses was found here as well. Given that
the former model included data from 60 carcasses, which were sampled
on 60 different days, this also suggests that other factors occurring on
the day of slaughter or between carcasses have contributed for Salmo-
nella contamination. The effect of the day of sampling on the Salmonella
prevalence was so large that the value of the EC to predict Salmonella
isolation in a daily basis was compromised. This fact may be illustrated
by the high levels of EC found on a given day (maximum of ~5 log CFU/
cm², slaughterhouse B) that were associated with a smaller probability
of Salmonella contamination than the average. On the contrary, on one
sampling day at slaughterhouse A the maximum of ~ 1.8 log CFU/cm²
found was associated with a higher probability of a carcass being
Salmonella-positive (Table 4). The size of the slaughterhouse was
also tested in the model MLR 3 to assess its effects on Salmonella
prevalence, and no association or change of estimates for EC fixed
effects was found. Baptista et al. (2010) also found no association
between slaughterhouse size and an increased probability of
Salmonella-positive pork carcasses.
Slaughterhouse and a day-to-day variation in prevalence are fre-
quently reported in Salmonella epidemiological studies performed in
pig slaughterhouses (Baptista et al., 2010; Botteldoorn et al., 2003; De
Busser et al., 2011; Delhalle et al., 2008; Hald et al., 2003; Prendergast
et al., 2008; Silva et al., 2012). Many of these studies reported a correla-
tion between seroprevalence, and the number of pigs shedding Salmo-
nella in their feces, with carcass contamination (Botteldoorn et al.,
2003; Duggan et al., 2010; Silva et al., 2012). Based on a previous sero-
logical survey conducted in 188 finishing farms in Brazil (Schwarz
et al., 2010), exposure to the pathogen on farm seems to be high, since
142 farms (75.5%) had a seroprevalence greater than 70% in their

Table 5
An example of the predicted probabilities estimated for carcasses i in the mixed logistic regression (MLR 2) on two days of sampling nested with slaughterhouse j. Date of sampling was
sorted by increasing intercept estimates values.

Subject

Slaughterhouse Day of sampling Estimate Carcass i Salmonella Fitted log CFU/cm² Enterobacteriaceae Predicted probability Logit

B 12/03/2012 −0.65 1 Neg 2.58 11% −2.10

2 Neg 2.92 13% −1.92
3 Neg 3.34 15% −1.70
4 Neg 3.94 20% −1.38
5 Neg 5.10 32% −0.77
. . . . . . . .
A 30/08/2011 1.98 6 Pos −3.42 9% −2.32
7 Pos 0.63 45% −0.19
8 Pos 1.10 52% 0.06
9 Pos 1.65 59% 0.35
10 Pos 1.76 60% 0.41
64 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66

Fig. 2. Graphic showing predicted values in the final linearized model obtained in MLR 1 and MLR 2 models that included carcasses and slaughterhouses/day of sample as a random effect
respectively. MLR 3 was omitted because it contains very similar results to MLR 2. Predictions based on the fixed effects alone (the population-averaged prediction) illustrate the linear
relationship between Enterobacteriaceae count (EC) and Salmonella in pork carcasses. The high variability in the probability of being Salmonella-positive according to EC (log CFU/cm²)
can be observed in the plots that show the predicted values including random effects.

