EXPERIMENTAL AND MATHEMATICAL MODELS FOR LISTERIA
MONOCYTOGENES TRANSFER BETWEEN DELICATESSEN MEATS AND
CONTACT SURFACES
By

Amanda Nicole Benoit

A THESIS

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

Biosystems Engineering – Master of Science

2013

ABSTRACT
EXPERIMENTAL AND MATHEMATICAL MODELS FOR LISTERIA
MONOCYTOGENES TRANSFER BETWEEN DELICATESSEN MEATS AND
CONTACT SURFACES
By
Amanda Nicole Benoit
Currently, there is limited ability to follow Listeria monocytogenes through food-service
environments, and there remain knowledge gaps concerning the underlying mechanisms that
govern surface-to-surface transfer processes. Therefore, the objectives were to: (1) develop a
method to quantify the spread of a physical surrogate, GloGerm™ powder (GGP), between
delicatessen meats and contact surfaces, and mathematically relate that to L. monocytogenes
transfer under equivalent conditions; and (2) aggregate existing transfer data and compare
candidate models across different transfer scenarios to elucidate phenomenological differences
attributable to contact surface or product type. GGP and L. monocytogenes inoculated pieces (8
x 8 cm), of turkey, ham, stainless steel, or high-density polyethylene were placed in contact with
an uninoculated surface of an opposite material (i.e., meat vs. equipment surface) for 12
sequential transfers. The GGP curves (photographed and quantitatively analyzed via image
processing) were fit to the corresponding L. monocytogenes data by a linear adjustment, noting
that parameters for recipient surfaces did not differ significantly (P > 0.05). Linear, Weibull, and
two-phase linear- Weibull models were fitted to transfer curves encompassing the major readyto-eat meat products, equipment surfaces, and contact events. The most likely models were
determined using the Akaike Information Criterion. For slicing data, model choice was product
dependent, but static contact data yielded the same model regardless of the meat type or surface,
indicating fundamental differences among transfer responses.

To my parents for their continuous support and never letting me give up
To my brothers for teaching me to be strong and to rise above all challenges

iii

ACKNOWLEDGEMENTS
This work would not have been possible without Dr. Bradley Marks, my major advisor
and mentor. I am truly appreciative of the time, advice, and support he graciously offered me
regarding this project and overwhelming personal decisions. He kept me optimistic and gave me
direction. His patience taught me patience. He pressed me to improve my problem solving and
critical thinking skills. I recognize him for my professional and individual growth during my
time at Michigan State.
I also want to give a special thanks to my committee members: Dr. Elliot Ryser and Dr.
James Steffe. Dr. Ryser welcomed me into his lab with open arms and ensured that I had access
to all necessary materials and equipment. His expertise in microbiology, past work, and unique
suggestions, were invaluable throughout my project. His editing skills were also much
appreciated. Dr. Steffe was the first professor I had during my graduate studies. His friendly and
welcoming disposition helped me transition into my new home. I learned so much from both his
courses and him as an individual. He is truly an educator. I am grateful for his door always being
open.
I want to recognize Dr. Sanghyup Jeong for his time researching and programing
algorithms as well as working with me to continuously tweak them. I also want to give a special
thanks to Dr. Dan Guyer for sharing his expertise in imagery, and Dr. Lei Zhang for teaching me
how to work in a microbiology lab especially with Listeria. Thank you to Dr. Ryser’s lab team
for allowing me to use your space and answering my endless questions and Dr. Marks’ research
team and graduate colleagues for their continuous support.

iv

I can never thank the wonderful friends I have made here enough for sharing smiles,
laughter, tears, and frustrations with me. You have consistently been there supporting for
everything. You girls are the sisters I never had, and I am so grateful that I met and shared this
time with each one of you.
And finally, my family, I would never have been able to do this without your love and
support. You made me believe that I could be anything I wanted to be and accomplish anything I
put my heart into. You made me strong when I was at my weakest, and would never let me quit.

This material is based upon the work supported by the Cooperative State
Research, Education, and Extension Service, U.S. Department of Agriculture, under Award No.
2010-51110-21004 and the United States Department of Agriculture, National Institute for Food
and Agriculture, Agriculture and Food Research Initiative, Grant #2012-67017-3018.

v

TABLE OF CONTENTS

LIST OF TABLES………………………………………………………………………..........vii
LIST OF FIGURES ..................................................................................................................... xi
1.

INTRODUCTION.................................................................................................................. 1
1.1 Background/problem statement.......................................................................................... 1
1.2 Goals and objectives ........................................................................................................... 2

2.

LITERATURE REVIEW ..................................................................................................... 4
2.1 Listeria monocytogenes and food safety ............................................................................ 4
2.1.1 Listeriosis outbreaks and recalls .................................................................................. 4
2.1.2 Regulatory agency guidelines ...................................................................................... 5
2.2 Obstacles with Listeria monocytogenes and RTE products ............................................... 5
2.2.1 Prevalence and viability............................................................................................... 6
2.2.2 Adhesion ...................................................................................................................... 9
2.2.2.1 Environmental factors ......................................................................................... 10
2.2.2.2 Bacterial physicochemical factors ...................................................................... 11
2.2.2.3 Material surface characteristics........................................................................... 12
2.2.3 Transfer and cross-contamination ............................................................................. 15
2.2.3.1 Inoculum size influence ...................................................................................... 15
2.2.3.2 Physicochemical factors...................................................................................... 16
2.2.3.3 Material characteristics ....................................................................................... 16
2.2.3.4 Transfer prevention ............................................................................................. 18
2.2.3.5 Modeling transfer ................................................................................................ 19
2.2.3.6 Aggregated transfer data ..................................................................................... 22
2.2.4 Experimental and recovery methods effects on quantitative results ......................... 24
2.3 Conclusion ........................................................................................................................ 26

3. UTILIZING IMAGE ANALYSIS OF A FLUORESCENT PHYSICAL SURROGATE
TO QUANTIFY LISTERIA MONOCYTOGENES TRANSFER BETWEEN COMMON
DELICATESSEN MEATS AND CONTACT SURFACES.................................................... 27
3.1 Introduction ...................................................................................................................... 27
3.2 Materials and Methods ..................................................................................................... 29
3.2.1 Listeria monocytogenes preparation .......................................................................... 30
3.2.2 Physical surrogate preparation................................................................................... 31
3.2.3 Delicatessen meat ...................................................................................................... 31
3.2.4 Stainless steel and high density polyethylene surfaces ............................................. 31
vi

3.2.5 Stainless steel and high density polyethylene inoculation......................................... 31
3.2.6 Turkey and ham inoculations .................................................................................... 32
3.2.7 Transfer of L. monocytogenes from turkey/ham piece to stainless steel/ high-density
polyethylene coupons ............................................................................................................ 32
3.2.8 Transfer of L. monocytogenes from stainless steel/ high-density polyethylene
coupons to turkey/ham piece ................................................................................................. 33
3.2.9 Cleaning and disinfecting coupons ............................................................................ 34
3.2.10 Transfer of GGP from turkey/ham piece to stainless steel/ high-density
polyethylene coupons ............................................................................................................ 34
3.2.11 Transfer of GGP from stainless steel/ high-density polyethylene coupons to
turkey/ham piece.................................................................................................................... 35
3.2.12 Processing images to determine amount of GGP on surfaces ................................. 35
3.2.13 Statistical analysis and model development ............................................................ 38
3.3 Results and Discussion ..................................................................................................... 39
3.3.1 Sequential transfer of L. monocytogenes inoculated surfaces to their counterparts . 39
3.3.2 Calibration curves for GGP on all surfaces ............................................................... 41
3.3.3 Relating transfer of L. monocytogenes and GGP to and from delicatessen meats and
surfaces .................................................................................................................................. 43
3.4 Conclusions ...................................................................................................................... 51
4. APPLYING MULTIPLE MODELS TO IDENTIFY COMMON
CHARACTERISTICS IN LISTERIA MONOCYTOGENES TRANSFER DATA
BETWEEN READY-TO-EAT MEAT PRODUCTS AND CONTACT SURFACES .......... 52
4.1 Introduction ...................................................................................................................... 52
4.2 Materials and Methods ..................................................................................................... 54
4.2.1 Data classification...................................................................................................... 55
4.2.2 Modeling pathogen transfer between meat and surfaces during mechanical slicing,
knife slicing, and sequential static contacts ........................................................................... 57
4.2.3 Fitting the equations to experimental data ................................................................. 59
4.2.4 Statistical comparisons .............................................................................................. 61
4.3 Results and Discussion ..................................................................................................... 62
4.3.1 Interpretation of models on transfer data ................................................................... 62
4.3.2 Mechanical slicing transfer characteristics ................................................................ 69
4.3.3 Knife slicing data ....................................................................................................... 74
4.3.4 Sequential static contact data ..................................................................................... 74
4.3.5 Effects on transfer due to product composition ......................................................... 76
4.3.6 Future utility of database ........................................................................................... 78
4.4 Conclusion ........................................................................................................................ 79
5.

OVERALL CONCLUSIONS ............................................................................................. 81
5.1 GGP comparisons by means of image analysis ............................................................... 81
5.2 Aggregation of foodborne pathogen transfer data............................................................ 81
5.3 Analysis of aggregated delicatessen test data .................................................................. 83
vii

5.4 Utility of a Pathogen Transfer Database .......................................................................... 84
5.5 Future work ...................................................................................................................... 84
5.5.1 Standardization of image capturing ........................................................................... 84
5.5.2 Creation of a simple analysis tool.............................................................................. 85
5.5.3 GGP Model validation, use in other projects ............................................................ 85
5.5.4 Enhancement of the TranBase database .................................................................... 85
5.5.5 Comparisons with other research groups................................................................... 86
5.5.6 The need for standardized testing methods and reporting results ............................. 86
APPENDIX .................................................................................................................................. 87
REFERENCES .......................................................................................................................... 124

viii

LIST OF TABLES

Table 2.1 List and approximate quantity of each variable considered in aggregated
manuscripts on food pathogen transfer………………………………………………………….23
Table 3.1 Intensity vs. concentration calibration curve equations and goodness of fit for
stainless steel, high density polyethylene, turkey, and ham recipient surfaces (y =
concentration in ppm; x = sum of pixel intensities)………………………………………….….43
Table 3.2 Model parameters and statistics for predicting L. monocytogenes transfer by GGP
for each transfer case……………………………………………………………………......…...45
Table 3.3 Model parameters and statistics for predicting L. monocytogenes transfer by GGP
for each recipient case……………………………………………………………………...…....47
Table 3.4. Model parameters and statistics for predicting L. monocytogenes transfer by GGP for
ham to HDPE at different inoculation levels……………………………………………………49
Table 4.1 Classification of L. monocytogenes data aggregated and considered for transfer
analysis based on transfer method and meat product type…………………………………....... 57
Table 4.2 Summary of L. monocytogenes transfer and model statistics for (a) slicer data, (b)
kitchen knife, and (c) static contact data ………………………..…………………………...….62
Table 4.3 The number of datasets in each transfer scenario most likely to be correct in each
model case based on (a) AICc and (b) RMSE…………………………………………………...70
Table 4.4 The number of mechanical slicer datasets for each product type most likely to be
correct in each model case based on (a) AICc and (b) RMSE…………………………………..71
Table 4.5 The number of sequential static transfer datasets for each product type most likely to
be correct in each model case based on (a) AICc and (b) RMSE……………………………….76
Table 4.6 The number of datasets for each product type most likely to be correct in each model
case based on (a) AICc and (b) RMSE………………………………………………………….77
Table 6.1 Recovery of known concentrations of peanut allergen by means of sonication or
composite tissue……………….…………………………………………………………………92
Table 6.2 Cut-off for background values for each surface and color…………………………100
Table 6.3 Minimizing differences between average L. monocytogenes and GGP transfer using of
an offset and multiplier for (a) high density polyethylene to ham, (b) ham to HDPE, (c) turkey to
stainless steel, (d) stainless steel to turkey, (e) turkey to HDPE, and (f) HDPE to
turkey………………………………………………………………………………………….107
ix

9

Table 6.4 Sequential transfer of L. monocytogenes from 10 CFU inoculated turkey to stainless
steel coupons…….………………………………………………………………………...…..110
9

Table 6.5 Sequential transfer of L. monocytogenes from inoculated 10 CFU stainless steel to
turkey coupons ……………………………………………………………………………..…111
9

Table 6.6 Sequential transfer of L. monocytogenes from inoculated 10 CFU ham to stainless
steel coupons ….………………………………………………………………………………111
9

Table 6.7 Sequential transfer of L. monocytogenes from inoculated 10 CFU stainless steel to
ham coupons …………….…………………………………………………………………....112
9

Table 6.8 Sequential transfer of L. monocytogenes from inoculated 10 CFU turkey to HDPE
coupons …..…………………………………………………………………………………...112
9

Table 6.9 Sequential transfer of L. monocytogenes from inoculated 10 CFU HDPE to turkey
coupons …………………………………………………………………………………….…113
9

Table 6.10 Sequential transfer of L. monocytogenes from inoculated 10 CFU ham to HDPE
coupons…...……………………………………………………………………………….......113
9

Table 6.11 Sequential transfer of L. monocytogenes from inoculated 10 CFU HDPE to ham
coupons…………………………………………………………………………………….….114
9

Table 6.12 Sequential transfer of L. monocytogenes from inoculated 10 CFU stainless steel to
turkey coupons………………………………………………………………………..……....114
7

Table 6.13 Sequential transfer of L. monocytogenes from inoculated 10 CFU turkey to stainless
steel coupons ………………………………………………………………………………….115
7

Table 6.14 Sequential transfer of L. monocytogenes from inoculated 10 CFU turkey to HDPE
coupons …………………………………………………..…………………………………...116
7

Table 6.15 Sequential transfer of L. monocytogenes from inoculated 10 CFU HDPE to turkey
coupons …………………………………..…………………………………………………...117
Table 6.16 Bacterial transfer curve publications identified from the ISI Web of
Knowledge……………………..…………………………………………………………...…118

x

LIST OF FIGURES

Figure 3.1 Population of L. monocytogenes on the receiving surface following sequential
9
contacts when the donor surface was inoculated at ~10 CFU (data are means of triplicates)….40
Figure 3.2 Stainless steel coupons, under ultraviolet light, after sequential contacts with GGP
inoculated turkey samples………………………………………………………………………..42
Figure 3.3 Calibration curves for image intensity vs. concentration of GGP for: (a) Stainless
steel calibration, (b) Turkey calibration, (c) Ham calibration, (d) HDPE calibration…….. ……42
Figure 3.4 Image-predicted vs. actual L. monocytogenes surface counts for: (a) Turkey to
stainless steel, (b) Stainless steel to turkey, (c)Turkey to HDPE (d) HDPE to turkey, (e) Ham
to stainless steel, (f) Stainless steel to ham, (g) Ham to HDPE, and (h) HDPE to ham…...........46
Figure 3.5 Image-predicted vs. actual L. monocytogenes surface counts for recipient surfaces:
(a) Stainless steel, (b) HDPE, (c) Turkey, and (d) Ham………………………………………...48
Figure 4.1 A screen-shot of the prototype database (TranBase) cataloging bacterial transfer data.
Note the data fields and drop-down menus for sorting (e.g., as shown for choosing the type of
transfer process)…………………………………………………………………………….…...56
Figure 4.2 Examples of evaluated L. monocytogenes data sets and fits of the most likely models:
(a) Linear, (b) Linear-Weibull, and c) Weibull……………………………………………….....68
Figure 4.3 An example of a single evaluated dataset fitted to of all three candidate models…..69
Figure 6.1 Absorbance readings of Glogerm serial dilutions on RTE turkey at 370nm..………89
Figure 6.2 Calibration curve for known concentrations of peanut allergen measured using a
Microtiter plate reader……………………………………………………………………..…….90
Figure 6.3 Comparing dilution number to concentration by use of calibration curve…….……91
Figure 6.4 Concentrations of peanut allergen recovered from stainless steel by using a KimWipe
and sonication ……………………………………………………………………………….…..92
Figure 6.5 Concentrations of peanut allergen recovered from turkey by stomaching……….....93
Figure 6.6 Photograph of GGP on stainless steel coupons taken with a point-and-shoot camera
……………………….………………………………………………………….……………..…94
Figure 6.7 Set-up of digital camera with tube light. Camera stands on a tripod directly above the
sample being photographed. The tube light is 10 degrees above the sample……………….…..96
Figure 6.8 Tripod with camera directly above sample with bundle of UV-LED lights attached to
tripod stand…………………………………………………………………………………........97
xi

Figure 6.9 Calibration curve comparing intensity to concentration between the tube light and
bundle LED’s……………………………………………………………………………….……98
Figure 6.10 Known concentrations of GGP on stainless steel, high density polyethylene, and
turkey surfaces………………………………………………………………………………......99
Figure 6.11 Graphical comparison of Cannon and Nikon digital SLR camera sensitivities during
transfer from turkey to stainless steel surfaces……………………………………………..……99
Figure 6.12 Average number of green pixels for each surface containing a known concentration
of GloGerm………………………………………………………………………………....….101
Figure 6.13 Comparing the green component value of known concentrations on different
surfaces…………………………………………………………………………………….......102
Figure 6.14 Images changed from original (left), to red, blue, green, and grey components…103
Figure 6.15 Threshtool that is used to subjectively choose a threshold for each image……....103

xii

1. INTRODUCTION

1.1

Background/problem statement
Listeria monocytogenes is a pathogen that originates in soil and water and is often found

in various food products, which raises considerable public safety concern. Vegetables can
become contaminated because of their contact with soil or improperly composted manure (CDC
2010). Animals can also carry Listeria, leading to contaminated raw meats and dairy products.
Once the meat is cooked or milk is pasteurized, these products are essentially Listeria-free, but
can be recontaminated at the manufacturing or food service level (CDC 2010).
Ingestion of Listeria contaminated food can lead to an illness called listeriosis in
consumers. Pregnant women, newborns, persons with weakened immune systems, and the
elderly are most susceptible to listeriosis (CDC 2010). Outbreaks and listeriosis cases have
declined over the last two decades, with a 36% decrease from 1996 to 2006; however, there are
still an estimated 1,600 illnesses and 290 deaths annually from listeriosis in the United States
(CDC 2013).
Listeria monocytogenes is the number one target in ready-to-eat (RTE) meat and poultry
products due to its ubiquitous presence. A risk ranking last completed in 2003, indicated that deli
meats have the highest risk for illness and cases on a per serving basis, relative to other food
categories, causing more than half the reported listeriosis cases each year (FSIS 2010a). In risk
modeling, pathways of exposure are used as inputs including the retail, growth, consumption,
and response stages (FSIS 2010a). Risk models predict the likelihood of illness or death resulting
from Listeria contaminated deli meats based on industry data and consumer surveys. The risk
assessment evaluating illness from manufacturer-packaged versus delicatessen-sliced meats
1

determined that 83% of cases resulted from deli meat sliced at the retail level (FSIS 2010a).
Studies are still needed to assess how contamination occurs at the retail level, in order to create
effective solutions to reduce the outbreaks of listeriosis.
An outbreak not only impacts public health, but also has large effects on the economy.
Economic research has estimated that total sales of RTE deli meats each year in the United States
are ~$25.2 billion, with 24.4% from pre-packaged meat and the remaining 75.6% sliced at retail
establishments (Gombas and others 2003). From these two sources, about 0.4% and 2.7% of the
product, respectively, contains Listeria monocytogenes (FSIS 2010b; Gombas and others 2003).
If these products are recalled, the economic loss would be hundreds of millions of dollars.
Currently, predictive modeling tools and microbial-risk assessments are being used as
guides to decisions relating to Listeria monocytogenes prevention and control. However, little is
known about cross-contamination routes in food facilities leading to contaminated products
(Ivanek and others 2004). Food processing and food service equipment and contact surfaces are
recognized harbors for pathogens, which increase the risk of transfer to RTE foods and the
resulting potential for foodborne illness. Very few tools and studies are available to accurately
and quantitatively evaluate cross-contamination; therefore, several assumptions are being used to
fill gaps, making it difficult to validate risk models.

