WWW

WNW

\
h
I

 

5 Q

 

LIBRARY
MiG? 159:1! I State
University

 

 

 

This is to certify that the
thesis entitled

VALIDATION OF SALMONELLA THERMAL LETHALITY IN
WHOLE MUSCLE MEAT PRODUCTS DURING PILOT-
SCALE SLOW ROASTING PROCESSES

presented by

Tasha Joy Breslin

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in Food Science

 
    
 

ro essor’s Signature

7266 2&7)?

Date

MSU is an Afﬁrmative Action/Equal Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:lProj/Acc&Pres/ClRC/DateDue.indd

 

VALIDATION OF SALMONELLA THERMAL LETHALITY IN WHOLE
MUSCLE MEAT PRODUCTS DURING PILOT-SCALE SLOW ROASTING
PROCESSES
By

Tasha Joy Breslin

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

MASTER OF SCIENCE
Food Science

2009

ABSTRACT
VALIDATION OF SALMONELLA THERMAL LETHALITY IN WHOLE
MUSCLE MEAT PRODUCTS DURING PILOT-SCALE SLOW ROASTING
PROCESSES
By

Tasha Joy Breslin

Previous research has shown that sub-lethal heating can increase subsequent thermal
resistance of bacteria. If this phenomenon occurs during slow roasting of meat products,
it might compromise the validity of thermal process validations. Therefore, this research
evaluated the accuracy of a traditional log-linear inactivation model, developed via prior
laboratory-scale, isothermal tests, applied to pilot-scale, slow cooking of whole muscle
roasts. Irradiated turkey breast, beef rounds, and pork loin were inoculated with an 8-
servovar Salmonella cocktail via vacuum tumble marination, in a salt/phosphate
marinade. The resulting initial Salmonella population in the geometric center (core) was
7.0, 6.3, 6.3 log CFU/g for turkey, beef, and pork respectively. The experimental design
consisted of seven different cooking combinations representing industry practices, in a
pilot-scale, moist-air convection oven. Core temperature was recorded during cooking,
and was used to calculate lethality real-time via the log-linear model. Calculated lethality,
using the log-linear model, was greater (P < 0.05) than the actual lethality for turkey and
beef. A path-dependent model accounting for sub-lethal history of Salmonella reduced
the lethality error by 2.6 and 1.4 log CFU/g in turkey and beef, respectively, but did not
reduce the error in pork. Results demonstrate that slow-cooked roasts, processed to a
lethality (as calculated by a state-dependent model) at or near that required by the

regulatory performance standards may be under-processed.

ACKNOWLEDGEMENTS

Heart felt and sincere appreciation to my major adviser, Dr. Bradley Marks, and my
committee members, Dr. Alden Booren, and Dr. Elliot Ryser for your guidance, and
expertise. My thanks are also expressed to Nicole Hall, who was always willing to take
the time to guide me, and kindly taught me how to be an effective problem solver.
Sanghyup Jeong and Michael James, were always a reliable resource, and took the time
to help me resolve equipment issues. Undergraduate students in Dr. Marks’ laboratory,
and fellow graduate students; we all worked together to accomplish our goals. Finally and
most importantly, I thank my entire family and all of my friends for always being there,

and helping me to become the person that I am today.

iii

TABLE OF CONTENTS

 

 

 

 

 

LIST OF TABLES vi
LIST OF FIGURES viii
1. INTRODUCTION - l
2. LITERATURE REVIEW - 4
2.1. Foodbome Pathogens ................................................................................................... 4
2.1.1 Salmonella .......................................................................................................... 4
2.1.2 Pathogens in Meat .............................................................................................. 5
2.1.2.1 Salmonella ............................................................................................. 6

2.1.2.2 Mechanisms and modes of transfer ....................................................... 7

2.2. Thermal Resistance ...................................................................................................... 8
2.2.1 Product Factors Affecting Thermal Resistance ................................................. 8
2.2.2 Process Factors ................................................................................................. 10
2.2.2.1 Sub-lethal Injury, Non thermal ........................................................... 11

2.2.2.2 Sub-lethal Injury, Thermal .................................................................. 11

2.2.3 Inactivation Models ......................................................................................... 13
2.2.3.1 Primary Models ................................................................................... 13

2.2.3.2 Secondary Models ............................................................................... 15

2.2.4 Modeling Effects of Sub-lethal History ........................................................... 16

2.3. Commercial Meat Cooking ........................................................................................ 17
2.3.1 Validation of Process ....................................................................................... 17
2.3.2 Validation of Oven Conditions ........................................................................ 18
2.3.3 Humidity .......................................................................................................... 19
2.3.3.1 F SIS Regulations of Humidity Use ..................................................... 19

2.3.3.2 Cook-in—bag ........................................................................................ 20

3. MATERIALS AND METHODS 21
3.1 Isothermal Research .................................................................................................... 21
3.2 Acquisition of Meat .................................................................................................... 21
3.3 Preparation of Salmonella ........................................................................................... 22
3.4 Preparation of Marinade ............................................................................................. 23
3.5 Inoculation .................................................................................................................. 24
3.6 Inoculation Veriﬁcation .............................................................................................. 25
3.7 Sample Preparation ..................................................................................................... 27
3.8 Oven ............................................................................................................................ 27
3.9 Cooking ....................................................................................................................... 28
3.10 Cooking Schedules .................................................................................................... 29
3.11 Model Computed Lethality ....................................................................................... 34

iv

3.11.1 State-Dependent Model ................................................................................. 34

 

 

 

 

3.11.2 Path-Dependent Model .................................................................................. 36
3.12 Data Analysis ............................................................................................................ 37
4. RESULTS AND DISCUSSION 38
4.1 Isothermal Results ....................................................................................................... 38
4.2 Product Composition .................................................................................................. 38
4.3 Initial Salmonella Concentration in Whole Muscle Roasts ........................................ 38
4.4 End Point Temperature ............................................................................................... 39
4.5 Lethality Error ............................................................................................................. 41
4.6 Replication Error ......................................................................................................... 43
4.7 Sub-lethal History vs. Model Error ............................................................................. 46
4.8 Comparison of State-Dependent and Path-Dependent Models .................................. 49
4.9 Safe Harbors ................................................................................................................ 54
4.10 Challenges of Whole Muscle Pilot-Scale Research .................................................. 56
5. CONCLUSIONS 60
5.1 Implications of Process Scale-Up Variability ............................................................. 60
5.2 Lethality Error vs. Sub-lethal History ......................................................................... 60
5.3 Implications of Using a State-Dependent vs. Path-Dependent Model ....................... 61
6. FURTHER WORK- 63
7. APPENDICES - 67
Appendix 7.A: Isothermal tests on whole and ground beef .............................................. 67
Appendix 7.B: Inoculation veriﬁcation summary data ..................................................... 78
Appendix 7 .C: Effect of the roast mass on the initial inoculums concentration ............... 79
Appendix 7.D: Veriﬁcation of coring method following cooking ................................... 80
Appendix 7.B: Alternative inoculation veriﬁcation for pork ............................................ 82
Appendix 7.F : Summarized cook schedules and spreadsheets ......................................... 85
8. BIBLOGRAPHY 96

 

LIST OF TABLES

Table l. D and 2 values calculated from isothermal testing of the 8-serovar Salmonella
cocktail in turkey, beef, and pork (Tuntivanich and others 2008; Velasquez and
others 2009). ............................................................................................................. 34

Table 2. State-dependent model parameters for whole turkey, beef, and pork (Marks and
others 2009). ............................................................................................................. 35

Table 3. Path-dependent model parameters and goodness of ﬁt obtained from
nonisothermal calibration data sets for turkey, beef, and pork (Tenorio-Bemal and
others 200X) .............................................................................................................. 37

Table 4. Initial core Salmonella population following inoculation, the USDA established
regulatory lethality target (FSIS-USDA 1999), and the experimental lethality target
for turkey, beef, and pork. ......................................................................................... 39

Table 5. The average Salmonella population (CF U/g) recovered after duplicate plating on

aerobic PetriﬁlmTM plates for turkey, beef and pork roasts cooked to 71 .1°C. Limit
of detection 0.4 log CFU/g ........................................................................................ 40

Table 6. Replication error for measured process lethality at three experimental scales log
CFU/ g (Tenorio-Bernal and others 200X, Jones and others 200X). ......................... 45

Table 7. Replication error (log CFU/ g) for 21 randomly selected samples from the 1 g
laboratory experiments, three replication for each species. ...................................... 45

Table 8. Prediction error and mean residual of the state-dependent and path-dependent
models for whole muscle turkey, beef, and pork. ..................................................... 49

Table 9. Documented times (8) that roasts cooked to a targeted lethality were held at a
speciﬁc temperature, for turkey and beef .................................................................. 56

Table 10. Pros and Cons of alternative whole muscle pathogen inoculation methods ..... 65

Table 11. D and 2 values calculated from beef isothermal inactivation tests of 8-serovar
Salmonella cocktail. .................................................................................................. 67

Table 12. Salmonella survivors (CFU/g) in ground beef during 55°C isothermal
inactivation. ............................................................................................................... 68

Table 13. Salmonella survivors (CF U/ g) in ground beef during 58°C isothermal
inactivation. ............................................................................................................... 69

vi

Table 14. Salmonella survivors (CFU/ g) in ground beef during 60°C isothermal
inactivation. ............................................................................................................... 70

Table 15. Salmonella survivors (CFU/ g) in ground beef during 62°C isothermal
Inactivation ......................................... 71

Table 16. Salmonella survivors (CF U/ g) in ground beef during 63°C isothermal
inactivation. ............................................................................................................... 72

Table 17. Salmonella survivors (CF U/ g) in ground beef during 55°C isothermal
inactivation. ............................................................................................................... 73

Table 18. Salmonella survivors (CF U/ g) in ground beef during 58°C isothermal
inactivation. ............................................................................................................... 74

Table 19. Salmonella survivors (CFU/ g) in ground beef during 60°C isothermal
inactivation. ............................................................................................................... 75

Table 20. Salmonella survivors (CFU/ g) in ground beef during 62°C isothermal
inactivation ................................................................................................................ 76

Table 21. Salmonella survivors (CFU/ g) in ground beef during 63°C isothermal

inactivation. ............................................................................................................... 77
Table 22. Summary of inoculation veriﬁcation data for turkey roasts. ............................ 78
Table 23. Summary of inoculation veriﬁcation data for beef roasts ................................. 78
Table 24. Summary of inoculation veriﬁcation data for pork roasts. ............................... 78

Table 25. Validation of the post-cook sampling method, coring for whole muscle roasts.
Purge from the cook-in bag was collected immediately after cooking Salmonella
inoculated whole muscle roasts in a pilot—scale moist air convection oven and again
after coring the whole muscle roasts with a sterile stainless steel corer. .................. 80

Table 26. Post-cook, comparison of the lethality error (observed - predicted) for the
biopsy and electrosurgical unit methods following pork inoculation. The standard
deviations of the lethality error post cook were compared to determine the most

repeatable process. .................................................................................................... 83
Table 27. Turkey roast spread sheet summary of all cook schedules. .............................. 87
Table 28. Beef roast spread sheet summary of all cook schedules. .................................. 90
Table 29. Pork roast spread sheet summary of all cook schedules. .................................. 94

vii

LIST OF FIGURES
Figure 1. Whole muscle roast cutting diagram. Segment A was considered the center cut.
of the roast ................................................................................................................. 26
Figure 2. Thermocouple wire threaded through a sterile stainless steel corer. ................. 28

Figure 3. Predetermined cook schedules utilized to process whole muscle roasts in a

modiﬁed Cres Cor® moist air convection Oven. ...................................................... 31‘
Figure 4. Sample cooking proﬁle for constant temperature cook schedule. ..................... 33
Figure 5. Sample cooking proﬁle for step-up temperature cook schedule. ...................... 33

Figure 6. Salmonella lethality error (observed - predicted) vs. sub-lethal history for turkey
breast. Lethality calculations based on laboratory derived isothermal D and 2 values
(Tuntivanich and others 2008). ................................................................................. 46

Figure 7. Salmonella lethality error (observed - predicted) vs. sub-lethal history for beef
roasts. Lethality calculationsbased on laboratory derived isothermal D and 2 values,
Appendix 7 .A ............................................................................................................ 47

Figure 8. Salmonella lethality error (observed - predicted) vs. sub-lethal history for pork
roasts. Lethality calculations based on laboratory derived isothermal D and 2 values
(Velasquez and others 2009). .................................................................................... 48

Figure 9. Lethality errors for the state-dependent and path-dependent model applied to
the whole muscle turkey data set. ............................................................................. 51

Figure 10. Lethality errors for the state-dependent and path-dependent model applied to
the whole muscle beef data set .................................................................................. 52

Figure 11. Lethality errors for the state-dependent and path-dependent model applied to
the whole muscle pork data set. ................................................................................ 52

Figure 12. Salmonella survivors (CFU/ g) in ground beef during 55°C isothermal
inactivation. ............................................................................................................... 68

Figure 13. Salmonella survivors (CFU/g) in ground beef during 58°C isothermal
inactivation. ............................................................................................................... 69

Figure 14. Salmonella survivors (CFU/ g) in ground beef during 60°C isothermal
inactivation. ............................................................................................................... 70

viii

Figure 15. Salmonella survivors (CPU/g) in ground beef during 62°C isothermal
inactivation. ............................................................................................................... 71

Figure 16. Salmonella survivors (CF U/g) in ground beef during 63 °C isothermal
inactivation. ............................................................................................................... 72

Figure 17. Salmonella survivors (CFU/ g) in ground beef during 55°C isothermal
inactivation. ............................................................................................................... 73

Figure 18. Salmonella survivors (CF U/g) in ground beef during 58°C isothermal
inactivation. ............................................................................................................... 74

Figure 19. Salmonella survivors (CFU/ g) in ground beef during 60°C isothermal
inactivation. ............................................................................................................... 75

Figure 20. Salmonella survivors (CFU/ g) in ground beef during 62°C isothermal
inactivation ................................................................................................................ 76

Figure 21. Salmonella survivors (CFU/ g) in ground beef during 63 °C isothermal
inactivation. ............................................................................................................... 77

Figure 22. Whole muscle turkey roast weight (g) vs. core Salmonella concentration (log
CFU/g). ..................................................................................................................... 79

ix

1. INTRODUCTION

With today’s fast paced society, Americans demand convenience. The ever evolving
on-the-go lifestyle is a delicate balance of long work days and full social schedules. In
order to make these busy schedules work, Americans must optimize their time in many
ways. One way Americans create more time in their days is by consuming a
“prepackaged” diet. With the lack of time to prepare meals, consumers rely on processors
to execute some or all food preparation steps in order save them time. These ready-to-eat
(RTE) products have many obvious advantages and many drawbacks. The intricate
supply chain that exists in the food industry adds to the food safety concern. Once food
leaves the processing facility, the processor loses control of the product, yet ultimately is
held accountable for its safety through consumption. Possibilities of risk factors are
broad, spanning from inadequate processing, temperature abuse in distribution, and cross-
contamination or non-compliance with cooking recommendations by consumers. The
average consumer has minimal knowledge about food safety principles. Therefore, when
processors rely on the consumer to perform the ﬁnal kill step at home, there is an inherent
risk of inadequate pathogen inactivation (Moss 2009).

Salmonella is a bacterial pathogen that originates in the intestinal tracts of poultry and
other livestock, and as a result is introduced to the environment through fecal
contamination Salmonella poses a health risk to humans when contaminated food or
water is ingested. The pathogen causes the disease salmonellosis; symptoms include
diarrhea, fever, and abdominal pain. Populations most at risk for experiencing potentially
life threatening salmonellosis infections are the elderly, newborns, and

immunosuppressed. Based on the FoodN ET surveillance data from the CDC (2005),

42% of all foodbome bacterial infections (conﬁrmed in the laboratory) were due to
Salmonella. Salmonella was second to norovirus, as the most common source of
outbreaks in 2005 (CDC 2008a).

Due to the danger of Salmonella, the United States Department of Agriculture
(USDA) and Food and Drug Administration (FDA) have introduced processing standards
for products that are most vulnerable to contamination. Meat products have a particularly
high risk of fecal contamination, because of the proximity of muscle tissue to the

intestinal tract during processing. Lethality requirements have been developed for RTE

meat products. The USDA Food Safety Inspection Service requires a 7- or 6.5- loglo

reduction of Salmonella in fully cooked poultry and beef products, respectively (F SIS-
USDA 1999). Setting standards based on log reduction of pathogens, as opposed to
minimum cook time or temperature, ensures that meat products receive adequate heating
to inactivate Salmonella, while allowing processors the ﬂexibility to determine their own
speciﬁc time and temperature parameters for a variety of products.

The ﬂexibility of processing to a calculated lethality is possible because the USDA
and FDA have established lethality standards using mathematical models that predict the
destruction of pathogens based on time-temperature calculations. Many of the models
utilized by the industry today are ﬁrst-order kinetic models, which presume log-linear
inactivation of bacteria under isothermal conditions. However, these model and lethalin
requirements typically have been developed using controlled, laboratory-scale studies,
and not validated in commercial processing environments.

However, the inactivation of bacteria in full scale processing environments is much

more complicated than a simple laboratory experiment; the processing environment,

conditions, and sample size, composition, and structure all impact the process outcomes.
Slow cooking processes, in particular, challenge the modeling approach, because
traditional models fail to account for sub-lethal injury of bacteria, which can occur
approximately between 38 and 52°C (see section 2.2.2.1). Sub-lethal injury occurs when
a bacterial cell is exposed to a temperature range that injures but fails to kill it, thereby
allowing the cell to adapt and become more heat resistant. Subsequently, these injured
cells require a higher temperature and/or a longer cook time to ensure complete
inactivation. The application of current models to whole muscle foods that are cooked
slowly to a calculated log reduction does not account for sub-lethal heating; therefore, the
total kill might be overestimated. Stasiewicz and others (2008) demonstrated the error
that occurs when applying a state-dependent inactivation model during slow cooking of 1
g samples of ground turkey thigh, subjected to varying sub-lethal heat treatments.
However, such a modeling approach has not been tested in or applied to pilot or
commercial — type products or processes.

The implications of overestimating the efﬁcacy of cooking processes is potentially
life threatening; therefore, the overall goal of this research was to evaluate the effect of
slow cooking on the thermal resistance of Salmonella in whole muscle meat and poultry
products. The hypothesis was that slow cooking of whole muscle meat products increases
the thennal resistance of Salmonella. Evaluation of the overall goal and hypothesis
included two main objectives: (1) Determine the D and 2 values for Salmonella in whole
and ground beef via laboratory isothermal lethality tests, and (2) Evaluate the accuracy of
traditional and alternative lethality models via pilot-scale, inoculated challenge studies

with whole muscle products in a moist-air convection oven.

2. LITERATURE REVIEW

2. 1. F oodborne Pathogens

Foodbome pathogens contaminate food products, and cause those whom consume
a contaminated food product to become ill. It is “estimated that foodbome disease causes
approximately 76 million illnesses, 325,000 hospitalizations, and 5,000 deaths in the
United States each year” (Mead and others 1999). Foodbome bacterial pathogens
responsible for the highest number of deaths each year include Salmonella, Listeria, and
Toxoplasma (Mead and others 1999). The FDA (Food and Drug Administration) and
USDA (United States Department of Agriculture) have developed regulations, and
standards in the processing of food products to help reduce illness caused by foodbome
pathogens. Food related illness results in billions of dollars in lost product, medical bills,

and law suits (Bubzy and Frenzen 2001).

2.1.1 Salmonella

Salmonella was the foodbome bacterial pathogen with the highest incidence of
infection in 2005 (CDC 2008a). Salmonella causes salmonellosis, which can result in
diarrhea, fever, and abdominal cramping for most healthy individuals (CDC 2008b).
Health complications can occur, such as reactive arthritis, Reiter’s syndrome, and
ankylosing spondylitis for susceptible populations, such as young children, the elderly,
and immunocompormised individuals (Doyle 2001). I

Salmonella spp. are members of the Enterobacteriaceae family, and are
characterized as facultative anaerobic, Gram negative rods. Salmonella is motile with a

peritrichous ﬂagella, and optimal growth occurs at 37°C (Doyle 2001). Although several

thousand Salmonella serotypes exist, Salmonella Typhimurium and Salmonella
Enteritidis are the two prevalent serotypes in the US. (FDA 2009). Salmonella Enteritidis
is predominantly associated with poultry, its primary reservoir. Salmonella Typhimurium
is a more widespread serotype, and is of great importance because of its multiple
antibiotic resistance (FDA 2009).