slaughter batches. In a scenario of high exposure on farm, it is possible
that on some days, batches with a high prevalence of Salmonella carriers
shedding a high number of Salmonella were slaughtered. On these days,
the hazard of contamination/cross-contamination of carcasses maybe
so high that even a hygienic slaughter, confirmed by the low EC on
carcasses, will not be able to avoid the presence of Salmonella on some
carcasses.
Including the original EC by adding 0.001 to all the data resulted in
similar estimates to those obtained with the fitted distribution of EC.
Smaller parameter estimates and standard errors were expected due
to the smaller variation in the data set when the same value is used
for the zero counts. Both sets of models were adjusted according to
the fit statistics, where no over-dispersion was detected. One possible
reason for the small difference observed is that on average, 40.8%
(MLR 2 and MLR 3) and 79.2% (MLR 1) of samples were positive for En-
terobacteriaceae and that most of the Salmonella was isolated from those
samples. However, the most reliable model is hard to define, as no infor-
mation regarding the level of contamination can be determined for
carcasses with very low levels of contamination. On the other hand, it
is reasonable to assume that the zero counts do have a distinct level of
contamination and that the data fitted to a normal distribution better
represents the reality.
The results of this study reinforce the importance of hygiene pro-
cedures during slaughter for the control of Salmonella in pork, since
on average Salmonella and Enterobacteriaceae detection was related.
Microbial testing is a useful tool for monitoring hygiene of the
slaughtering process over time (FSIS, 1996). However, even in
slaughterhouses with high standards of hygiene measured as EC,
the delivery of pig batches with a high level of Salmonella shedding
may result in a slaughter day with higher than average Salmonella-
contaminated carcasses. This fact should be taken into account
when Enterobacteriaceae are used as an indicator of food safety in
pork. It also stresses that hygienic processing alone cannot control
Salmonella contamination and that a whole of chain approach is like-
ly required, taking into account cost effectiveness of the various
control options.
Conflict of interest statement
No conflict of interest exists for any of the authors.
Appendix A. SAS code and model dimensions for mixed logistic
regression (MLR 1 and MLR 2)
In this model, for each subject (carcass_id) SAS sets up 1 column in
the Z matrix for each carcass. The options DIST = BINARY and
LINK = LOGIT specify that the response variable is binarily distributed
and the link function is logit. The residual degree of freedom was divid-
ed into between-subjects and within-subjects to determine whether a
fixed effect (that is, EC) changes within any subject. The option
DDFM = BW was used for this. SOLUTION option in the MODEL state-
ment requests a listing of the fixed-effects parameter estimates. The
RANDOM statement specifies that the linear predictor contains an inter-
cept term that randomly varies at the level of the carcass_id effect (or
slaughterhouse_id and day of sample nested with slaughterhouse_id ef-
fect in MLRs 2 and 3). Two types of predicted values are requested in the
OUTPUT statement for each observation in the data set. The first type
uses the solutions for the random effects [the estimated best linear un-
biased predictors (BLUPs) in the final linearized model], and the second
type uses predictions based on the fixed effects alone (NOBLUPs, i.e., the
population-averaged prediction). The ILINK suboption of the PRED key-
word requests that the inverse link function be applied to the resulting
predictions. This yields predictions of probabilities. SOLUTION option in
the random statement requests that the solution for the random-effects
parameters be produced. The output for the solution for random
carcass_id effects is as follows (from the smallest to the highest esti-
mates of the intercepts):

Effect Subject Estimate Std Err Pred DF t Value Pr N |t|

Acknowledgements
Intercept Id_carcassC49 −0.92 0.75 650 −1.24 0.22
Intercept Id_carcassC8 −0.76 0.77 650 −0.99 0.32
Financial support was provided by the CNPq/Brazil through the Post- Intercept Id_carcassC32 −0.75 0.78 650 −0.96 0.34
doctoral Training Program (process # 241904/2012-9). Special thanks Intercept Id_carcassC28 −0.71 0.77 650 −0.92 0.36
are given to all the official veterinarians from MAPA (Ministério da Intercept Id_carcassC38 −0.69 0.79 650 −0.88 0.38
Agricultura, Pecuária e Abastecimento) enrolled in the microbiological Intercept Id_carcassC30 −0.69 0.78 650 −0.88 0.38
Intercept Id_carcassC25 −0.67 0.78 650 −0.86 0.39
baseline study.
 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66 65