1.2

Goals and objectives
With the importance of food safety, and the knowledge of problems related to cross-

contamination of RTE deli products, the research presented in this thesis was conducted to

2

improve tools to quantify and describe Listeria transfer during handling and slicing of RTE meat
products. The objectives were to:
1. Develop a method to quantify the spread of a physical surrogate, GloGerm™ powder
(GGP), between delicatessen meats and contact surfaces.
2. Mathematically relate GGP to Listeria monocytogenes transfer under equivalent
conditions.
3. Compare multiple models across different transfer scenarios, to elucidate
phenomenological differences attributable to contact or product type.

3

2. LITERATURE REVIEW

2.1

Listeria monocytogenes and food safety
Listeria monocytogenes is a Gram-positive, non-spore forming, rod-shaped bacterium

(Swaminathan and others 2007). This pathogen is of particular concern because of its ability to
attach to surfaces and persist in low temperature environments. It is also resistant to acidic and
low moisture conditions. The capability of this pathogen to survive in harsh conditions, and its
potentially low infectious dose has led to RTE food safety concerns (Lunden and others 2002).

2.1.1

Listeriosis outbreaks and recalls
Numerous outbreaks and recalls have put L. monocytogenes in the headlines. For

example, in 2007, there was a recall of ready-to-eat chicken products due to the presence of
Listeria monocytogenes (FSIS 2010b). This confirms that sanitary procedures were not being
practiced in this particular establishment. In 2000, ~17 million pounds of country ham, turkey,
and chicken were recalled in Texas; over 900,000 pounds of wieners and deli meats were
recalled in Tennessee, 19,000 pounds in New York, and 13,500 pounds of roast beef recalled in
Utah (FSIS 2010b). In 2010, several ready-to-eat turkey breasts and smoked ham were recalled
in Pennsylvania and Minnesota (FSIS 2010b). The continuing outbreaks and recalls over recent
years support the need for preventative procedures to be established and enforced.

4

2.1.2

Regulatory agency guidelines
In 2003, the Food Safety and Inspection Service (FSIS) amended regulations to include

Listeria as a hazard to address through HACCP plans. New regulations require that all
establishments that produce RTE meat products must prevent any Listeria monocytogenes
contaminated products or direct contact with Listeria contaminated surfaces (zero-tolerance) by
adhering to one of three USDA alternatives (FSIS 2009). To prevent outbreaks, FSIS issued
guidance to establishments, including resources and recommendations on: (1) validating and
applying microbial agents, (2) effective control processes, and (3) sanitary practices (FSIS 2009).
These compliance guidelines include information regarding challenge studies, shelf life
assessments, and critical factors affecting microbial growth (time, temperature, pH, and
concentration), as well as Listeria reduction due to different interventions (FSIS 2009). It also
includes a Listeria sanitation program and procedures that provide information on water
exposure challenges, and the importance of personal hygiene, separation of products,
temperature, facilities, and equipment (FSIS 2009). FSIS also provided an updated list of
microbial agents that are accepted for use in the production of meat and poultry products.
Although the FSIS made efforts to regulate, inform, and educate these establishments, Listeria
continues to persist and affect consumers.

2.2

Obstacles with Listeria monocytogenes and RTE products
The present obstacles related to RTE meat consumption are largely due to L.

monocytogenes causing product recalls, foodborne illness, and challenges with control and

5

regulatory compliance (Sofos and Geornaras 2010). Over time, pathogens can increase in
virulence and become are more resistant to antimicrobials and environmental stresses. They can
emerge from cross-contamination from other foods, surfaces, water, and soils. To reduce
Listeria outbreaks, meat safety must be improved through plant and retail establishment sanitary
practices, and education to service workers and consumers (CDC 2010). Safety and training can
be significantly improved once bacterial prevalence, adhesion, and transfer characteristics are
further understood.

2.2.1

Prevalence and viability
The presence of bacteria on RTE meat and poultry products is likely a result of

recontamination after processing. Foods and materials that support pathogen growth should be a
focus when studying food safety (Chen and others 2003; Norrung and others 1999). The
prevalence of bacteria appears to be dependent on the attachment, food type, manufacturing
process, storage, and intrinsic factors. One strain may be capable of adapting and surviving on a
specific surface, while others may not (Johnson and others 1990).
It is well known that planktonic bacteria are more susceptible to sanitizers,
antimicrobials, and temperature treatment than attached bacteria biolfilms, which tend to be
much more resistant (Krysinski and others 1992; Mosteller and Bishop 1993; An and Friedman
1996; Gray and others 1984). For example, Staphylococcus epidermidis secretes slime after
adhering to implant surfaces making it more resistant to the body’s defense system as well as
antibiotics (Gristina and others 1987; Gristina and others 1989; Naylor and others 1990; Pascual

6

and others 1993; Sheth and others 1985). Attached biofilms of L. monocytogenes and Salmonella
Typhimurium showed high tolerance when treated with trisodium phosphate, while their
planktonic cells showed much lower resistance (Somers and others 1994). In addition to sanitizer
resistance, studies have found that attached cells of Listeria (Oh and Marshall 1995) and
Salmonella Enteritidis (Dhir and Dodd 1995) are less susceptible to heat.
Foods characteristics affect the growth and survival of bacteria. Higher L. monocytogenes
populations have been documented in ready-to-eat products with high pH, water activity, and
low nitrite concentration; however, such populations are seldom found in fermented meats (Grau
and Vanderlinde 1992; Wang and Muriana 1994). This is supported by Lin and others (2006),
who found that L. monocytogenes can survive and grow better on oven roasted turkey than on
fermented salami or bologna, because of the lower moisture content and pH compared to turkey.
High salt concentration also slows growth (De Reu and others 2002; Vorst and others 2006a).
However, products that do not support the growth of pathogens can still serve as a means for
transmission (Vorst and others 2006a).
Environmental factors also affect the viability and growth of a pathogen. Elevated
temperatures and increased storage time often favor proliferation (Perez-Rodriguez and others
2008; Norrung and others 1999). However, Listeria has the ability to proliferate during cold
storage (Lin and others 2006). In one study, the number of L. monocytogenes positive samples
increased during storage of turkey, but, decreased during storage of salami and bologna(Lin and
others 2006).

7

Low relative humidities limit growth and injure cells (Jackson and others 1993). The
effect of dryness on viability differs between Gram-negative and Gram-positive bacteria. Grampositive bacteria, such as L. monocytogenes, have a thicker peptidoglycan layer that enhances
resistance to drying (Takahashi and others 2010). L. monocytogenes populations have been found
to be the highest when on wet surfaces, including sponges, dish clothes, and faucets (Chen and
others 2001); however, bacteria were found to survive for hours or days on dried surfaces
(Pether and Gilbert 1971; Scott and Bloomfield 1990; Kusumaningrum and others 2003).
Survival of L. monocytogenes on stainless steel with food residues was greater than without food
residues (Takahashi and others 2010). The pre-slicing of meat also has shown greater
populations compared to unsliced chubs during refrigerated storage (Elson and others 2004;
Gillespie and others 2000).
Bacteria can be injured due to environmental factors, such as low relative humidity and
cold temperatures, but mechanical slicing may also cause injury. Mechanical shear during slicing
has been determined to injure L. monocytogenes due to the edge of the blade having greater
friction and therefore increasing stress (Sheen and others 2010). The number of dead cells was
found to increase with slicing. During slicing, most microbes were found on the receiving
surface under the slicer blade (dead or alive). Although the mechanism was not elucidated,
mechanical shear was reported to have a significant impact on bacterial death during slicing
(Sheen and others 2010; Hoffman and Wiedmann 2001).

8

2.2.2

Adhesion
Use of additional mechanical equipment increases food production but also increases

instances where food can come into contact with surfaces (McEldowney and Fletcher 1988). If
these surfaces are not sanitized properly (with approved microbial agents) and regularly,
bacterial attachment can occur, which can lead to cross-contamination during processing and
handling. Reducing bacterial attachment on equipment has been a major concern due to the risk
of foodborne illnesses that can occur when bacteria come into contact with food (Gibbs and
others 1978; Adams and Mead 1983; Franco and others 1995; Smith and Fratamico 1995).
It has been noted for several decades that bacteria prefer to grow on a nutrient rich
surface-liquid interface rather than moving freely in nutrient poor liquids (Costerton 1995;
Zobell 1943). Adhesion is described as a two-phase process in which the first phase is physical,
thus making it instantaneous and reversible, while the second phase is molecular and cellular,
which is time-dependent and irreversible (Marshall and others 1971). In addition to modes of
cellular motility, freely moving planktonic bacteria can move by physical forces, such as
Brownian motion, van der Waals attraction forces, gravitational forces, surface electrostatic
charge, ionic forces, chemical bonds, and hydrophobic interactions (Dankert and others 1986;
Krekeler and others 1989; van Oss 2003; McEldowney and Fletcher 1988; Arnold and others
1993; Ong and others 1999). Once these interactions occur, initial attachment has been
completed and phase two can begin, in which bridging between the bacteria and surface occurs
by bacterial structures, such as capsules, fimbriae, or pili and slime (An and Friedman 1998).
The anchored bacteria produce exopolysaccharides to further anchor the bacteria and create a

9

favorable environment to allow more growth, attachment, and potential biofilm formation (Mafu
and others 1990a).
Adhesion of bacteria to a surface is a complex process that depends on several factors
including bacterial species, the environment, bacteria and material surface characteristics, and
intrinsic physicochemical factors. Having a deeper understanding of each of these factors can
assist in finding solutions to reduce product contamination, in many food service industries
(Zottola 1994; Wirtanen and others 1996).

2.2.2.1 Environmental factors
Several environmental factors affect bacterial adhesion, including temperature, time of
exposure, pH, bacteria concentration, nutrient availability, shear force, fat, proteins, and
chemicals. For example, in one study, the number of bacteria adhering to a surface increased as
time increased until a point of saturation was reached (Satou and others 1988). Listeria can
attach to equipment surfaces and survive for extended periods, confirmed by positive meat and
surface samples at the end of slicing (Hoffman and Wiedmann 2001). Higher inoculum levels of
E. coli O157:H7 show stronger attachment to leafy greens and beef tissue, and L. monocytogenes
has a strong attachment to glass when inoculum levels are high, due to nutrient availability and
surface hydrophobicity. (Takeuchi and Frank 2000; Dickson and Daniels 1991; Fratamico and
others 1996; Arnold and Silvers 2000). Attachment to surfaces has been found to be dependent
on nutrient availability - the more available the nutrient, the greater the attachment strength
(Takahashi and others 2010). High concentrations of electrolytes also increase attachment, while

10

the presence of protein can often inhibit adhesion (Abbott and others 1983). These factors all
affect the adhesion of bacteria to surfaces by changing interactions that occur that allow
adhesion, or, by changing bacterial or material characteristics.

2.2.2.2 Bacterial physicochemical factors
Physicochemical factors influencing adhesion are unique to each type and strain of
bacteria. Two of the main factors are hydrophobicity and surface charge, which are both affected
by growth medium, bacteria age, and bacteria surface structure (Krekeler and others 1989;
Dankert and others 1986). It is well established that hydrophobic bacteria adhere more favorably
to hydrophobic surfaces, and hydrophilic bacteria adhere better to hydrophilic surfaces (Hogt
and others 1983; Satou and others 1988). Also, hydrophobic bacteria generally adhere to surfaces
more strongly than hydrophilic bacteria (Vanloosdrecht and others 1987). This concept is
supported by a study that found that S. epidermidis and S. aureus adhere better to polymers and
metals, respectively (Barth and others 1989).
Listeria monocytogenes expresses low surface hydrophobicity and is classified as
hydrophilic, but this is only apparent at low pH and high ionic strength (Mafu and others 1991).
Often, a chemical treatment can change hydrophobicity and therefore attachment strength of
bacteria (Hogt and others 1983; Westergren and Olsson 1983). The presence of proteins also can
affect adhesion by changing bacteria surface physiochemistry (Reynolds and Wong 1983). When
in solution, bacterial surface charge is negative (Hogt and others 1982) thus attracting ions of
opposite charge (An and Friedman 1998). It has been determined that hydrophilic bacteria have a

11

high surface charge, while the surface charge is usually low for hydrophobic bacteria (Hogt and
others 1982). However, adhesion of bacteria to surfaces could not always be attributed to the
relative surface charge of the bacteria (Abbott and others 1983).

2.2.2.3 Material surface characteristics
The structure, chemical composition, physicochemistry, and electronic properties of
material surfaces control reactivity that influences bacterial attachment. In a study by Arnold and
Silvers (2000), susceptibility of many common equipment surfaces in a poultry processing plant
to bacterial attachment and growth was tested. Results showed that, depending on the surface,
the affinity for bacteria differed (Arnold and Silvers 2000). It was determined that a picker finger
(made of rubber) inhibited bacterial attachment, while stainless steel, polyethylene, and conveyor
belting had increased populations over time, stainless steel having the most attachment (Arnold
and Silvers 2000). The study also discussed the effect of surface elemental composition, such as
sulfur and zinc, two commonly known antimicrobials that are found in rubber and polyethylene
(Arnold and Silvers 2000). This supports the picker finger results, but does not explain the
susceptibility of polyethylene to bacterial contamination.
Chemical composition of surfaces affects the initial attraction forces as well as the
strength and mechanism of adhesion between bacteria and surface materials. Several studies
have shown that when a material surface is treated with microbial coating, attachment decreases
(Bridgett and others 1992; Farber and Wolff 1992; Speier and Malek 1982). Coating surfaces

12

with protein alters surface hydrophobicity, also affecting attachment (Fletcher and Marshall
1982).
The hydrophobicity of surfaces that come into contact with bacteria has a large influence
on attraction and strength of attachment. Metals are hydrophilic, have a negative charge, and a
high surface energy, while polymers are hydrophobic, have a lower charge, and low surface
energy (An and Friedman 1998). These low charged hydrophobic surfaces are however more
susceptible to adhesion than hydrophilic materials, which resist adhesion (Hogt and others 1983;
Ludwicka and others 1984). (Mafu and others 1990a) tested L. monocytogenes attachment ability
on low energy surfaces of polypropylene and rubber and found no correlation, but in another
study found L. monocytogenes to attach to both low and high energy surfaces after a short
contact time. Another study came to the conclusion that large numbers of bacteria attached to
hydrophobic plastics, a moderate number attached to hydrophilic surfaces with positive or
neutral charge, and few to hydrophilic negatively charged surfaces (Fletcher and Loeb 1979).
This result agrees with the idea that hydrophobic surfaces are more susceptible to contamination.
Additionally, L. monocytogenes is more resistant to sanitizers when attached to hydrophobic
polypropylene and rubber than when attached to hydrophilic stainless steel and glass, due to
biofilm formation once multiple colonies are strongly attached (Mafu and others 1990b). This is
in agreement with Fletcher and Leob’s findings that greater numbers, and thus larger biofilms,
are supported by hydrophobic materials, making the bacteria more resistant to chemical
treatment (Fletcher and Loeb 1979).
Surface roughness and physical configuration also are contributing factors to adhesion
after bacteria come in contact with a material surface. A medical implant study found that the

13

ultra-smooth surface Vivathane supported minimal bacterial attachment throughout the duration
of the study, while other tested surfaces with irregularities promoted adhesion, biofilm
formation, and accumulation (McAllister and others 1993). When surfaces of glass and
polystyrene were roughened with a grindstone and placed on a river bottom, increased rates of
colonization were also found (Baker and Greenham 1988). Rough surfaces increased effective
surface area for attachment to occur, and surface depressions are favorable growth sites for
bacteria (Baker and Greenham 1988; Arnold and Bailey 2000; Rodriguez and McLandsborough
2007). However, An and others (1995) found that roughening commercially pure titanium with
sandpaper did not affect attachment of S. epidermidis. Sandpaper was also to treat stainless steel
surfaces in a study by Arnold and Bailey (2000), who found that sanding, sandblasting, and
electro-polishing smoothed the surface, thus increasing resistance to bacterial contamination,
while the same finishes tested by Hilbert and others (2003) showed no differences in Listeria
attachment.
Additionally, porous and grooved surfaces support greater bacterial growth and
attachment (Merritt and others 1979; Locci and others 1981; Vorst and others 2006b). For
example, braided medical sutures have increased attachment compared to non-braided sutures
(Sugarman and Musher 1981). With non-absorbable materials, the physical configuration
contributes more to bacteria attachment than does surface finish (An and Friedman 1998).
Stainless steel grade (304/316) also has an effect on attachment due to the difference in pore
number and size (Vorst and others 2006b). These findings agree with the idea that irregularities
on material surfaces promote adhesion.

14

2.2.3

Transfer and cross-contamination
Multiple food-to-surface contacts during industrial processing of food are inevitable.

Other sources of contact in the retail environment include workers, customers, gloves, utensils,
slicing equipment, other food products, counter tops, and cutting boards. Often, many of these
surfaces can be overlooked or poorly sanitized, leading to the potential for cross-contamination.
Knowledge of bacterial sources, transmission routes, and rates (even at low levels) can help in
finding effective methods for control in food service and retail environments.

2.2.3.1 Inoculum size influence
The number of bacteria on a surface may influence the number of bacteria transferred and
the amount of time bacteria remain on the surface. Some authors determined that higher
inoculation levels result in lower transfer rates, and lower inoculation levels result in higher
transfer rates (Montville and Schaffner 2003), where “rates” are defined as the fraction of
bacteria transferred from a donor to a recipient surface during a contact event. However, the
number of bacteria transferred may remain constant as inoculum levels increase (Montville and
Schaffner 2003). Some authors, however, have reported more L. monocytogenes positive deli
meat slices when inoculum size on the blade was larger (Lin and others 2006; Rodriguez and
others 2007b, Vorst and other 2006a). However, the efficiency of transfer, as calculated by
Rodriguez and others (2007b), was not affected by initial inoculation size. It has also been noted
that the extent of contamination depended on bacteria levels, as well as the product being sliced

15

(Lin and others 2006). Because of these contradicting results, more research must be conducted
to further understand whether inoculum size does in fact affect transfer rates of bacteria.