Salmonella outbreaks are widespread, having occurred in numerous food
products, including meat, poultry, peanut butter, vegetables, and nuts (FDA 2001, FDA
2009a, FDA 2009b). Now that the government has implemented new surveillance
techniques, such as FoodNET, PulseNET, and NARMS, detection of foodbome illness is
becoming more efﬁcient (Richard and Arkin 2007). However, there still remains a need
for more reliable control methods that will reduce the transmission of Salmonella (J uneja

and Eblen 2000).

2.1.2 Pathogens in Meat

The predominant meat safety concerns of the let century will likely revolve

around microbial pathogens. This is due to alterations in the production and distribution
of our intricate food system (Sofos 2008). Consumers want convenient, high quality,
inexpensive foods, and these demands will come with a price (Sloan 2009). The dangers
associated with pathogens in meat products exist because fresh muscle tissue has a high
nutrient value, and a high water activity of about 0.99 and a near neutral pH, all of which
provide an environment ideal for bacterial proliferation (Doyle 2001). Temperature abuse
of the product will also encourage pathogen growth. Pathogens readily exist within the
environment of a slaughterhouse; therefore, monitoring and control of theses pathogens is

crucial. HACCP (Hazard Analysis and Critical Control Points) is a program originally

developed by NASA, and now implemented in the food industry by the FDA and USDA,
to assist in reducing product contamination during processing (van der Fels-Klerx and
others 2008). The Food Safety and Inspection Service (F SIS), a division of the USDA,
has labeled Salmonella as a target organism in major meat species (Duffy and others

2001).

2. I . 2.1 Salmonella

Salmonella is a pathogen that occurs naturally within the intestinal tract of
animals and is often present on animal hide due to fecal contamination. Salmonella is
transmitted via the fecal-oral route (CDC 2008b). Raw meat products are a conﬁrmed
vehicle of Salmonella contamination and transmission ( Stopforth and others 2006;
Kegode and others 2008; Sofos 2008). Salmonella can attach to the surface of meat tissue
from cross contamination, and is capable of migrating into the interior of a whole muscle
product (Warsow and others 2008). Currently the limit of detection for microbial
contamination is about one cell per 25 g of meat, therefore making it possible for a
pathogen or other microorganism to go undetected, and multiply during periods of
temperature abuse (Shimoni and Labuza 2000). Salmonella is capable of growing
throughout a wide temperature range (4 to 54°C) and pH range (4.5 to 9.5), because of its
ability to adapt to environmental conditions (Doyle 2001).

The possibility of meat product contamination exists throughout processing, with
many opportunities for cross contamination. This danger, linked with the potential for
insufﬁcient processing, is an important concern for manufacturing ready-to-eat meat

products. Salmonella is an adaptive pathogen, capable of surviving in many

environmental conditions. While Salmonella does not proliferate at freezing
temperatures, it is not destroyed when subjected to freezing temperatures on beef
trimmings (Dykes and Moorhead 2001). Dykes and Moorhead (2001) attributed the lack
of reduction in Salmonella during freezing to the protective effect of the fat layer.
Salmonella is generally salt susceptible, but increased salt tolerance has been observed at
10 to 30°C (Doyle 2001 ). At low water activity, Salmonella exhibits increased thermal
resistance ( Pearson and Dutson 1986; Carlson and others 2005). The ability of
Salmonella to adapt to meat processing stressors, such as heat, cooled, acid, and salt

causes it to be a major safety issue in regards to foodbome illness.

2.1.2.2 Mechanisms and modes of transfer

Mishandling of meat products is almost inevitable, given the complexity of the
food distribution system. Cross-contamination of meat products can occur within the
processing facility via equipment or handlers. For this reason, it is crucial to be sure that
all RTE meat is properly cooked before distribution, therefore inhibiting pathogen growth
during temperature abuse.

Currently, marination is used by processors as a method to increase meat
tenderness, enhance water content, and amplify ﬂavor and color. Most processors utilize
marinades consisting of a combination of salt and phosphate. Salt functions as a ﬂavor
enhancer, while phosphate enhances marinade uptake and retention (Xiong and Kupski
1999).

The ability for bacteria to penetrate into the interior of whole muscle products

that have been surface contaminated or marinated has been demonstrated, and poses a

safety issue during processing (Gill and Penney 1982; Gupta and others 1983). Vacuum
tumbling aids in the penetration of marinade into intact muscle. If the marinade is
contaminated with Salmonella, vacuum tumbling increases the migration of Salmonella
into the interior of whole muscle turkey by forcing bacteria toward the center of the
product (Warsow and others 2008). Vacuum tumbling facilitates the uptake of liquid
marinade into whole muscle, but Salmonella can penetrate into whole muscle turkey
without external intervention (Breslin and others 2007). The migration of Salmonella
without intervention or a liquid marinade was demonstrated by inoculating whole muscle
turkey using only the pellet from an 8 - serovar Salmonella cocktail following

centrifugation (Breslin and others 2007).

2. 2. Thermal Resistance

2.2.1 Product Factors Affecting Thermal Resistance

Regulatory standards for the thermal processing of muscle foods, based on a
calculated lethality, have been established to allow processors ﬂexibility in processing
conditions (FSIS-USDA 1999). Measured lethality as a determination of product safety
can prevent over processing, therefore increasing product yield, proﬁts, and consumer
acceptance. Contrary to the positive implications of this standard, safety concerns are
associated with heating products to a lethality requirement, when determined by a model
that fails to account for the multiple factors affecting thermal resistance of muscle foods.

Product composition and structure have been shown to affect the thermal
resistance of Salmonella ( Murphy and others 2000; Smith and others 2001; Orta-
Ramirez and others 2005). Juneja (2001b) attributed the different D-values in ground

beef, pork, turkey, and chicken to differences in product composition that affect bacterial

inactivation. In addition to the product variability, bacterial strains can vary widely in
their heat resistance (Juneja and Eblen 2000). Thermal resistance of foodbome pathogens
is typically reported as D and 2 values, (Smith and others 2001). D and 2 values are used
in the log-linear model to calculate the lethality of pathogens within food products.
Microorganisms and food matrixes similar to the product of interest must be used in order
to accurately evaluate the safety of a ﬁnal product (Murphy and others 1999; Smith and
others 2001).

Fat is a major factor in the disparity of thermal resistance (Juneja and others
2001). Fat globules within the food matrix act as protective barriers for Salmonella cells,
thereby increasing thermal resistance (Juneja and Eblen 2000).

Product structure, ground versus whole muscle, also has an impact on the thermal
inactivation of Salmonella (Orta-Ramirez and others 2005; Tuntivanich and others 2008).
Orta-Ramirez and others (2005) and Tuntivanich and others (2008) found that Salmonella
was more resistant in whole muscle beef and turkey than in ground products of equivalent
species, composition, and history. However, the fundamental mechanisms for those
differences are not yet known.

The moisture composition of a food product has also been shown to have a major
impact on Salmonella survival. If Salmonella exists in a food product with a low moisture
content, it has an increased resistance to heat, as compared to the same product with a
higher moisture content (Murphy and others 2001b; Carlson and others 2005; Naphapom
and others 2007). Water activity also profoundly affects the effectiveness of processing
conditions on bacterial lethality. It needs to be evaluated when validating a process,

because there is a high risk for Salmonella survival following the heating step.

Many researchers have found that pH of the product affects Salmonella survival.
Despite these ﬁndings, pH is not a factor included in lethality calculations (Leguerinel
and others 2007). Failing to account for the pH of the raw and processed product is likely
to contribute to inaccurate lethality estimations. Currently pH is not incorporated into
lethality models; however, this product factor is minimal when compared to other product
factors (Leguerinel and others 2007).

Product composition and structure have .a large inﬂuence on the thermal
resistance of bacteria; therefore, in order to develop accurate models for evaluating the

safety of RTE meat products, all of these intrinsic factors need to be considered.

2.2.2 Process Factors

In addition to product composition, processing factors also affect thermal
inactivation of Salmonella. Dry cooking conditions increase the heat resistance of
Salmonella on the surface during processing (Blankenship 1978; Goodfellow and Brown
1978; Murphy and others 2001b). However, in practice, even moderate added humidity
(e. g. ~30% rh) can enable sufficient lethality during cooking to effectively eliminate
Salmonella (Mann and Brashears 2007). Humidity during processing has been found to
signiﬁcantly affect the survival of Salmonella in meat (Murphy and others 2001b).
However, Carlson (2002) concluded that the water activity of the meat product has a
larger effect on the Salmonella thermal resistance than does the process humidity. Since
process humidity signiﬁcantly affects water activity, process humidity must be
considered, not only for yield and heat transfer purposes, but more importantly for its

impact on product safety.

10

2. 2. 2.1 Sub-lethal Injury, Non thermal

Processing intervention steps also are capable of causing injury to pathogens.
These can include a broad range of factors including temperature, drying, irradiation,
pressure, acid, sanitizers, preservatives, and antimicrobials (Wu 2008). Cells that have
experienced stress injury, and repair under favorable conditions, are deﬁned as sub-
lethally injured (Wu 2008).

There are three broad categories of microbial reduction steps: thermal, chemical,
and physical. Causative agents of non-thermal sub-lethal injury include acid washes,
hydrostatic pressure, sanitizers, preservatives, and antimicrobials (Wu 2008). Bacteria are
capable of adapting to their environment, and sub-lethally injured cells adapt to the
causative agent that injured them, as well as cross protection to different stressors based
on stress proteins. For example, E. coli has exhibited the capability to adapt to sanitizer
stress on produce, thereby potentially decreasing the effectiveness of the cleaning process
of fresh produce (look and others 2001). Sub-lethal injury of bacterial cells creates a
safety risk in the food industry, but hurdle effects, When applicable, aid in eliminating the

danger of sub-lethal injury.

2. 2. 2.2 Sub-lethal Injury, Thermal

Thermal sub-lethal injury has been investigated thoroughly because of the
popularity of value-added food products and outbreaks that have occurred after a
perceived lethality treatment. Thermal inactivation of Salmonella is signiﬁcantly affected
by prior thermal treatments (Wesche and others 2005). The range of values reported as

the heat shock region are between 42 and 48 °C (Mackey and Derrick 1986; Xavier and

11

Ingham 1997), indicating that bacterial cells held at a temperature between 42 and 48 °C
will not be destroyed, but rather injured, and capable of repairing. Following thermal
sub-lethal injury, bacterial cells will require a longer exposure, or higher temperature for
inactivation (Wesche and others 2005). Only mild temperature abuse is necessary to
cause sub-lethal injury (Xavier and Ingham 1997), and this is quite possible, given the
potential for improper temperature control during distribution (O'Bryan and others 2006).
Heat injured cells are likely to display a lag period that has the potential to be days long
before growth on nutrient rich media occurs (Juneja 2007). For this reason, knowledge of
pathogen growth is crucial for analyzing the safety of RTE products following processing
(Murphy and others 2001b); otherwise, a product may be deemed safe when in fact it is
contains injured cells, which could result in ill consumers.

Storage conditions, temperature, and length of storage following processing
greatly impact the growth of heat injured Salmonella (Murphy and others 2001b).
Processing conditions also impact the survival of bacterial pathogens. Packaging
conditions have been shown to affect Listeria monocytogenes thermal resistance in raw
meat, with L. monocytogenes to be more heat resistant in vacuum packaged as compared
to air-packaged products (Kim and others 1994).

Though many conclusions have been made about the thermal resistance of
bacterial cells in food products, a gap exists because the majority of studies have been
conducted in controlled laboratory and small-scale environments. Small samples are
much easier to control, because they eliminate the variability associated with larger
sample sizes, such as non-uniform heating and sampling variability. However, in order

for the industry to apply laboratory data to its processes, they must be validated in larger

12

commercial studies. The majority of thermal inactivation research has been conducted
using isothermal tests to determine D values for various microorganisms in liquid media,
or samples weighting less than 10 g, which are cooked for various times. These data then
are applied directly by industry to their models, which can pose problems due to the
length of time it takes to process a larger product. Hence, there is a necessity to develop
lethality information that is directly applicable to industry processing conditions.
Currently, no process validation method exists that quantiﬁes the sub-lethal injury of
bacterial cells during heat treatment. A processing model that accounts for sub-lethal
injury of bacterial cells would assist the industry in ensuring the safety of ready-to-eat

products.

2.2.3 Inactivation Models

In predictive microbiology, mathematical models are used to estimate changes in
microbial populations during storage or processing (Doyle 2001). However, due to the
complexity of model applications, and the difficultly in obtaining precise data, no gold

standard model exists for all applications (Doyle 2001).

2. 2. 3. 1 Primary Models

A primary model describes the correlation of microbial response as a function of
time variation and cooking conditions (Metris and others 2008). Due to the various
factors affecting thermal resistance, use of predetermined thermal inactivation data from
other unrelated studies has been proven to be unreliable (Murphy and others 2000).
Computer-based microbial pathogen growth/inactivation models can not be solely

depended on to determine food safety (FSIS-USDA 2005). More sophisticated models

13

that account for a range of product and processing parameters need to be developed. This
can be done by creating models that account for speciﬁc processing variables that most
closely relate to the product in question. There is a need for inactivation models to be
validated using actual processing conditions. Commercial processors desire models that
incorporate the factors affecting thermal resistance, with a user friendly interface (Peleg
and others 2005), in order to simply validate the safety of their ﬁnal products.

Log-linear, ﬁrst-order kinetic models are primarily utilized during or prior to most
thermal treatments (Mattick and others 2001). Such models do not necessarily account
for the vast difference between fast, high temperature cooking, and slow cooking at low
temperatures.

The log-linear model does not account for exposure extent (Peleg and others
2005). The log-linear model utilizes D-values, deﬁned as the time required at a given
temperature to kill the initial population of bacteria by 90%. D-values are typically
obtained from laboratory studies using broth cultures or small model food systems that do
not represent true processing conditions. Using these data can result in inaccuracy in the
lethality calculations during processing. Shoulders or lag periods can exist during
processing (Peleg 2000). Hence, estimations made by a log-linear model using only data
that ﬁt the most linear trend can result in over or under processing (Peleg 2000; Peleg and
Penchina 2000; Juneja and Eblen 2000; Juneja and others 2001). Despite the potential
inaccuracy of the log-linear model, it is still highly utilized in the food industry via the
USDA Pathogen Modeling Program (PMP) (USDA 2006) and other modeling tools, such
as the American Meat Institute (AMI) lethality spreadsheet (AMI Foundation 2009).

Modiﬁcations to the log-linear model that would account for changing heat resistance of

14

vegetative cells during thermal processing would provide more reliable lethality data
(Mackey and Derrick 1986). Validation of modiﬁcations and new models should be
conducted using industry relevant conditions, variability, and sample size.

An alternative primary model, the Weibull model, is described by the power law

model LogS(t) = -b(T)tn(T) (Corradini and Peleg 2004). For situations where n(T) > 1 the

curve has an upward concavity, which indicates that the microbial kill increases with
time; for n(T) < 1, the survival curve is concave downward, indicating that resistant
survivors remain persistent following death of weaker cells; if n(T) = 1, the survival
curve is that of ﬁrst-order log-linear kinetics (Corradini and Peleg 2004). The Weibull
model is more ﬂexible than the traditional log-linear model; however, it alone does not

account for the effects of prior sub-lethal history (Stasiewicz and others 2008).

2. 2. 3.2 Secondary Models

Secondary models describe how primary model parameters depend on changing
environmental factors (Doyle 2001; Metris and others 2008). Examples include the
Arrhenius model, the square-root (Béleradek) model (primarily utilized for bacterial

growth), and the Bigelow / z-value model, which relates the D-value to the temperature

via the z-value, such that D(T) = Dref ware“ T)”.

Additional secondary models include polynomial models, which rely on the best
ﬁt assumption when comparing the independent and modeled variables (Doyle 2001).
When recently applied to microbial thermal inactivation data, a 2-term fractional
differential equation (F DE) model could not account for the shoulders and tails in the

inactivation data (Kaur and others 2008). However, the model did account for concavity,

15

and remained accurate during extended processing (Kaur and others 2008). This
alternative model seems appealing, but further research is needed, because these ﬁndings
are based only on the application of the model to lab-scale pathogen inactivation data.

There is a need to evaluate the 2-term F DB in large-scale, diverse processing conditions.

2.2.4 Modeling Effects of Sub-lethal History

Several researchers, including Corradini and Peleg (2009) and Valdramidis and
others (2007), proposed lethality models that incorporated the effect of sub-lethal history

on an organism during thermal processes. However, both of these studies modeled the

Inactivation parameters as functions of temperature and heatmg rate (2’7 ), even through

the physiological response of heat shock has not been shown to be related to gr? Also,

neither of these models was validated using an actual food product. Valdramidis and
others (2007) heated tubes containing suspended bacterial cells in a circulating waters
bath, whereas Corradini and Peleg (2009) estimated their model parameters using
previous thermal inactivation data generated from other laboratories.

In contrast, the thermal inactivation model proposed by Stasiewicz (2008)
accounts for sub-lethal injury as an integral of the sub-lethal history during thermal
processing. This model is referred to as a path-dependent model, and quantiﬁes the sub-
lethal history as the integral of the temperature proﬁle within the deﬁned heat shock
region (Stasiewicz and others 2008). The model was applied to slow cooking conditions,
and the error between the actual and predicted log reductions was evaluated. The error for

a traditional Weibull / Arrhenius model was large during slow cooking, because it failed

16

to account for sub-lethal injury (Stasiewicz and others 2008). However, Stasiewicz’s
path-dependent, Weibull / modiﬁed-Arrhenius model accounts for sub-lethal history, and
eliminated the systematic error that occurred using the state-dependent model.
Additional research validating the path-dependant model needs to be conducted. The
rationale for this model was evaluated using one product (turkey thigh meat). Another
drawback of the research was that it was in a controlled laboratory environment, with
very small samples sizes (1 g), which does not directly correlate to industrial processing

situations.

2. 3. Commercial Meat Cooking

2.3.1 Validation of Process

The United States Department of Agriculture (USDA) has implemented safety
regulations for the processing of meat and poultry products, which concentrate on an
endpoint internal temperature. Hence, a meat roast cooked to a speciﬁed core
temperature, would be deemed fully cooked and safe. The USDA speciﬁes an internal
temperature of 71.1°C for the product to be labeled fully cooked (F SIS-USDA 1999).
These standards have been developed for customers at home and as a reference
temperature in cooking trials (Pittia and others 2008).

Endpoint lethality requirements are enforced by the USDA to assist processors in
determining the safety of their product, without over-processing (FSIS-USDA 1999). End
point temperature requirements can lead to over processing, which causes an increase in
energy cost, and a decrease in product yield, both of which directly impact processors

proﬁts. Proﬁts drive the industry; therefore, lethality requirements that maximize product

17

yield and minimize energy expenditure are desirable as long as safety of the product can

be ensured.

2.3.2 Validation of Oven Conditions

Larger meat roasts are often cooked at lower temperatures for extended periods in
order to achieve uniform heating, a more uniform color throughout, and a more tender
and juicy end product. Several common cooking “schedules” can be used, depending on
the quality goals.

Step-up oven conditions consist of a gradual increase in the oven temperature
throughout the cooking process. Step-up cooking increases the surface and internal
temperature at a similar rate, because a smaller temperature gradient exists, which helps
to eliminate crust formation. This processing condition can provide a ﬁnal product that
has a rare center appearance, and is primarily utilized in roast beef processing.

