Predictive Microbiology Approach for Enumeration of Salmonella on Chicken Parts

1
Predictive Microbiology Approach for Enumeration
of Salmonella on Chicken Parts Thomas P. Oscar,
Agricultural Research Service, USDA, Room 2111,
Center for Food Science and Technology,
University of Maryland Eastern Shore, Princess
Anne, MD 21853 410-651-6062 410-651-8498
(fax) Thomas.Oscar_at_ars.usda.gov
INTRODUCTION A data gap identified in risk
assessments for Salmonella and chicken is lack of
quantitative data. Enumeration of Salmonella on
chicken parts is difficult because Salmonella are
often present in low numbers. However, during
the pre-enrichment phase of Salmonella isolation
from chicken parts, there is a mathematical
relationship between the initial number of
Salmonella on the chicken part and the
concentration of Salmonella in the pre-enrichment
broth at early times of incubation1. Thus, it
should be possible to develop a mathematical
model that predicts the number of Salmonella on
chicken parts as a function of the concentration
of Salmonella in isolation broth at an early time
of incubation during pre-enrichment of whole
chicken parts. OBJECTIVE To develop a
predictive model for enumerating Salmonella on
chicken parts during pre-enrichment. MATERIALS
AND METHODS Salmonella. Predictive models for
enumerating Salmonella on chicken parts were
developed with four isolates of Salmonella S.
Typhimurium DT104 ATCC 700408 s165 (sT165), S.
Typhimurium ATCC 14028 s2 (sT2), S. Kentucky s361
(sK361), and S. 8,20-z6 s362 (sz362). Chicken
part preparation. Fresh, whole chickens were
purchased at retail. A sterile cutting board and
knife were used to partition the whole chicken
into two wings, two breasts, two drumsticks, and
two thighs. A sterile cooked chicken breast was
then cut into two equal-size portions using the
cutting board and knife used to partition the
whole raw chicken this was done to study
transfer of Salmonella from raw chicken to cooked
chicken during meat preparation. Chicken
part inoculation and incubation. Chicken parts
were pre-enriched in 400 ml of buffered peptone
water (BPW) for 6 h at 42C and 80 rpm. For
predictive model development, chicken parts were
spot inoculated (5 µl) with 0.36 to 4.86 logs of
Salmonella before pre-enrichment in BPW. Model
development. At 6 h of pre-enrichment, the
concentration of Salmonella in the BPW
pre-enrichment (Y log CFU/ml) was determined by
spiral plating onto XLT4 agar, graphed as a
function of the log number of Salmonella
inoculated onto the chicken parts (X log10
CFU/part) and then the data were fitted to a
linear model using GraphPad Prism for Windows
version 5.02 Y a bX where a was the
Y-intercept and b was the slope of the best-fit
line. Pair-wise comparisons of the parameters of
the linear models among chicken parts and
Salmonella isolates were made using an F-test in
Prism.
Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts Table 1. Standard curve parameters for different isolates of Salmonella inoculated onto chicken parts
Isolate Code Parameter BFV SE CI R2
S. Typhimurium DT104 ATCC 700408 s165 sT165 Y-int 2.75 0.10 2.47 to 3.03 0.994
S. Typhimurium ATCC 14028 s2 sT2 2.50 0.06 2.25 to 2.75 0.999
S. Kentucky s361 sK361 1.93 0.06 1.69 to 2.17 0.999
S. 8,20-z6 s362 sz362 1.96 0.12 1.47 to 2.46 0.997