2.2.3.2 Physicochemical factors
Meat composition, including fat, moisture, preservatives, pH, and nitrite concentration,
affects the number of bacteria transferred, as well as growth and survival. There was greater L.
monocytogenes cross-contamination when processed meats had a high pH, high water activity,
and low nitrite concentration (Wang and Muriana 1994; Grau and Vanderlinde 1992). This is
supported by the finding that there is more bacterial transfer from an inoculated blade to turkey
with high moisture, no nitrite, and high pH than to bologna and salami with lower pH and
moisture (Lin and others 2006). A film of fat was left on the slicer blade after cutting salami,
while a clear-liquid film was left after cutting turkey (Lin and others 2006). Therefore, although
salami is less supportive of L. monocytogenes growth, its fat transfer to the slicer may prolong
cross-contamination (Vorst and others 2006a).

2.2.3.3 Material characteristics
Transfer rates are also affected by hydrophobicity, surface roughness, porosity, texture,
finish, and hydration level of the surface material. One study found that although stainless steel
has weaker attachment strength to bacteria compared to polyvinylchloride and polyurethane,
stainless steel had a lower transfer rate (Midelet and Carpentier 2002). However, greater

16

numbers of L. monocytogenes were transferred from bologna to stainless steel than from bologna
to polyethylene (Rodriguez and others 2007b). These two findings suggest that transfer between
two surfaces depends on which surface contains the pathogen. This is supported by Harrison and
others (2003), who reported that transfer from contaminated hands to a paper towel dispenser
was lower than from a contaminated dispenser to hands. Additionally, the strength of
hydrophobic interactions affects transfer, and conditioning of surfaces can alter hydrophobicity
and thus attachment strength and transfer ability (Midelet and Carpentier 2002). Chen and others,
assessed the transfer between many common surfaces in a kitchen and found that the transfer
rates from highest to lowest were: chicken to hand, cutting board to lettuce, spigot to hand, hand
to lettuce, , and hand to spigot (Chen and others 2001). Although both inoculation level and
material affect transfer, Rodriguez and others (2007b) did not detect interactions between
inoculation level and material type.
The effects of surface configuration and roughness on transfer are topics that have shown
conflicting results across studies. In one instance, surface roughness and finish on stainless steel
were reported to have no effect on L. monocytogenes transfer (Rodriguez and others 2007b);
however, several other studies determined that finishes had an impact on transfer (Arnold and
Silvers 2000; Whitehead and others 2006). These differing results could be due to the direction
and amount of pressure applied when simulating the transfer event. In a study of wood and
plastic cutting boards and their ability to transfer Salmonella, Cliver (2006) reported that results
depended on the experimental method, varying from: plastic having more transfer, wood having
more transfer, and both having the same amount of transfer. Variation also likely was due to the
condition of the cutting boards.

17

Material hydration or wetness of a surface and relative humidity are factors that may need
consideration in transfer studies. However, Rodriguez and others (2007a) reported that transfer to
a wetted surface did not differ from that of a dried surface. A study comparing transfer of wet
attached and dry attached cells of L. monocytogenes on stainless steel to bologna also found no
difference in transfer (Montville and Schaffner 2003). When decreasing relative humidity of
Listeria biofilms on stainless steel, transfer was shown to increase to bologna and hard salami
(Rodriguez and others 2007a). Transfer was however greater to bologna, probably due to
physicochemical properties of the meat. Rodriguez and McLandsborough (2007) suggested that
adhesion forces between cells weaken during drying; therefore, increasing their chances for
transfer upon contact.

2.2.3.4 Transfer prevention
Transfer prevention techniques, such as the use of gloves or hand washing, must be done
appropriately to prevent contamination adequately. When working with raw meat and continuing
onto to RTE products, high transfer is observed when gloves are not changed (Perez-Rodriguez
and others 2008). It was determined that transfer from chicken to hands is reduced by using
gloves, but not prevented (Montville and Schaffner 2003). Transfer from contaminated hands in
gloves to food has been observed as well (Montville and Schaffner 2003).
In many transfer studies, high variability is reported between runs and often depends on
the volunteer and hand contacts. Chen and others (2001) reported greater variability in crosscontamination rates and in levels of contamination when hands were included in a transfer study,

18

with the greatest variability in post-washed hands to surface. Montville and Schaffner (2003)
also found large variability in results when studying transfer from hand to lettuce through a
glove. In a model by Ivanek and others (2004), it was found that contamination within a slicing
area is influenced mainly by contacts that employees’ gloves make with the food and the
duration the glove is worn.

2.2.3.5 Modeling transfer
Quantitatively modeling L. monocytogenes transfer can be useful in creating microbial
risk-assessments for improved food safety. To date, most such models are empirical, created
simply by plotting variables and fitting that data with an arbitrary function (Sheen 2008; Sheen
and Hwang 2010; Aarnisalo and others 2007). Many also use the transfer-ratio based model,
TR(%)/100×N colony-forming units (CFU), to determine the number of bacteria transferred
(Perez-Rodriguez and others 2008). A simple model based on a contaminated source,
intermediate source, and food that contacts the intermediate source has been established by (den
Aantrekker and others 2003). The transfer ratio can easily be determined for each step. The
problem with this model is that transfer rates vary greatly due to multiple factors that affect
bacterial transmission as well as experimental differences and recovery method used (PerezRodriguez and others 2008). Because of this, probability distributions should be created to
account for variability in data. Often, transfer rates are log transformed to create a normally
distributed bell-shaped curve (Chen and others 2001; An and Friedman 1998). Hoelzer and
others (2012) analyzed a broad set of Listeria transfer data by creating empirical coefficient-

19

based mathematical models and determining the shape of the distribution of transfer coefficients,
finding that transfer coefficients varied considerably and after a log10 transformation were often
best described by normal distributions.
Studies that use transfer ratios assume independence between cross-contamination events,
allowing for transfer events to be multiplied to obtain a final number of bacteria once a crosscontamination process is complete (Perez-Rodriguez and others 2008). However, high
inoculation levels are used for transfer studies to allow for detection, which can affect transfer
rates leading to less accurate results, especially because in reality, when inoculum levels are very
low, fewer transfer events may occur (Vorst and others 2006a). Another assumption in modeling
is the state of bacteria being used in the transfer events. Stress, biofilms, and survival ability due
to these can greatly affect transfer and thus should be considered when modeling (PerezRodriguez and others 2008, Keskinen et al, 2008).
Transfer ratio simulations, which are used to quantitatively describe microbial transfer,
have been created that consider data gathered from other studies. Schaffner (2004) used a MonteCarlo simulation to model Listeria cross-contamination in food-processing plants by tracking
sub-types and prevalence. The model was based on prevalence of L. monocytogenes entering the
plant on raw products, which contaminate the environment at a specific rate and then become a
part of the environmental reservoir (Schaffner 2004). The strains that persist continue to be
transferred to a contact surface, become part of the reservoir and, if persistent, continue to the
finished product (Schaffner 2004). Strains with low numbers, but ability to persist, are dominant
in the final product; however, different strains can have greater persistence in different areas of

20

the plant (Schaffner 2004). A beta-distribution was used to represent prevalence uncertainty in a
population based on the number of positive samples (Schaffner 2004).
Growth models are also necessary to determine whether low numbers of pathogens can
proliferate to levels unsafe for consumption. Ivanek and others (2004) quantified transfer and
proliferation of L. monocytogenes through a fish processing plant by simulation. A
compartmental model was used that included food, gloves, surfaces, and the environment as
reservoirs. A difference-equation system created from the Reed-Frost model to simulate transfer
was based on assumptions and several parameters, which included durations, rates, frequencies,
contaminated units, and prevalence. The simulation showed that prevalence increased over a
work shift and that the confidence intervals became wider over time as a result of uncertainties in
the system; however, it did not model loss of contamination. Therefore, numbers were higher
than expected, leading to a conservative estimate. Their simulation also indicated that crosscontamination in a slicing area is greatly influenced by glove contact and the duration of wearing
gloves (Ivanek and others 2004). This model, as well as many others, is limited in application
due to many assumptions that must be made due to the current lack of sufficient data in this field.
Sheen (2008) empirically modeled L. monocytogenes transfer to salami during slicing,
with two cases: 1) when the blade was inoculated and 2) when the chub was inoculated and
sliced, followed by slicing of an uncontaminated chub. Plots were created using CFU per slice as
the dependent variable and slice number as the independent variable (Sheen 2008). When
inoculation levels were high (>log CFU/g), exponentially decaying empirical models fit the
transfer curves relatively well; however, when the inoculation levels were low, there was a
random transfer pattern, but the same empirical models were used (Sheen 2008). Sheen and

21

Hwang (2010) used similar procedures to model transfer of E. coli O157:H7 on ham during
slicing. However, they reported that when the blade was inoculated directly, a power law model
fit the data best. But, when the inoculated chub was sliced, followed by an uncontaminated chub,
the exponential decay model was the best fit (Sheen and Hwang 2010). This showed that for
E.coli O157:H7 on ham, case 1 results in a higher probability of transfer when the inoculation
level is low (< 5 log CFU/g), but case 2 shows longer persistence of transfer (Sheen and Hwang
2010). (Aarnisalo and others 2007) ran similar tests, but also considered slicing temperature and
attachment time. Transfer of L. monocytogenes from an inoculated blade to uncontaminated
gravad salmon and from inoculated gravad salmon showed exponential decay as the slice
number increased (Aarnisalo and others 2007). Low inoculation level, colder temperatures, and
longer attachment time led to lower total L. monocytogenes transfer (Aarnisalo and others 2007).

2.2.3.6 Aggregated transfer data
The number of papers published on bacterial transfer to/from food has increased
approximately ten-fold over the past two decades, reflecting increasing attention to this
important issue. However, there has been no standardization of methods or aggregation of data
in this field. In order to quantitatively evaluate the research trends in this area over the past 40
years, a comprehensive keyword search of the literature was conducted via the ISI Web of
Science database, to identify all studies including transfer data for key foodborne pathogens
(Listeria, Salmonella, Escherichia coli, and Campylobacter). Published transfer data were
characterized in terms of pathogen, product type, surface, and other variables. The total analysis
yielded ~55 distinct publications on foodborne pathogen transfer, of which 43 contained

22

numerical data, with a total of 756 data sets, ~1,194 individual replicate curves, and over 14,456
individual observations quantifying pathogen transfer between food products and contact
surfaces, including: meat/poultry (n = 29) and produce (n = 11), with transfer to metals (n = 19),
plastics (n = 16), hands/gloves (n =10), and several other surfaces (Table 2.1).
Table 2.1 List and approximate numbers of prior publications including each variable in food
pathogen transfer studies (References categorized here are listed in Appendix, Table 6.16).
Variable

Number of Papers

Contact/Attachment time
Inoculation level

10
6

Temperature

12

Inoculated surface

7

Wet vs. dry

6

Pressure

6

Weak/strong biofilm

5

Friction

2

Hydrophobicity

4

Roughness

4

Sample size

2

Relative humidity

2

Composition

9

This growing body of literature is almost exclusively empirical, without any theoretical
connection to the underlying mechanisms that govern these processes. There is a critical need

23

for a modeling framework that bridges micro- and macro-scale knowledge in order to move
toward a unifying theory of bacterial transfer between foods, processing media, and relevant
food contact surfaces. A thorough analysis of existing published transfer data and aggregation of
that data as such, is an important first step toward that overall long-term goal.
The papers described above have been organized into a preliminary database defining the
characteristics of each study, which will serve as the foundation for a broader, publicly-available
database in this domain. A unified database that aggregates pathogen transfer data, and becomes
a repository for future data, will help advance linkages between fundamental research and the
observed transfer outcomes, while also improving the design of future studies to fill critical data
gaps.

2.2.4

Effect of experimental design and recovery methods on quantitative transfer
When performing transfer studies, the experimental design and recovery methods are

crucial to the results. Transfer studies will yield different results based on the variables
considered during transfer and the methods used. Some researchers use single strain cultures,
while others use multi-strain cocktails (Arnold and Silvers 2000). However, bacterial behavior
can change due to interspecies competition, thus affecting study results (Okabe and others 1995).
Likewise, in prior studies on the effects of surface roughness, some protocols used perpendicular
force, while others used shear force, and the amount of force varied as well. The wood and
plastic cutting board studies also differed due to the methods of inoculation, preparation,
disinfection, and how samples were recovered (Cliver 2006). In another study, a sample of 25 g

24

and an entire package of sliced turkey were assayed and found that a higher percentage of
samples were positive for L. monocytogenes when whole packages were analyzed (Lin and
others 2006). This result, however, was material specific because the same test was performed on
bologna and yielded approximately the same percentage in both cases.
The method of recovery also greatly influences the number of pathogens found on a
surface. There are multiple methods used to assess surface contamination; however, there is not
an accepted standard (Moore and others 2007). Often, sporadic data collection is documented
due to the insensitivity of enumeration methods, which can yield misleading results (Kang and
others 2007). Techniques to study bacteria attachment and biofilm formation need to be
improved by finding ways to accurately quantify counts on a surface (Arnold and Bailey 2000).
A challenge is that characteristics of each bacterium, food surface, and contact surface are
different; therefore, each method has a different efficiency depending on the case (Kang and
others 2007). Efficiency is also affected by the method and mechanics the researcher uses during
sampling (Kang and others 2007). The most common recovery methods are swabbing, rinsing,
and direct agar contact. It has been determined by several authors that swabbing does not give an
accurate assessment of bacteria population on a surface (Kang and others 2007; Harrison and
others 2003; Richard and Piton 1986). Kang found that rinsing had the highest recovery out of
the three common methods and introduces the idea of using a sonicating toothbrush, which was
found to show the greatest recovery. Recovery is important in quantifying bacteria transfer and
the method used must be accurate and suitable to the equipment surface being evaluated so that
data can be validated and models created.

25

2.3

Conclusion
Although there has been significant recent work conducted in the area of bacterial

transfer and modeling in food facilities, this field is still in its infancy, and several critical
knowledge gaps exist. These topics need to be addressed so that outbreaks can be further
understood and prevention measures can be improved. Bacteria attachment and biofilm
formation is well understood; however, knowledge concerning how adhesion and other factors
affect transfer is lacking. Several studies have tested factors influencing transfer but have found
differing results. Various models quantifying transfer have also been created, but nearly all are
empirical, and others are simulations that have limitations.
In identifying research priorities related to L. monocytogenes and public health, the
USDA previously highlighted food safety education and training, filling knowledge gaps,
applying control measures, and monitoring the impact on food safety (USDA-NIFSI 2010).
Consequently, the overall goal of this work is to understand cross-contamination by: identifying
variables and potential phenomenological factors affecting transfer characteristics, as well as to
improve the means to evaluate routes of Listeria transfer in a food service environment. In doing
so, new control measures can be developed, and sources and incidence ultimately can be
minimized through improved worker training interventions.

26

3. UTILIZING IMAGE ANALYSIS OF A FLUORESCENT PHYSICAL SURROGATE TO
QUANTIFY LISTERIA MONOCYTOGENES TRANSFER BETWEEN COMMON
DELICATESSEN MEATS AND CONTACT SURFACES

3.1

Introduction
Listeria monocytogenes is a ubiquitous pathogen found in various food products, including

vegetables, raw meats, and dairy products. Once fully cooked, meat becomes essentially Listeriafree, but can be recontaminated during packaging or slicing in manufacturing facilities and retail
establishments (CDC 2010). Outbreaks and listeriosis cases have declined over the last two
decades; however, there are still an estimated 1,600 illnesses and 290 deaths annually from
listeriosis in the United States (CDC 2013). This has raised a considerable public safety concern,
leading to regulations created by the U.S. Department of Agriculture-Food Safety and Inspection
Service (USDA-FSIS). These include the requirement that all establishments that produce readyto-eat (RTE) meat products must: (1) apply a post-lethality treatment and an anti-microbial agent
or process, (2) apply a post-lethality treatment or an anti-microbial agent or process, or (3)
follow sanitary practices so that direct contact surfaces are tested Listeria-free (FSIS 2009).
A risk ranking, last completed in 2003, indicated that deli meats are the highest risk food
category for illness and cases on a per serving basis, being potentially responsible for more than
half the reported listeriosis each year (FSIS 2010b). The risk assessment compared illness from
RTE meats prepared in processing plants to those in retail establishments and determined that
more cases resulted from deli meat sliced at the retail level than from pre-packaged meat (FSIS
2010b). The food industry, together with regulators, are continuously working to reduce the

27

incidence of Listeria, through improved plant and retail establishment sanitary practices, and
education of processing workers, service workers, and consumers.
Currently, predictive modeling and microbial risk assessments are used as tools in
evaluating food safety policies and practices. However, little is known about cross-contamination
routes in food facilities leading to contaminated products (Ivanek and others 2004). This is due to
the limited ability to follow specific pathogens through food processing facilities. Multiple
factors, such as inoculum level, physicochemical factors, and material characteristics, also make
predictive modeling inherently difficult by influencing transfer dynamics (Lin and others 2006;
Montville and Schaffner 2003; Chen and others 2001; Midelet and Carpentier 2002; Rodriguez
and others 2007a). Researchers who have studied Listeria transfer therefore have reported some
conflicting results, likely due to variability in these factors across studies (Arnold and Bailey
2000; Rodriguez and others 2007b). A quantitative understanding could provide new insight and
further reduce the risk of RTE contaminated products. Thus, mathematical cross-contamination
models for L. monocytogenes transfer would be valuable tools that would enable assessment of
interventions, particularly those related to sanitation and working training practices.
Previous work has assessed L. monocytogenes transfer between product and equipment
surfaces in controlled settings (Sheen and Hwang 2010; Vorst and others 2006b; Vorst and
others 2006a; Sheen 2008). These data, however, are difficult to translate into actual crosscontamination events that occur in food processing and food service environments where there is
a more complex network of surfaces, including the contribution of human decisions and contacts.
The use of a non-pathogenic surrogate would allow for evaluation of cross-contamination events
in a “real-world” setting, as well as rapid assessment of transfer without the risk of foodborne

28

illness. Physical surrogates have been used previously to qualitatively identify high risk locations
on slicers (Vorst and others 2006a) and leafy green processing lines (Buchholz and others 2012);
however, a quantitative evaluation of a surrogate would be valuable in validating environmental
cross-contamination models and quantifying transfer risk in facilities where RTE foods are
handled/served.
In identifying research priorities related to L. monocytogenes and public health, the
USDA previously highlighted food safety education and training, filling knowledge gaps,
applying control measures, and monitoring the impact on food safety (USDA-NIFSI 2010). In
order to create a useful and novel tool for studies in this area, the specific objective of this study
was to develop and validate an image-based analysis method for a physical surrogate (Glo
Germ™ Powder, GGP) to simulate the movement and population changes of L. monocytogenes
to and from RTE delicatessen meats and common food contact surfaces.

3.2

Materials and Methods
GGP (Glo Germ™, Inc., Moab, UT) was chosen as a physical surrogate to simulate L.

monocytogenes cross-contamination in a food service environment due to its small particle size
(5 µm, compared to 1-2 µm for L. monocytogenes), fluorescence under ultraviolet lighting, and
the potential to rapidly quantify the amounts of GGP transferred among surfaces. The materials
used in this study are commonly found in delicatessens: stainless steel representing countertops
and slicer surfaces, and high-density polyethylene representing typical cutting board materials.
These materials also can exhibit high levels of bacterial attachment and growth over time

29

(Arnold and Silvers 2000). Delicatessen ham and turkey were chosen based on the amount of
prior data available as well as their popularity with consumers. Each surface and meat served as
a donor and recipient during separate tests.