Constant oven temperature processing is utilized to consistently increase the
internal temperature over a short period. During this type of cooking, the oven remains at
a steady temperature throughout cooking (e. g., ~93.3°C), which is higher than the desired
product end temperature (~71 . 1°C), to encourage pathogen destruction (Murphy and
others 2001a). The larger initial temperature gradient causes the internal temperature to
increase more quickly during cooking. The result in turn yields a more time-efﬁcient
cook, as compared to the step-up process. Beef roasts cooked at a constant temperature
will have a less pink center as compared with the step-up process. This is more appealing
to some consumers, because it results in a product that “looks” cooked, but is less

appealing to consumers who prefer rare or medium beef.

18

2.3.3 Humidity

Oven conditions inﬂuence the sensory characteristics of a processed product
(James and Calkins 2008); therefore, depending on the desired end product, commercial
processors may choose to control humidity in different manners. Introducing moisture
into a hot air convection oven increases the heat transfer rate, resulting in a decreased
cook time (Murphy and others 2001a). The increase in heat transfer rate is partly
responsible for the more efﬁcient killing of microorganisms when compared with dry
heat (N aphapom and others 2007). Also, Salmonella thermal resistance is affected by
product moisture within different cooking environments (Carlson 2002). In addition to
decreasing the processing time, and increasing lethality, increased humidity also
minimizes water loss during cooking, which increases product yield, and thereby proﬁts

(V ittadini and others 2005).

2. 3. 3. I F SIS Regulations of Humidity Use

Humidity requirements “apply only to those processes in which the surface
moisture of the product can evaporate, and surface drying can occur, prior to destruction
of the microorganisms”; exceptions include processes where humidity is inherently
sustained around the product (FSIS-USDA 2005). Humidity requirements were
developed to ensure complete inactivation of foodbome pathogens in RTE meat. Surface
bacterial pathogens are more thermally resistant when the moisture level is low

(Blankenship 1978; FSIS-USDA 1999b).

l9

2. 3. 3.2 Cook-in-bag

In-bag cooking is used to trap humidity in the bagged product during cooking, and
thereby enhance product yield, making this process more economical as compared to no-
bag even cooking. According to the FSIS humidity regulations (2005), “cooking the
product in a sealed, moisture impermeable bag” avoids surface drying. Speciﬁcally
engineered bags that withstand the cooking environment and are impermeable to
moisture are used. Advances in cook-in—bag technology have provided processors with
numerous bag options, depending on the desired end product. If a cooked roast with
golden color is desired (e. g., turkey), then the bag remains closed throughout cooking,
then just before ejection from the oven the bag splits open, to allow for top-surface color
development. Other advantages of using cook—in-bags during processing include their
ability to act as a barrier to the environment, preventing cross-contamination, decreasing
purge loss, increasing yields, and retaining protein, mineral, and pigments components
(Qiaofen and Da-Wen 2007).

Commercial processors choose cooking schedules that fulﬁll the end goal of
yield, color, taste, and texture. Now, consumers and the government are putting the
responsibility of providing safe ready to eat products on the processors; therefore, the

industry needs to take appropriate steps to ensure the safety of their products.

20

3. MATERIALS AND METHODS

3. 1 Isothermal Research

Isothermal laboratory studies were conducted with ground and whole beef and
pork. The meat was acquired as stated below in section 3.2, and the Salmonella inoculum
was the same as in section 3.3. Methods were the same as those performed by (Orta-
Ramirez and others 2005), except experiments were performed at ﬁve different

temperatures: 55, 58, 60, 62, and 63°C.

3.2 Acquisition of Meat

Fresh, skin-off turkey breast (Pectoralis major and minor muscles, acquired ﬁ'om
a federally inspected commercial processor), top beef round (Semitendinosus, Adductor,
and Pectineus muscles, commercial supplier from a federally inspected source), and
center cut pork loin (Longisimus Dorsi muscle, commercial supplier from a federally
inspected source) were used. All products were acquired directly from the
processor/supplier, and shipped to Michigan State University’s meat processing facility.
Product temperature at time of receipt was veriﬁed to be < 4.4°C and documented. Meat
was sectioned into approximately 0.68 kg roasts, vacuum packaged in double plastic
bags, and frozen (-20°C).

Frozen roasts were then packaged in 57 L plastic storage boxes and shipped eight
days after arrival to Tampa, FL for irradiation (F TSI, FDA registration 1054811).
Products were shipped in a refrigerated truck at -20°C. Gamma rays were used to

irradiate boxes (~10 kGy) to eliminate indigenous microﬂora, then transported frozen

21

back to Michigan State University’s meat laboratory. Frozen irradiated samples
remained frozen (-20°C) until use.

Sterility of the samples was conﬁrmed by randomly testing 3 roasts of each
species. A 25 g core was removed aseptically from the roast using a sterile scalpel blade
(Rib-Back Carbon Steel Surgical Blades Bard-Parker, Becton Dickinson AcuteCare,

. Franklin Lakes NJ), placed in 8 oz Whirl-Pak® bags, diluted 1:10 in sterile triptic soy
broth (Difco Laboratories, Sparks, MD) containing yeast extract (Difco Laboratories,
Sparks, MD) (TSBYE), homogenized in a masticator (N eu-Tec Group Inc, Barcelona,

Spain) for 180 s, and incubated for 24 h at 37°C. Following incubation, duplicate samples

' were plated on aerobic PertriﬁlmTM (3M Microbiology Products, St. Paul, MN).

The moisture and fat composition of all species was determined using Association

of Ofﬁcial Analytical Chemists (1990) methods 950.46B and 991.36, respectively.

3.3 Preparation of Salmonella

Triptic soy broth with yeast extract (TSBYE) was prepared by combining 30 g
triptic soy broth, 6 g Yeast extract, and 1 L of deionized distilled water. The solution was
dispensed into 250 (turkey), and 500 ml (beef, pork) capped, autoclavable bottles. The
solution was autoclaved (Castle WC 3522 Sterilizer, Getinge, Rochester, NY) for 20 min
at 121°C.

The inoculum consisted of eight Salmonella serovars: S. Thompson FSIS 120
(chicken isolate), S. Enteritidis H3527 and H3502 (clinical isolates phage types 13A and
4 respectively), S. Typhimurium DT 104 H3380 (human isolate), S. Hadar MF60404

(turkey isolate), S. Copenhagen 8457 (pork isolate), S. Montevideo FSIS 051 (beef

22

isolate), and S. Heidelberg F 503 8BGI (human isolate), previously obtained from V.K.
Juneja (Agricultural Research Service, Eastern Regional Research Center, USDA-ARS,
Wyndmoor, PA). Each serovar was separately maintained at -80°C in vials containing
tryptic soy broth and 10% glycerol. Cultures were started by transferring one loop of
frozen culture into 9 mL of TSBYE, incubating at 37°C for 18-24 h. Cultures were
maintained using consecutive daily transfers for up to one week, with a minimum of two
consecutive transfers prior to use. One day prior to use, the strains were transferred

separately into 250 mL (turkey) or 500 mL (beef, pork) bottles of TSBYE and incubated

at 37°C for 24 h.

3.4 Preparation of Marinade

A typical commercial marinade containing 11.5% (w/v) salt and 3.7% (w/v)
phosphate was prepared by adding 169 g of liquid phosphate (50% food grade liquid
potassium phosphates, Butcher and Packer Supply Company, Detroit, M1) to 1,778 g of
deionized distilled water, then stirring on a stir plate (Corning PC-420, Corning, NY)
until the phosphate dissolved. Once the phosphate and water were completely combined,
253 g of NaCl was added to the solution, while stirring continued. Once the NaCl was
dissolved into the mixture, aliquots of 520 mL and a small stir bar were added into 500
mL twist capped autoclavable bottles and autoclaved for 20 m on the liquid cycle at

121°C.

On the day of experimentation, the 8-serovar Salmonella cocktail was prepared by
pipeting 16 mL of each strain into four individual 250 mL centriﬁrge bottles (turkey) or
20 mL of each strain into 18 (beef) or 22 (pork) 250 mL centrifuge bottles. The bottles

were centrifuged at 6000 x g for 15 min, the supemant was poured off from each

23

centrifuge bottle, and the remaining pellets were transferred into 500 mL of sterile
marinade, using a sterile spatula. The 500 mL bottle of marinade contained a sterile
stirrer bar, which was added before autoclaving, in order to dissolve the pellet into the
marinade. The pellet and marinade were stirred about 10 min, until the pellet had

dissolved in the marinade to ensure a homogeneous suspension. Thereafter, the inoculum

was plated onto duplicate aerobic PertriﬁlmTM plates (3M Microbiology Products, St.

Paul, MN) and incubated for 24 h at 37°C before enumeration. The 8-serovar Salmonella

inoculum consistently had a Salmonella population ~109 CFU/ml.

3.5 Inoculation

Irradiated meat products were inoculated with the 8-serovar Salmonella cocktail.
For inoculation, each roast was pre-weighed and placed in a lab-scale, sterile vacuum
tumbler (T-15 Vacuum Meat Tumbler, Kent Butcher Supply, Grandville, MI) that was
modiﬁed with a stainless steel bafﬂe insert. The targeted uptake was 15% (w/w) (turkey),
or 10% (w/w) (beef, pork). Marinade uptake was calculated based on the following
equation:
Where:

X = amount of marinade

Y = weight of meat
For 15% uptake:

X / (X+Y) = 0.15

X = 0.15X + 0.15Y

X=0.18Y

24

For 10% uptake:

X / X+Y = 0.10

X = 0.10X + 0.10Y

X = 0.11Y

Following the addition of the proper amount of marinade and the roast(s) to the
sterile tumbler, a vacuum pump (Welch Vacuum model 2534B-01, Thomas Compressors
and Vacuum Pumps, Skokie, IL) was used to pull a vacuum of ~84.65 kPa on the
tumbler. Roast(s) were then vacuum tumbled at 8 rpm for 20 min, rested for 5 min, and

then tumbled at 8 rpm for an additional 20 min.

3.6 Inoculation Verification

The initial Salmonella inoculation level for turkey and beef was validated by

removing the center core (16.4 cm3) from the inoculated roast(s) with an electrosurgical

unit (Valleylab SurgiStat II, Boulder CO). Tuntivanich and others (2008) previously
conﬁrmed that this electrosurgical unit allowed for the interior of the muscle to be
sampled without exterior contamination for similar work with turkey breast.

To verify the inoculation process, the top 1.27 cm of the roast was sterilely

removed using the electrosurgical knife, to avoid contamination from the exterior to the
interior. Following removal of the contaminated surface, ﬁve 2.54 x 2.54 x 2.54 cm3
samples (Figure 1) from near the center of each roast were removed with a sterile scalpel
(Becton Dickinson AcuteCare, Franklin Lakes, NJ). The t-shaped ﬁguration was used to

conﬁrm that the core sample had the lowest population of Salmonella, as compared to the

surrounding sample locations.

25

 

Figure 1. Whole muscle roast cutting diagram. Segment A was considered the center cut
of the roast.

Once extracted from the roast, the sample cubes were diluted 1:5 with 0.1%
buffered peptone water (Difco Laboratories, Sparks, MD) and homogenized in a
masticator (NEUTEC Group Inc., Barcelona, Spain) for 180 3. Samples were then
serially diluted with 0.1% buffered peptone water, plated on duplicate PetriﬁlmTM
aerobic plates (3M Microbiology Products, St. Paul, MN), and enumerated following 24
h of incubation (Cenco, Central Sciences Co. Chicago, IL) at 37°C. Six roasts from each
species were used to verify the inoculation procedure. The mean value from the six center
cores (segment A) was used as the initial Salmonella population for the remainder of the

study.

26

3. 7 Sample Preparation

Before cooking, roasts were thawed and inoculated as described previously. For
in-bag cook schedules, two septums (SSP 134, Spa, NY) were glued (Silicone II, General
Electric Company, Huntersville, NC) to the exterior of the bag at least 24 h before use, to
ensure adequate adhesion. Inoculated samples that were cooked in-bag were inserted
aseptically into boil-in-bags, which were vacuum sealed (V acMaster, Kansas City, MO).
The septums allowed for thermocouples to be inserted into the roast through the bag,
while maintaining the vacuum. Except for the bagging step, the in-bag and out-of-bag

samples were prepared identically.

3.8 Oven

A commercial, moist air roasting oven (C01 51FWUA12B2083, Cres Cor®,
Mentor, OH) was modiﬁed to control the cooking temperature proﬁle and to log the time-
temperature data. The control and recording system consisted of a LabVIEW® (National
Instruments, Austin, TX) data acquisition unit (CompactDAQ, National Instruments,
Austin, TX), thermocouple signal conditioning modules (NI 9211), a digital output
module (NI 9401), and a universal relay module (URM-800, Omega Engineering, Inc.,
Stamford, CT). Thermocouples consisted of 12 short needle probes (type K, PA1454B,
length 1.5 m, diameter 1.6 mm, barb end, accuracy 3: 1.1°C, max temperature 265°C,
Datapaq, Inc., Wilmington, MA), which were used to measure core temperature of the
samples. The control system was used to generate the step-up temperature proﬁles. Oven
humidity ranging from 20-100 % RH was controlled by a built-in dial on the oven, but

the humidity was measured by using a humidity sensor (Hydro Clip, serial number 36737

27

009, Rotronic, Huntington, NY) and a data logger (Datapaq, Inc., Wilmington, MA).
However, the humidity control was not a feedback system, based on the continuous

humidity measurement, as would be typical in an industrial system.

3.9 Cooking

Prior to cooking, stainless steal corers (2.54 cm diameter) were wrapped in
aluminum foil, and autoclaved for 20 min at 121°C to ensure sterility. Thereafter, the
wires for each of the 12 thermocouples were sterilized using sanitizing wipes (Sani-Cloth
Plus, PDI, Orangeburg, NY). The sterile corers were then strung onto the sterile
thermocouple wires, which recorded product core temperature in real time throughout the

cook (Figure 2).

Corer

Thermocouple wire

    

 

Figure 2. Thermocouple wire threaded through a sterile stainless steel corer.

After the thermocouple wire had been strung through the corers, two
thermocouple probes were inserted near the center of each roast. For in-bag samples,
these probes were inserted through the septum, to ensure integrity of the vacuum during
cooking. The temperature data from the colder of the two probes for each roast were used

to evaluate the roast’s endpoint temperature or predicted lethality, which was calculated

28

in real-time, based on the log-linear model with parameters obtained ﬁ'om isothermal
research (section 3.1 1 . 1).

When the software program indicated that the cold spot of the roast neared the
targeted end point, 71 .1°C, or calculated 7.0 / 5.5 / 3.0 log reductions, the roast was
immediately removed from the oven and placed into a sterile stainless steel pan. A core
was quickly and aseptically taken of the recorded cold spot using the 2.54 cm sterile corer
that was strung onto the thermocouple wire. After the entire core was removed from the
roast, about 2.54 cm were removed from the top and bottom of the sample, using a sterile
scalpel. The remaining center of the core, ~8 g sample (“cold spot”), was then quickly
inserted into a Whirl Pak bag containing 18 g of sterile refrigerated (4°C) 0.1% buffered
peptone water to cool the sample below 15°C in less than 10 s after coring. The computed
lethality used in the subsequent analysis included all of the temperature history of the
core through cooling to < 15°C.

Following cooling, the ~8 g samples removed from the treated roasts were diluted

1:5 with 0.1% buffered peptone water, homogenized for 180 s in a masticator, serially

diluted, plated on duplicate PetriﬁlmTM aerobic plates, and enumerated for Salmonella

survivors following 48 h of incubation at 37°C. The limit of detection for Salmonella on

PetriﬁlmTM aerobic plates following dilution was 0.4 log CFU/ g.

3. 10 Cooking Schedules

Following inoculation and preparation for cooking, roasts were processed using a
variety of pre-determined, industry relevant cooking schedules. Cook schedules varied in

containment, time, temperature, and humidity (Figure 3). Each schedule included two

29

roasts cooked to a target end point temperature of 71.1°C, and three roasts cooked to a

target endpoint for Salmonella lethality.

For turkey, the targeted end point lethality was 7 logm , based on the USDA
Salmonella lethality standards (FSIS - USDA 1999). For beef roasts, the targeted end
point lethality was 5.5 loglo rather than the USDA standard of 6.5 loglo, This was
because our inoculation process achieved a core Salmonella population of only 6.27
loglo CFU/ g. It was necessary to choose a lethality endpoint lower than our initial
concentration in order to quantify the actual lethality of Salmonella during the treatments.

Had the sample been cooked to the USDA Salmonella lethality standard of 6.5 loglo for

beef, survivors would have been recovered only from cooks that over-predicted lethality,

which would have biased the study results.

For pork, the targeted end point lethality was 3.0 loglo. The reason for the low

targeted lethality for pork was again due to the low initial Salmonella population within

the core (6.27 i: 1.04 log CF U/g), and a high level of variability between roast cores. It

was concluded that if the targeted lethality was lowered to (3.0 loglo), survivors could be

consistently quantiﬁed, because the initial population was always going to be larger than

3.0 loglo. The experimental design (Figure 3) consisted of seven different cooking

combinations representing typical industry conditions.

30

7 Schedules

I I
[ 4 In-bag Samples ][ 3 Out of Bag Samples]

 

 

4 In-bag Samples

 

 

 

 

l
Constant Temperature Step-up
933°C 600°C 90 min, 683°C 90 min, 67.7'C

 

 

 

No Humidity High Humidity No Humidity High Humidity
20% RH 78% RH 20% RH 78% RH

 

3 Out of Bag
Samples

 

 

Constant Temperature Step-up
° 600°C 90 min, 683°C 90 min, 67.7‘C

I
High Humidity High Humidity Changing Humidity
78% RH 78% RH No, Medium, High
(20%, 50%, 78% RH)

 

 

 

 

 

 

Figure 3. Predetermined cook schedules utilized to process whole muscle roasts in a
modiﬁed Cres Cor® moist air convection oven.

31

Of the seven cook schedules, four consisted of roasts that were contained within
vacuum packaged boil-in-bags (Smurﬁt-MBI, Butcher and Pack Supply, Detroit, MI).
Two of the in—bag schedules were cooked at a constant temperature of 933°C, one with
high humidity, and the other with no humidity. The two other in-bag schedules used a
step-up temperature condition, 600°C for 90 min, 683°C for 90 min, and 767°C until
the product reached its targeted end point. This condition included one treatment with
high humidity and one with added humidity.

Humidity was also a variable. For each of the temperature proﬁles, roasts were
processed with the humidity control set to no humidity, medium humidity, high humidity,
or a combination of the three. Relative humidity was measured by using a humidity
sensor and a data logger, following completion of the cook, the relative humidity data
was copied into a spreadsheet for analysis. The terms no, medium, and high refer to
settings on the oven control panel. The high humidity setting on the oven resulted in a
relative humidity >98%, but the average relative humidity throughout the course of an
entire cook was 78%, because the door was opened to remove samples. The average
relative humidity for the medium setting was 50%. The no humidity setting only
inhibited the oven from producing steam, but the environmental humidity had an impact,
making the relative humidity average 25%.

In the three remaining cook schedules, product was processed out-of-bag,
exposing the roast to the oven environment. For out-of-bag cooking, one schedule was at
a constant temperature with high humidity. Two others were step-up cook schedules, one

with high humidity, and the other with changing humidity. The changing humidity

32

condition increased the humidity every hour, from none to medium to high, based on the
settings of the oven control panel.

Figure 4 shows the oven air temperature and product core temperature for a
constant temperature cook schedule, whereas Figure 5 shows the oven air temperature

and product core temperature for a step-up temperature cooking proﬁle.

 

 

 

 

 

100
90 V
8 80 Oven Arr
2., 7o 1
0
5 60 r
N -
g 50
GE, 40 -
'— 30 ‘
20 I Product Core
10 r
o l I l I
0 20 4O 60 80 100
Time (min)

Figure 4. Sample cooking proﬁle for constant temperature cook schedule.

 

 

 

 

90
30 -
A 70 - Oven Ai
o
1, 60 -
2
a 50 -
E 40
i
'9 30 -
20 —
10 ‘ Product core
0 . . . . e . . . .
o 20 4o 60 80 100 120 140 160 180 200
Time (min)

Figure 5. Sample cooking proﬁle for step-up temperature cook schedule.