Appendix A (continued)
(continued) There are two G-sides random effects in this model, which are
Effect Subject Estimate Std Err Pred DF t Value Pr N |t|
slaughterhouse and date nested within slaughterhouse. In this model,
for each subject (slaughterhouse_id) SAS sets up 20 columns in the Z
Intercept Id_carcassC40 −0.66 0.78 650 −0.84 0.40
Intercept Id_carcassC5 −0.65 0.79 650 −0.81 0.42
matrix for each slaughterhouse, which corresponds to the maximum
Intercept Id_carcassC59 −0.64 0.79 650 −0.81 0.42 number of sampling days in any slaughterhouse.
Intercept Id_carcassC3 −0.62 0.79 650 −0.79 0.43
Intercept Id_carcassC50 −0.60 0.80 650 −0.75 0.45
Intercept Id_carcassC24 −0.60 0.80 650 −0.75 0.45
References
Intercept Id_carcassC23 −0.59 0.81 650 −0.73 0.47
Intercept Id_carcassC51 −0.59 0.80 650 −0.73 0.46 ABIPECS, 2012. Available in: http://www.abipecs.org.br/uploads/relatorios/relatorios-
Intercept Id_carcassC14 −0.58 0.80 650 −0.72 0.47 associados/ABIPECS_relatorio_2012_pt.pdf (Accessed on November 5, 2013).
Intercept Id_carcassC55 −0.57 0.81 650 −0.70 0.48 Arguello, H., Alvarez-Ordonez, A., Carvajal, A., Rubio, P., Prieto, M., 2013. Role of
Intercept Id_carcassC45 −0.56 0.81 650 −0.69 0.49 slaughtering in Salmonella spreading and control in pork production. J. Food Prot.
Intercept Id_carcassC44 −0.56 0.81 650 −0.69 0.49 76, 899–911.
Intercept Id_carcassC19 −0.56 0.80 650 −0.69 0.49 Baptista, F.M., Dahl, J., Nielsen, L.R., 2010. Factors influencing Salmonella carcass preva-
Intercept Id_carcassC4 −0.55 0.81 650 −0.67 0.50 lence in Danish pig abattoirs. Prev. Vet. Med. 95, 231–238.
Intercept Id_carcassC29 −0.50 0.82 650 −0.61 0.54 Berends, B.R., Van Knapen, F., Snijders, J.M., Mossel, D.A., 1997. Identification and quanti-
Intercept Id_carcassC20 −0.49 0.82 650 −0.60 0.55 fication of risk factors regarding Salmonella spp. on pork carcasses. Int. J. Food
Intercept Id_carcassC52 −0.48 0.83 650 −0.58 0.56 Microbiol. 36, 199–206.
Bohaychuk, V.M., Gensler, G.E., Barrios, P.R., 2011. Microbiological baseline study of beef
Intercept Id_carcassC18 −0.39 0.84 650 −0.47 0.64
and pork carcasses from provincially inspected abattoirs in Alberta, Canada. Can.
Intercept Id_carcassC12 −0.27 0.70 650 −0.39 0.70
Vet. J. 52, 1095–1100.
Intercept Id_carcassC47 −0.27 0.70 650 −0.39 0.70
Bolton, D.J., Pearce, R.A., Sheridan, J.J., Blair, I.S., McDowell, D.A., Harrington, D., 2002.
Intercept Id_carcassC60 −0.25 0.72 650 −0.35 0.73 Washing and chilling as critical control points in pork slaughter hazard analysis
Intercept Id_carcassC10 −0.19 0.72 650 −0.26 0.79 and critical control point (HACCP) systems. J. Appl. Microbiol. 92, 893–902.
Intercept Id_carcassC11 −0.15 0.72 650 −0.22 0.83 Borch, E., Nesbakken, T., Christensen, H., 1996. Hazard identification in swine slaughter
Intercept Id_carcassC1 −0.15 0.73 650 −0.21 0.83 with respect to foodborne bacteria. Int. J. Food Microbiol. 30, 9–25.
Intercept Id_carcassC33 −0.11 0.73 650 −0.14 0.88 Botteldoorn, N., Heyndrickx, M., Rijpens, N., Grijspeerdt, K., Herman, L., 2003. Salmonella
Intercept Id_carcassC43 −0.08 0.72 650 −0.11 0.91 on pork carcasses: positive pigs and cross contamination in the slaughterhouse.
Intercept Id_carcassC42 0.03 0.76 650 0.04 0.97 J. Appl. Microbiol. 95, 891–903.
Intercept Id_carcassC22 0.04 0.74 650 0.06 0.95 Brown, M.H., Gill, C.O., Hollingsworth, J., Nickelson II, R., Seward, S., Sheridan, J.J.,