3.2.1

Listeria monocytogenes preparation
Strain selection and preparation closely followed those reported by Vorst (Vorst and

others 2006a). A six-strain cocktail of L. monocytogenes was used for surface transfer and crosscontamination studies. The strains were previously obtained from Dr. Catherine W. Donnelly
(University of Vermont, Burlington). L. monocytogenes CWD 205 (source unknown), CWD 578
(dairy plant environment), CWD 701 (cheese), CWD 730 (dairy plant environment), CWD 845
(dairy plant environment), and CWD 1002 (pork sausage) were chosen based on their ability to
form biofilms in a microtiter plate assay (Keskinen and others 2008). Strong (CWD 730 and
CWD 845), medium (CWD 701 and CWD 1002), and weak (CWD 205 and CWD 578) biofilm
formers were selected to consider all possible biofilm scenarios that may affect attachment and
transfer. All stains were kept at –80°C in trypticase soy broth (TSB; Difco, Becton Dickinson)
with 10% (vol/vol) glycerol. A loopful of each frozen stock culture was inoculated in separate
tubes of TSB containing 0.6% (wt/vol) yeast extract (TSB-YE; Difco, Becton Dickinson, Sparks,
Md.) and incubated for 24 h at 37°C. After 24 h a loopful of each TSB-YE culture was
transferred to new tubes of TSB-YE and incubated overnight at 37° C. The optical density at 600
nm was taken for each suspension to initially verify cell concentration, and equal volumes of
9

each culture were mixed to create a six-strain cocktail containing ~10 CFU/ml. Cell

30

concentration was also quantified by plating an appropriate dilution on modified Oxford agar
(MOX; Difco, Becton Dickinson), followed by incubation at 37°C for 48 h.

3.2.2

Physical surrogate preparation
A 1:10 w/v suspension of the physical surrogate GloGerm polymer powder (GGP) was

prepared by adding 1 g of GloGerm to 10 ml of 70% ethanol, followed by vortexing for 1 min.

3.2.3

Delicatessen meat
A chub each of restructured roasted turkey breast and ham were purchased from a local

food store (Gordon Food Service, Okemos, Mich.). The chubs were held at 4°C and used within
20 days of purchase. The package label listed the turkey ingredients as turkey breast, turkey
broth, and less than 2% of salt, dextrose, and sodium phosphates; the ham ingredients included
fully cooked ham, water, salt, sodium phosphate, sodium erythorbate, and sodium nitrite.

3.2.4

Stainless steel and high density polyethylene surfaces
A series of electropolished grade 304 stainless steel (8 x 8 x 0.2 cm) and HDPE (8

x 8 x 1 cm) coupons were fabricated by the Michigan State University Department of
Biosystems and Agricultural Engineering Shop (East Lansing, MI) and used for transfer.

3.2.5

Stainless steel and high density polyethylene inoculation
For the tests where stainless steel or HDPE was the donor surface, the L. monocytogenes

inoculum (300 µl) was spread evenly over the surface of one of the coupons with an inoculating
8

2

needle, so as to yield ~10 CFU/cm (subsequently verified). A duplicate coupon of the same

31

material was inoculated with 300 µl of the concentrated GGP suspension. The inoculated
coupons were held at ambient temperature in a laminar flow cabinet for approximately 30 min
until visibly dry.

3.2.6

Turkey and ham inoculations
For tests where the meat products were the donor surfaces, two turkey or ham pieces, ~1

cm thick, were aseptically sliced from the chub. The meat was cut into squares measuring 8 by 8
cm using a flame-sterilized knife blade. The L. monocytogenes inoculum (300 µl) was spread
evenly over the surface of one of the meat surfaces with an inoculating needle, so as to yield
8

2

~10 CFU/cm (subsequently verified). A second meat sample was surface inoculated with 300
µl of the GGP suspension. The inoculated deli meat slice cutouts were refrigerated at 7°C for 15
min prior to the transfer tests.

3.2.7

Transfer of L. monocytogenes from turkey/ham piece to stainless steel/ high-density
polyethylene coupons
At room temperature (~27ºC), uninoculated coupons (n = 12) were placed side-by-side,

and the inoculated meat sample (50-55 g) was placed, inoculated side down, on the first coupon.
Another coupon (36 g) was placed on top of the meat sample, with a 280 g weight on top of this
second coupon. This was done to add an equal and consistent normal contact force of ~5.8 g/cm
to each coupon. After 3 s of contact, the meat sample, coupon, and 280 g weight were removed
and placed on top of a second uninoculated coupon. This process was repeated for a total of 12
consecutive coupons. Each recipient coupon was put into a plastic Whirl-Pack® bag (Nasco

32

2

Sterile Whirl-Pack® Sampling Bags) with 25 ml of Phosphate Buffered Saline (PBS). The bags
were sonicated in a sonicating water bath (Model FS 20, 40 kHz sonic cleaner, Fisher Scientific,
Pittsburgh, Penn) for 10 min at 30°C. The liquid was plated (100 µl) on MOX, and incubated at
37°C for 48 h. All Listeria colonies were counted to determine CFU per coupon. Each test series
was conducted in triplicate.

3.2.8

Transfer of L. monocytogenes from stainless steel/ high-density polyethylene
coupons to turkey/ham piece
At room temperature (~27ºC), uninoculated 8 x 8 x 1 cm meat samples (n = 12) were

placed side-by-side, and a Listeria-inoculated coupon (36 g for SS and 81 g for HDPE) was
placed, inoculated side down, on the first recipient meat sample. A 280 g weight was placed on
2

2

top of the test coupon, to give a normal force of ~5 g/cm or 5.6 g/cm onto the sample. After 3 s
of contact, the coupon and 280 g weight were removed from the first meat recipient surface and
placed on top of a second uninoculated meat sample. This process was repeated for a total of 12
consecutive contacts. Each recipient sample was put into a plastic Whirl-Pack® Filter bag
(Nasco Sterile Whirl-Pack® Filter Sampling Bags) with 25 ml of PBS. The bags were stomached
for 1 min (Seward Stomacher 400, Worthing, West Sussex, England), and the liquid was then
spiral plated (100 µl) on MOX, which was incubated at 37°C for 48 h. Each test series was
conducted in triplicate. Populations were determined as the CFU per meat sample.

33

3.2.9

Cleaning and disinfecting coupons
Between tests, coupons were sanitized, rinsed with deionized water, and dried with a

composite tissue. The coupons then were autoclaved, followed by washing with detergent and
drying before use.

3.2.10 Transfer of GGP from turkey/ham piece to stainless steel/ high-density polyethylene
coupons
At room temperature (~27ºC), uninoculated test coupons (n = 12) were placed side-byside, and a GGP-inoculated meat sample was placed on the first recipient coupon. A 36 g coupon
2

and 280 g weight were placed on top of the meat sample for a normal force of ~5.8 g/cm . After
3 s of contact, the meat sample, coupon, and 280 g weight were removed from the first recipient
surface and placed on top of a second uninoculated surface. This process was repeated for a total
of 12 consecutive contacts. These studies also were conducted in triplicate. All recipient
coupons were photographed individually using a digital SLR camera positioned 31 cm above and
normal to the sample surface, and set to manual settings (Nikon-D3100 14.2-Megapixel
Camera,18-55 mm f/4-5.6 VR Image Stabilization Lens; shutter speed = 1/5 s, aperture = f-stop
5.6, ISO = 400). The only light source was a long-wave (365 nm) ultraviolet lamp (UVP Inc,
Model XX-15, 15 W bulb) positioned 25 cm to the side and 10º above the sample surface plane
being imaged. Settings were determined by trial for the most uniform and brightest results in
preliminary tests.

34

3.2.11 Transfer of GGP from stainless steel/ high-density polyethylene coupons to
turkey/ham piece
At room temperature (~27ºC), uninoculated meat samples (n = 12) were placed side-byside, and a GGP-inoculated coupon was placed on the first sample with a 280 g weight on top of
2

2

the coupon for a normal force of ~5 g/cm or 5.6 g/cm . After 3 s of contact, the coupon and 280
g weight were removed and placed on top of a second uninoculated meat sample. This process
was repeated for a total of 12 consecutive samples. Each recipient sample was photographed
individually, as described above.

3.2.12 Processing images to determine amount of GGP on surfaces
Image processing tools in MATLAB® (v7.6, The MathWorks, Natick, MA) were used to
quantify the amount of GGP (ppm) on each surface. The amount of GGP transfer was quantified
by an algorithm that determined specific thresholds for each meat type or coupon material
(changing the image to binary) to filter out background noise, followed by multiplying the binary
image pixel values by the original image pixel values and summing the pixel intensity in the
sample area. Based on preliminary testing, this method yielded the most accurate pixel count
and a useful sum of the intensity values that included only those pixels encompassing the GGP
visible in the image.
The images first were read and imported into the program and cropped to the size of the
sample surface. To correct for background noise in the images, red, blue, and green components
were evaluated. The red component was used to identify the GGP area, the blue component was

35

used to distinguish the background noise, and the green component was used to determine the
concentration of GGP on the surface.
Rimg=I_crop(:,:,1);
Gimg=I_crop(:,:,2);
Bimg=I_crop(:,:,3);
The blue component in the background was chosen to best distinguish the background from the
GGP, and therefore was converted into its binary form, using the following algorithm.
Bimg_bw=Bimg;
for i=1:(dx+1)
for j=1:(dy+1)
if (Bimg(i,j) == 255), Bimg_bw(i,j)=1; else Bimg_bw(i,j)=0;
The area of background illumination, sum of the pixels in the red component image, and
threshold to identify the effective green image were then determined.
B_area=sum(sum(Bimg_bw));
R_sum=sum(sum(Rimg));
R_thresh=R_sum/B_area;

36

By using the red component to identify the area containing GGP, the green component could be
isolated from the background surface. Doing so changed the image into its binary form.
Gimg_bkg=Gimg;
for i=1:(dx+1)
for j=1:(dy+1)
if (Rimg(i,j)>= R_thresh), Gimg_bkg(i,j)=0;
By blurring the green component image, as detailed below, the background noise was removed
without compromising the GGP affected area.
se=strel('disk',3); %Create morphological structuring element
h=fspecial('average',3); % Create predefined 2-D filter
[row, col]=size(Gimg_bkg);
sImg=imresize(Gimg_bkg, 0.4); % image reduction 1=full
bkg=imopen(sImg,se);
blurred=imfilter(bkg,h,'replicate'); % blurring background
blurred=imresize(blurred, [row,col]); % back to original image size
nGimg=Gimg-blurred; % background corrected image
This binary green component was then multiplied by the original image intensity values in order

37

to yield an intensity value that functioned as a good representation of the amount and intensity of
fluorescent area in the original image.
G_sum=sum(sum(nGimg));
I=G_sum;
The GGP intensity was related to the concentration by creating calibration curves for each
recipient surface. This was done by spreading 7 known serially diluted GGP concentrations
ranging from 49 to 100,000 ppm of GGP on the four surfaces (stainless steel, HDPE, ham, and
turkey) and taking images as previously described. The computed intensities were plotted against
concentration. By visual observation of the data, and preliminary tests, a second-order calibration
equation yielded sufficiently good results for this application.

3.2.13 Statistical analysis and model development
The sequential Listeria transfer results (mean values, LMobs, log CFU) and GGP transfer
results (log ppm) were related via a linear adjustment by offsets and multipliers. The offsets and
multipliers were determined by linear regression as follows:

LMpred = GGP × Multiplier + Offset

(3.1)

SSE =

(3.2)

𝑛
∑ 𝑖=1(LM 𝑝𝑟𝑒𝑑 − LM 𝑜𝑏𝑠 𝑖 )2
𝑖

RMSE =

�

𝑆𝑆𝐸
𝑛

; n = 12

(3.3)

38

3.3
3.3.1

Results and Discussion
Sequential transfer of L. monocytogenes inoculated surfaces to their counterparts
Sequential transfer of L. monocytogenes from an inoculated surface containing ~10

8

2

CFU/cm to uninoculated surfaces followed an approximately logarithmic decline in all cases
(Figure 3.1). According to the student t-test, there were no significant differences (P<0.05)
between the average recovery for the turkey and stainless steel surface transfer pair; however,
other pairs are significantly different from one-another. There were, based on an analysis of
covariance (ANCOVA), significant differences (P<0.05) between the average slopes of L.
monocytogenes recovery from the four surfaces: stainless steel, HDPE, ham, and turkey;
suggesting that knowledge of surfaces involved in a transfer event is critical to determining
amount of transfer. The figure shows the mean of triplicate experiments, with all standard errors
of replication ≤ 0.59 log CFU.

39

Number of L. mono. (log
CFU)

0
9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0

2

0

2

SS to T

4
6
8 10
contact number

T to HDPE

Number of L. mono. (log
CFU)

T to SS

H to SS

0

HDPE to T

4
6
8 10
contact number

9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0

12

Number of L.mono. (log
CFU)

Number of L. mono. (log
CFU)

9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0

12

9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0

2

4
6
8 10
contact number

H to HDPE

0

2

SS to H

12

HDPE to H

4
6
8 10
contact number

12

Figure 3.1 Population of L. monocytogenes on the receiving surface following sequential
8
2
contacts when the donor surface was inoculated at ~10 CFU/cm (data are means of
triplicates).

In a similar study, E. coli O157:H7 was sequentially transferred between raw beef and
identical stainless steel and high-density polyethylene coupons. Similar logarithmic trends were
illustrated for each case; as well as surface pairs exhibiting different average transfer rates
depending on which served as the donor and recipient (Campos and others 2007). Zhinong and

40

others (2006c) also performed experiments evaluating sequential transfer from L. monocytogenes
inoculated ham to common conveyor belt materials: high density polyethylene, acetyl, and
polypropylene. Trends were logarithmic as well; however, there were no differences in the
transfer rates between materials, which was inconsistent with the data in this study. This could be
due to the relative similarity in properties between the conveyor materials within that study. Both
of these prior studies also reported more rapid reduction in transferred populations (3.0-4.0 log
and 2.0-2.5 log in 10 contacts) than in this study, likely due to the use of a greater normal force
(1800 g and 340 g versus 280 g used in this study) (Campos 2007; Zhinong. 2006).

3.3.2

Calibration curves for GGP on all surfaces
Cross-contamination studies have increased over the past few decades; however,

quantification of bacterial transfer has not been well reported due to the inability to intentionally
contaminate a food facility with a pathogen to follow its transfer in an actual food handling
environment. The use of GGP as a quantitative transfer surrogate could make these types of
studies feasible, at least within pilot-scale or simulated processing and retail environments. In
order to quantify the amount of GGP (Fig. 3.2), calibration curves were created for each unique
surface (Fig. 3.3, Table 3.1) by depositing dots of known concentrations on each surface and
photographing under controlled ultraviolet light. This allowed for quantification of GGP
regardless of image background. The intensities determined by image processing were related to
the known concentrations on each surface to create the curves.

41

(a)

0

Concentration (ppm)

Concentration (ppm)

3500
3000
2500
2000
1500
1000
500
0

3500
3000
2500
2000
1500
1000
500
0

50
100
150
Intensity x100000

(c)

0

3500
3000
2500
2000
1500
1000
500
0

(b)

0

Concentration (ppm)

Concentration (ppm)

Figure 3.2. Stainless steel coupons, under ultraviolet light, after sequential contacts of GGP
inoculated turkey samples.

3500
3000
2500
2000
1500
1000
500
0

(d)

0

100
200
X10000
Intensity

200
400
Intensity X1000

200
400
Intensity X10000

Figure 3.3 Calibration curves for image intensity vs. concentration of GGP for (a) Stainless
steel, (b) Turkey, (c) Ham, (d) HDPE.

42

Table 3.1. Intensity vs. concentration calibration curve equations and goodness of fit for
stainless steel, high density polyethylene, turkey, and ham recipient surfaces (y =
concentration in ppm; x = sum of pixel intensities).

Recipient Surface

2

R

Equation
-11 2

-05

-12 2

-05

-12 2

-06

-12 2

-05

Stainless Steel

y = 1.1 × 10

High Density Polyethylene

y = 3.7 × 10

Turkey

y = 2.4 × 10

Ham

y = 1.1 × 10

x – 1.4 × 10

x – 2.4 × 10

x + 6.2 ×10

x + 7.3 ×10

x

0.94

x

0.96

x

0.99

x

0.99

Factors that will affect the feasibility of this image-based method for quantifying GGP
include imaging distance, angle, light, and camera settings, because the amount of GGP being
quantified is based solely on the images. Therefore, it is critically important to control the image
acquisition variables

3.3.3

Relating transfer of L. monocytogenes and GGP to and from delicatessen meats and
surfaces
Sequential transfer of L. monocytogenes from an inoculated material containing ~10

8

CFU/cm2 to uninoculated material followed an approximately logarithmic decline in most cases.

43

GGP also exhibited a logarithmic trend that was adjusted upward, with offsets and multipliers
determined by minimizing the root mean square error for each case; this was done first for each
individual donor-recipient surface pair and then by aggregating the data based on just the
recipient (i.e., imaged) surface. When treating each donor-recipient pair separately (Figure 3.4, ah), the offset and multiplier values (Table 3.2) were significantly affected by the pairing (P <
0.05; by ANCOVA); however, when comparing the values among different surface pairings with
a common recipient surface, there were no significant differences (P > 0.05). When analyzing
the data pooled by recipient surface (Table 3.3), RMSE generally increased, and R2 generally
decreased, as would be expected; however, the predictive ability of the method (Figure 3.5) was
still reasonable. Clearly, predictions based on specific donor-recipient pairs (Table 3.2 and
Figure 3.4) were statistically superior to those based solely on the recipient surface (Table 3.3
and Figure 3.5). However, in practical application, the image will be taken only of the recipient
surface, with the donor surface not necessarily known; therefore, the utility of the latter method
is the more important analysis.

44

Table 3.2. Model parameters and statistics for predicting L. monocytogenes transfer with
GGP for each transfer case.
Transfer
surfaces
T to SS

Offset
(log CFU)
2.29

Multiplier
(log CFU/ log ppm)
1.49

RMSE
(log CFU)
0.156

R
0.91

SS to T

5.01

0.768

0.216

0.68

T to HDPE

4.99

0.504

0.107

0.94

HDPE to T

5.07

0.997

0.208

0.71

H to SS

2.18

1.27

0.207

0.81

SS to H

3.07

0.848

0.201

0.79

H to HDPE

4.74

0.512

0.149

0.74

HDPE to H

3.63

0.653

0.237

0.29

45

2

(a)

Actual L. mono.
(log CFU)

Actual L. mono.
(log CFU)

9.0
8.0
7.0
6.0

7.0
6.0
5.0

6.0

9.0

5.0
6.0
7.0
8.0
9.0
GGP Predicted L. mono (log CFU)
9.0
(d)
8.0
7.0
6.0
5.0
6.0
7.0
8.0
9.0
GGP predicted L. mono (log CFU)

(e)

Actual L. mono.
(log CFU)

Actual L. mono.
(log CFU)

7.0

5.0

5.0
6.0
7.0
8.0
9.0
GGP predicted L. mono (log CFU)

8.0
7.0
6.0

9.0

(f)

8.0
7.0
6.0
5.0

5.0

5.0(h) 6.0
7.0
8.0
9.0
GGP Predicted L. mono (log CFU)

5.0
6.0
7.0
8.0
9.0
GGP Predicted L.mono (log CFU)
9.0 (g)

Actual L. mono.
(log CFU)

Actual L. mono.
(log CFU)

Figure 3.4.