33

3.11 Model Computed Lethality
3.11.1 State-Dependent Model

The state-dependent model calculated the predicted lethality of the samples in
real-time, based on the log-linear model, using D and 2 values obtained from isothermal
research. Data obtained from the data acquisition system at the end of each cook were
used to verify the real-time lethality calculations using the log-linear (Bigelow) model:

No D(T(f))

where S is the survivor ratio, N is the population of microorganisms at time t, and No is
the initial microbial population. The D-value, D(T) is the time required at a certain
temperature to achieve a one log reduction. The z-value is the temperature change
required to achieve a one log reduction in the D-value.

The state-dependent model parameters, D and z, for turkey were obtained from
Tuntivanich and others (2008) (Table 1). For beef, the parameters were obtained from
unpublished work (Appendix 7.A) based on the methods of Orta-Ramirez and others
(2005) (Table l). The pork parameters were obtained from Velasquez and others (2009)
(Table 1).

Table 1. D and 2 values calculated from isothermal testing of the 8-serovar Salmonella

cocktail in turkey, beef, and pork (Tuntivanich and others 2008; Velasquez and others
2009)

 

 

 

 

 

 

 

 

 

Turkey Beef Pork
2 (°C) 5.38 5.52 5.29
Dref(S) 114.7 111.5 120.8
Tier (°C) 60 60 6O

 

34

For a direct comparison of the state-dependent and path-dependent model (section
3.11.2), a log-linear equation was used, but with a different formulation, such that:

N
logS = log-N— = —b(T(t)) t (2)

0
where S is the survivor ratio, N is the population of microorganisms at time t, and N0 is

the initial microorganism population. Parameter b is the temperature-dependent rate of
inactivation.
To be consistent with Stasiewicz and others (2008), the effect of temperature on b

was modeled as a modiﬁed Arrhenius dependency:

1 1
( ) ref p ﬂl T ( t) Tref (3)

where Bl determines the effect of temperature on b, using the same data as described
below (Table 2).

Table 2. State-dependent model parameters for whole turkey, beef, and pork (Marks and
others 2009).

 

 

 

 

Product bre f ,6] (K)
Whole-Muscle 0.5350 46,395
Turkey
Whole-Muscle 0.5632 44,854
Beef
Whole-Muscle 0.5705 34,859
Pork

 

35

3.11.2 Path-Dependent Model

The path-dependent model accounts for the effect of sub-lethal history

(Stasiewicz and others 2008):

1 1
b(T)=bref -exp _ﬂl T—(B—E— ‘ﬂzT (4)

where B21: accounts for sub-lethal thermal injury of the microorganism within the

organism’s region of sub-lethal injury. Sub-lethal history was quantiﬁed as an integral of
the temperature vs. time curve when the temperature was within the heat shock region

(38 and 52°C) (Stasiewicz and others 2008):
tT=HSu
r — l (Tm HS... )dt ,
tT=HS lower ( )
The path-dependent model was calibrated with 30 cook schedules using 1 g of

ground product (Tenorio-Bemal and others 200X), to obtain the model parameters (bref,

B], and [32, table 3), by minimizing the sum of squared errors, where error refers to the

difference between the observed and predicted lethality (Marks and others 2009). These
previously determined parameters were then applied to the cook schedules of the whole
muscle roasts in this study, in order to determine whether the path-dependent model

reduced the error between the observed and predicted lethality as compared to the state-

dependent model, at the pilot-scale.

36

Table 3. Path-dependent model parameters and goodness of ﬁt obtained from
nonisothermal calibration data sets for turkey, beef, and pork (Tenorio-Bemal and others
200X).

 

 

 

 

RMSE Bias

Product Br m (K) ﬁzﬂCl minj) (log CFU/g) (log CFU/g)

Turkey 0.91 50,752 0.0018 0.67 0.09
Beef 0.96 44,243 0.0019 0.87 0.11
Pork 0.82 53,952 0.0032 1.27 0.55

 

3.12 Data Analysis

T-tests using R-2.7.2 software were performed to determine whether a difference
existed between the mean predicted and observed log reductions. The null hypothesis was
that no difference existed between the predicted and observed log reductions, and the
alternative hypothesis was that a difference did exist.

An Analysis of Variance (AN OVA) using Excel was also’performed to evaluate
the relationship between the error (observed — predicted log reductions) vs. the sub-lethal
history. The potential for an error bias existed for the turkey and beef tests where no cells
were recovered following cooking. In this case, it is not possible to conclude Whether the
state-dependent model under-predicted the lethality; therefore, the result of the ANOVA
could be skewed negatively.

The root mean square error (RMSB) and mean residual for each species and the
lethality prediction model were calculated to compare the accuracy and average error, .

respectively, of the state-dependent and path-dependent model.

37

4. RESULTS AND DISCUSSION

4.1 Isothermal Results

The D and 2 values used for whole turkey and pork were reported by Tuntivanich
and others (2008), and Velasquez and others (2009). The beef isothermal D and 2 values
were new for this study and were reported in Table 1 in section 3.11.1. The individual
experimental replication averages (log CFU/ g), and log N vs. time (min) graphs for these

results are reported in Appendix 7.A.

4.2 Product Composition

The moisture content of the turkey, beef, and pork was 74.3 i 0.8%, 73.6 d: 0.2%,

and 68.5 :1: 0.8%, moisture respectively. The fat content of the turkey, beef, and pork was

1.09 i 0.1%, 8.08 i 2.8%, and 9.99 :l: 2.0%, respectively.

4.3 Initial Salmonella Concentration in Whole Muscle Roasts

Veriﬁcation of the inoculum level (Table 4) was performed as described in
section 3.6.

Unfortunately, there is a large standard deviation associated with research using
microorganisms, and the large standard deviation associated with the initial Salmonella
population introduced an aspect of variability into these experiments. This was because
the exact initial population for every sample (roast) could not be quantiﬁed before
cooking (as that would have destroyed the sample before cooking), which thereby

affected the accuracy of the measured log reductions for each individual sample.

38

Table 4. Initial core Salmonella population following inoculation, the USDA established
regulatory lethality target (FSIS-USDA 1999), and the experimental lethality target
(calculated) for turkey, beef, and pork.

 

 

 

 

 

Initial Regulatory Experimental
Salmonella Target Target
concentration Lethality Lethality
Turkey

log CFU/g 6.96 d: 0.53 7 7.0
Beef

log CPU/g 6.27 :t 0.89 6.5 5.5
Pork

log CFU/g 6.27 i 1.06 6.5 3.0

 

 

 

 

 

The marinade uptake and Salmonella population varied between meat species,
due to differences in product structure, protein, fat content, location of fat, connective
tissue, and moisture content. In addition to product composition, other factors can also
contribute to the high variability between samples, such as sampling location, sampling
technique, plating, and enumeration methods. Larger sample sizes inherently introduce
variability into this process, and it is more difﬁcult to achieve uniform bacterial

concentrations in whole muscle as compared to ground (See section 4.9).

4.4 End Point Temperature

For all cook schedules, duplicate roasts of each species were cooked to 71.1°C.
FSIS states that 71 .1°C “. . .is the minimum (temperature) that must be achieved...” in
order to eliminate all existing Salmonella for all of the species used in this research
(F SIS-USDA 1999).

Cooking the roasts to an end point temperature of 71 . 1°C resulted in the near
elimination of all countable cells (Table 5). For turkey and beef, three of the total 14

roasts cooked to 71 .1°C had quantiﬁable numbers of Salmonella. For both turkey and

39

beef, the maximtun surviving Salmonella population was 0.7 log CF U/ g which

corresponds to 6.3 and 5.6 log reductions for the turkey and beef, respectively. Complete
elimination of all Salmonella was not conﬁrmed, because enrichment was not performed
on samples following cooking. None of the 14 pork roasts cooked to 71 .1°C yielded any

surviving salmonellae after plating.

Table 5. The average Salmonella population (CFU/ g) recovered after duplicate plating on

aerobic PetriﬁlmTM plates for turkey, beef and pork roasts cooked to 71 . 1°C. Limit of
detection 0.4 log CFU/g.

Turkey Beef Pork
Average Average Average
Plate Plate Plate

Cook Count Count Count

Schedule (CFU) (CFU) (CFU)
93.30 In Bag, High
humidity
Replication 1 0.0 0.0 0.0
Replication 2 0.0 0.5 0.0
93.30 In Bag, No Humidity
Replication 1 0.0 0.0 0.0
Replication 2 0.0 0.5 0.0
93.3C No Bag, High
humidity
Replication 1 0.0 0.0 0.0
Replication 2 0.0 0.0 0.0
Step Up, In Bag, High
Humidity
Replication 1 1.0 0.0 0.0
Replication 2 1.0 0.0 0.0
Step Up, In Bag, No
Humidity
Replication 1 0.0 0.0 0.0
Replication 2 0.5 0.0 0.0
Step Up, No Bag, High
Humidity
Replication 1 0.0 0.0 0.0
Replication 2 0.0 0.0 0.0

Step-up, No Bag, Changing

Humidity (No, Med, High)

Replication 1 0.0 1.0 0.0
Replication 2 0.0 0.0 0.0

40

The turkey, beef, and pork roasts were inoculated to contain an initial Salmonella
population in the center core of 6.96 i: 0.53, 6.27 :l: 0.89, and 6.27 i 1.06 log CFU/g. Of
course, an initial Salmonella population this high within the core of whole-muscle
meat/poultry products is very unlikely in normal commercial processing. A product that
contained an initial bacterial load greater than 6.0 log CFU/ g is likely to have a tainted
aroma, and be labeled rotten. Following processing to 71.1°C, the products had achieved
at least a 5.6 log reduction. Given the high initial population of Salmonella and the
negligible number of total recovered cells supports the safety of processing to the USDA
suggested internal temperature of 71 .1°C.

Cooking roasts to an end point temperature of 71 .1°C is very likely to eliminate
Salmonella. This has positive implications for meat processors, because it reiterates that a
product cooked to an internal temperature of 71 .1°C is likely to yield a safe product,

therefore providing processors with a reliable slow cooking technique.

4.5 Lethality Error

The difference between the calculated and experimentally observed Salmonella
log reductions are referred to here as the lethality error, or error.

The mean predicted lethality for Salmonella in turkey and beef roasts was
_ signiﬁcantly greater than the observed lethality (P<0.05), reported in section 4.8. This
result indicated that the state-dependent model resulted in a large negative (dangerous)
error (-3.0 and -1.2 log CF U/ g for turkey and beef, respectively), and inaccurately over-
predicted the lethality of Salmonella Within whole muscle turkey and beef roasts during

pilot-scale, slow cooking processes.

41

The mean predicted lethality for Salmonella in whole muscle pork roasts was not
signiﬁcantly different from the mean observed lethality (P=0.36), reported in section 4.8.
Because no signiﬁcant difference was seen between the predicted and observed lethality,

the state-dependent model had a small lethality error, and was appropriate for predicting

the lethality of Salmonella within whole muscle pork roasts, cooked to a targeted 3 loglo

reduction, during pilot-scale slow cooking.

The lethality error for whole turkey roasts was signiﬁcantly greater for samples
that were cooked in-bag compared to those out-of-bag (P=0.026). In addition, the
samples processed in-bag without humidity had a signiﬁcantly greater negative error
(P<0.05) than those processed in-bag with humidity for both the constant and step-up
temperature schedules.

The lethality errors for whole beef roasts were not signiﬁcantly different for the
in-bag vs. out-of-bag cook schedules (P=0.29). The lethality error for the cook schedules
processed at a constant temperature, in-bag, with humidity was signiﬁcantly greater than
the lethality error for the constant temperature samples processed in-bag without
humidity (P=0.04). No statistically signiﬁcant difference in errors was seen for the step-
up temperature proﬁles in-bag with and without humidity (P=0.8). In addition, no
signiﬁcant differences were evident between the lethality errors among any of the pork
roast cook schedules.

The lethality error was inconsistent across the three species. Statistically
signiﬁcant interactions based on the cook schedules could not be tested, due to the small

number of samples available for analysis (8 for turkey and 11 beef), because only

42

samples with recoverable Salmonella cells were used in the analysis. In contrast, all 20
pork samples yielded quantiﬁable Salmonella for analysis.

Turkey and beef were processed to a targeted lethality near the initial Salmonella
concentration, therefore making it more difficult to quantify positive lethality errors
(section 4.8). Positive lethality errors are when the experimental log reductions exceed
the values predicted by the state-dependent model. The inability to quantify positive
lethality errors negatively skews the data, and may attribute to the inconsistency in
statistical signiﬁcance between the three species.

Inconsistencies in the inoculation process could have also led to irregularities in
the results between species. The standard deviation for the initial Salmonella population
was as large as 1 log CFU/g, with this high variability perhaps impacting the observed
lethality, thereby affecting the lethality error.

There are no fundamental explanations for why humidity would have impacted
turkey, but not beef or pork. The statistically signiﬁcant results could be attributed to the
small number of samples with recoverable cells in turkey and beef samples that yielded

surviving salmonellae.

4.6 Replication Error

The experimental replication errors for the recovered Salmonella in turkey, beef,
and pork were 1.42, 0.93, and 1.03 log CFU/g, respectively. The issue of high variability
was consistent across all three species, suggesting that there was substantial variability
inherent in the method of scaling-up to pilot-scale processing. A study considered to have

low variability would have a replication error less than 0.2 log CFU/ g.

43

Many factors within the experimental methods had the potential to contribute to
the overall experimental error. These factors included, but were not limited to, inaccurate
determination of the initial Salmonella population within the core, ability to insert the
thermocouple probe directly into the geometric center of the product, uncertainty in the
state-dependent model parameters, sampling of the roast core, enmneration of survivors,
and the recovery of sub-lethally injured cells.

Replication error (i.e., the standard deviation among replicates) increased with
increasing sample size. The increase previously occurred when the experiment was
scaled-up from 1 g ground samples cooked in a therrnocycler (Tenorio-Bemal and others
200X), to 25 g ground and whole muscle samples cooked in a bench—top convection oven
(Jones and others 200X), to 500-1000 g whole muscle samples of turkey, beef, and pork
cooked in a pilot scale moist air convection oven (this study). The small, controlled,
laboratory samples had a very low replication error, while the larger bench-top and pilot-
scale samples both had substantially larger replication errors (Table 6), even though all
three studies were conducted with the same basic laboratory procedures in the same
research group. This suggests that considerable variability in the processing of larger
products is unavoidable, due to the various causes mentioned above.

The laboratory scale experiments included a larger number of treatments, 30 for
every species, as compared with ﬁve treatments for bench-top experiments and seven
treatments for pilot-scale experiments. All experiments were performed with three

replicates per treatment per species.

44

Table 6. Replication error for measured process lethality at three experimental scales log
CFU/g (Tenorio-Bemal and others 200X, Jones and others 200X).

 

 

 

 

1g 25 g 25 g 500 - 1000 g

ground ground whole whole
Turkey 0.15 0.56 0.99 1.37
Beef 0.15 0.93 1.16 0.93
Pork 0.08 NA 1.51 1.02

 

 

 

 

 

 

 

In order to conﬁrm that the low replication error for the 1 g samples was not due
to the larger sample set, survivor (log CF U/g) data from 21 of the l g samples were
randomly selected to match the sample set size with the present study. The replication
error for these samples was computed and compared to the replication error for all of the
samples. This process was performed three times for every species (Table 7). The
average replication error including all 90 samples and the average replication error with
only 21 samples did not differ more than 0.02 loglo CF U/g (Table 7 vs. Table 6),
indicating that the increase in replication error scaling up the experiments was not due to
differences in the number of samples, but rather differences inherent in the experimental

system.

Table 7. Replication error (log CFU/ g) for 21 randomly selected samples from the 1 g
laboratory experiments, with three replication for each species.

 

 

 

 

 

 

 

 

Turkey Beef Pork
Replication 1 0.13 0.14 0.10
Replication 2 0.10 0.16 0.07
Replication 3 0.17 0.12 0.07
Average 0.13 0.14 0.08

 

 

 

45

4.7 Sub-lethal History vs. Model Error

The sub-lethal history of Salmonella was quantiﬁed during every cook according
to equation 5. As a simple test of any systematic failure of the traditional, state-dependent
model for slow cooking, a linear regression was conducted for error vs. sub-lethal history.
The sub-lethal history was compared with the model error (observed — predicted log
reductions) to determine whether the error of the state-dependent model changed linearly
with sub-lethal history (Figures 6, 7, and 8). Based on the variability in the data collected
for the pilot-scale research, another method of analysis may have been more suitable for
data with this amount of dispersion. Clearly, the data points do not follow a linear trend,
or any obvious alternative trend. However, when a linear regression was used, error did
not increase (P = 0.25) with sub-lethal history, which was inconsistent with the

hypothesis of this study and the prior results of Stasiewicz and others (2008).

 

F’
c:

 

 

 

“a
8
0 .
:8 -1.0
2 A o o
A: S) -2.0 - e
E E3 °.
3 -30
an
8 3 0
3 V -4.0
d‘
e -5.0- ’ 0
In
El
-6,0 I I I I I
0 100 200 300 400 500 600
Sub-lethal history (min K)

Figure 6. Salmonella lethality error (observed - predicted) vs. sub-lethal history for turkey
breast. Lethality calculations based on laboratory derived isothermal D and 2 values
(Tuntivanich and others 2008).

46

 

0.5

 

 

 

E 0.0 . .
.2 0 ,
g -0.5 - °
1.. 9
ﬁ- ’33 -1.0 - O
I \ .
B E -1.5 - e
5 en -2.0 - e
3 E -2.5 - .
E -3.0 -
a -3.5 - e
4.0 l T l l T
0 1 00 200 300 400 500 600
Sub-lethal history (min K)

Figure 7. Salmonella lethality error (observed - predicted) vs. sub-lethal history for beef
roasts. Lethality calculations based on laboratory derived isothermal D and 2 values,
Appendix 7.A

A linear regression evaluating the relationship between the error and sub-lethal
history for beef roasts was statistically signiﬁcant (P=0.04). This indicated that the
lethality error was signiﬁcantly changed with sub-lethal history; however, the linear
relationship was negative, giving a result that again was inconsistent with our hypothesis.
According to the statistical analysis for this experiment, the absolute error decreased with

an increase in sub-lethal history for beef roasts.

47

 

 

 

 

1.5
'8
g 1.0 ~ 9 .
E A 0.5 - e
- g 00‘ 0
'° in e e
E u 05 .
0 NJ . . . O
.8 .3 -1.0 r 9’ 9
O V ’
e -1.5 -
s . r
t. e
I. -2.0 ‘ e
“a e e

’2.5 1 I I I I

0 100 200 300 400 500 600
Sub-lethal history (min K)

Figure 8. Salmonella lethality error (observed - predicted) vs. sub-lethal history for pork
roasts. Lethality calculations based on laboratory derived isothermal D and 2 values
(Velasquez and others 2009).

For pork roasts, no statistically signiﬁcant linear relationship was seen between
the lethality error estimated by the state-dependent model and the sub-lethal history

(P=0.80).

Slow cooked roasts processed to a targeted lethality in a pilot-scale convection
oven demonstrated an error between the observed and predicted log reductions when the
state-dependent model was used to estimate Salmonella lethality, but this error did not
linearly increase with an increase in the sub-lethal history. These particular results lack a
common trend between the error and sub-lethal history, perhaps because of the large
amount of variability associated with the pilot-scale data as referred to in section 4.6
and/or the inconsistencies associated with recovering usable data following lethality

treatments.

48

4.8 Comparison of State-Dependent and Path-Dependent Models

A comparison of the lethality error (observed — predicted log reductions) for the
state-dependent and path-dependent models was performed only on cook schedules that
had quantiﬁable Salmonella (n=8 turkey, n=11 beef, n=20 pork). The path-dependent
model parameters were obtained by minimizing the sum of squares from the laboratory 1
g, ground data calibration set. The root mean square error (RMSE) was used to validate
the accuracy of each model, by comparing the RMSE of the state-dependent and path-
dependent models to determine which model provided a smaller RMSE. Reported below,
the RMSE was reduced by applying the path-dependent model for whole muscle turkey

and beef, but not for whole muscle pork (Table 8).

Table 8. Prediction error and mean residual of the state-dependent and path-dependent
models for whole muscle turkey, beef, and pork.