Intercept Id_carcassC54 0.08 0.75 650 0.10 0.92 Stevenson, T., Summer, J.L., Theno, D.M., Usborne, W.R., Zinf, D., 2000. The role of mi-
Intercept Id_carcassC7 0.10 0.65 650 0.15 0.88 crobial testing in systems form assuring the dafety of beef. Int. J. Food Microbiol. 62,
7–16.
Intercept Id_carcassC46 0.10 0.75 650 0.13 0.89
Busschaert, P., Geeraerd, A.H., Uyttendaele, M., Van Impe, J.F., 2010. Estimating distribu-
Intercept Id_carcassC58 0.13 0.66 650 0.20 0.84
tions out of qualitative and (semi)quantitative microbiological contamination data
Intercept Id_carcassC48 0.16 0.68 650 0.24 0.81
for use in risk assessment. Int. J. Food Microbiol. 138, 260–269.
Intercept Id_carcassC39 0.21 0.78 650 0.27 0.79 Commission regulation (EC) No 2073/2005, 2005. Commission regulation (EC) No. 2073/
Intercept Id_carcassC2 0.23 0.77 650 0.30 0.76 2005 of 15 November on microbiological criteria for foodstuffs. Off. J. Eur. Union
Intercept Id_carcassC35 0.30 0.71 650 0.43 0.67 331–338.
Intercept Id_carcassC6 0.49 0.68 650 0.73 0.47 De Busser, E.V., Maes, D., Houf, K., Dewulf, J., Imberechts, H., Bertrand, S., De Zutter, L.,
Intercept Id_carcassC36 0.53 0.71 650 0.74 0.46 2011. Detection and characterization of Salmonella in lairage, on pork carcasses and
Intercept Id_carcassC21 0.61 0.71 650 0.87 0.39 intestines in five slaughterhouses. Int. J. Food Microbiol. 145, 279–286.
Intercept Id_carcassC26 0.63 0.70 650 0.90 0.37 De Busser, E.V., De Zutter, L., Dewulf, J., Houf, K., Maes, D., 2013. Salmonella control in live
Intercept Id_carcassC13 0.73 0.63 650 1.15 0.25 pigs and at slaughter. Vet. J. 196, 20–27.
Intercept Id_carcassC41 0.75 0.62 650 1.20 0.23 Delhalle, L., De Sadeleer, L., Bollaerts, K., Farnir, F., Saegerman, C., Korsak, N., Dewulf, J., De
Intercept Id_carcassC27 0.75 0.66 650 1.14 0.25 Zutter, L., Daube, G., 2008. Risk factors for Salmonella and hygiene indicators in the 10
Intercept Id_carcassC16 0.84 0.64 650 1.32 0.19 largest Belgian pig slaughterhouses. J. Food Prot. 71, 1320–1329.
Duggan, S.J., Mannion, C., Prendergast, D.M., Leonard, N., Fanning, S., Gonzales-Barron, U.,
Intercept Id_carcassC57 0.89 0.66 650 1.35 0.18
Egan, J., Butler, F., Duffy, G., 2010. Tracking the Salmonella status of pigs and pork from
Intercept Id_carcassC56 0.93 0.67 650 1.38 0.17
lairage through the slaughter process in the Republic of Ireland. J. Food Prot. 73,
Intercept Id_carcassC34 0.97 0.67 650 1.44 0.15 2148–2160.
Intercept Id_carcassC53 1.05 0.64 650 1.63 0.10 EFSA, 2008. Report of the task force on zoonoses data collection on the analysis of the
Intercept Id_carcassC17 1.16 0.60 650 1.93 0.05 baseline survey on the prevalence of Salmonella in slaughter pigs, in the EU, 2006–
Intercept Id_carcassC15 1.17 0.60 650 1.94 0.05 2007. EFSA J. 135, 1–111.
Intercept Id_carcassC37 1.19 0.61 650 1.96 0.05 FSIS, 1996. Pathogen reduction: hazard analysis and critical control point (HACCP) sys-
Intercept Id_carcassC31 1.21 0.59 650 2.07 0.04 tems; final rule. Fed. Regist. 61 (144), 38,806–38,989.
Intercept Id_carcassC9 1.56 0.63 650 2.48 0.01 Ghafir, Y., Daube, G., Dierick, K., De Zutter, L., Cornelis, M., Jouret, M., 2003. Assessment of
microbiological criteria for regular checks of faecal contamination and general hy-
giene in Belgian establishments producing meat. Sci. Aliment. 23, 104–106.