8.0

5.0

5.0
6.0
7.0
8.0
9.0
GGP predicted L. mono (log CFU)
9.0
(c)
8.0

Actual L. mono.
(log CFU)

Actual L. mono.
(log CFU )

5.0

9.0 (b)

8.0
7.0
6.0

9.0

(h)

8.0
7.0
6.0
5.0

5.0

5.0
6.0
7.0
8.0
9.0
5.0
6.0
7.0
8.0
9.0
GGP Predicted L. mono (log CFU)
GGP Predicted L. mono (log CFU)
Figure 3,4 Predicted vs. actual L. monocytogenes surface counts for: (a) Turkey to stainless
steel, (b) Stainless steel to turkey, (c) Turkey to HDPE (d) HDPE to turkey, (e) Ham to
stainless steel, (f) Stainless steel to ham, (g) Ham to HDPE, and (h) HDPE to ham.

46

Table 3.3. Model parameters and statistics for predicting L. monocytogenes transfer by
GGP for each recipient case.

Stainless Steel

Offset
(log CFU)
2.23

Multiplier
(log CFU/ log ppm)
1.38

RMSE
(log CFU)
0.436

R
0.54

High Density
Polyethylene

4.62

0.59

0.157

0.91

Turkey

5.98

0.53

0.345

0.29

Ham

1.91

1.05

0.265

0.76

Recipient Surface

47

2

Actual L. mono. (log
CFU)
9.0

9.0 (c).
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
5.0
6.0
7.0
8.0
9.0
GGP Predicted L. monocytogenes
(log CFU)

Actual L. mono. (log
CFU)

Actual L. mono. (log
CFU)
Actual L. mono. (log
CFU)

9.0
8.5 (a)
8.0
7.5
7.0
6.5
6.0
5.5
5.0
5.0
6.0
7.0
8.0
GGP Predicted L.
monocytogenes (log CFU)

9.0
8.5 (b)
8.0
7.5
7.0
6.5
6.0
5.5
5.0
5.0

9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0

6.0
7.0
8.0
GGP Predicted L.
monocytogenes (log CFU)

9.0

(d)

5.0
6.0
7.0
8.0
9.0
GGP Predicted L. monocytogenes
(log CFU)

Figure 3.5 Image-predicted vs. actual L. monocytogenes surface counts for recipient
surfaces: (a) Stainless steel, (b) HDPE, (c) Turkey, and (d) Ham.

Although Listeria numbers in actual food handling environments would likely be much
lower, very high concentrations of both L. monocytogenes and GGP were used in this study so
that numbers were easily detectable. GGP transfer curves were therefore compared to L.
monocytogenes transfer data from a prior study (Zhinong and others 2006a) that evaluated the

48

7

sequential contact from ham to HDPE surfaces at lower inoculation levels of 10 and 10

4

2

CFU/cm . To accomplish this, the model slope determined above (for GGP transfer from ham to
HDPE, 0.512 log CFU/ log ppm) was held constant, but the model intercept determined above
was reduced by the appropriate value to compensate for the difference in inoculation levels, as
8

7

4

compared to the 10 used in this study (i.e., reduced by 1 for the 10 data and by 3 for 10 data.
2

The resulting prediction errors were < 0.55 log CFU/cm . Although greater than the calibration
errors (Table 3.2), this is a reasonable value, relative to the desired application, thereby
validating the utility of GGP as a surrogate for L. monocytogenes transfer when initial
concentrations are known.

Table 3.4. Prediction error (RMSE) when comparing GGP based predictions to LM
4
7
2
transfer from ham to HDPE at two different inoculation levels (10 and 10 CFU/cm )
(Zhinong and others, 2006a). Model parameters and statistics for predicting L.
monocytogenes transfer by GGP for ham to HDPE at different inoculation levels.

multiplier
(log CFU/ log ppm)
7

10
51% mc
62% mc
73% mc

0.512

4

10

0.512

offset
2
(log CFU/cm )
7

10

3.74

4

10

0.74

RMSE
2
(log CFU/cm )
7

4

10
10
0.49
0.54
0.39
0.51
0.53
0.47

In the present study, transfer of L. monocytogenes yielded a smaller slope than did GGP
in all cases, meaning that the bacteria persisted longer on the donor surface, which was

49

accounted for by statistically relating L. monocytogenes to the corresponding GGP transfer data.
This demonstrated that GGP could be used as a physical surrogate for Listeria transfer when
considering the change in populations from an initial contamination level (rather than an absolute
value).
As an image-based method for using physical surrogate, GGP, for the transfer of Listeria
monocytogenes will enable subsequent studies that will contribute to the improvement of risk
assessments as well as worker training. Currently, microbial risk in a commercial facility is
based on probabilities, assumptions, and computational simulations that rarely can be validated
with in situ observations (Ivanek and others 2004; Schaffner 2004). A simple model based on a
contaminated source, intermediate source, and food that contacts the intermediate source has
been established by (den Aantrekker and others 2003). In that work, transfer rates varied greatly
due to multiple factors that affect bacterial transmission, as well as experimental differences and
recovery methods used (Perez-Rodriguez and others 2008). Ivanek and others (2004) quantified
transfer and proliferation of L. monocytogenes through a fish processing plant by simulation.
That simulation showed prevalence increased over a shift, and the confidence intervals became
wider overtime as a result of uncertainties in the system; however, it did not model loss of
contamination; therefore, numbers were higher than actual numbers, making it an overly
conservative estimation. Using a physical surrogate and the associated technique described here
for microbial transfer is a relatively simple and rapid method to predict the change of L.
monocytogenes populations in an actual food handling system after multiple contacts, which was
the goal of this project.

50

3.4

Conclusions
Surface-to-surface transfer of L. monocytogenes and the proposed physical surrogate,

GGP, was related by regression analysis, based on image analysis of GGP transfer. This was
shown to be a feasible method to rapidly quantify risk, based on the change in populations of L.
monocytogenes, given a known initial concentration and surface type.
Quantifying this surrogate during experimental food handling scenarios and using it to
predict pathogen transfer risk provides the capability to quantitatively map actual crosscontamination routes in food facilities, thus giving a method to fill gaps in the data for risk
assessments and validate risk models. This can be used as a tool to educate food service workers
in food safety while determining the effects of worker training interventions in a retail foodhandling environment. GGP image analysis is a valuable approach to assess the strength of
control measures, and monitor their impact on food safety.

51

4.
APPLYING MULTIPLE MODELS TO IDENTIFY COMMON
CHARACTERISTICS IN LISTERIA MONOCYTOGENES TRANSFER DATA
BETWEEN READY-TO-EAT MEAT PRODUCTS AND CONTACT SURFACES

4.1

Introduction
The number of papers published on bacterial transfer to/from food has increased

approximately ten-fold over the past two decades (Benoit 2013), reflecting an increasing
attention to this important issue. Listeria monocytogenes is of particular concern for prepared
foods, due to its ability to survive and proliferate in cold temperatures (Lunden and others 2002).
A recent risk assessment determined that listeriosis resulting from delicatessen products sliced in
retail establishments is a significant concern, and therefore a high priority target for improvement
(FSIS 2010b).
Multiple contacts between food and equipment/environmental surfaces are inevitable in
processing, retail, and/or foodservice environments. At the retail level, sources of contact can
include workers, utensils, slicing equipment, other food products, countertops, and cutting
boards, which are often overlooked or poorly sanitized (Perez-Rodriguez and others 2008). Many
factors have been suggested to affect transfer rates between these surfaces, including inoculum
size, physicochemical factors, and material characteristics (Lin and others 2006; Montville and
Schaffner 2003; Chen and others 2001; Midelet and Carpentier 2002; Rodriguez and others
2007a). However, since most research studies are conducted independently, using different
methodologies and test variables, conflicting results and conclusions have been reported (Arnold
and Bailey 2000; Rodriguez and others 2007b). This approach inherently leaves uncertainty and
therefore knowledge gaps in this domain.
52

There is a critical need for standardization of methods and/or aggregation of data in this field.
Literature addressing bacterial transfer to/from food products and contact surfaces is almost
exclusively empirical, without any theoretical connection to the underlying mechanisms that
govern these processes. Vorst (2005) and Sheen and Hwang (2008) modeled Listeria transfer
between delicatessen slicers and ready-to-eat meat products, mentioning a variety of variables
that impact transfer rates, including: direction of transfer, inoculation level, product composition,
product surface characteristics, blade characteristics, and blade speed/force. Although transfer
curves were analyzed differently based on some of these variables, phenomenological
explanations for how they affect rates of transfer were not discussed (Sheen and Hwang 2008;
Vorst 2005). These empirical models consider few parameters for transfer rate estimation, and
each author noted that additional parameters that affect bacteria transfer are needed. Hoelzer and
others (2012) aggregated a broad set of transfer data including the data described above and
applied similar coefficient-based mathematical models, but additionally assessed probability
distributions of transfer coefficients across studies. This is an important first step in evaluating
cross-contamination dynamics; however, underlying factors affecting transfer still must be
understood to explain why transfer coefficients and probability curves differ with differing
parameters.
Utilizing a modeling framework that bridges micro- and macro-scale knowledge would
support a unifying theory of bacterial transfer between foods and relevant food contact surfaces.
Biologically meaningful parameters that help elucidate phenomenological explanations are
necessary to further understand and analyze transfer occurring between meat and contact
surfaces. Studies that generally characterize surfaces are limited and only briefly discuss the

53

possible effects on bacteria transfer. For example, Aarnisalo and others (2007) and Vorst and
others (2006a) characterized roughness and wear of slicer blades throughout experimentation,
and Rodriguez and others (2008) used atomic force microscopy to evaluate topography of
different finishes on stainless steel coupons. Midelet and Carpentier (2002) measured contact
angles of all surfaces used to determine material hydrophobicity. Identifying these material
characteristics and if/how and the extent to which they affect bacterial transfer is an important
part of studying cross-contamination and must be done to properly establish transfer models.
Given that the public safety concerns surrounding Listeria in delicatessen environments and
the limited published data assessing its transfer, gaps in phenomenological understanding need to
be assessed. Therefore, the objectives of this work were to: (i) develop a foodborne pathogen
transfer database for comparisons across studies, and (ii) quantitatively compare three candidate
models in order to identify phenomenological differences in pathogen transfer response between
common meat products and equipment surfaces, including: mechanical slicers, kitchen knives,
cutting boards, conveyor belts, and countertops. The overall goal is to demonstrate that
aggregation of foodborne pathogen transfer data from many sources can significantly improve
current understanding of the factors affecting transfer.

4.2

Materials and Methods
The general approach was to compile Listeria transfer data from prior studies that evaluated

static contact, slicer, and knife slicing, and then analyze the aggregated data sets. The analysis
(detailed below) entailed fitting three different models to each transfer curve, and quantitatively

54

evaluating the resulting parameters and goodness of fit. The core purpose was not to simply fit
multiple models, but rather to use the models to identify any phenomenological characteristics of
the transfer curves that were consistent for a particular transfer scenario (i.e., product and contact
type) across multiple studies.

4.2.1

Data classification

Published bacterial transfer data were characterized in terms of product type, contact surface,
and means of transfer, and included only data for ready-to-eat (RTE) meat products and common
food contact surfaces. The variables considered included: means of transfer (mechanical slicing,
knife slicing, sequential static contacts) and meat type (RTE ham, turkey, salami, and bologna).
The available data encompassed approximately 75 data sets, with 253 total replications, and
5,838 individual observations from 6 publications, by 5 different authors in a single laboratory,
distributed as described in Table 4.1. Although additional published transfer studies exist, the
ability to access the associated data is currently limited. The data acquired were organized into a
database (TranBase, Figure 4.1), reflecting the construct of ComBase, a well-known database for
microbial response (i.e., growth, survival, inactivation) data (Baranyi and Tamplin 2004).

55

Figure 4.1 A screen-shot of the prototype database (TranBase) cataloging bacterial transfer data. Note the data fields and
drop-down menus for sorting (e.g., as shown for choosing the type of transfer process).

56

Table 4.1 Classification of L. monocytogenes data aggregated and considered for transfer
analysis based on transfer method and meat product type.

Contact type:

n

Sequential Static Contacts

30

Kitchen Knife

8

Mechanical Slicer

37

Product type:
Turkey
Ham

29

Salami

14

Bologna

4.2.2

30

2

Modeling pathogen transfer between meat and surfaces during mechanical slicing,
knife slicing, and sequential static contacts
In the previously published transfer studies, a range of experimental and analytical

methods were used, and different test variables were considered. Additionally, each study
tended to apply different models to describe “transfer curves” (i.e., number of bacteria on
receiving surfaces, such as meat slices, after repeated contacts with a donor surface).
Consequently, it is difficult to draw broad inferences across these prior studies. Therefore, using
the data available, three empirical models were chosen as follows:

57

Log-Linear: logN = -(k×n) + logNo

(4.1)

p

Weibull: logN = -(k×n ) + logNo

(4.2)

Linear-Weibull: if n < nc , then logN = logNo - k×n,

(4.3)

p-1

else logN = logNo – (k/(p × nc

p

) × n + (1-p)/p × k × nc

Where n is the slice or contact number; N is the bacterial concentration of slice n; No is a
regression parameter (the intercept); k is a regression parameter related to the slope of the Log N
versus n curve; p is a shape factor, c is a second regression parameter related to the slope of the
Log N versus n curve; and nc is the critical value where the curve shape changes for the twophase model (described below).
The log-linear model has been used in previous studies to describe attachment strength of
bacteria on surfaces during successive transfers to plates by contact (Eginton and others 1995).
The Weibull curve was also used to describe pathogen cross- contamination during common
food service tasks (Chen and others 2001) . Perez-Rodriguez and others (2007) used both a loglinear model and a Weibull model to describe transfer of Escherichia coli during slicing at two
different inoculation levels. To our knowledge, a linear-Weibull two-phase model has not been
previously used to explain bacterial transfer. However, Vorst and others (2006a) noted that,
during mechanical slicing, most transfer occurs during the first 10-15 slices, with the transfer rate
decreasing thereafter. Campos and others (2009) previously noted that such transfer curves had a
distinctly log-linear decline for the first ~10 slices, followed by tailing. They theorized that this

58

distinct change in curve shape was due to the accumulation of meat residues on the slicer blade
(Campos and others 2009); however, this hypothesis was not previously tested quantitatively.
Therefore, a two-phase linear-Weibull curve with a distinct critical value (i.e., slice number)
where the transfer response curve changes from linear to non-linear was considered for this
analysis. In cases where the linear-Weibull fit was forced via regression and yielded a low
critical value (e.g., 2), that result indicated that the two-phase model should not be considered to
be the most likely model, as such a result actually indicates a rejection of this model in
describing the response curve.
Several publications have generated transfer coefficients based on empirical models
describing transfer during slicing of ham, salami, and ‘gravad’ salmon under exclusive
conditions (Aarnisalo and others 2007; Sheen 2008; Sheen and Hwang 2010), but were not
considered here due to the specificity of certain parameters on transfer rates. By applying
quantitative, statistical comparisons of the three models across the aggregated data sets, the goal
was to elucidate inherent differences in the general nature of transfer for the various cases, not
just re-fit curves to the transfer data.

4.2.3

Fitting the equations to experimental data

Replicate data from individual studies were pooled, and the three models were fit by
minimizing the root mean squared error (RMSE) in Excel using Solver (Version 14.2.4,
Frontline Systems, Incline Village, NV), where:

59

2

SSE= Σ(Model predicted value - Experimental value)

(4.4)

RMSE = �

(4.5)

𝑆𝑆𝐸
𝑛

where n is the total number of observations in the analyzed set.

The Akaike Information Criterion (AICc) (Motulsky and Christopoulos 2004) was used in
addition to the RMSE to determine the best model fit. AICc , which quantifies model correctness
by taking into account both the error and the number of parameters, includes the models’ sum of
squared residual errors (SSE), number of parameters (K), and number of data points (n). The
Akaike model used for small samples (n/K<40) is:

AICc = -n×ln(SSE/n) + 2×K + (2×K×(K+1))/(n-K-1)

(4.6)

The Akaike values for the three models were normalized based on measuring distances from
the minimum Akaike value of the set (Eq. 4.7), which is the best fitting model, followed by
determining the Akaike weights, wi , of each candidate model (Eq. 4.8). These weights represent
the probability that a model is the ‘most likely’ fit out of the tested candidate models and is based
on the absolute difference between AICc scores, Δi.:

60

Δi = AICc,i – min AICc

𝑤𝑖 =

(4.7)

𝑒 −0.5∗𝚫 𝑖

Σ 𝑒 −0.5∗Δ𝑖

(4.8)

Δi = AICc difference between best fitting model and other candidate models i
AICc.i = AICc for candidate model i
min AICc = minimum AICc when all candidate models are compared

Having done this for each pooled data set and each model, the most likely model was chosen
for each data set based on the minimum RMSE and the highest weighted AICc. Subsequently,
the fraction of data sets for which a given model was selected as best was determined for each
transfer scenario and product type.

4.2.4

Statistical comparisons
All Listeria transfer curves considered had a minimum of three replicate datasets. Data

were evaluated by pooling the replications and fitting a single curve in each model case.

61

Analyses of variance (ANOVA) were conducted to determine significant differences between
two-phase critical contact values (nc) due to product type, transfer scenario, and contact surface
material, to establish whether there were consistent characteristics of the transfer curves for a
given scenario that might relate to inherent, phenomenological explanations for the observed
transfer responses.