 

 

 

 

 

 

 

 

 

 

 

 

Mean
Model parameters RMSE Residual
11 used (log CPU/g) (log CFU/g)
Whole State-dependent 3.3 -3.0
Turkey 8 Path-dependent 1 . 1 0.4
Whole State-dependent 1 .7 -l .2
Beef 1 1 Path-dependent 0.8 0.2
Whole State-dependent 0.8 -0.07
Pork 20 Path-dependent 2.3 1 .8

 

 

A deﬁnitive explanation for why the state-dependent model had a lower RMSE as
compared with the path-dependent model for whole pork is unknown. A possible

explanation for the increased RMSE for only pork could be attributed to the experimental

methods, because the pork was processed to a lower targeted lethality (3.0 loglo CFU/ g)

49

than turkey (7.0 loglo CF U/ g) and beef (5.5 loglo CFU/g) roasts, which allowed only the

pork cooks to quantify positive errors. A positive error occurs when the predicted log
reductions were smaller than the observed log reductions.

The mean residual (bias) for each species and lethality prediction model was
calculated in order to compare average errors. The state-dependent model for turkey and
beef over predicted lethality for the ﬁnal product. The large negative error could be due
to an experimental bias, because these products were cooked to targeted lethalities near
the initial Salmonella inoculum level, making it difﬁcult to quantify positive errors.

If the hypothesis that slow cooking increases the thermal resistance of Salmonella
was true for every observation, then every individual treatments would have resulted in
quantiﬁable Salmonella following cooking. But if the hypothesis was not true for an
individual observation, then that treatment would result in un-useable data. Unfortunately
this appears to have been the case; therefore, quantiﬁcation of cells when the predicted
log reductions were less than the observed log reductions was inconsistent for turkey and
beef, because the targeted lethality was too near the initial Salmonella concentration.
Ideally, if the initial Salmonella population was 10 log CFU/ g, then survivors could have
been consistently recovered following every lethality treatment.

In the situation with pork, unlike turkey and beef, roasts were cooked to a targeted
lethality 3.0 log below the initial Salmonella population, which allowed for quantiﬁcation
of recoverable Salmonella when the observed log reductions exceeded the predicted log
reductions. Turkey roasts did not yield any positive errors, with the maximum positive

errors being 0.62, and 2.33 log CFU/ g for beef and pork, respectively.

50

The mean residual for turkey and beef showed that the state-dependent model
over-predicted Salmonella lethality in the ﬁnal product, which is dangerous, because it
falsely indicates that the product is safe. The average errors of the state-dependent model
were -3.0 and -1.2 log CFU/ g for turkey and beef, respectively. A negative error means
that the model over-predicted the actual lethality. Over-predicting the ﬁnal product safety
can be dangerous in regard to food safety.

The largest over-predictions of lethality for individual roasts, using the state-
dependent model were -5.07, -3.31, and -0.96 log CPU/g for turkey, beef, and pork,
respectively (Figures 9, 10, and 11). A negative value indicates that the lethality model
over-predicted product safety, and a positive value indicates that the lethality model

under-predicted the lethality of the product, making it more safe.

 

 

 

 

 

re 3
0 O
3.3. 2 ‘
g 1 4 O O
9‘: A O 00
. Q” 0 t t f t o
E E4 -
E err-2 e . ° .
.3 ,3 ’0
C v-3 .
g .4 - 0 State-dependent model
i: _5 _ .
iii 0 Path-dependent model

-6

0 100 200 300 400 500 600
Sublethal history (min K)

Figure 9. Lethality errors for the state-dependent and path-dependent model applied to
the whole muscle turkey data set.

51

 

 

 

 

 

3
E 2 4 °
.2
a
i: 1 0 ° °
a: 3” ° 0 e
a O I I I I T
E5 o , . o ,

o

3 v ‘2 _ . 0
g 3 _ o o State-dependent model
”-1 , 0 Path-dependent model

.4

0 100 200 300 400 500 600
Sublethal history (min K)

Figure 10. Lethality errors for the state-dependent and path-dependent model applied to
the whole muscle beef data set.

 

 

 

 

 

6
“G
g 5 < o o
’6' 4 ‘ o
i.’ 3 . o
e- a, . ,
D ‘ 0 o
B 5 1 a; C20 0
E ' Q o 0
a E 0 a. s ‘ T I #7
.o e
O V -1 ~ . 9 O o ’ ’
. e
‘2' -2 . 0 State-dependent model
In
i-I-l '3 ' ° Path-dependent model
.4
o 100 200 300 400 500 600
Sublethal history (min K)

Figure 11. Lethality errors for the state-dependent and path-dependent model applied to
the whole muscle pork data set.

52

Using the state-dependent model to predict product safety of slow cooked whole
muscle turkey and beef roasts can be dangerous, because it over estimates the safety of
the ﬁnal product.

When the path-dependent model was applied to the experimental data from turkey
and beef, the average error decreased from -3.0 to 0.4 and -l .2 to 0.2 log CFU/g,
respectively. The path-dependent model resulted in a decreased error for turkey and beef,
making the model a more accurate prediction of Salmonella lethality during slow cooking
of whole muscle. Furthermore, the average error of the path-dependent model was
positive, indicating that the model under-predicted the safety of the ﬁnal product,
therefore resulting in a ﬁnal product that has achieved, or slightly exceeded, its predicted
log reductions.

Application of the path-dependent model to the experimental data for pork species
yielded different results than for turkey and beef. For pork, the state-dependent model
had a small average error of -0.07 log CF U/ g. The path-dependent model had an average
error that was greater than that of the state-dependent model, 1.8 log CFU/g. Application
of the path-dependent model to pork increased the average lethality error, but the increase

resulted in a larger under-prediction of lethality. For pork roasts cooked to targeted

lethality of 3.0 loglo, the state-dependent model was more accurate than the path-

dependent model, but the path-dependent model produced a safer product.

While for pork the state-dependent model had a lower error (-0.07 log CF U/ g) as
compared with the error of the path-dependent model (1.8 log CFU/ g), the state-
dependent model still had a negative error. The state-dependent model over-predicted the

average safety of the ﬁnal product for all turkey, beef, and pork. In contrast, the path-

53

dependent model always resulted in an average positive error, showing that it under-
predicted the safety of the ﬁnal product.

The state-dependent model more accurately estimated Salmonella lethality for
only slow-cooked whole muscle pork as compared with the path-dependent model. As
mentioned previously, a possible explanation for the disparity between the species could

have been because pork was cooked to a lower targeted lethality.

4.9 Safe Harbors

Safe harbors are a regulatory paradigm that speciﬁes the amount of time that
cooked meat need to be held at or above a given temperature in order to be acceptable
(F SIS-USDA 1999 and 2005). These time-temperature conditions are speciﬁed by FSIS
in tables that help processors determine the conditions of their process. The safe harbors
published by FSIS were based on data generated in laboratory-scale studies, with small

sample sizes (Juneja and others 2001; F SIS-USDA 2005). Examples of FSIS speciﬁed

safe harbor temperatures and times for beef to achieve a 6.5 loglo reduction include 60°C

for 12 min, or 656°C for 67 sec, or 71.1°C for 0 sec (FSIS-USDA 1999).

As another evaluation of the cooking treatments in this study, the present data
were compared to the FSIS-speciﬁed safe harbors. The amount of time that the roasts
were held at a certain temperature was evaluated based on the time and temperature data
recorded during each cook. Only the roasts cooked to a targeted lethality were evaluated
(Table 9).

From these evaluations, none of the cook conditions (Table 9) complied with the

safe harbors as deﬁned by FSIS (1995 and 2005). None of the products were held at a

54

single temperature for the amount of time speciﬁed by the safe harbors; all samples failed

to satisfy the time requirement.

Although all of the turkey cooking treatments achieved a calculated 7.0 loglo

reduction based on the state-dependent model parameters, a 7.0 loglo reduction was not

actually achieved in all cases. Therefore, given that the processes did not comply with the

safe harbors, but not all samples achieved a 7.0 loglo reduction, this analysis does not

allow any general conclusions about the validity of the safe harbors for these processes.

55

Table 9. Documented times (s) that roasts cooked to a targeted lethality were held at a
speciﬁc temperature, for turkey and beef.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Turkey Beef
Temp Temp

(°C) Time (s) at Temp (°C) Time (s) at Temp
Cook Required Actual Required Actual
60 1512 171 60 720 181
Constant Temp, in bag, 60 1512 181 61 480 I 88
high humidity 60 1512 182 61 480 l 16
60 1 5 12 64 60 720 170
Constant Temp, in bag, 60 1512 I 82 59 900 212
no humidity 60 l 5 12 191 60 720 85
6O 1 5 l 2 l 69 60 720 96
Constant Temp, no bag, 60 1512 23 5 61 480 107
high humidity 60 1512 l 17 61 480 149
60 1512 215 60 720 117
Step-up, in bag, high 60 1 5 l 2 203 60 720 202
humidity 60 1 5 12 256 60 720 106
59 1896 404 58 1380 342
Step-up, in bag, no 59 1896 364 58 I 3 80 372
humidity 59 1896 364 58 13 80 254
60 1512 171 59 900 192
Step-up, no bag, 60 I 5 12 171 60 720 128
changing humidity 60 15 12 223 59 900 160
60 1512 213 60 720 116
Step-up, no bag , high 60 15 I 2 244 60 720 I38
humidity 60 1 5 12 202 59 900 l 70

 

 

 

 

 

 

 

4.10 Challenges of Whole Muscle Pilot-Scale Research

In scientiﬁc studies, the method of experimentation has an inﬂuence on the
results. During this research, many challenges centered around developing an
experimental method that would be relevant to industry processing techniques. Very

little prior thermal inactivation research has been done using inoculated whole muscle

56

products (particular at the pilot or commercial scale); therefore, a standard method of
experimentation has not been developed.

The critical challenge associated with whole muscle, pilot-scale, thermal
inactivation studies is getting the pathogen into and out of the product, while still
replicating industry practices as closely as possible. While other studies have reported on
the migration of pathogens into whole muscle (Warsow and others 2008; Luchansky and
others 2009), a reliable inoculation method for achieving a high number of the target
pathogen in the core of whole muscle products does not yet exist. Warsow and others
(2008) did show penetration of Salmonella into Whole muscle, but at core concentrations
less than 3.0 log CFU/g.

In this study, an inoculation marinade was used to achieve a Salmonella
population greater than nearly 7.0 log CFU/ g within a large (500-1000 g) sample. The
vacuum tumble marination method used in this experiment worked relatively well, but it
was difﬁcult to consistently achieve the targeted core population for our experimental

protocol. Using needle injection and a 30 min rest period, Pradhan and others (2007)

. . . . . . . . . 6 7
were able to achreve an Imtral Internal Listeria Innocua populatron of 10 to 10 log

CF U/g in chicken breast. Though a standard deviation for the initial population was not

reported, based on their reported inoculation level of 106 to 107 log CFU/g, one might

assume that similar results would have been obtained using a method similar to vacuum
tumble marination. However, chicken breasts are substantially smaller than the roasts
used in this cooking study.

Luchansky and others (2009) showed Escherichia coli 0157:H7 has the capability

to move from the surface to the interior of beef roasts during blade tenderization.

57

Contamination levels were 2 3.4 log”) CFU/ g within the interior of the meat. High

numbers of internalized E. coli 0157:H7 were not achieved because blade tenderizing
does not inoculate uniformly within a roast. Following blade tenderization, E. coli
01572H7 reductions were quantiﬁed for steaks cooked over a commercial open-ﬂame
grill (Luchansky and others 2009). Though steaks cooked over an open ﬂame differ in
multiple ways from roasts cooked in a pilot-scale convection oven, both studies
experienced challenges. A limitation to the study conducted by Luchansky and others
(2009) was that post-cook samples were not rapidly cooled in order to inhibit carry-over
cooking lethality. Similar to the whole muscle pilot-scale cooking research, quantifying
the post-cooking pathogen reduction proved to be a challenge for Luchansky and others
(2009), as evident from their large and varying standard deviations, ranging from 0.16 to
2.13 log CFU/g.

Most researchers do not perform pathogen inactivation research using intact
whole muscle products. This is likely due to the challenges associated with inoculation
and sample collection after processing. When intact whole muscle products are used, the
sample sizes are typically small, and not applicable to industry relevant processing. In a
bench-top study, small (5 g) samples were used to evaluate the effect of grinding on the
thermal resistance of Salmonella in beef (Mogollon and others 2009). The marination
process for smaller whole muscle samples was also quite different to that used for larger

sample sizes. After Mogollon and others (2009) immersed their whole beef muscle

sample in an inoculated marinade (108 CFU/mL); the reported initial Salmonella

population was ~ 7.8 loglo CFU/g. For whole muscle beef roasts used in pilot-scale

58

Salmonella inactivation research, the inoculated marinade had a concentration of 9.54 i

0.56 loglo CFU/mL, but the average Salmonella population at the center of the roast

following vacuum tumble marination was 6.27 loglo CFU/ g, lower than that in the

smaller samples used by Mogollon and others (2009).

In order for this research to be both applicable to industry standards and
investigate our objectives, it was necessary to overcome several experimental challenges,
such as achieving a high initial Salmonella population within the interior of a large,
whole muscle roast, quantifying the initial Salmonella population, quickly cooling the
roast post-cooking, and aseptically removing a sample from the interior of the roast.
These challenges impacted the experimental methods used, and the variability of the
results. There are few scientiﬁc studies that have cooked large, whole muscle roasts in
pilot-scale ovens, because this type of research introduces a great deal of variability into
the process and results, as compared with laboratory studies. While the variability of
results can be discouraging, such research is critically important in validating industry
processes. This study evaluated the accuracy of two microbial process lethality models,
and demonstrated that prior results from laboratory-scale research with slow-cooked
product were not necessarily well replicated at the pilot scale. However, a modiﬁed,
path-dependent model generally yielded improved predictions of process'lethality for

slow cooked roasts.

59

5. CONCLUSIONS

5.1 Implications of Process Scale-Up Variability

Currently, the industry is validating the safety of their ﬁnal product using
predictive lethalities calculated with state-dependent models developed from isothermal
laboratory studies. Most isothermal thermal inactivation tests have been conducted using
liquid media, slurries, or ground meat, rather than whole muscle products. Depending on
laboratory-scale data to validate the safety of industrial processes can be risky, because
large commercial processes are far more variable compared to controlled laboratory
studies. The increase in replication error associated with scaling up from a small,
controlled laboratory study to bench-top and pilot-scale slow cooking experiments was
demonstrated in the results (section 4.6). This suggests that commercial applications of
state-dependent models validated with laboratory studies may need to be re-evaluated at
the processing level. Clearly, pathogen inactivation studies can not be conducted in
industrial processing facilities; therefore, pilot plants are the next best option in terms of
applicability to validating models and accounting for the variability associated with

industrial processing conditions.

5.2 Lethality Error vs. Sub-lethal History

There was no obvious trend in the relationship between sub-lethal history and
calculated lethality error, even though prior laboratory-scale research had shown this to
be true. However, application of the path-dependent model did reduce the prediction error

(RMSE) and the risk of over predicting the Salmonella lethality in the cooked product.

60

For whole muscle roasts cooked in a pilot-scale convection oven, the path-dependent

model under-predicted lethality, thereby increasing the safety of the ﬁnal product.

5.3 Implications of Using a State-Dependent vs. Path-Dependent Model

The path-dependent model did not decrease the RMSE for every species of meat
processed, but it did always result in a positive mean error, or under-prediction of
lethality. The state-dependent model had a negative mean error, which over-predicts
lethality, making the product less safe.

The path-dependent model was applied to whole muscle data sets using
parameters previously developed from 1 g samples of ground muscle. These parameters
were previously estimated using 30 randomly selected treatments, and veriﬁed with 15
randomly selected treatments from the experimental protocol of each species. When
applying parameters that have been developed in a controlled laboratory experiment, it
would be ideal to apply ground parameters to ground product, and whole parameters to
whole product. Unfortunately it is not feasible to conduct an experiment using 1 g whole
muscle samples, because samples of this size lose much of their whole muscle structure.

In order to understand the usefulness of the path-dependent model in pilot-scale
processes, ground muscle parameters were applied to the whole muscle data in order to
demonstrate the path-dependent models potential to account for the sub-lethal history of
Salmonella during slow cooking. To evaluate this, the RMSE for the state-dependent and
path-dependent models were compared. The path-dependent model had a lower RMSE
for turkey and beef roasts, suggesting that the path-dependent model may be an improved
model for predicting the lethality of Salmonella in slow cooked, whole muscle roasts, as

compared with the state-dependent model.

61

The actual product used for the three different experiments — 1 g laboratory
(Tenorio-Bemal and others 200X), 25 g bench-top convection oven (Jones and others
200X), and 500-1000 g whole muscle pilot-scale — were all from the same original lot of
meat for each species. Only the structure (whole vs. ground) and size of the samples
varied between the three experiments.

Based on previous research, Salmonella is 2-3 times less heat resistant in ground
than in whole muscle products (Orta-Ramirez and others 2005; Tuntivanich and others
2008; Velasquez and others 2009). In this study, the path—dependent model (based on
data from ground product) under-predicted the lethality of slow-cooked, whole muscle
roasts. Based on the effect that meat structure has on heat resistance, it could be
suggested that if the path-dependent model were applied to ground product and the same
thermal conditions as for the samples in this study, then the path-dependent model (using
the existing parameters) might have under-predict the Salmonella lethality as compared

with the whole muscle roasts. This needs ﬁrrther investigation.

62

6. FURTHER WORK

Additional research and analysis of the path-dependent model need to be
conducted in order to make these ﬁndings more applicable for industry use. Currently,
the path-dependent model is being applied to whole muscle data, based on ground meat
parameters. Additional whole muscle data from the methods performed by Tenorio-
Bemal and others (200X), with larger whole muscle samples (~ 10 g) needs to be
collected in order to estimate the whole muscle parameters of the path-dependent model.
These parameters can then be applied to a randomly selected data set for validation. This

will provide an apples-to-apples comparison of the state-dependent and path-dependent

 

models for whole muscle products in pilot-scale slow cooking applications.

The various sources of error contributing to the large experimental variability
associated with the scale-up of cooking processes needs to be evaluated. All possible
sources of error should be evaluated individually and as a group to determine whether
any have a synergistic effect. For example, if it is determined that the inoculation method
and sampling location do not have a signiﬁcant impact on the overall error when
evaluated individually, but together these two sources of error provide a signiﬁcant
increase in experimental variability, then it could be concluded that the two sources have
a synergistic effect. Individually they do not have an impact, but together they interact
with each other to have a signiﬁcant impact.

The average lethality error should be evaluated with alternative end point lethality

values. The pre-determined, industry relevant cook schedule used in this research can be

re-run with turkey and beef cooked to obtain 3.0 loglo reductions, in order to determine

63

whether the end point lethality has an impact on the accuracy of the state-dependent

model. Cooking these products to a 3.0 loglo reduction will allow the quantiﬁcation of

any positive errors, thereby eliminating possible experimental bias. Also, after the present

results for cooking pork roasts to a targeted 3.0 loglo reduction, it would also be

beneﬁcial to evaluate the results from cooking pork roasts to a targeted 6.5 loglo

reduction, to determine whether cooking to increased end point lethality has an impact on
the average lethality error. Tenorio-Bernal and others (200X) demonstrated that for 1 g
laboratory experiments, there was an increase in the lethality error when the targeted
lethality was increased. I

Alternative inoculation methods should also be investigated. Various methods
should be compared to the vacuum marination technique to determine whether alternative
marination methods facilitate a higher initial Salmonella population, lower variability of
the initial pathogen level, and/or lower variability of the ﬁnal population. Various
inoculation methods need to be tested; a few examples include injection, injection
followed by vacuum tmnbling, injection followed by a rest period, a “dry” surface
inoculation, or a “dry” surface inoculation on an exposed interior surface that is then

closed into the core prior to cooking (Table 10).

64

Table 10. Possible pros and cons of alternative whole muscle pathogen inoculation
methods.