Ghafir, Y., China, B., Dierick, K., De Zutter, L., Daube, G., 2008. Hygiene indicator microor-
ganisms for selected pathogens on beef, pork, and poultry meats in Belgium. J. Food
Prot. 71, 35–45.
Giovannacci, I., Queguiner, S., Ragimbeau, C., Salvat, G., Vendeuvre, J.L., Carlier, V., Ermel,
G., 2001. Tracing of Salmonella spp. in two pork slaughter and cutting plants using
serotyping and macrorestriction genotyping. J. Appl. Microbiol. 90, 131–147.
Gomes, B.C., Franco, B.D.G., de Martins, E.C.P., 2013. Microbiological food safety issues in
Brazil: bacterial pathogens. Foodborne Pathog. Dis. 10, 197–205.
Hald, T., Wingstrand, A., Swanenburg, M., von Altrock, A., Thorberg, B.M., 2003. The occur-
rence and epidemiology of Salmonella in European pig slaughterhouses. Epidemiol.
Infect. 131, 1187–1203.
Kich, J.D., Coldebella, A., Mores, N., Nogueira, M.G., Cardoso, M., Fratamico, P.M., Call, J.E.,
Fedorka-Cray, P., Luchansky, J.B., 2011. Prevalence, distribution, and molecular
characterization of Salmonella recovered from swine finishing herds and a slaughter
facility in Santa Catarina, Brazil. Int. J. Food Microbiol. 151, 307–313.
MAPA, 2003. Diário Oficial da União. Instrução Normativa, SDA N° 62, de 26 de
agosto de 2003. Available in: http://extranet.agricultura.gov.br/sislegis-
consulta/consultarLegislacao.do?operacao=visualizar&id=2851 (Accessed
on November 5, 2013).
Pearce, R.A., Bolton, D.J., Sheridan, J.J., McDowell, D.A., Blair, I.S., Harrington, D., 2004.
Studies to determine the critical control points in pork slaughter hazard analysis
and critical control point systems. Int. J. Food Microbiol. 90, 331–339.
Prendergast, D.M., Duggan, S.J., Fanning, S., Cormican, M., Gonzales-Barron, U., Butler, F.,
Duffy, G., 2008. Prevalence and numbers of Salmonella spp. and Enterobacteriaceae
66 L.G. Corbellini et al. / International Journal of Food Microbiology 228 (2016) 58–66
on pork cuts in abattoirs in the Republic of Ireland. J. Appl. Microbiol. 105,
1209–1219.
Ray, B., 2005. Fundamental Food Microbiology. third ed. CRC Press Boca Raton, Florida.
Schwarz, P., Kich, J.D., Coldebella, A., Seyboth, L., Romeiro, C., Corbellini, L.G., Cardoso, M.,
2010. Frequência de suínos soropositivos para Salmonella sp. em granjas afetadas em
diferentes níveis de severidade pela síndrome multissistêmica de definhamento do
leitão desmamado. Acta Sci. Vet. 38, 127–132.
Silva, L.E., Dias, V., Ferronatto, A., Guerra, P., Berno, L., Triches, N., Kich, J.D., Corbellini, L.G.,
Cardoso, M., 2012. Longitudinal dissemination of Salmonella enterica clonal groups
through the slaughter process of Salmonella-positive pig batches. J. Food Prot. 75,
1580–1588.
Smid, J.H., Heres, L., Havelaar, A.H., Pielaat, A., 2012. A biotracing model of Salmonella in
the pork production chain. J. Food Prot. 75, 270–280.
USDA, 1998. Nationwide sponge microbiological baseline data collection program: swine.
Baseline Data Reports. United States Department of Agriculture, United States of
America, p. 15.
Van Hoek, A.H., de Jonge, R., van Overbeek, W.M., Bouw, E., Pielaat, A., Smid, J.H., Malorny,
B., Junker, E., Lofstrom, C., Pedersen, K., Aarts, H.J., Heres, L., 2012. A quantitative ap-
proach towards a better understanding of the dynamics of Salmonella spp. in a pork
slaughter-line. Int. J. Food Microbiol. 153, 45–52.