4.3

Results and Discussion

4.3.1

Interpretation of models on transfer data

A total of 225 curve fits were performed, fitting each of the three models to the pooled replicate
data sets. Utilizing replicates accounted for experimental errors as well as model errors when
selecting the most likely model to explain the curve. A summary of the RSME and AICc results
for each curve fit are organized in Table 4.2. Examples of most-likely model curve fits are
illustrated in Figures 4.2 and 4.3.
Table 4.2 Summary of L. monocytogenes transfer and model statistics for (a) slicer
data, (b) kitchen knife, and (c) static contact data considered in this study.
(a)

Surf. 1

Surf. 2

Weighted AICc

RMSE
Weibull

Linear

LinearWeibull

Weibull

Linear

nc

LinearWeibull

1 Slicer

Ham

*0.53

0.54

0.54

*0.78

0.20

0.02

6

2 Slicer

Ham

0.48

0.49

*0.48

*0.72

0.03

0.25

11

3 Slicer

Ham

0.49

0.56

*0.49

0.33

0.00

*0.67

12

62

Table 4.2a (cont’d)
3
4 Slicer

Ham

0.56

0.57

*0.56

*0.68

0.07

0.25

5

5 Slicer

Ham

0.51

0.54

*0.51

*0.50

0.00

0.50

14

6 Slicer

Ham

0.50

0.53

*0.49

0.28

0.00

*0.72

8

7 Slicer

Ham

0.53

0.54

*0.53

*0.57

0.16

0.27

6

8 Slicer

Turkey

0.53

0.63

*0.41

0.00

0.00

*1.00

6

9 Slicer

Turkey

0.51

0.54

*0.51

*0.73

0.00

0.26

11

10 Slicer

Turkey

0.58

0.62

*0.57

*.63

0.00

0.37

16

11 Slicer

Turkey

0.60

0.70

*0.58

0.05

0.00

*0.95

5

12 Slicer

Turkey

0.66

0.67

*0.65

0.16

0.30

*0.54

18

13 Slicer

Turkey

0.48

0.55

*0.45

0.02

0.00

*0.98

8

14 Slicer

Turkey

0.63

0.68

*0.61

0.22

0.00

*0.78

8

15 Slicer

Turkey

0.56

0.63

*0.56

*0.61

0.00

0.39

4

16 Slicer

Turkey

*0.53

0.59

0.53

*0.82

0.00

0.18

3

17 Slicer

Turkey

*0.43

0.45

0.43

*0.65

0.14

0.20

4

18 Slicer

Turkey

0.50

0.57

*0.49

0.31

0.00

*0.69

17

19 Slicer

Turkey

0.62

0.69

*0.59

0.09

0.00

*0.91

5

20 Slicer

Salami

0.34

0.37

*0.34

0.33

0.00

*0.67

2

21 Slicer

Salami

0.59

0.60

*0.58

0.23

0.36

*0.41

2

22 Slicer

Salami

0.47

0.52

*0.46

0.26

0.00

*0.74

2

23 Slicer

Salami

0.73

0.79

*0.62

0.00

0.00

*1.00

2

24 Slicer

Salami

0.45

0.44

*0.44

0.08

*0.83

0.09

0

25 Slicer

Salami

0.46

0.46

*0.46

0.22

*0.69

0.10

2

26 Slicer

Salami

0.56

0.56

*0.56

0.23

*0.69

0.08

3

63

Table 4.2a (cont’d)
26
27 Slicer

Salami

0.60

*0.59

0.59

0.10

*0.81

0.09

2

28 Slicer

Salami

0.69

0.69

*0.69

0.21

*0.64

0.15

2

29 Slicer

Salami

*0.51

0.51

0.51

0.47

*0.49

0.04

2

30 Slicer

Salami

0.60

0.61

*0.60

*0.51

0.24

0.26

2

31 Slicer

Salami

0.52

0.52

*0.52

0.22

*0.70

0.08

0

32 Slicer

Turkey

*0.75

0.79

0.75

*0.76

0.02

0.21

9

33 Slicer

Salami

0.35

0.35

*0.34

0.22

*0.61

0.18

0

34 Slicer

Bologn
a

*0.69

0.69

0.69

0.24

*0.69

0.07

2

35 Slicer

Turkey

0.69

0.71

*0.68

0.37

*0.37

0.26

4

36 Slicer

Salami

0.92

0.93

*0.92

0.27

*0.62

0.10

1

37 Slicer

Bologn
a

0.47

0.47

*0.47

0.21

*0.72

0.07

3

64

Table 4.2 (cont’d)

(b)

Surf.
1

Surf. 2

Weighted AICc

RMSE
Weibull

Linear

LinearWeibull

Weibull

Linear

nc

LinearWeibull

38 knife

Turkey

0.77

0.80

*0.74

0.28

0.19

*0.53

10

39 knife

Turkey

*1.06

1.12

1.06

*0.66

0.17

0.17

7

40 knife

Turkey

1.47

1.53

*1.46

*0.52

0.26

0.23

9

41 knife

Turkey

1.01

1.08

*1.00

*0.59

0.08

0.33

7

42 knife

Turkey

0.74

0.82

*0.71

0.31

0.01

*0.69

7

43 knife

Turkey

0.95

0.99

*0.94

*0.52

0.24

0.25

8

44 knife

Turkey

0.66

0.72

*0.66

*0.66

0.04

0.30

9

45 knife

Turkey

0.95

0.98

*0.94

*0.54

0.25

0.21

5

65

Table 4.2 (cont’d)

(c)

Surf. 1

Surf. 2

Weighted AICc

RMSE
Weibull

Linear

LinearWeibull

Weibull

Linear

nc

LinearWeibull

46 T

SS

0.67

0.86

*0.58

0.00

0.00

*1.00

1

47 SS

T

*0.66

0.74

0.66

*0.77

0.01

0.23

1

48 H

SS

*0.20

0.50

0.21

*0.91

0.00

0.09

1

49 SS

H

0.60

0.63

*0.60

*0.68

0.11

0.21

1

50 T

HDPE

0.58

0.77

*0.39

0.00

0.00

*1.00

0

51 HDPE

T

0.59

0.67

*0.52

0.01

0.00

*0.99

0

52 H

HDPE

*0.23

0.62

0.25

*1.00

0.00

0.00

0

53 HDPE

H

0.39

0.39

*0.32

0.00

0.00

*1.00

0

54 SS

T

0.38

0.43

*0.38

*0.74

0.03

0.23

0

55 T

SS

0.31

0.31

*0.30

0.19

*0.72

0.09

0

56 T

HDPE

*1.46

1.49

1.46

0.33

*0.58

0.08

1

57 HDPE

T

0.22

0.25

*0.22

*0.92

0.02

0.06

0

58 Ham

PP

0.23

0.24

*0.23

*0.61

0.25

0.14

0

59 Ham

PP

0.21

0.22

*0.21

*0.53

0.32

0.15

0

60 Ham

HDPE

1.07

1.07

*1.07

0.23

*0.71

0.06

0

61 Ham

HDPE

0.29

0.29

*0.29

0.29

*0.65

0.06

0

62 Ham

AC

*0.32

0.33

0.33

*0.52

0.40

0.08

0

63 Ham

AC

0.43

0.46

*0.43

*0.69

0.09

0.23

0

64 Ham

PP

0.38

0.41

*0.38

*0.74

0.04

0.22

2

65 Ham

AC

0.27

0.27

*0.27

0.32

*0.50

0.18

0

66 Ham

HDPE

0.40

0.40

*0.40

0.22

*0.71

0.06

2

66

Table 4.2c (cont’d)

67 Ham

PP

0.32

0.45

*0.30

0.05

0.00

*0.95

0

68 Ham

AC

0.74

0.74

*0.73

0.24

*0.67

0.10

0

69 Ham

HDPE

0.69

0.70

*0.48

0.42

*0.44

0.14

0

70 Ham

AC

*0.53

0.55

0.53

0.03

0.02

*0.96

1

71 Ham

HDPE

0.20

0.37

*0.17

0.00

0.00

*1.00

0

72 Ham

PP

0.28

0.45

*0.25

0.00

0.00

*1.00

0

73 Ham

AC

0.46

0.67

*0.40

0.00

0.00

*1.00

3

74 Ham

HDPE

0.32

0.46

*0.30

0.00

0.00

*1.00

21

75 Ham

PP

0.43

0.46

*0.43

*0.63

0.05

0.33

4

*Most likely model based on lowest RMSE or highest weighted AICc.
Datasets 1-7 from (Zhinong and others 2006b), 8-31 from (Keskinen and others 2008), 32-37
from (Vorst and others 2006a), 38-45 from (Keskinen and others 2008), 46-57 from (Benoit
2013), 58-75 from (Zhinong and others 2006a).

67

5.0
4.0
3.0
2.0
1.0
0.0
0
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0

5
(b)
(b)

0
L. monocytogenes transfer
(log CFU/slice)

10 15 20 25 30
contact number

8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0

5
10
contact number

15

(c)

0

5

10 15 20 25 30
contact number

L. monocytogenes transfer
(log CFU/slice)

6.0

L. monocytogenes transfer
(log CFU/cm^2)

7.0

L. monocytogenes transfer
(log CFU/cm^2)

L. monocytogenes transfer
(log CFU/slice)
L. monocytogenes
transfer (log CFU/slice)

(a)

8.0

3.0
2.5
2.0
1.5
1.0
0.5
0.0
0

5

10 15 20 25
contact number

0

5

0

5

30

7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
10 15 20 25
contact number

30

7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
10 15 20 25 30
contact number

Figure 4.2 Examples of evaluated L. monocytogenes data sets and fits of the most likely
models: (a) Linear, (b) Linear-Weibull, and c) Weibull.

68

L. monocytogenes transfer
(log CFU/slice)

8

Data

7

linearWeibull
Weibull

6
5
4
3
2
1
0
0

10
20
contact number

30

Figure 4.3 An example of a single evaluated dataset fitted to all three candidate models.

4.3.2

Mechanical slicing transfer characteristics
The two-phase (linear-Weibull) model has the lowest (best) RMSE for ~80% of the slicer

data (Table 4.3b). However, when considering the AICc of the aggregated slicer data, no model
was the superior choice for the majority of the cases (Table 4.3a). Because the two-phase model
would be expected to have the lowest RMSE due the greater number of parameters, the AICc is
the better criterion for evaluating correctness of the models, and will be the primary measure
considered throughout the analysis.

69

Table 4.3 The number of datasets in each transfer scenario most likely to be correct in each
model case based on (a) AICc and (b) RMSE.
AICc

(a)

Knife

Slicer

Contacts

Weibull

6

12

17

Linear

0

12

8

Linear-Weibull

2

13

13

SUM

8

37

38

Knife

Slicer

Contacts

Weibull

1

6

9

Linear

0

1

2

Linear-Weibull

7

30

29

SUM

8

37

38

RMSE

(b)

Inconsistency in selection of the best model for the slicer data, based on AICc, suggests
that other factors may affect bacterial transfer during the mechanical slicing of delicatessen
meats. The authors of the original studies noted that the amount of soiling on the slicer blade was
dependent on product composition, thus affecting the transfer results. (Vorst and others 2006a;
Lin and others 2006; Keskinen and others 2008). Vorst (2006a) reported the percent moisture
and fat of each product before slicing, and noted that salami, which was relatively high in fat,

70

created a visible film, while the low fat turkey had a washing effect. Consequently, sliced
products were categorized to determine whether there was a noticeable trend between transfer
curves and product characteristics (Table 4.4). According to the AICc, the Weibull model best
described turkey and ham transfer, while the linear curve best described salami and bologna
transfer, thus supporting the hypothesis that product composition affects the overall transfer
response of Listeria from product to slicer.

Table 4.4 The number of mechanical slicer datasets for each product type most likely to be
correct in each model case based on (a) AICc and (b) RMSE.
Slicer AICc

(a)

Turkey Ham Salami Bologna
Weibull

12

5

1

0

Linear

1

0

9

2

Linear-Weibull

7

2

4

0

SUM

20

7

14

2

Slicer RMSE

(b)

Turkey Ham Salami Bologna
Weibull

7

2

2

0

Linear

0

0

3

0

Linear-Weibull

13

5

9

2

SUM

20

7

14

2

71

Although the Weibull model was the most likely for most of the mechanical slicer data,
the linear-Weibull critical value (i.e., critical slice number) where the shape of the transfer
response changes from linear to non-linear was analyzed to determine whether the value was
affected by product type. The critical value for the aggregated slicer data ranged from 0 (for
linear data) to 18, with an average of 7.4 and standard deviation of 4.9. The variability was
largely due to the salami data that were more linear, having a low critical value of ~2, meaning
that the two-phase model was forced, and therefore the simpler linear model better describes this
type of transfer. This linear trend (Figure 4.2a) in salami data has been previously recognized in
work that commented on the layer of fat the salami left on the slicer blade surface, which
contributed to constant but lower transfer rates (Vorst and others 2006; Lin and others 2006).
Ham and turkey, on the other hand, had statistically similar critical values (P > 0.05), with an
average of ~9, which, due to their compositional similarities, as compared to salami, is plausible
(Figure 2b). Previous work considered the effect of product residue on the blade as a contributing
factor of curve-shape change. Campos and others (2009) qualitatively observed a bi-phasic
tendency that began tailing after ~10 slices, which roughly corresponded to meat product residue
accumulation of ~5 µm thickness. They noted that this critical value was product-specific,
decreasing with increasing fat content of the food (Campos and others 2009).
Some model fit variability in slicer data may be due to methodology differences between
studies and laboratories, as well as variables tested within a given study, which included biofilm
forming ability, culture health, inoculation level, desiccation time, force, blade type, and blade
age/wear. Different limits of detection and the manner in which this was reported also may be a
factor affecting the results. For example, Vorst (2006a) noted that at lower inoculation levels,

72

salami yielded significantly different transfer curves than at higher inoculation levels, unlike
turkey and bologna.
Empirical (curve fitting) models previously have been used in several commercial
delicatessen slicer studies, including some of the studies evaluated here. A coefficient-based
model developed by Vorst (2005) predicted L. monocytogenes transfer during slicing of RTE
delicatessen meats at different inoculation levels, assuming: (1) the number of Listeria cells
transferred from the blade to the meat during slicing was a fraction (f1) of the number of Listeria
cells on the blade just before each sequential slice, (2) the number of Listeria cells transferred to
surrounding areas was a different fraction (f2) of the number of cells on the blade just before
each sequential slice, and (3) the number of cells on the blade before any slicing begins was N0.
The resulting exponential decay model predicted the number of CFU transferred to any given
slice, as well as the number of CFU lost to the environment (including potential aerosols and
bacterial death), therefore predicting the number of CFU transferred to slice X. This model was
validated by Keskinen (2007) to accurately describe transfer to RTE delicatessen meats when
considering biofilm forming ability, time, and injury, noting changes in variance and total
cumulative transfer between studies without explanation. Curve fits also differed based on meat
and inoculation levels (Vorst 2005); however, both Vorst and Keskinen noted that most transfer
to the meat occurred during the first 10-15 slices. Sheen and Hwang (2008) also used exponential
empirical models to describe bacteria transfer during mechanical slicing of deli meats; however,
they also suggested that transfer results depended on inoculation level and contamination route.
These prior studies mentioned model limitations and recommended that other factors also be

73

included in the models; however, they did not investigate underlying factors that influence shape
of the transfer curves or use the model forms to quantitatively evaluate such differences.

4.3.3

Knife slicing data
According to the AICc of the transfer data analyzed, when meats were sliced with a knife

(i.e., single blade action, rather than the continual spinning action of a mechanical deli slicer), the
Weibull model was the best model in 6 of 8 cases (Table 4.3a). These data could not be
compared across meats, as done for the slicer studies, because insufficient datasets were
unavailable for this comparison. Therefore, it is uncertain whether the curve would be different
based on product composition. However, when the turkey data from the knife studies were
compared to those of the slicer studies, a statistically similar critical value (p > 0.05) of ~8 was
observed, as well as the same conclusions based on AICc and RMSE findings (Table 4.2),
suggesting that product composition had a greater impact on the transfer than the means of
cutting.

4.3.4 Sequential static contact data
Analysis of sequential static contact data between various types of meats and contact
surfaces was best described by the Weibull model according to the AICc (Table 4.3a). Both the
AICc and RMSE implied that selection of the best model was unaffected by product composition

74

and contact material (Table 4.5). Zhinong and others (2006) also observed no significant
difference in transfer rate between high density polyethylene, polypropylene, and acetylene
surfaces and ham. The average critical value from all of the static contact data analyzed in this
study was 1.5, indicating that the linear-Weibull was forced, and that a single-phase curve better
described static sequential static contact transfer. These critical contact values differed from
those of ham and turkey subjected to mechanical slicing and contact surfaces; however, in those
cases, there was likely accumulation of residue on the blade, causing the transfer curve to change
shape. In contrast, during sequential static contacts, excessive accumulation of product residue is
unlikely, and therefore does not affect the shape of the transfer curves.

75

Table 4.5 The number of sequential static transfer datasets for each product type most
likely to be correct in each model case based on (a) AICc and (b) RMSE.

Contacts AICc

(a)

Turkey Ham
Weibull

4

7

Linear

2

6

LinearWeibull

1

6

SUM

7

21

Contacts RMSE

(b)

Turkey Ham
Weibull

1

5

Linear

0

0

LinearWeibull

6

16

SUM

7

21

4.3.5

Effects on transfer due to product composition
When all slicer, kitchen knife, and sequential contact data were pooled by product type,

the effect on transfer became apparent. According to the AICc, transfer to turkey and ham, which
are similar in moisture and fat content, was generally best described by the Weibull model; while

76

transfer to salami and bologna, with much greater fat and lower moisture, were best described by
the linear model (Table 4.6a).

Table 4.6 The number of datasets for each product type most likely to be correct in each
model case based on (a) AICc and (b) RMSE.

(a)

AICc
Turkey Ham Salami Bologna

Weibull

17

12

1

0

Linear

3

6

9

2

Linear-Weibull

7

11

4

0

29

28

14

2

SUM

RMSE

(b)

Turkey Ham Salami Bologna
Weibull

8

17

2

0

Linear

1

1

3

0

Linear-Weibull

20

10

9

2

SUM

29

28

14

2

When comparing the critical contact values within products, the transfer response during
cutting or slicing was different from that of sequential contacts. Turkey, for example, had an
average critical contact value of ~9 for slicing, ~8 for kitchen knives, but ~1 for sequential static

77

contacts. This trend was similar for ham, which had a critical contact value of ~9 for slicing and
~1 for static contacts. This suggested that the build-up of product residue on the blade surfaces
during mechanical slicing or cutting changed the transfer response as hypothesized above.

4.3.6

Future utility of database
This preliminary database provides a compilation of published transfer data that will be

useful to scientists and risk assessors evaluating food pathogen cross-contamination. It contains
data published over the past 8 years in an accessible format and provides further details
concerning each study. This information previously was not available in this format, and offers a
baseline for understanding the mechanisms that govern these processes. Currently, models and
simulations for bacterial transfer through food processing environments are of macro-scale,
focused on probabilities of contact events and/or amount of transfer.
Schaffner (2004) studied cross-contamination by using Monte Carlo simulation to track
the number and prevalence of certain Listeria strains through a food processing plant. This
method is probabilistic and based on transfer rate distributions that are assumed to be related to
persistence and prevalence only. Ivanek and others (2004) described a compartmental model for
Listeria cross-contamination that used a difference equation system to model transfer through a
fish processing plant. This simulation focused on prevalence, frequency, number of contaminated
units, and duration probabilities and assumptions. These models were not based on mechanisms
of attachment and transfer, which leaves several uncertainties remaining

78

A comprehensive database compiling foodborne pathogen transfer data would be a
valuable addition for the scientific community, enabling easier comparisons across studies, and
broader analyses of transfer phenomena. The long-term goal is to build the current database into
a publically-available tool. Such a database could then serve as a valuable resource for model
development and risk assessments.