 

 

Inoculation

Method Pros Cons

Needle injection High pathogen population in Uncertain uniformity
interior of distribution

 

Needle injection
followed
by vacuum tumbling

High pathogen population in
interior, and potentially
improved uniformity of
distribution

Potential for bacterial
migration out of
product

 

Needle injection
followed
by a rest period

High pathogen population in
interior, and potentially
improved uniformity of
distribution

Uncertain uniformity
of distribution and
potential for bacterial
migration out of

 

 

product
“Dry” inoculation High pathogen population in Integrity of the whole
on interior of speciﬁc interior location muscle is sacriﬁced,
muscle surface and unknown

 

 

potential for
migration away from
inoculated interior
surface

 

All alternative inoculation techniques will need to be veriﬁed for uniform

 

distribution of Salmonella throughout the roast. An alternative inoculation process should
be evaluated, because if it does decrease the large variability in initial numbers of
Salmonella, then it could more accurately account for the log reductions obtained after
cooking.

Additionally, bacterial transport during whole muscle cooking should be assessed.
During this research, it was assumed that the initial bacterial population within the center
of the roast was altered only by heating; bacterial transport during cooking was not
assessed. An experiment evaluating bacteria transport during cooking, relative to liquid

(purge) transport, would be useful. A proposed method to evaluate bacteria movement

65

would be to remove a core from a roast, then burry a small metal tube containing
inoculated whole muscle product into the middle, cored location of the roast prior to
cooking, and cook the roast according to the methods in section 3.9. Following cooking,
the contents of the product within the tube cooked inside the roast could be evaluated and

compared with an inoculated roast cooked without a metal insert.

66

7. APPENDICES

Appendix 7.A: Isothermal tests on whole and ground beef

Initial cooking trials revealed that the previously published inactivation
parameters for Salmonella in beef (Orta-Ramirez and others 2005) were insufﬁcient and
yielded extremely large errors. Therefore, the methods of Orta-Ramirez and others (2005)
were repeated, using the same beef as in this study, and with more temperatures (55, 58,
60, 62, and 63°C) in an attempt to generate more robust model parameters. The results
from the isothermal tests are reported below (Table 11). Raw, average data from the

isothermal tests are reported in Tables and Figures 12 -21.

Table 11. D and 2 values calculated from beef isothermal inactivation tests of 8-serovar
Salmonella cocktail.

 

 

 

 

 

 

 

 

 

 

 

Ground Whole
2 5.63 °C 2 5.52 °C
Dref 64.22 sec Dref 1 1 1.52 sec
Tref 60 0C Tref 60 °C

 

67

Table 12. Salmonella survivors (CFU/g) in ground beef during 55°C isothermal

inactivation.

log N

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   
 

 

 

Repl Rep2 Rep3
Time (sec) average average average
0 17,250,000 18,250,000 1 1,000,000
308 8,575,000 10,250,000 3,100,000
616 5,600,000 2,850,000 702,500
924 1,550,000 130,000 74,750
1232 235,000 14,750 10,250
1540 14,000 10,875 3,775
1848 15,750 2,425 3,375
2156 6,150 2,897.5 500
8
7 1 y = -0.002x + 7.2634
R2 = 0.8995
6 _
5 a
4 , e
O
3 ..
o
2 _
1 ..
0 T I i I
0 500 1000 1500 2000
Time (sec)

 

2500

Figure 12. Salmonella survivors (CF U/ g) in ground beef during 55°C isothermal

inactivation.

68

Table 13. Salmonella survivors (CF U/ g) in ground beef during 58°C isothermal

inactivation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Repl Rep2 Rep3
Time (sec) average average average
0 41 ,000,00 1 1,500,000 3,975,000
65 1,075,000 7,400,000 772,500
130 242,500 2,825,000 1 10,000
195 17,500 860,000 30,000
260 10,500 83,500 7,250
325 7,225 43,750 13,150
390 3,700 23,750 5,125
455 6,050 10,750 32,750
8
0 y = ~0.0065x + 6.4709
7 , 2
4) . R = 0.7429
6 ..
5 4
Z
.3? 4 ”
3 _
2 4
1 J
o . . T A
0 100 200 300 400 500

Time (sec)

Figure 13. Salmonella survivors (CFU/ g) in ground beef during 58°C isothermal
inactivation.

69

Table 14. Salmonella survivors (CF U/ g) in ground beef during 60°C isothermal

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

inactivation.
Repl Rep2 Rep3 Rep4 Rep5
Time (sec) average average average average average
0 495,000 1,725,000 495,000 775,000 172,500
14 2,325,000 6,250 2,550 587,500 285,000
28 NA 24,000 97,250 1,375,000 1,525
42 75 11,025 4,650 307,500 97.5
56 38,300 175,000 7,175 3,250 20
70 15 292.5 161,000 310,000 40
84 0 1,100 1,565 37.5 7.5
98 0 1,050 1,225 780 90
8
y = -0.0183x + 6.8186
7 1) . 2
R = 0.8157
0
6 ..
5 -l
Z
E0 4
3 -
o
2 a
1 _
o I T i I
0 50 100 150 200 250
Time (see)

Figure 14. Salmonella survivors (CFU/ g) in ground beef during 60°C isothermal

inactivation.

70

 

Table 15. Salmonella survivors (CF U/ g) in ground beef during 62°C isothermal

inactivation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time Repl Rep2 Rep3 Rep4 Rep5
(sec) avertﬂ average average average average
0 495,000 1,725,000 495,000 775,000 172,500
14 2,325,000 6,250 2,550 587,500 285,000
28 NA 24,000 97,250 1,375,000 1,525
42 75 1 1,025 4,650 307,500 97.5
56 38,300 175,000 7,175 3,250 20
70 15 292.5 161,000 310,000 40
84 0 1,100 1,565 37.5 7.5
98 0 1,050 1,225 780 90
7
0 e y = -0.0268x + 5.2823

6 4 R2 = 0.3722

Z

en

3
2 ‘ e
1 . e
0 I I I i i

0 20 40 60 80 100 120
Time (see)

Figure 15. Salmonella survivors (CFU/ g) in ground beef during 62°C isothermal

inactivation.

71

 

Table 16. Salmonella survivors (CFU/ g) in ground beef during 63 °C isothermal

inactivation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Repl Rep2 Rep3 Rep4 RepS
Time (sec) average average average_ average averﬁage
0 165,000 2,525,000 577,500 2,250,000 1,700,000
10 685,000 950,000 10,250 975,000 21,250
20 NA 975,000 6,450 100,000 950
30 17,250 582,500 310 270,000 9,700
40 NA 625 152.5 11,250 9,800
50 525 46,000 155 308,750 NA
60 375 19,950 210 665 4850
70 60 680 47.5 797.5 2.5
7
3 y = -0.0506x + 5.7967
5 1 8 ’ . R2 = 0.5824
0
5 ..
z 4 i
on
.9.
3 4
2 J
8
1 _.
e
0 T I I
0 20 40 60 80
Time (see)

Figure 16. Salmonella survivors (CFU/ g) in ground beef during 63°C isothermal

inactivation.

72

 

Table 17. Salmonella survivors (CPU/g) in ground beef during 55°C isothermal
inactivation.

 

 

Time Repl Rep2 Rep3
(sec) average average average
0 19,750,000 1.25E+08 9,625,000

 

760 4,950,000 2,300,000 1 ,225,000
1520 812,500 402,500 305,000
2280 107,500 52,500 12,500
3040 25,250 2,375 14,000

 

 

 

 

 

 

 

 

 

 

 

3 800 6,975 1,025 600
4560 700 142.5 165
5320 225 622.5 940

 

 

y=-0.001x+7.1005
7 t R2= 0.9192

 

 

 

O I I I I I
0 1 000 2000 3000 4000 5000 6000

 

Time (see)

Figure 17. Salmonella survivors (CFU/ g) in ground beef during 55°C isothermal
inactivation.

73

Table 18. Salmonella survivors (CFU/ g) in ground beef during 58°C isothermal
inactivation.

 

 

Time Repl Rep2 Rep3
(sec) averag: avergge average
0 2,600,000 12,175,000 4,150,000

 

140 507,500 2,275,000 720,000
280 255,000 760,000 537,500

 

 

 

 

 

 

 

 

 

 

 

420 4,575 43,000 55,500
560 6,175 6,000 16,375
700 877.5 4,775 2,625
840 292.5 1,375 2,537.5
980 160 1,100 160

 

 

y = -0.0044x + 6.591
R2 = 0.9265

 

 

 

0 I I I l I
O 200 400 600 800 1000 1 200

 

Time (see)

Figure 18. Salmonella survivors (CF U/g) in ground beef during 58°C isothermal
inactivation.

74

Table 19. Salmonella survivors (CFU/ g) in ground beef during 60°C isothermal
inactivation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time Repl Rep2 Rep3 Rep4
(sec) average average average average
0 13,000,000 21,250,000 1,925,000 17,125,000
65 875,000 2,152,500 165,000 672,500
130 217,500 220,000 12,275 50,750
195 230,000 15,750 11,250 18,000
260 575 2,450 1,000 5,500
325 3,725 232.5 115 900
390 110 30,500 130 1,175
455 375 12.5 2.5 367.5
8
7 I y = -0.0109x + 6.5655
2
, , R = 0.834
6 ..
5 _
Z
3.” 4‘
3 A
2 .4
1 ~ 0
O
0 I I I I
0 100 200 300 400 500
Time (sec)

Figure 19. Salmonella survivors (CF U/g) in ground beef during 60°C isothermal
inactivation.

75

Table 20. Salmonella survivors (CFU/g) in ground beef during 62°C isothermal

inactivation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time Repl Rep2 Rep3 Rep4 Reps
(sec) average avera e average average average
0 232,500 6,750 3,075,000 3,475,000 2,550,000
30 840,000 59,750 895,000 4,750,000 52,750
60 830,000 6,900 217,500 1,107,500 13,750
90 1 1,300,000 520 24,250 92,500 6,525
120 1,450 615 2,450 142,500 2,925
150 350 25 822.5 1,250 1,900
180 NA 50 1,075 525 5,450
210 850 7.5 2.5 12.5 320
8
7 - = -0.0206x + 6.0972
c R2 = 0.669
Z
en
.2
o
O T l l l
O 50 100 150 200 250
Time (see)

Figure 20. Salmonella survivors (CPU/g) in ground beef during 62°C isothermal
inactivation.

76

Table 21. Salmonella survivors (CFU/ g) in ground beef during 63°C isothermal

inactivation.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time Repl Rep2 Rep3 Rep4 Rep5 Rep6
(sec) average average average average average average
0 6,150,000 10,750,000 3,450,000 747,500 3,350,000 762,500
20 63,750 5,875,000 657,500 3,500 2,510,000 1,825,000
40 1,100,000 2,875,000 141,000 2,250 NA 54,000
60 200,000 667,500 632,500 1 1,500 1 1,600 20,500
80 15,250 395,000 875 525 2,875 1,302.5
100 532.5 765,000 2,200 15 ‘ 8,750 832.5
120 70 405,000 8250 2.5 252.5 287.5
140 NA 6,050 625 NA 395,000 10
8
y = -0.0266x + 6.2227
7 0 2
it ; , R = 0.4832
Z
on
.2.
1 — ° 0
O . .
0 50 100 150
Time (see)

Figure 21. Salmonella survivors (CF U/ g) in ground beef during 63 °C isothermal

inactivation.

77

 

Appendix 7.B: Inoculation veriﬁcation summary data

24). Only the aerobic PetriﬁlmTM data from the core labeled “A” were used. The average

Inoculation levels were veriﬁed as described in section 3.6 (Tables 22, 23, and

of the entire A data set on aerobic PetriﬁhnTM was deemed the initial Salmonella

population. *TNTC indicates plates that where too numerous to count.

Table 22. Summary of inoculation veriﬁcation data for turkey roasts.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample Replicates
St.
1 3 4 5 6 7 8 10 Averagg Dev
Marinade 8.95 8.65 8.95 NA 9.45 9.28 9.53 9.53 9.61 9.61 9.28 0.35
A 6.71 6.90 7.83 6.38 7.56 6.24 7.57 6.82 6.62 6.97 6.96 0.53
B 7.59 7.52 7.63 6.43 7.54 6.02 7.77 6.66 7.16 7.61 7.19 0.61
C 6.79 6.87 7.58 6.89 7.24 6.56 7.14 7.07 7.04 7.15 7.03 0.28
D 6.96 7.12 7.69 6.89 7.67 6.96 8.00 5.89 6.33 6.94 7.04 0.63
E 7.14 7.85 7.82 6.18 7.24 6.15 6.99 7.80 6.95 NA 7.13 0.65
Table 23. Summary of inoculation veriﬁcation data for beef roasts.
Sample Replicates
St.
1 2 3 4 5 7 8 9 Average_ Dev
Marinade 9.31 9.31 9.17 9.17 9.17 9.31 9.04 9.04 9.19 0.11
A 6.11 6.69 6.75 6.99 7.14 6.62 4.70 5.18 6.27 0.89
B 6.45 6.59 6.76 6.99 7.61 6.63 4.70 5.54 6.41 0.90
C 6.25 6.95 8.09 7.10 7.11 6.86 5.54 4.88 6.60 1.01
D 6.44 7.11 6.93 6.33 7.25 6.21 4.70 4.70 6.21 1.00
E 6.05 6.74 6.73 7.19 6.74 7.38 4.88 6.42 6.52 0.78
Table 24. Summary of inoculation veriﬁcation data for pork roasts.
Sample Replicates
St.
1 2 3 4 5 6 Average Dev
Marinade TNTC“ TNTC“ 9.68 9.68 9.68 9.68 9.68 0.00
A 6.63 6.55 4.46 5.71 6.74 7.56 6.27 1.06
B TNTC“ 6.15 4.39 5.87 6.60 7.04 6.01 1.01
C 6.35 6.47 4.63 5.53 6.19 7.26 6.07 0.90
D 6.20 6.25 4.49 6.00 6.83 7.99 6.29 1.14
E 7.12 7.16 5.05 7.33 7.52 7.16 6.89 0.91

 

 

 

 

 

 

 

 

 

 

 

78

 

 

Appendix 7.C: Effect of the roast mass on the initial inoculums

concentration

A linear regression comparing the turkey roast mass to the core Salmonella

population was performed (Figure 22). The mass span of 500 — 1200 g is representative

of the variation in roast mass throughout all experimental trials. No signiﬁcant

relationship between the roast mass, and the inoculation level at the core of the roast was

found (P=0.76). This conclusion indicates that using a mean value as the initial core

inoculation level is appropriate to determine the log reductions occurred during cooking.

 

8.00

.‘1
01
C?

9’
01
C?

 

log CF U/g
\l
O
O

 

y = -0.0003x + 7.1644
R2 = 0.0121

 

 

 

6.00
500

Core Salmonella concentration

600

700

800

l

900

1

1000 1100 1200 1300

Mass of Turkey Roast (9)

Figure 22. Whole muscle turkey roast weight (g) vs. core Salmonella concentration (log

CPU/g).

79

Appendix 7.D: Verification of coring method following cooking

Large roasts (500-1000 g) were cooked to a calculated log reduction in a pilot-
scale oven. In order to achieve the targeted log reduction, and stop carry-over cooking
after the roast is removed from the oven, it needs to be cooled quickly. It was impractical
to cool an entire roast quickly enough with an ice bath to stop heating at the center.

A rapid cooling method was needed. The method proposed was to core the sample
after it had been cooked to its targeted end point, aseptically cut off the top and bottom
surfaces of the sample core, and then submerge the sample into chilled peptone water.
Before utilizing the alternative method, it had to be validated that the mechanical pressure
of coring a hot roast was not pushing Salmonella from the sample. The method was
veriﬁed by sampling the turkey purge following cooking (when it was sterile), before and

after coring the sample. The results in CFU/ g are shown in Table 25.

Table 25. Validation of the post-cook sampling method, coring for whole muscle roasts.
Purge from the cook-in bag was collected immediately after cooking Salmonella
inoculated whole muscle roasts in a pilot-scale moist air convection oven and again after
coring the whole muscle roasts with a sterile stainless steel corer.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Plate Average
Count Plate Plate Amount
A Count Count Plated Initial Log
Sample Media (CFU/g) B (CFlm (CFU/g) (ml) Dilution CFU/g_
PF 0 0 0 1 1 <04
4 Plate 0 0 0 10 1 <1.4
PF 0 0 0 l 1 <0.4
Purge 5 Plate 0 0 0 10 1 <1.4
Before PF 0 0 0 1 1 <0.4
Coring 6 Plate 0 0 0 10 1 <1 .4
PF 0 0 0 1 1 <04
4 Plate 0 0 0 10 1 <1 .4
PF 0 0 0 l 1 <0.4
Purge 5 Plate 0 0 0 10 1 <1 .4
After PF 0 0 0 1 1 <0.4
Coring 6 Plate 0 0 0 10 I <1 .4

 

 

80

The recovered purge data veriﬁes that the coring method did not change the
amount of Salmonella in the purge. The result demonstrated that the coring method was
not forcing Salmonella out of the sample; therefore, the coring method was used to

rapidly chill the samples core.

81

Appendix 7.B: Alternative inoculation veriﬁcation for park

An alternative method for verifying the initial Salmonella inoculation level in
pork prior to cooking was tested in order to attempt an improved direct measure of the
initial level of Salmonella contamination for each roast, rather than using a mean value
(as was done for turkey and beef). When the electrosurgical knife was used to determine
a mean initial inoculation value for pork roasts (section 3.6), the standard deviation was
very large (1.06 log CFU/g); therefore, a biopsy method was investigated.

The alternative, biopsy method consisted of determining the initial Salmonella
population within the pork roasts by extracting ﬁve 0.02 g samples from the near center
of each roast. The extraction was performed with an E-Z core single action biopsy needle,
2.7 mm x 9 cm (Products Group International, Lyons, CO). Before extracting the samples
using the biopsy needle, the portion of the surface to be penetrated (4 cm diameter) was
ﬂamed using the electrosurgical knife, in order to prevent external contamination of the
interior of the roast. After the roasts were inoculated via vacuum tumbling, the biopsy
needle was sterilized by immersing it in 250 ml of ethyl alcohol (Cecon Labs, Inc, King
of Prussia, PA) for 20 s. The sterilized needle then was placed into 250 ml of sterile 0.1%
buffered peptone water, in order to remove any excess ethanol.

The biopsy was performed ﬁve consecutive times per roast, aseptically removing

each sample independently from the needle, and collecting them into a single Whirl-

PakTM bag (N asco, Fort Atkinson, WI). Samples were then diluted, stomached, plated on

aerobic PetriﬁlmsTM , and incubated for 24 h at 37°C before enumeration. The biopsy

needle was sterilized after each roast, as mentioned previously with ethyl alcohol.

82

After the samples were cooked, plated, and enumerated, the average and standard
deviation of the error (observed - predicted) of the biopsy and electrosurgical knife (ESU)

methods were compared (Table 26).

Table 26. Post-cook, comparison of the lethality error (observed - predicted) for the
biopsy and electrosurgical unit methods following pork inoculation. The standard
deviations of the lethality error post cook were compared to determine the most
repeatable process.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Error of Error of
Biopsy Standard ESU Standard
Cook Method Deviation method Deviation
-0.71329 -0.32815
93.3C In-bag, High -0.0635 -2.05474
humidity -1.47631 0.707162 -1.74199 0.919953
-1.85408 -2.30453
-0.41037 -1.18206
93.3C In-bag, No Humidity 0.611733 1.238897 -0.74445 0.804703
93.3C No Bag, High -0.26895 -0.82714
humidity -0.47897 0.148505 -2.32572 1.059655
-0.66655 0.002974
140F 90m, 155F 90m, 170 0.031757 -0.86167
In-bag, High Humidity 0.259171 0.482406 -1.0263 0.552889
-1 .12742 0.451134
140F 90m, 155F 90m, 170 -1.4984l 0.816101
In-bag, No Humidity -1.528l3 0.223264 0.955023 0.260258
0.848762 -0.97408
140F 90m, 155F 90m, 170F -0.07181 -1.89465
No Bag, High Humidity 0.223681 0.470015 -1 .59916 0.470015
140F 90m, 155F 90m, 170F 0.446033 -0.35562
No Bag, Changing Humidity -0.7282 -0.73813
(No, Med, High) -1.31046 0.894717 -0.94397 0.298564

 

Based on the comparison of the standard deviations for the biopsy and

 

electrosurgical knife methods, it was determined that the methods resulted in similar
variability. Due to the fact that the biopsy method did not reduce experimental variability

by improving the prediction of the initial level of Salmonella contamination it was not

83

used for analysis. The method using the average core value from the electrosurgical unit
sampling method was used in order to maintain inoculation validation uniformity

between species.