S. Typhimurium DT104 ATCC 700408 s165 sT165 slope 0.95 0.04 0.85 to 1.04  
S. Typhimurium ATCC 14028 s2 sT2 1.00 0.02 0.90 to 1.10  
S. Kentucky s361 sK361 1.00 0.02 0.93 to 1.08  
S. 8,20-z6 s362 sz362   1.03 0.04 0.86 to 1.19  
Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination. Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination. Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination. Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination. Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination. Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination. Abbreviations BFV best fit value SE standard error CI 95 confidence interval and R2 coefficient of determination.
Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella Table 2. Pair-wise comparison of standard curve parameters among isolates of Salmonella
  Y-intercept Y-intercept Y-intercept Y-intercept Slope Slope Slope Slope
Comparison F-value P-value DFn DFd F-value P-value DFn DFd
sT165 vs sT2 2.98 0.128 1 7 0.80 0.406 1 6
sT165 vs sK361 80.51 lt 0.001 1 7 0.82 0.399 1 6
sT165 vs sz362 46.22 0.001 1 7 1.67 0.244 1 6
sT2 vs sK361 343.9 lt0.001 1 5 0.00 1.000 1 4
sT2 vs sz362 93.01 lt0.001 1 5 0.39 0.566 1 4
sK361 vs sz362 6.43 0.052 1 5 0.44 0.542 1 4
Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part Table 3. Serotype of number of Salmonella per chicken part
Chicken Date Brand Part Weight (g) Isolate Serotype log CFU/ml log CFU/part CFU/part
4 12/14/2010 C thigh 173 s361 Kentucky 2.34 0.41 3
15 4/12/2011 A wing 74 s362 8,20-z6 1.30 0.00 1
15 4/12/2011 A thigh 129 s363 8,20-z6 1.78 0.00 1
15 4/12/2011 A cooked 102 s364 8,20-z6 1.30 0.00 1
15 4/12/2011 A wing 78 s365 8,20-z6 1.60 0.00 1
Prevalence of Salmonella among edible parts from
whole chickens obtained at retail was 3 (4/132),
whereas incidence of cross-contamination from raw
chicken to cooked chicken breast during simulated
meal preparation was 1.8 (1/57). The positive
chicken parts (Table 3) were thigh from chicken
4, which was contaminated with 3 CFU of serotype
Kentucky, and both wings, one thigh and one
cooked breast portion from chicken 15, which
were all contaminated with 1 CFU of serotype
8,20-z6. DISCUSSION The reason for acquiring
data for initial contamination of chicken parts
with Salmonella and cross-contamination of other
foods with Salmonella from raw chicken during
meal preparation was to fill important data gaps
in risk assessments for Salmonella and chicken
that are being performed by regulatory agencies
in the United States, Europe and throughout the
world. These risk assessments are being used as
the scientific basis to inform policy decisions
aimed at protecting public health from this
important foodborne pathogen. For risk
assessment purposes it is not only useful to know
the number of pathogens on and in a food sample
but also their ability to grow and cause
infection. Thus, an enumeration method, such as
the one used in the current study, that is based
on the growth kinetics of Salmonella during
incubation of the food sample under favorable
conditions for growth, provides the type of
quantitative data that is highly relevant for
risk assessments. These results indicated that
a predictive microbiology approach can be used to
enumerate low numbers of Salmonella on chicken
parts during pre-enrichment of samples in BPW.
However, because of the low prevalence of
Salmonella on the chicken parts examined it was
not possible, at this time, to fill the data gap
for quantitative data identified in recent risk
assessments for Salmonella and chicken.

Fig. 1. Pair-wise comparisons of linear models
developed for different types of chicken parts
combined data from all isolates of Salmonella.
RESULTS All linear models had high
goodness-of-fit (Table 1) regardless of the
serotype of Salmonella used. Parameters of the
linear models were not affected (P gt 0.05) by
type of chicken part (Fig. 1) but were affected
(P lt 0.05) by isolate of Salmonella (Fig. 2).
Isolates sT165 and sT2 from ATCC were found to
grow faster (higher Y-intercept but similar
slope) than isolates sK361 and sz362, which were
isolated from chicken parts in the present study
(Table 2).
worst-case scenario
Fig. 2. Pair-wise comparisons of linear models
developed for different isolates of Salmonella
combined data from all types of chicken parts.