4.4

Conclusion
The goal regarding common types of food pathogen transfer events was to compare

simple general models to identify phenomenological attributes of particular transfer scenarios,
consistent across multiple studies. Comparing models based on an equivalent criterion (i.e.,
AICc) enabled quantitative evaluation of transfer response curve characteristics, and illustrated
that the effects of product and contact type were consistent across multiple studies.
This analysis illustrated that there are considerable data already published in this area, but
most such studies have been designed, executed, and reported independent of one another,
without connecting data across studies. This is consistent with a recent report by Hoelzer and
others (2012), which aggregated published transfer data from 37 studies (825 data points) to
determine the probability distributions of transfer coefficients, finding that transfer coefficients
varied widely across studies. Studies that have proposed and/or developed system-level modeling
frameworks require prior knowledge or assumptions about the rates of transfer; however, given
the current state of knowledge, such information is available only as product/pathogen/processspecific data or parameters. There is a significant need and opportunity for a unified database
that aggregates all of these prior data and becomes a repository for future data. Such a database

79

will help advance linkages between fundamental research on factors controlling attachment and
the observed transfer outcomes, while also improving the design of future studies to fill data
gaps.

80

5. OVERALL CONCLUSIONS

5.1

GGP comparisons by means of image analysis
•

Image-based analysis of GGP concentration on food contact surfaces is a means to
rapidly quantify the risk of L. monocytogenes transfer occurring in a given scenario, for a
known recipient surface and initial contamination level.

•

An ANCOVA was performed to compare the slopes of the different GGP regressions.
There was a significant difference among the slopes between surface transfer pairs;
therefore, different thresholding levels were necessary for each surface, indicating that
use of this approach likely requires knowledge of the surface type. However, when an
ANCOVA was performed on recipient surfaces, there was no significant difference (P >
0.05) between the slopes; therefore, only knowledge of the recipient surface type (i.e., the
surface being imaged) is necessary to quantify the amount of transfer.

•

Transfer of L. monocytogenes and the proposed physical surrogate, GGP, by sequential
contacts to and from ready-to-eat roast turkey, ham, 304 stainless steel, and high-density
polyethylene can be related by regression analysis.

5.2

Aggregation of foodborne pathogen transfer data
•

There is a growing body of literature addressing bacterial attachment and transfer to/from
food products and contact surfaces, but most of the reports are essentially observational.

81

•

Meta-analysis illustrates that there are considerable data already published in this area,
but every study has been designed, executed, and reported with little consideration for its
role in the larger body of data. This is consistent with a recent report by Hoelzer and
others (2012), which aggregated published transfer data from 37 studies (825 data points)
and showed that transfer coefficients varied widely across studies.

•

Most transfer models are probabilistic or empirical curve fitting and do not consider
underlying mechanisms that govern these processes.

•

Studies that have proposed and/or developed system-level modeling frameworks require
prior knowledge or assumptions about the rates of transfer; however, given the current
state of knowledge, such information is available only as product/pathogen/processspecific data or parameters.

•

There is a significant need, and opportunity, for a unified database that aggregates all of
these prior data and becomes a repository for future data. Such a database (here described
as TranBase) will help advance linkages between fundamental research on factors
controlling attachment and the observed transfer outcomes, while also improving the
design of future studies to fill data gaps that are revealed by meta-analyses of the
compiled data.

82

5.3

Analysis of aggregated delicatessen test data
•

Analysis revealed that the action of slicing and chopping gives similar critical contact
values (~9), and therefore similar curves for ready-to-eat turkey.

•

During slicing, high moisture and low fat turkey and ham transfer Listeria in a similar
Weibull manner, while transfer to salami is more typically linear due its high fat and low
moisture creating a product residue film on the blade.

•

Sequential static contacts contrarily have the same shape of transfer regardless of the
meat or surface.

•

Transfer curves are dependent on several variables that must be reported for accurate
comparisons.

•

In this case, several differing transfer rates were observed, but the shape of the curve
between the different types of transfer and food was the factor evaluated and compared.

•

Aggregating data from different studies and research groups revealed transfer
characteristics that may not have been evident within individual studies.

•

Critical variables, such as temperature, inoculation methods, and sampling methodology,
should be controlled when conducting transfer experiments, so that resulting data can be
better compared across studies.

83

5.4

Utility of a Pathogen Transfer Database
•

This preliminary database (i.e., TranBase) provides comprehensive, published transfer
data that will be useful to the scientific community and risk assessors for evaluating food
pathogen cross-contamination.

•

This information previously has not been available in a compiled form that offers a
baseline for exploring the factors that govern these processes.

•

The information within such a database can be a valuable tool for the scientific
community, in that it can allow for easier comparisons across studies, determination of
the types of studies that are needed, and different or new types of analyses.

•

The ability to analyze and identify problematic areas and understand transfer mechanisms
will help improve risk models and safety measures, and thus prevent foodborne illness
and outbreaks.

5.5
5.5.1

Future work
Standardization of image capturing
Quantification of GGP as a surrogate for L. monocytogenes in an actual delicatessen will

require greater photographing technology than the one reported in this research. Since coupons
were being photographed under controlled conditions, a standard tabletop setup was used;
however, in a delicatessen it would be desirable to have an all-inclusive and mobile setup: a
camera with an ultraviolet light attachment at a fixed distance and angle from the surface being
imaged. This would allow for images to be rapidly taken of several surfaces without re-setup.

84

5.5.2

Creation of a simple analysis tool
In this research, a MatLab® image analysis code was used to quantify the amount of

GGP surrogate on surfaces. This code requires manual input of individual images, cropped area x
and y values, different thresholds for recipient surfaces, and different equations for
transformation from intensity to concentration. If these steps were put into a simple format,
where all images could be uploaded, the cropped area could be recognized, a selection box for
recipient surface can be checked, and concentration on these surfaces were outputted into a file,
the tool would be much easier for others to use. Automating the code as described would be
possible and should be done to increase the utility of the tool.

5.5.3

GGP Model validation, use in other projects
Quantifying this surrogate during experimental food handling scenarios and using it to

predict pathogen transfer risk provides the capability to quantitatively map actual crosscontamination routes in food facilities, thus giving a method to fill gaps in the data for risk
assessments and validate risk models. This can be used as a tool to quantify the impact of worker
training interventions in a retail food-handling environment. GGP image analysis is a valuable
approach to assess the strength of control measures, and monitor their impact on food safety.

5.5.4

Enhancement of the TranBase database
This database was intended to be a baseline and serve as a depository for future data. It

can then serve as a foundation for model building and risk assessments, which often note that
assumptions have to be made because of data unavailability or inaccessibility.

85

5.5.5

Comparisons with other research groups
Aggregated existing food pathogen transfer data in an easily accessible format allows for

easy data comparisons across research groups. This needs to be done on past and current work to
maximize understanding of the information available, and should lead to standardizations of
methods and reporting so that data comparisons will be simpler for future transfer studies. This
important step will help advance linkages between fundamental research and the observed
transfer outcomes, while also improving the design of future studies to fill critical data gaps.

5.5.6

The need for standardized testing methods and reporting results
Standard test methods and/or control measures should be used in pathogen transfer

studies; in this way, data can be directly compared across studies without uncertainty due to
differences in methods. A standard transfer data reporting method should also be suggested, to
maximize the future utility of transfer studies.

86

6. APPENDIX

87

Trials for determining GGP concentration on surfaces
In attempt to model the transfer of Listeria through a food service environment, several
physical surrogates and quantification methods were considered. The use of GGP polymer
powder was first considered due to its size (5 um), approximately the size of a bacterium, and
past work identifying high risk locations on slicers (Vorst and others 2006a). Tests to quantify
GGP included creating calibration curves on several materials including stainless steel and RTE
turkey. These were stomached/sonicated or swabbed, and absorbances were taken on a
spectrophotometer. Both methods were unsuccessful either due to the presence of protein
residues in the slurry or poor recovery. The use of a peanut allergen was then considered because
of the high sensitivity of tests to determine its presence. Although there was high sensitivity, the
recovery process of the protein was time-consuming and the protein was undesirable to spread.
GGP had the qualities desirable: correct size, visibility, and rapid quantification. Therefore,
capturing images of the indicator was attempted followed by image analysis algorithms.

GGP absorbance on surfaces
Methods and Results:
1. A GGP suspension containing 1g GGP in 10 mL 70% ethanol was created.
2. Six serial dilutions of 1:1 (vol/vol) were made from the suspension.
3. Six ten by ten square cut of turkey were inoculated with 100 μl of each of the dilutions of
GGP suspension and spread with a hockey stick.
4. The suspension was allowed to soak in for 5 minutes at 7°C.

88

5. Each turkey piece was placed in a filter bag with 15 ml of 70% ethanol and massaged to
release the GGP while keeping it in suspension.
6. Specimen from each bag was collected for absorbance readings with a spectrophotometer

Absorbance

at 370 nm.
2.0
1.8
1.6
1.4
1.2
1.0
0

2
4
dilution number

6

Figure 6.1 Absorbance readings of GGP serial dilutions on RTE turkey at 370nm.

7. Since this did not work due to soils in the solution from the turkey, the specimen was
centrifuged for 10 minutes at 2000 rpm to separate the turkey solids.
8. It was visible that both the turkey pieces and GGP were stuck to the wall, not in solution,
therefore absorbance could not be read.
9. The specimen was also filtered, however the filter caught mixtures of solids and GGP as
well.
Conclusions: When attempting to determine absorbance readings from GGP on food surfaces, a
spectrophotometer cannot properly estimate the correct value. This is likely due to the presence
of meat pieces and/or fats and oils that alter the readings. Attempts to isolate the GGP for
readings failed as well, therefore a new method for quantification of GGP on RTE surfaces must
be considered.

89

Peanut allergen on RTE surfaces
Methods and Results:
1. A peanut allergen test-kit was obtained from Neogen Corporation. The kit including 5
solutions of different known concentrations of peanut allergen, as well an unknown spike
solution, and extraction materials to determine concentrations in unknown amounts.
2. The vials of known concentrations were measured by pipetting 150 μl into a microtiter
plate, and using the Microplate reader (Molecular Devices Corp. 2001) to relate readings

Mean Value

to concentrations for a calibration curve.

y = 0.001x2 + 0.0628x + 0.3433

3
2.5
2
1.5
1
0.5
0

R² = 0.99574

0

10
20
Concentration (ppm)

30

Figure 6.2 Calibration curve for known concentrations of peanut allergen using a
Microtiter plate reader.

3. The unknown spike solution was diluted serially 5 times and was measured by pipetting
150 μl into the microtiter plate.

90

Concentration (ppm)

20
15
10
5
0
3.0

4.0
Dilution Number

5.0

Figure 6.3 Comparing dilution number to concentration by use of calibration curve.

4. Once the curve was created, the spike solution was serially diluted 5 times and used to
inoculate 5 stainless steel coupons with 100 μl of teach of the dilutions. The inoculated
stainless steel coupons sat for 10 minutes to dry. A KimWipe was then wetted with 1 ml
of PBS and half the plate was whipped, followed by putting each coupon in a Whirlpack
bag with the addition of 25 ml of PBS. The bags were then sonicated. The extraction
procedures of the Neogen Veratox kit were then followed and the final solution was
measured using the Microplate reader.

91

Table 6.1 Recovery of known concentrations of peanut allergen by means of sonication or
composite tissue.
Spike solution on Stainless steel and recovered
KimWipe
Dilution Mean Value

Sonication

KimWipe

Sonication

Mean Value

Conc. (ppm)

Conc. (ppm)

1.62

1.73

16.1

17.3

1

0.757

1.58

5.86

15.6

2

0.614

0.710

3.97

5.25

3

0.659

0.337

4.57

0.166

4

0.300

0.503

-0.357

2.47

5

0.314

0.323

-0.159

-0.031

6

0.535

0.466

2.908

1.96

Concentration (ppm)

0

20
18
16
14
12
10
8
6
4
2
0
-2 0

KimWipe

2

4
Dilution Number

Sonication

6

8

Figure 6.4 Concentrations of peanut allergen recovered from stainless steel by using
KimWipe and sonication.

92

5. The method above was repeated for turkey. Five pieces were inoculated with dilutions of
the spike solution and allowed to soak followed by additional of 25 ml of PBS in a
whirlpack bag and stomached. The Neogen kit was used to extract the peanut allergen

Concentration (ppm)

and and the final solution was measured using the Microplate reader.

25
20
15
10
5
0
0

2
4
Dilution number

6

Figure 6.5 Concentrations of peanut allergen recovered from turkey by stomaching.

Conclusions:
Using peanut allergen as an indicator of Listeria monocytogenes on common RTE food
and contact surfaces is not ideal. The extraction process is extensive and there is much room for
error. As seen in the graphs above, simple recovery from stainless steel and turkey surfaces are
not representative of serial dilutions inoculated onto the surfaces. Because these simple tests did
not yield accurate results, furthering tests with actual transfer events was not attempted.

93

Trials and final protocol for imaging
Types of cameras considered
A point-and-shoot camera was utilized to take a single image of all coupons that served
as transfer events. Since the camera could not be set to manual, capturing all the coupons at once
would eliminate variability between photo, and therefore would give a more accurate account of
indictor on the surfaces. Variability was found however due to the position of the coupon relative
to the ultra-violet light source.

Figure 6.6 Photograph of GGP on stainless steel coupons taken with a point-and-shoot
camera.
Because of the importance of consistent lighting, it was determined that a Digital SLR
camera with manual settings should be used, and each coupon that served as a transfer event

94

should be photographed separately in the same location to the ultra-violet source to control
variability.

Light sources considered
A bench-top 385 nm ultraviolet light was used for most of the images. The long
wavelength UV, however, allows more background illumination than a shorter wavelength
source, such as a 360 nm light. Background illumination/reflection can be a problem during
image analysis because the algorithm may not recognize a difference between the indicator and
background noise. Shorter wavelength LED flashlights were later considered due to this. The
smaller size of the LED’s also allowed for mobility in comparison to the bench top tube UV
source.

Imaging protocol
Camera settings
Since GGP is sensitive to ultra-violet light controlled imaging is required. The digital SLR
camera should have manual settings of the following:
Shutter speed: 1/5
Aperture: fsstop 5.6
ISO speed: 400
Each surface must be photographed separately but under the same conditions.

95

Tube lights
The black light should be at a constant or nearly constant distance from the surfaces
being photographed. A distance of 10+-1 inch is preferred. The more variability in distance
however, the less accurate the outcome, therefore choosing a distance is best. The light should
also be angled at approximately 80 degrees from the camera (170 from the surface being
photographed), which must be directly normal to the surface being imaged so that the surface
appears to be 2-D. The distance from the camera lens to the surface being photographed was held
consistently at 18 inches.

90°

10°

Figure 6.7 Set-up of digital camera with tube light. Camera stands on a tripod directly
above the sample being photographed. The tube light is 10 degrees above the sample.

Bundle of UV LED flashlights
The UV- LED lights should be at a constant or nearly constant distance from the surfaces
being photographed. A distance of 8+-1 inch is preferred. The more variability in distance
however, the less accurate the outcome, therefore choosing a distance is best. The light should

96

also be angled at approximately 45 degrees from the surface being photographed. The distance
from the camera lens to the surface being photographed was held consistently at 18 inches.

Figure 6.8 Tripod with camera directly above sample with bundle of UV-LED lights
attached to tripod stand.

97

Comparing UV light sources
Both light sources with each of the mentioned conditions above demonstrated to not have
a significant difference when determining the concentration of GGP on surfaces, therefore both
methods can be used.

3500
Concentration (ppm)

3000

Tube

2500
Bundle 4
2000
1500
1000
500
0
0

500

1000
Intensity

1500
x 10000

Figure 6.9 Calibration curve comparing intensity to concentration between the tube light
and bundle LED’s.

Calibration before imaging
For calibration before imaging, drip dots of the suspension at different known
concentrations onto the surface and photograph with the settings described above. Take images
of the surfaces without the GGP so that these can be used for calibration as well. Calibration
will relate intensity to the known concentration.

98

Figure 6.10 Known concentrations of GGP on stainless steel, high density polyethylene, and
turkey surfaces.

Challenges with using different cameras:
Different cameras have different sensitivities. Therefore, a Nikon and a Cannon digital
SLR Camera were used to photograph images under the same conditions and compared. The
curves below demonstrate that with the same camera settings and conditions (distances, angles,

Intensityx 100000

light source etc.), the Nikon camera is more sensitive.

1400
1200
1000
800
600
400
200
0

Nikon
Cannon

0

2

4
6
contact number

8

Figure 6.11 Graphical comparison of Cannon and Nikon digital SLR camera sensitivities
during transfer from turkey to stainless steel surfaces.

99

MatLab trials and final code for image analysis
Trial Code 1:
The first algorithm considered filters a single or multiple color components: red, blue,
green, or grey. The image is read and cropped to the size of the coupon, followed by calculating
reference values of GGP on black and stainless steel surfaces for calibration. This calibration is
manual however and has different wavelength cut-offs for each color and for each surface. The
cropped image is then converted to grey, then to a binary image. The average amount of red,
green, and/or blue pixels can then be determined.
From Figure 6.12, it is apparent that calibration curves for HDPE and turkey are limiting
compared to stainless steel. The code also does not recognize the value of intensity of each color
component, it is binary, simply counting if a pixel is illuminated (a value of 1), or not illuminated
(a value of zero). It therefore, is not the most accurate account of amount of GGP on surfaces.

Table 6.2 Cut-off for background values for each surface and color.
Dilutions
ppm
r
100000
50000
25000
12500
6250
3125
1562.5
781.25
390.625
195.3125
97.65625
48.828125

Stainless Steel
g
140
133
125
115
91
72
61
55
39
25
10
10

HDPE
r

b
255
240
228
215
195
170
155
125
80
61
30
15

255
255
255
255
255
255
255
255
202
195
180
166
100

g
145
140
135
120
90
77
75
75
65
53
34
25

b
255 255
255
255
235
220
190
168
143
116
77
60

concentration (ppm)

7000
SS g pixels
Turkey g pixels
HDPE g pixels

6000
5000
4000
3000
2000
1000
0
0

50

100
green pixels

150
x 100000

Figure 6.12 Average number of green pixels for each surface containing a known
concentration of GGP.

Trial Code 2:
The second code uses black and white plates for calibration to bring background noise to
zero. It only considers the green component of each surface. Then, depending on the surface, a
threshold can be chosen. The problem with this code is that it is binary as before, and since the
threshold is chosen by the user, it is subjective.

101

7000

Concentration (ppm)

6000
5000
4000

SS
T
HDPE

3000
2000
1000
0
0

50
100
150
200
250
Average Green Component Value

300

Figure 6.13 Comparing the green component value of known concentrations on different
surfaces.

Trial Code 3:
The third code works similarly to the second code, however, it is more user friendly in
that it downloads all the images in the file at once and crops them based on recognition. It also
has a thresholding toolbox that can be used to appropriately choose the correct threshold for each
surface. This however, as before, changes the images to binary when determining intensity and is
subjective because the threshold is chosen by the user.

102

Figure 6.14 Images changed from original (left), to red, blue, green, and grey components.

Figure 6.15 Threshtool that is used to subjectively choose a threshold for each image.

103

Trial Code 4, Final code:
The final code mixes ideas from each of the codes above. The images are imported into
the program and cropped to the size of the surface. By using the red component, the green
component, used to track the GGP, was isolated from the background surface. Doing so will
change the image into binary, as before. This binary green component is then multiplied by the
original image intensity values so as to give the correct intensity value. The intensity can then be
related the concentration by creating a calibration curve and inputting it into the end of the code.