84

Appendix 7.F: Summarized cook schedules and spreadsheets

Using Excel, master spreadsheets for each species were made in order to compile
a summary of each experimental mm (Tables 27-29). These sheets do not include all of
the raw data, but summarize it with averages, making the data easier to interpret. Column
labels were used help compare the cooks to each other. This allowed for the

determination of similar trends. An explanation to the table labels is below.

DE: Indicates the date that they experiment was performed.

£9913: Identiﬁes the cook schedule performed.

Target End Point: Identiﬁes when the sample was removed from the oven - either at an
end point temperature of 71 .1°C or a speciﬁed log reduction. Species were removed at
different log reductions; therefore, the targeted log reductions for each sample are
labeled.

Weight (g): This column indicates the initial weight of the raw, irradiated, thawed roast

before marination or cooking.
Media: Speciﬁes the growth media used. The data from the aerobic PetriﬁlmsTM (PF)
were the only data used in analysis. The modiﬁed Triptic Soy Agar plates (plates) were

only used to assure that Salmonella was being recovering on the PetriﬁlmsTM. In all

cases, the plates only contained colonies with the characteristics of Salmonella. Samples
plated on modiﬁed Triptic Soy Agar plates were not used in the analysis, because it was

difﬁcult to consistently recover cells, because the level of detection (25 CFU/g), was too

high; therefore, the aerobic PetriﬁlmsTM were used.

85

M This indicates a number label for the sample. All samples with a label of 1 or 2
were cooked to an end point temperature of 71 .1°C. Samples labeled 4, 5 and 6 were
cooked to targeted end point lethality.

Total time (hr): This is the time in hours, of the cook, from placing the roasts in the oven
to when they were completely cooled.

Marinade inoculum level 102( CFU/ g): The amount of Salmonella within the salt,

 

phosphate marinade.

Initial inoculum level 102(CFU/g): The initial concentration of Salmonella within the core
of the roast before cooking. This was determined using the procedure explained in section
3.6.

Survivors 102(CFU/g): The amount of recovered cells following enumeration and
cooking.

Observed log reductions: The difference between the initial level of Salmonella
concentration, and the enumerated cells following cooking.

Predicted log reductions: The log reductions estimated in real-time via the log-linear

 

model, using isothermal D and 2 values.
Max sub-lethal histom: The total sub-lethal history of the Salmonella cells during
cooking. A detailed explanation of the quantiﬁcation of this value is provided in section

3.11.2.

86

 

Hugo Na. 2523. Sam" €83 mung 95:55 0». w: ooow moronic?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

la 29%|
2.2%" 2.9»: 33:8,” 585:8 2.3.2522 magma».
moor m3 (<12: 25a 5855. r95. r22 msméa 033% row row 2.580.
can margin wow: E 32:» 3an 2.5 282993: romAnmcﬁv Pom 925 woacomoa 269.30: 3:: 5
Sun 5 En. Em: 2:555.
2598 2.. 0 m8.” 2.. 2 mm: «.92 98 A933 v93 maven
2._ a mo: 2” N ”.2 a? 98 A933 v93 33%.
@338“ So am a: 2U A wt 0.3 90o NS #2 a» 8.2.3
So Em 82a 2.. m N2 9% 93 ASE v93 92 8mg
Ho .om 3o.» 2” a wt 9.3 93 2.8 92 93 89;
8.8 2: En. 2° 25.83.
258m 2.. n a: 2.. 2 NE «.2 93 A; v93 $9.3
2.2 n 35 23... NE «.2 an .3 MS
2._ 0 8mm 2.. N NE «.2 93 A; v93 22.2
2.2 o 83 23“ NE «.2 98 A: v93
$3.28 So am $2 2.. a N8 0.8 as who go 93 :5“
So am 2: 2.. m Now 0.8 98 PS :6 9% Beam
So 8m 23 2.. a New 98 98 A.» v92 30 No.3
8mm Zo mum. Em: 2555
$38 2.. 0 M23 2.. 2 NE 30 93 AA v3.2 38.:
2.2 0 M26 23“ N3 9% 9% ATS v93 38.:
2._ n :85 2” N V; «be 93 A.» v92 33.8
2._ n :86 2%.. w; go 9% ATS v93 38.8
2288 2.30m 23 2.. A am who 93 A... v93 3o 39%
So _om ES 2% 2.8 who 9% ATS v93 93 can.
So Em 85 2U m _.2 who 98 A; v93 9.: :93
So 6m 83 2% _.2 «be 9% ATS VaS 92 :93
.5 5m ﬁt 2.. a :5 who 93 A; v93 .3» .33
so am if 2% Ta go 98 ATE v98 8% SUN

 

87

 

2.3.0 N2 003.22.

«2 :33

232.8

2323

«9.523

«8.. C? .= w»«. Em:
2:335.

«82. c... .a mum. 2o
2:583.

 

q . .. 0
q . .. 0
Ho .o«
u... .o«
a... _o«
a... .o«
a... .o«
q... .o«

2.. n
2.. o
2.. o
2.. o
No.3
..Som
2:3
No.3
Scam

 

q... .o«

 

3....
23..
.3...
3w.
«-.~
«N~.~

3e

30

«3b
«3b
«5...
«.Nm.
«mm.
««P.
m. ..u
m. ..w
2..;
..««

2“.
2..
2..
23a
2”
2%
2..
2%

2”
298
20
22a
2..
2%
2..
220
2..
298

 

 

@ON'JIMh-BN~

O‘GMMA-bNN—u—

 

N3
«.3
N..~
N..~
NS
~..~
N..«
N..«

«.2
«.«.
«a.
«.2
P...
P...
Po.
8.....
Po.
B..e.

 

0.3
v.3
«.AN
«.3
«.3
«.3
«.AN
«.3

«.««
«.3
«.3
«.3
can
v.3
on...
on...
33
PM...

 

«.3
«.«N
«.3
«.«N
«.3
«.«N
«.3
«.«N

«.3
«.«N
«.3
«.«N
«.3
«.«N
«.3
«.«N
«.3
«.«N

 

«.2.
«.2.

A...
A. .A
NNN
A. .A
~.«.
A. .A

A...
A...
ex...
A. .2
who
..««
PB
«.3

A...
A. .A

 

«.3
«.3
v93
v«.«~
92.
v«.«~
P3
v«.«~

v93
v«.«~
«.3
v«.«~
P3
Pf
NA...
«.3
v93
v«.«~

 

. .«A...
~39;
«.2.
«.3
SN.
..N.
a. . N
2.5

.«3.««
.«3.««
33.3
33.3
8...
2o
8:.
2....
«.«u
«.««

 

 

~53
«.93
No.3».
~33»
ENS
BPS

3N. .c
3N3
«3.3
«3.3
«2.3
3 . .2.

 

 

88

 

 

, HmEo Samoa-maﬁa

 

 

 

 

89

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mNavaNZo mum.
LEmNN-m-EENQ- : - ...: . . .- .. -- :3... N.
----- ..----.:e.-. - .N_-.._-n----:NNN 2%- , -._----N.Nm ...mNN-N.-.;. ...N-NM ...AE ...wN.N.N-:--~NNN-.._-N -
.. ,. .N._N_--.N-N-.---N-NN.N-N._NNN..-.-.--NNNN-ﬂ-NN-N- NNN . 2N: VNNN NNNN,.NN.-_---.-
...NNNNNNNN..---:-3 : --.NN-..Nm--:NNNNN_NNN-::- N. --N-..NN - NNN N-NN ANVNNN -N-N.--.---N-_-N-.NN-
HEN-um 0.3.» 280. :.- m. www.- - amNa; . - :93 . . A: Noun 9.3 58.:
..- . -i NN_NN_NNN_ -N,_.NNN ;.-.-- N N..:.N.N - - N-NN N-NN ANN. - -wN...N-N,--::N-.NN. :iNNN-NN
maNuCPzO mum.
OENNWIENNNEQ
: 9.02333 -: - . .- ‘ . .. - : ---
.NNNONNBM----:-._--. -N.-_-.-_.-n:-:--N-N-N...N. :N_-N.NN-.NN:NN New: 2 vNNNNILSNNN
---.:--.-:..-.-.---.-:.-.:--,---:-.NN-_-.-_:niiiN-N-N-MN-NNNNNN-Eii-..LB «.3 3N :----.-M.....NN- v93 NoNN.N-N..T-;--:--
:--:--. -- N-_-N_:n-..:-..SNNN wm.--.-------N-iwais: ..-:mNN-w-.-:-----NN-N.a. .1.-.-.--..---m-N.N.-----.-_---..wN.N-.MNN--.Lbopﬂ- --:-.--.---
:2: N-E-m- 15%-..N. N._NNN:---.-:----- ﬁw-ﬁi; 3N.- - Nam- ...- ...... 1A....» 7.... v93 38.2
NNSNNSN - ; -- _ $2.0m MN: NNm--.-..-E.N NB _,_.....-:N-.-N.N. :---.-.--.-..N...N..N. ..N-No. ...--.-:MN-N ----,-:-.N.N.NN---N.N.N.NN.
$8.0m . ---_.N§.-N 3% - SN .-..-m.-N-.N..---: :-..-N-.N~. ---..._...._.-,_.A-.N..-N .,--.-.-...-----wn.N--.§------N..NN ,-!N.NN..NN-N-
N-Som - NNNN 2N- .-..,.N.---N-NN-:- --N-.-N.N. 398.. -- - ..--.-A..N .-...,--.v.N.,.wNN.-:-.--N.-.N.N - N-NN-NN.
-- - N-.o-_-0m .--NNN.N NN_NNN.-----:..--.-.---N.NN.--.-.-_--..-3N. arm-N73 -..mN-NN .-.-..--..----wE-N-sii-m-Nd --.N-NN...N-N-
- --.:.:: z----:--N.-o-._mm..-::imwo-wmzt a PB «.3 98 A1 v93 9%- mow-ow
Noam 3o Ema PNN «.mw 93 A: v93 93 398.

 

 

 

HmEm Mm. womm 88H mung—h wraﬂ man—85 om 9: Sex 8:85am.

QSDSSS

S\e\Noow

QQNSSS

Sh§oow

 

8.8 N:
mum.
Ema
€535

ew.w0 5 mum.
20 :53th

 

Q: 0
.2.— 0
q: 0
3.. O
Sh .om
uh .om
Sh .om
uh 5m
Sh 8m
Sh Em

2.. O
q: 0
q: n
q: 0
uh 5m
Sh .om
Sh 8m
Sh 5m
Sh .om
Sh 5m

 

 

Ema
3..
Eur“
2.-
ENE"
3..
330
E...
38o
3..

32¢
3.-
33a
2.-
2&0
3..
23¢
2..
38a
3...

 

 

qwhho
QNSQS
hooho
hooho
Smmho
Smmho
em _ .eo
dB .eo
Shmho
Shuho

Schho
SoS.wo
Swwbo
Swuho
SShS
macho
S333
Segue
Soho
QSobo

 

N3
Noe
~hq
“he
:3
73
.ho
_.ho
_.hu
Thu

Nhu
whw
Nhh
Nwh
N:
N:
P:
N:
N:
P:

 

ehw
ehm
eh—
ehm
She
She
She
She
She
She

ehh
eho
ehh
eho
e. 5
ehq
e. 3
e.:
e.:
ehq

 

Shh
Sb.»
Shh
She
Shh
qu
Shh
Sb.»
Shh
qu

Shh
She
Shh
She
Shh
She
Shh
She
Shh
qu

 

A_ .h

A.h
A_ .h
ohc
hhm
uhc
A_ .h
N. S
h; _
wbS

A_h
cho
A..h

Ah
Nhh
wh_
A—h
73
A— .h

Ah

 

vShh
VS..:
vShh
93
NS
th
vShh
hhe
uhw
ub—

vShh
She
vShh
VSh—
h. S
NhS
VShh
hem
vShh
VS..:

 

www.mm
www.mw
393
Euhw
hbm
hbu
Seq
See
uh—
uh—

Nmm. _h
Nam. _h
aha-SN
.3qu
Sea
Mew
hhw
hhw
h..:
h..:

 

eS.eo
eS.eS
Suhc
Suhc
— 3:3
_ 3.3

SShe
Shhe
Nuhhc
Nuhhc
Buhm
Buhm

 

90

Huh—o mm. 005333

E _ :Noom

SDSRSSS

 

 

8.8 20
gm.
Ems

25.35

 

 

Q: 0
Q: Q
Q: 0
q: 0
Sh .om
Sh 3m
Sh 8m
uh .om
Sh .om
Sh _om

 

e58
EN
Suﬁ
2..
Ema
3..
Eva
EN
398
3..

 

 

Su. S
Sw. S
Smho
.3th
Shake
mhqbo
mhchc
whoho
ehebo
Qhﬂwo

 

few
Few
rem
few
ES—
fu—
ah—
_.h_
Phw
hhu

 

58,535
5.85:3?
58582...
585565
ehe
e.we
ewe
ehe
ehe
ewe

 

Shh
qu
Shh
She
Shh
SNQ
Shh
Sb.»
Shh
She

 

 

A_ .h

Ah
A_ .h

Ah
A .h

Ah
A .h
hum
_.ho
me

 

vShh
VS..:
VShh
VS..:
VShh
VS..:
VShh

hhe

Shh

wee

 

Nhuhh
Nhuhh
NSobw
NSobw
Sew
Sew
hQS
hQS
Sew
Sm;

 

Suh—
Suh—
Shh—
Shhh
hNeNS
huebS

 

91

 

Hugo mm. 00:32am

«\SDcom

_ o\ _ Swoom

«\Nwhoom

ho:h\moom

 

mac Cw. E
mum. 3m:
$555

mnow C? 5
mum. Zo
:5:me

 

q: 0
q: 0
3.. n
q: 0
Sh .om
Sh 8m
Sh .om
uh 5m
Sh 3m
Sh 3m

4: 0
q: 0
.2.— 0
q: 0
SS Em
Sh Em
Sh Em
SS 8m
Sh 5m
Sh 8m

 

Ema
3..
28a
2..
Ema
Eu
23a
3.,
330
Sn

Ema
3..
$58
3...
Ema
Eu
Ema
2.,
Spa
3”

 

O‘O‘MMAkNN—u—o

amMMAANNn—u—n

 

thhc
thho
Shwho
Shuho
Swmbc
www.mc
Sumac
Sqwho
hS _ .wo
hm _ .8

San»:
Scwhc
Suqbo
qubo
mhwho
whwho
Sacha
Smoho
$98
$98

 

..wu
_.wu
Nhh
Nwh
...3
f3
_.mh
fwh

._ ho

..So

Shh
Sbc
M:
M:
and
mud
wa
wwq
whw
whw

 

«.3
who
«.3
who
5.:
3.3
3.3
3.3
5.3
8.3

whw
whw
whw
whw
wh—
wh—
wh—
wh—
wh—
wh—

 

Shh
qu
Shh
Sb.»
Shh
qu
Shh
qu
Shh
qu

Shh
qu
Shh
Sb.»
Shh
qu
Shh
qu
Shh
Sb.»

 

A_ h
Ah
A. .h
Ah
A_ .h
Ah
A. .h
A.h
A_ .h
A.h

A_.h
Ah
Ahh
Ah
A_.h
Ah
A_.h
Ah
A_.h
F5

 

vShh
VS..:
VShh
VS..:
VShh
VS..:
vShh
VS..:
vShh
VS..:

vShh
VS..:
vShh
VS..:
VShh
v9.2
vShh
VSQ—
VShh

Shq

 

NowhN
mowhw
hho.wo
hho.wo
hhh
hhh
Shw
ShN
98
Mon

Nh.wm
Nh.wm
SN. 5
3N. 5
hhS
hhS
h.wu
h.ww
hhu
hhS

 

3::
3::
_S_.NS
_S_.NS
Embh
_hS.Oh

Sﬂhh
92.3
393
$93
hqwhh
hqwhh

 

92

 

w.wo—a wm. Conan—com

«\ﬁwoom

ho\ h .twoom

_o\ h S\woom

 

what Cw. Zo
mum. EB.
:EEEQ

mum—V Cw. Zo
mum.
Orngm
hhh—BEE

 

3h 0
2h O
3h 0
2h 0
Sh how
Sh hem
Sh .om
Sh hem
Sh ham
Sh hom

2h 0
3h 0
3h 0
3h 0
MS hom
Sh hom
Sh ham
Sh hem
SS hem
Sh hem

 

Ema
hum.
Ema
3..
298
3..
23¢
3..
Emma
2..

3m.»
2..
23a
3..
Ema
Eu
Ema
Eu
Ema
3..

 

O‘OUIMAANN—n—a

@GMU’IA-bNNu—u—n

 

Swhho
Swhho
Shwho
Shwho
Sum. ho
Stho
ShS.wo
ShS.wo
mmoho
«moho

23.3
23.8
29,8
298
$28
«2.8
395
398
5.8
$3

 

who
who
who
who
h.wS
h.wS
hhS
hhS
whh
whh

uhq
whq
whh
whh
uhw
uhw
uhS
whS
Pow
Pow

 

w.wm
w.wo
w.wS
w.wu
w.wo
w.wo
w.wo
w.wo
w.wo
w.wo

w.w—

w.wh

whh

w.wh

w.wh

 

Shh
S.wq
Shh
S.w.~
Shh
S.wq
Shh
S.wq
Shh
h.ww

Shh
S.wq
Shh
S.w.~
Shh
S.wq
Shh
S.ww
Shh
S.wq

 

Ah .h

Ah
A .h

Ah
A .h
h.ww
A .h
h.ww
A .h
oho

Z>
owo
Z>
A.h
A— h
Ah
w.oo
w.om
Ah .h
h.ww

 

VShh
VS..:
VShh
v9:
vShh

h.ww
VShh

h.wo
vShh

Sh.»

Z>
Shq
Z>
VS..:
VShh
VS..:
hhh
h. _w
VShh
h.hw

 

hhShw
hhShq
wwh .Sw
wwh .Sw
w. h u
q. hm
w.ww
w.ww
ShS
ShS

SSohh
Smohh
Shhhw
Shhhw
hho
hho
hhw
hhw
hhu
hhw

 

hmmhm
hmwhm
:Shm
:Shm
hSqu
hSwwS

wShwq
wuhwq
wahS
wahS
wwohh
wwohh

 

 

93

Hmhho ww. won» nommh "68mm whoa.“ 9585 o». m: noow moraazhom.

 

 

 

 

Q hh\woow

Qwﬂwoow

m\wm\woow

 

8mm 5
Em.
Em:

Eséq

wwhn h:
mum.

20
$53th

wwho
Zo wmm.

h.ww:
3:855.

 

3h 0
3h 0
9o how
Po ham
Po hom

3h 0
.2h 0
mo 8m
9o how
uh hem

3h 0
3h 0
Po hom
Po how
oh how

 

3H
3“.
3...
3...
3...

3..
3..
3“.
3...
3U

3U.
3H
3..
3U.
3U

 

O‘M-BNI— OMAN—I

ONUIANH

 

hhwho
uwhho
thhw
wmmwh
hmﬂwu

hhwhw
hhwhq
quhm
hm h .hS
Shu. 3

hShho
hSqu
Sww. hw
hawhw
hwwhm

 

h.ww
h.ww
hhw
hhw
hhw

NE
N8
h.ww
hhw
h.ww

hhw
hhw
h.wo
hhw
h.wo

 

whS
whS
whS
whm
whu

w.wq
w.wq
w.wq
w.wq
w.wq

wow
w.wq
w.wq
w.wq
w.wq

 

oho
th
whw
ohm
hhm

th
th
hho
h.ww
Pow

th
th
th
h . h m
_ .wh

 

Saw
V S.wq
wow
w.ww
Sow

V S.wq
v S.wq
Sh.»
h.hh
w.wh

v S.wq
V S.wq
v S.wu
Pow
Vow

 

huqhh
Soohq
wwS
w.wh
mum

:wShh
huwhh
whq
w.wS
whw

mwwww
hmmhm
whw
h.ww
w.wo

 

wmhw
ww.:
wSho
whhw
how.wm

hSmwS
hmwhS
wohhq
woohw

whwhm

hoShw
hohwo
h hwhu

wth
hoohw

 

94

 

Hwhhm ww. 003963

«\wowoow

Q h Qwoow

w\w\woow

w\ h \woow

 

man Cu
5 mum.
Em:
$5330.

mac Cw.
E mum.
Zo
3:355.

mac :9
Zo mum.

Em:
:53th

m5. Cu.
20 mum.
03533
$5535.

 

 

Eh 0
2h 0
Po hom
Po hom
9o how

2h 0
3h 0
Po how
Po hom
Po 3m

Eh O
.2.— 0
Po hem
Po Em
Po .om

3h 0
3h 0
uh how
9o hem
Po 3

 

3.
3..
3.
3n
3..

3..
3.
3..
3.
3..

.3.
3..
3..
3”
3H

3..
3..
3“.
3..
3.

 

GMAN—ﬁ ONUﬁ-BNI—

GMbN—

QUI-bNﬂ

 

hwo.oh
hSS. ho
Smwhm
hmmww
hwwhS

hwqwo
hwo.om
Swmhh
Sombo
hSoh h

wwShS
wwwwm
huwww
hSh .mw
hwhho

hSh .wS
hSShq
Sh hhh
hSSho
Shﬂoo

 

w.wh
w.wh
hhh
h.ww
h.hw

h.wS
hwm
w.wh
whw
“h.ww

h.ww
h.ww
h.ww
h.wh
h.ww

whw
whw
whS
whw
whw

 

who
who
who
who
who

whS
whS
whS
whS
whS

whw
whw
whw
whw
whw

w.ww
w.wh
«.ww
w.ww
w.wu

 

th
th
w.wh
www
w.ww

th
th

hho
h.ho

th
th
who
h.ww
w.oh

th
th
h.wh
ohm
w.wS

 

v S.wq
v S.w.u
whw
whw
whS

v S.wq
v S.wq
w.wS
h.ww
h.ww

v S.wq
V S.wq
whw
h.ww
h.ww

v S.w.~
v S.wq
..uS
Shw
whw

 

hhhwhw
Swuww
whw
whw
whw

wwmwhq
hhSwhw
w.wo
whw
whS

h hwwhw
Sqwow
whw
who
whw

mqohm
Soma—
_.o_
hhS
w.wm

 

hwhhu
hwwhw
hmhho
hwhwq
hho.w_

wwwho
h h w. 3
hwwhq
Sumow
hwoho

hwwhw
hwmwo
hSohw
huwww
hho.ww

wwqhh
uwhho
wSuhq
wwwhh
uohhh

 

 

95

8. BIBLOGRAPHY

AMI Foundation. 2009. AMI Foundation Facts and Figures Process Lethality
Spreadsheet (Microsoft Excel ﬁle).
http://www.amif.org/ht/d/sp/i/26870/pid/26870AM]. Accessed 12-3-2009.

Association of Ofﬁcial Analytical Chemists (AOAC). 1990. Ofﬁcial methods of analysis.
Association of Official Analytical Chemists, Arlington, Va.

Blankenship LC. 1978. Survival of a Salmonella-Typhimurium experimental contaminant
during cooking of beef roasts. Applied and Environmental Microbiology
35(6):]160-1165.

Breslin TJ, Marks BP, Hall NO, Booren AM & Orta-Ramirez A. 2007. Migration of
Salmonella into intact turkey breast muscle without intervention of external
factors. Michigan State University.

Bubzy J C, Frenzen PD. 2001 .Product Liability and Microbial F oodbome Illness. Food
and Rural Economics Division, Economic Research Service, US Department of
Agriculture. Agriculture Economic Report No.799.

Carlson TR. 2002. Effect of Water Activity and Humidity on the Thermal Inactivation of
Salmonella during the Heating of Meat. Food Science and Human Nutrition East
Lansing: Michigan State University. p. 111.

Carlson TR, Marks BP, Booren AM, Ryser ET & Orta-Ramirez A. 2005. Effect of water
activity on thermal inactivation of Salmonella in ground turkey. Journal of Food
Science 70(7):M363-M366.

CDC. 2008a. Foodbome Diseases Active Surveillance Network (FoodNET): FoodNET
Surveillance Final Report 2005. Atlanta, GA: US Department of Health and

Human Services, CDC.
http://www.ch.gov/foodnet/annual/2005/2005_AR_Report.pdf. Accessed 10-29-

2009.

CDC. 2008b. Salmonellosis.
http://www.ch.gov/print.do?url=http%3A//www.cdc.gov/nczved/dfbmd/disease_
listing/salmonellosis_gi.html. Accessed 1-16-2009.

96

Corradini MG & Peleg M. 2004. Demonstration of the applicability of the Weibull-log-
logistic survival model to the isothermal and nonisothermal inactivation of
Escherichia coli K-12 MG1655. Journal of Food Protection 67(1 1):2617-2621.

Corradini MG & Peleg M. 2009. Dynamic model of heat inactivation kinetics for
bacterial adaptation. Applied and Environmental Microbiology 75(8):2590-2597.

Doyle M. 2001. Food Microbiology Fundamentals and Frontiers, 2nd ed. Washington,
DC: ASM Press.

Duffy EA, Belk KE, Sofos JN, Bellinger GR, Pape A & Smith GC. 2001. Extent of
microbial contamination in United States pork retail products. Journal of Food
Protection 64(2): 172-178.

Dykes GA & Moorhead SM. 2001. Survival of three Salmonella serotypes on beef
trimmings during simulated commercial freezing and frozen storage. Journal of
Food Safety 21 (2):87-96.

FDA. 2009a. Peanut Butter and other Peanut Containing Products Recall List
Information.

httgﬂwww.accessdata.fdaw/scripts/peanutbutterrecall/index.cfm#FruitandVege
table. Accessed 11-23-2009.

FDA. 2009b. Pistachios and other Pistachio Containing Products Recall List.
http://www.accessdata.fdagov/scripts/piﬂchiorecall/index.cﬁn. Accessed 1 1-23-
2009.

FDA. 2001. Analysis and Evaluation of Preventive Control Measures for the Control and
Reduction/Elimination of Microbial Hazards on Fresh and F resh-Cut Produce,
Chapter IV. Outbreaks Associated with Fresh and Fresh-Cut Produce. Incidence,
Growth, and Survival of Pathogens in Fresh and F resh-Cut Produce.
http://www.fda.gov/Food/ScienceResearch/ResearchAreas/SafePracticesforFoodP
rocesses/ucm091265.htm. Accessed 1 1-23-2009.

FSIS-USDA. 1999. Appendix A Compliance Guidelines for Meeting Lethality
Performance Standards For Certain Meat And Poultry Products.
http://www.fsis.usda.gov/oa/fr/95033F-a.htm. Accessed 9-25-2009.

F SIS-USDA. 2005. Appendix A Guidance on Relative Humidity and Time/Temperature
for Cooking/Heating and Applicability to Production of Other Ready-to-Eat Meat
and Poultry Products. http:l/www.fsis.usda.gov/OPPDE/rdad/FRPubs/95-
033F/Appendix_A_guidance_95-O33F.pdf. Accessed 2-9-2009.

97

FSIS-USDA. 1999. Lethality and Stabilization Performance standards for certain Meat
and Poultry Products: Technical Paper. Fed Reg 64:(3) 732-49
http://www.fsis.usda.gov/OPPDE/rdad/FRPubs/95-033F.htm. Accessed 1-20-
2009.

FSIS-USDA. 2005. Time temperature table for cooking ready-to-eat poultry products.
http://www.fsis.usda. gov/OPPDE/rdad/FSISNotices/RTE_Poultry_Tables.pdf.
Accessed 11-25-2009.

F SIS-USDA. 2005. Use of Microbial Pathogen Computer Modeling in HACCP Plans.
http://www.fsis.usda.gov/regulations/notice_25-05/default.asp. Accessed 10-29-
2009.

Gill CO & Penney N. 1982. Bacterial penetration of muscle tissue. Journal of Food
Science 47(2):690-691.

Goodfellow SJ & Brown WL. 1978. Fate of Salmonella inoculated into beef for cooking.
Journal of Food Protection 41(8):598-605.

Gupta LK, Kalra MS, Ajit S & Balwant S. 1983. Penetration of bacteria into muscles of
goat, pork and poultry meat. Journal of Food Science and Technology, India
20(1):]3-16.

James JM & Calkins CR. 2008. The inﬂuence of cooking rate and holding time on beef
chuck and round ﬂavor. Meat Science 78(4):429-437.

Jones SL, Marks BP, Ryser ET, and Booren AM. 2009. Effect of sublethal heating on
Salmonella lethality during slow-cooking of turkey and beef products. Journal
article to be submitted to Journal of Food Protection.

J uneja VK. 2007. Thermal inactivation of Salmonella spp. in ground chicken breast or
thigh meat. International Journal of Food Science and Technology 42(12): 1443-
1448.

J uneja VK & Eblen BS. 2000. Heat inactivation of Salmonella typhimurium DT104 in
beef as affected by fat content. Letters in Applied Microbiology 30(6):461-467.

98

Juneja VK, Eblen BS & Marks HM. 2001. Modeling non-linear survival curves to
calculate thermal inactivation of Salmonella in poultry of different fat levels.
International Journal of Food Microbiology 70(1-2):37-51.

Kaur A, Takhar PS, Smith DM, Mann J E & Brashears MM. 2008. Fractional differential
equations based modeling of microbial survival and growth curves: Model

development and experimental validation. Journal of Food Science 73(8):E403-
E414.

Kegode RB, Doetkott DK, Khaitsa ML & Wesley IV. 2008. Occurrence of
Campylobacter species, Salmonella species and generic Escherichia coli in meat

products from retail outlets in the Fargo metropolitan area. Journal of Food Safety
28(1):111-125.

Kim KT, Murano EA & Olson DG. 1994. Heating and storage conditions affect survival
and recovery of Listeria monocytogenes in ground pork. Journal of Food Science
59(1):30-&.

Leguerinel I, Spegagne 1, Convert O, Coroller L & Mafart P. 2007. Quantifying the
effects of heating temperature, and combined effects of heating medium pH and

recovery medium pH on the heat resistance of Salmonella lyphimurium.
International Journal of Food Microbiology 116(1):88-95.

Luchansky JB, Porto-Fett ACS, Shoyer B, Phebus RK, Thippareddi H & Call JE. 2009.
Thermal inactivation of Escherichia coli 0157:H7 in blade-tenderized beef steaks

cooked on a commercial open-ﬂame gas grill. Journal of Food Protection
72(7): 1404-141 1.

Mackey BM & Derrick CM. 1986. Elevation of the heat resistance of Salmonella
typhimurium by sublethal heat shock. Journal of Applied Bacteriology 61(5):389-
393.

Mann J E & Brashears MM. 2007 . Contribution of humidity to the lethality of surface-
attached heat-resistant Salmonella during the thermal processing of cooked ready-
to-eat roast beef. Journal of Food Protection 70(3):762-765.

Marks B, Tenorio-Bemal I, Breslin T, Ryser E & Booren A. 2009. Validating a
Salmonella thermal lethality model that accounts for Prior sublethal Injury during
commercial meat and poultry cooking processes. American Society of
Agricultural and Biological Engineers (ASABE). ASABE Paper No. 096277.
Reno, Nevada.

99

Mattick KL, Legan JD, Humphrey TJ & Peleg M. 2001. Calculating Salmonella
inactivation in nonisothermal heat treatments from isothermal nonlinear survival
curves. Journal of Food Protection 64(5):606-613.

Mead PS, Slutsker L, Dietz V, McCaig LF, Bresee J S, Shapiro C, Griffin PM & Tauxe
RV. 1999. Food-related illness and death in the United States. Emerging
Infectious Diseases 5(5):607-625.

Metris A, George SM, Mackey BM & Baranyi J. 2008. Modeling the variability of
single-cell lag times for Listeria innocua populations after sublethal and lethal
heat treatments. Applied and Environmental Microbiology 74(22):6949-6955.

Mogollon MA, Marks BP, Booren AM, Orta-Ramirez A & Ryser ET. 2009. Effect of
beef product physical structure on Salmonella thermal inactivation. Journal of
Food Science 74(7):M347-M351.

Moss M. 2009. Food Companies Are Placing the Onus for Safety on Consumers. New
York Times. http://www.nytimes.com/2009/05/1S/business/l Singredientshtml.
Accessed 5-15-2009.

Murphy RY, Johnson ER, Duncan LK, Clausen EC, Davis MD & March JA. 2001a. Heat
transfer properties, moisture loss, product yield, and soluble proteins in chicken
breast patties during air convection cooking. Poultry Science 80(4):508-514.

Murphy RY, Johnson ER, Marcy JA & Johnson MG. 2001b. Survival and growth of
Salmonella and Listeria in the chicken breast patties subjected to time and
temperature abuse under varying conditions. Journal of Food Protection 64(1):23-
29.

Murphy RY, Marks BP, Johnson ER & Johnson MG. 1999. Inactivation of Salmonella
and Listeria in ground chicken breast meat during thermal processing. Journal of
Food Protection 62(9):980-985.

Murphy RY, Marks BP, Johnson ER & Johnson MG. 2000. Thermal inactivation kinetics
of Salmonella and Listeria in ground chicken breast meat and liquid medium.
Journal of Food Science 65(4):706-710.

Naphapom C, Warista P & Sakamon D. 2007. Effect of water activity on thermal
resistance of Salmonella Krefeld in liquid medium and on rawhide surface.
International Journal of Food Microbiology 114(1):43-49.

100

 

O'Bryan CA, Crandall PG, Martin EM, Griffis CL & Johnson MG. 2006. Heat resistance
of Salmonella spp., Listeria monocytogenes, Escherichia coli 0157 : H7 and
Listeria innocua M1, a potential surrogate for Listeria monocytogenes, in meat
and poultry: A review. Journal of Food Science 71(3):R23-R30.

Orta-Ramirez A, Marks BP, Warsow CR, Booren AM & Ryser ET. 2005. Enhanced
thermal resistance of Salmonella in whole muscle compared to ground beef.
Journal of Food Science 70(7):M359-M362.

Pearson AM & Dutson TR. 1986. Advances in Meat Research. Westport, Connecticut:
The AVI Publishing Company, INC.

Pearson AM & Dutson TR. 1987. Advances in Meat Research. New York: Van Nostrand
Reinhold Company Limited.

Peleg M. 2000. Microbial survival curves - the reality of ﬂat "shoulders" and absolute
thermal death times. Food Research International 33(7):531-538.

Peleg M, Normand MD & Corradini MG. 2005. Generating microbial survival curves
during thermal processing in real time. Journal of Applied Microbiology
98(2):406-417.

Peleg M & Penchina CM. 2000. Modeling microbial survival during exposure to a lethal
agent with varying intensity. Critical Reviews in Food Science and Nutrition
40(2):]59-172.

Pittia P, Furlanetto R, Maifreni M, Mangina FT & Rosa MD. 2008. Safe cooking
optimisation by F-value computation in a semi-automatic oven. Food Control
19(7):688-697.

Pradhan AK, Yanbin L, Marcy J A, Johnson MG & Tamplin ML. 2007. Pathogen kinetics
and heat and mass transfer-based predictive model for Listeria innocua in

irregular-shaped poultry products during thermal processing. Journal of Food
Protection 70(3):607-615.

Qiaofen C & Da—Wen S. 2007. Effect of cooking bag and netting packaging on the
quality of pork ham during water cooking. Meat Science 75(2):243-247.

101

Richard L & Arkin JD. 2007. PulseNet, FoodNet, NARMS; Tools to ﬁght disease,
Protect Public Health. FDA Veterinarian Newsletter.

Shimoni E & Labuza TP. 2000. Modeling pathogen growth in meat products: future
challenges. Trends in Food Science & Technology] 1(11):394-402.

Sloan E. 2009. Top 10 Food Trends. Food Technology. Chicago, IL: Institute of Food
Technologists. p. 112.

Smith SE, Maurer JL, Orta-Ramirez A, Ryser ET & Smith DM. 2001. Thermal
inactivation of Salmonella spp., Salmonella typhimurium DT104, and Escherichia
coli 0157 : H7 in ground beef. Journal of Food Science 66(8):1164-1168.

Sofos JN. 2008. Challenges to meat safety in the let century. Elsevier Sci Ltd. p. 3-13.

Stasiewicz MJ, Marks BP, Orta-Ramirez A & Smith DM. 2008. Modeling the effect of
prior sublethal thermal history on the thermal inactivation rate of Salmonella in
ground turkey. Journal of Food Protection 71(2):279-285.

Stopforth JD, Lopes M, Shultz JE, Miksch RR & Samadpour M. 2006. Microbiological.
status of fresh beef cuts. Journal of Food Protection 69(6): 1456-1459.

Tenorio-Bemal MI, Marks BP, Ryser ET, Booren AM. 200x. Obtaining a path-dependent
thermal inactivation model that accounts for the acquired thennotolerance of
Salmonella under sublethal heating in turkey, beef, and pork. Journal article to be
submitted to Journal of Food Protection.

Tuntivanich V. 2008. Effects of Marination on Salmonella Penetration and Muscle
Structure of Turkey Breast. Food Science and Human Nutrition East Lansing:
Michigan State University. p. 129.

Tuntivanich V, Orta-Ramirez A, Marks BP, Ryser ET & Booren AM. 2008. Thermal
inactivation of Salmonella in whole muscle and ground turkey breast. Journal of
Food Protection 71(12):2548-2551.

Valdramidis VP, Geeraerd AH & van Impe JF. 2007. Stress-adaptive responses by heat
under the microscope of predictive microbiology. Journal of Applied
Microbiology 103(5): 1 922- 1 930.

102

van der Fels-Klerx HJ, J acobs-Reitsma WF, Van Brakel R, Van Der Voet H & Van
Asselt ED. 2008. Prevalence of Salmonella in the broiler supply chain in The
Netherlands. Journal of Food Protection 71(10):]974-1980.

Velasquez A, Breslin T, Marks B, Orta-Ramirez A, Hall N, Booren A & Ryser E. 2009.
Enhanced thermal resistance of Salmonella in marinated whole muscle as
compared to ground pork. Journal of Food Protection in press.

Vittadini E, Rinaldi M, Chiavoro E, Barbanti D & Massini R. 2005. The effect of
different convection cooking methods on the instrumental quality and yield of

pork longissimus dorsi. Meat Science 69(4):749-756.

Warsow CR, Orta-Rarnirez A, Marks BP, Ryser ET & Booren AM. 2008. Single
directional migration of Salmonella into marinated whole muscle turkey breast.
Journal of Food Protection 71(1):]53-156.

Wesche AM, Marks BP & Ryser ET. 2005. Thermal resistance of heat-, cold-, and
starvation-injured Salmonella in irradiated comminuted turkey. Journal of Food
Protection 68(5):942-948.

Wu VCH. 2008. A review of microbial injury and recovery methods in food. Food
Microbiology 25(6):735-744.

Xavier IJ & Ingham SC. 1997. Increased D-values for Salmonella enteritidis following
heat shock. Journal of Food Protection 60(2): 1 81-184.

Xiong YL & Kupski DR. 1999. Monitoring phosphate marinade penetration in tumbled

chicken ﬁlets using a thin-slicing, dye-tracing method. Poultry Science
78(7):]048-1052.

Zook CD, Busta FF & Brady LJ. 2001. Sublethal sanitizer stress and adaptive response of
Escherichia coli OlS7:H7. Journal of Food Protection: 64 (6) 767-769 64(6):767-
769.

103

Mill

v
.I.
S
R
E
N
N"
U"

o
3
9

H)

ulmnuuumn

3 1

5