% Question to: Dr. Sangyup Jeong (Biosystems & Agricultural Engineering)
% December 22, 2011
% MICHIGAN STATE UNIVERSITY

clear all
%% Reading image file
% I=imread('C:\Documents and Settings\jeongsa1\My Documents\Academicmajor\Research\Image Processing\Florescence\image_1.bmp');
fp='C:/Users/abenoit/Pictures/THESIS WORK/824 uv ss to turkey/';
% fp_Ref='C:\Users\jeongsa1\Documents\Academic-Active\Research\Topics\Image
Processing\Image Processing\Amanda\Raw Files 100711\Reference Images\';
%% Variables to be CHANGED!
% File name
I_org=imread([fp,'DSC_0620.JPG']); % test image
% I_org=imread([fp_Ref,'DSC_0748_TURKEY.JPG']); % reference image

104

% Image crop parameters
dx=2200; dy=2200; % for test image
Xw=1054; Yw=358; % upper left corner of the crop image
figure, imshow(I_org)
%% Cropping image of interest
I_crop=imcrop(I_org,[Xw Yw dx dy]);
%% Nonuniform background correction
Rimg=I_crop(:,:,1);
Gimg=I_crop(:,:,2); % Extracting green component image
Bimg=I_crop(:,:,3);
% Identify backgroud illumination region
Bimg_bw=Bimg;
for i=1:(dx+1)
for j=1:(dy+1)
if (Bimg(i,j) == 255), Bimg_bw(i,j)=1; else Bimg_bw(i,j)=0; end; % convert blue image
into binary
end
end
B_area=sum(sum(Bimg_bw)); % area of background illumination
R_sum=sum(sum(Rimg)); % sum of red image
R_thresh=R_sum/B_area; % threshold for red image to identify effective green image area
% Isolate only background illumination in green image using red image and
% threshold
Gimg_bkg=Gimg;

105

for i=1:(dx+1)
for j=1:(dy+1)
if (Rimg(i,j)>= R_thresh), Gimg_bkg(i,j)=0; end; % 150 for steel, 100 for black
end
end
% Compensating green image with blurred background illumination
se=strel('disk',3); %Create morphological structuring element
h=fspecial('average',3); % Create predefined 2-D filter
[row, col]=size(Gimg_bkg);
sImg=imresize(Gimg_bkg, 0.4); % image reduction 1=full
bkg=imopen(sImg,se);
blurred=imfilter(bkg,h,'replicate'); % blurring background
blurred=imresize(blurred, [row,col]); % back to original image size
nGimg=Gimg-blurred; % background corrected image
% Total effective gfp intensity
G_sum=sum(sum(nGimg));
I=G_sum;
% Plotting images
% figure, imshow(I_org)
% figure, imshow(Rimg)
% figure, imshow(Bimg)
% figure, imshow(Gimg)
% figure, imshow(blurred)

106

figure, imshow(nGimg)
%% Calculating intensity per unit area
%Different equation for each surface. Insert here:
C=

Minimizing differences between surface transfer data for average L. monocytogenes and
GGP
Table 6.3 Minimizing differences between average L. monocytogenes and GGP transfer
using of an offset and multiplier for (a) high density polyethylene to ham, (b) ham to
HDPE, (c) turkey to stainless steel, (d) stainless steel to turkey, (e) turkey to HDPE, and (f)
HDPE to turkey.

1
2
3
4
5
6
7
8
9
10
11
12

LM AVG
6.79131572
6.713408068
7.191851594
6.871166141
6.868944035
6.814991337
6.61800081
6.67244164
6.395098585
6.300350348
6.136500636
6.434028181
multiplier
offset

(a) HDPE to H
GGP AVG
NEW GGP
5.139312246 6.985257642
4.537693194 6.592683197
4.764728927 6.740830811
4.709566741 6.704835832
4.581027553 6.620960164
4.678936994 6.684849006
4.691258358 6.692889065
4.471922154 6.549765624
4.308713777 6.443267271
4.609444995 6.639503396
4.219951086 6.385346958
4.806227472 6.767909854
0.652529942 sum
3.631702522 RMSE

107

Resid^2
0.037613469
0.014574494
0.203419747
0.027665772
0.061496
0.016937026
0.005608251
0.015049405
0.002320222
0.11502479
0.061924492
0.111476972
0.67311064
0.236838665

Table 6.3 (cont’d)

1
2
3
4
5
6
7
8
9
10
11
12

1
2
3
4
5
6
7
8
9
10
11
12

LM AVG
6.556012097
6.463700195
6.209548107
6.035103518
6.318017991
6.00E+00
6.073417047
5.96E+00
5.791971182
5.671296687
5.664476339
5.696979437
multipier
offset

(b) H to HDPE
GGP AVG
NEW GGP
Resid^2
3.347134783 6.457363794 0.009731488
3.396460892 6.482653304
0.00035922
2.876654053
6.21614819 4.35611E-05
2.955833693 6.256743614 0.049124332
2.303904611 5.922499388 0.156434965
2.438738172 5.991628593 1.84876E-05
2.386074809 5.964628072 0.011835041
2.48998954 6.017905184 0.003089151
2.173138173 5.855455398 0.004030246
2.13350889 5.835137453 0.026843797
1.90474529 5.717850291 0.002848779
1.910215527 5.720654883 0.000560527
0.512700285 sum
0.264919594
4.741286838 RMSE
0.148582074

LM AVG
8.083955419
7.752157502
7.311162429
7.175896768
6.934995612
6.690734643
6.673660275
6.498355705
6.42541185
6.418436734
6.450024598
6.57281984
multiplier
offset

(c). T to SS
GGP AVG
NEW GGP
Resid^2
3.893289067 8.093640644
9.38036E-05
3.582813664 7.630566579
0.014784353
3.154543318 6.991801359
0.101991493
3.337652681 7.264908974
0.007923173
3.205546996 7.067873345
0.017656492
2.98044376 6.732131868
0.00171373
2.95944094 6.700806161
0.000736899
2.91220328 6.630351192
0.017422809
2.924923652 6.649323627
0.050136484
2.854844064 6.544799922
0.015967655
2.644208167 6.230636481
0.048131146
2.79275087 6.452187923
0.014552059
1.4915 sum
0.291110097
2.2868 RMSE
0.15575357

108

Table 6.3 (cont’d)

1
2
3
4
5
6
7
8
9
10
11
12

1
2
3
4
5
6
7
8
9
10
11
12

LM AVG
7.90894182
7.806806924
7.459382641
7.416234964
7.382543669
7.392241527
7.337994299
6.854554309
6.976585086
6.75916159
6.874445704
6.731794453
multiplier
offset

(d) SS to T
GGP AVG
NEW GGP
Resid^2
3.904892358 8.081945393 0.029930236
3.183940289 7.515277067 0.084989657
2.88193127 7.277897978 0.032936683
2.768141072 7.188458883 0.051881943
2.782166689 7.199483018 0.033511202
3.090896547 7.442144686 0.002490325
2.568376296 7.031443769 0.093973227
2.773319604 7.192529209 0.114227033
2.788876906 7.204757248 0.052062535
2.465928005 6.950919412 0.036771062
2.451283026 6.939408458 0.004220159
2.37160505 6.876781569 0.021021264
0.786 sum
0.558015329
5.0127 RMSE
0.215641548

LM AVG
7.531819117
7.527348411
7.201466365
7.020973768
6.997440278
6.885310961
6.51059465
6.494154036
6.611926365
6.452357773
6.241202285
6.240855646
multiplier
offset

(e) T to HDPE
GGP AVG
NEW GGP
Resid^2
5.051792022 7.537883696 3.67791E-05
4.759225895 7.390395751 0.018756031
4.611487331 7.315918034 0.013099185
3.999215007 7.007260339 0.000188058
3.928295183 6.971508357 0.000672465
3.532351468 6.771905876 0.012860713
3.045207688 6.526327772 0.000247531
3.117912611 6.562979655 0.004736966
3.193465856 6.601067431 0.000117916
2.956495704 6.481606436 0.000855484
2.952427242 6.479555449 0.056812231
2.138070371 6.069023231 0.029526379
0.50411832 sum
0.137909738
4.991182787 RMSE
0.107202977

109

Table 6.3 (cont’d)

1
2
3
4
5
6
7
8
9
10
11
12

(f) HDPE to T
LM AVG
GGP AVG
NEW GGP
Resid^2
8.202987532 3.349149808 8.411797267 0.043601505
8.031408464 2.840678345 7.904777242 0.016035466
7.918641835 2.496468054 7.561549505 0.127514932
7.617175143 2.448067728 7.513287338 0.010792676
7.792099727 2.496736576
7.56181726 0.053030014
7.49E+00 2.313830912 7.379433703 0.011498888
7.207499723 2.199385812 7.265315288
0.00334264
7.19E+00 2.434564713 7.499822867 0.096509759
7.255272505 2.305917712 7.371543091 0.013518849
7.353788045 2.233972253 7.299803001 0.002914385
7.325652471 2.181678112 7.247658135 0.006083116
6.8162413 2.113620898 7.179795191 0.132171432
multiplier
0.997145486 sum
0.517013663
offset
5.072207654 RMSE
0.207567994

Replicate data for L. monocytogenes transfer between surfaces
9

Table 6.4 Sequential transfer of L. monocytogenes from 10 CFU inoculated turkey to
stainless steel coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12

8.653212514
8.101231387
7.903089987
7.740362689
7.329906123
7.246129126
7.015988105
6.935759104
6.278753601
6.545616333
6.574031268
7.197280558

8.200713734
7.882239848
7.217483944
6.821185883
6.68797462
6.602059991
6.68797462
6.301029996
6.896250562
6.70969387
6.382467322
6.193820026

7.397940009
7.273001272
6.812913357
6.966141733
6.787106093
6.224014811
6.317018101
6.258278015
6.101231387
6
6.393575203
6.327358934

110

9

Table 6.5 Sequential transfer of L. monocytogenes from 10 CFU inoculated stainless steel
to turkey coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12

7.966141733
7.84509804
7.431363764
7.804480189
7.342422681
7.449092531
7.482516287
7.395763089
7.015988105
7.161368002
7.065392962
7.041392685

7.419129308
7.860338007
7.373371817
7.130333768
7.494850022
7.761551989
7.005395032
7.258278015
6.769007871
6.759667845
7.417056299
7

8.281601444
8.342422681
7.740362689
7.916453949
7.595220567
7.478927056
7.267171728
6.942008053
7.06069784
6.769007871
6.665111737
6.447158031

7.829303773
7.67669361
7.404406051
7.37794338
7.303735889
7.157607853
7.875061263
6.317018101
7.764362966
6.860338007
7.026328939
7.311753861

9

Table 6.6 Sequential transfer of L. monocytogenes from 10 CFU inoculated ham to
stainless steel coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12

6.942008053
6.916453949
6.146128036
6.375663614
6.70969387
6.23363058

6.070037867
6.287241711
6.210853365
5.916453949

7.329906123
7.126293791
6.942008053
6.84509804
6.146128036
5.829303773
5.983400738
5.960232873
5.317018101
5.249198357
5.966141733
5.373371817

5.641474111
5.278753601
5.079181246
5.667452953
4.966141733
4.413299764
4.328379603
4.432969291
4.390935107
4.298853076
4.334453751
4.190331698

111

9

Table 6.7 Sequential transfer of L. monocytogenes from 10 CFU inoculated stainless steel
to ham coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12

8.096910013
7.84509804
7.574031268
7.496583734
7.261262869
7.138302698
7.408663874
6.994537104
6.90982337
6.795880017
6.386944624
7.105510185

8.118099312
8.026328939
7.860338007
7.556302501
7.458637849
7.410777233
7.817069316
7.270096281
6.977723605
6.948168362
7.16879202
7.101231387

8.68797462
8.16879202
7.889301703
7.740362689
7.566732029
7.264227348
6.882239848
7.157607853
7.382467322
6.829303773
7.06069784
7.146128036
9

Table 6.8 Sequential transfer of L. monocytogenes from 10 CFU inoculated turkey to
HDPE coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

7.210853365
7.172456974
6.90982337
6.821185883
6.837272703
6.804480189
6.477121255
6.347330015
6.45484486
6.342422681
6.157607853
6.092545208
5.787106093
5.896250562
6.227243782

7.852784869
7.882239848
7.49310936
7.220761654
7.157607853
6.966141733
6.544068044
6.640978057
6.769007871
6.562292864
6.324796718
6.389166084
6.239924813
6.433369747
6.146128036

112

9

Table 6.9 Sequential transfer of L. monocytogenes from 10 CFU inoculated HDPE to
turkey coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

8.122215878
7.875061263
7.971971276
7.337459261
7.711807229
7.204119983
7.303735889
7.190331698
7.389166084
7.516865761
7.643452676
6.498310554
6.89279003
6.799340549
6.8162413
6.953034457

8.281601444
8.157607853
7.759667845
7.67669361
7.581209852
7.16879202
7.130333768
7.314393957
7.284430734
7.176091259
6.90982337
6.90982337
6.68797462
6.571126277
6.67669361
6.412880358

8.190331698
8.015988105
7.989004616
7.740362689
7.983400738
7.787106093
7.16879202
7.010723865
7.010723865
7.298307137
7.055951405
6.922984816
6.942008053
6.821185883
6.431363764
6.395763089
9

Table 6.10 Sequential transfer of L. monocytogenes from 10 CFU inoculated ham to HDPE
coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12

7.270096281
7.118099312
7.079181246
7.036429266
7.070037867
6.515211304
6.611457766
6.759667845
6.458637849
6.301029996
0
6.319626484

6.301029996
5.698970004
5.896250562
4.875061263
5.640978057
5.375663614
5.795880017
5.439332694
5.458637849
5.544068044
5.419129308
5.896250562

6.096910013
6.574031268
5.653212514
6.193820026
6.243038049
6.096910013
5.812913357
5.68797462
5.458637849
5.16879202
5.90982337
4.875061263

113

9

Table 6.11 Sequential transfer of L. monocytogenes from 10 CFU inoculated HDPE to ham
coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12

6.954242509
7.051152522
7.088136089
6.929418926
7.193820026
6.989004616
6.67669361
6.867762025
6.769007871
6.462397998
6.574031268
6.48784512

6.62838893
6.375663614
7.2955671
6.812913357
6.544068044
6.640978057
6.559308011
6.477121255
6.021189299
6.138302698
5.698970004
6.380211242
7

Table 6.12 Sequential transfer of L. monocytogenes from 10 CFU inoculated stainless steel
to turkey coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

6.2955671
5.852784869
5.882239848
5.750122527
5.478927056
5.33243846
5.010723865
4.935759104
4.720159303
4.698970004
4.267171728
4.16879202
4.134336511
3.983400738
3.903089987
3.730378469
3.574031268
3.829303773

6.197280558
5.730378469
5.837272703
5.84509804
5.740362689
5.391376239
5.342422681
4.90982337
5.005395032
4.860338007
4.875061263
4.698970004
4.451018452
4.230448921
3.68797462
4.84509804
4.896250562
4.588271707

5.68797462
5.653212514
5.243038049
5.559308011
5.292809665
5.431363764
5.06069784
4.588271707
4.812913357
4.456745495
4.460521993
4.214181309
4.327358934
4.25224605
3.882239848
3.615423953
3.720159303
3.62838893

114

Table 6.13 Sequential transfer of L. monocytogenes from 107 CFU inoculated turkey to
stainless steel coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

5.640978057
5.278753601
5.079181246
5.667452953
4.966141733
4.412880358
4.327358934
4.433369747
4.391376239
4.298307137
4.334956116
4.190331698
4.787106093
4.462397998
5.005395032

6.264227348
6.065392962
5.867762025
6.139879086
5.105510185
5.031408464
4.942008053
4.916453949
4.67669361
4.740362689
4.948168362
4.994537104
4.339948062
4.698970004
5.031408464

5.837272703
5.435366507
5.421192468
6.01494035
5.77815125
5.267171728
5.070037867
4.309097617
4.456745495
4.04630002
5.339948062
5.412880358
5.015988105
5.45484486
4.339948062

115

7

Table 6.14 Sequential transfer of L. monocytogenes from 10 CFU inoculated turkey to
HDPE coupons
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

6.122215878
5.929418926
5.769007871
5.612783857
5.635986112
5.576916956
5.347330015
5.220761654
5.569666462
5.214181309
5.314393957
4.653212514
3.243038049
4.750122527
4.829303773
4.759667845
5.04630002
5.101231387
5.179695383
4.80106053

6.088136089
5.812913357
5.591064607
5.544068044
5.322219295
5.243038049
5.010723865
4.653212514
3.574031268
5.126293791
4.665111737
4.67669361
4.916453949
4.359361103
4.67669361
4.698970004
4.431363764
4.179695383
4.484299839
4.079181246

6.40654018
6.239924813
5.867762025
5.759667845
5.787106093
5.402261382
5.138302698
4.821185883
5.015988105
5.021189299
4.942008053
5.545616333
4.948168362
4.812913357
4.464265934
4.602059991
4.165095875
4.186815124
4.220761654
3.971971276

116

7

Table 6.15 Sequential transfer of L. monocytogenes from 10 CFU inoculated HDPE to
turkey coupons.
Log CFU on slice
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

5.942008053
5.821185883
5.70969387
5.477121255
5.588271707
5.537819095
5.270096281
5.031408464
5.27588696
5.349763044
5.06069784
4.867762025
5.005395032
4.929418926
4.261262869
4.787106093
4.306425028
4.966141733
4.653212514
4.544068044

5.439332694
5.301029996
5.109747238
5.68797462
5.134336511
5.200713734
4.67669361
5.220761654
4.948168362
5.197280558
5.179695383
4.916453949
4.559308011
4.875061263
4.77815125
5.055951405
4.615423953
4.539389782
4.041392685
4.698970004

5.615423953
5.730378469
5.544068044
5.544068044
5.301029996
5.193820026
5.010723865
5.026328939
4.804480189
4.615423953
4.439332694
4.375663614
4.489606966
4.23363058
4.179695383
4.309097617
4.142232992
3.875061263
4.230448921
4.230448921

117

TranBase
Table 6.16 Bacterial transfer curve publications identified from the ISI Web of Knowledge.
Date

Author

Pether and
1971 Gilbert

Boer and
1990 Hahne

1990 Dickenson

Scott and
1990 Bloomfield
Humphrey
1994 et al.
Bradford et
1996 al.

1998 Zhao et al.
2001 Chen et al.

Title
The survival of salmonellas on fingertips and transfer of the organism to
foods
Cross-contamination with
Campylobacter jejuni and Salmonella
spp. From raw chicken products during
food preparation
Transfer of Listeria monocytogenes and
Salmonella typhimurium between beef
tissue surfaces

Bacteria

Data Data
Sets points

Rep Equation Surfaces

Salmonella

No curve

No curve

Campylobact
er,
Salmonella
Listeria,
Salmonella
Escherichia,
Salmonella,
Staphylococc
us, Klebsiella

The survival and transfer of microbial
contamination via cloths, hands, utensils
Contamination of hands and work
surfaces with Salmonella enteritidis PT4
during the prepartion of egg dishes
Salmonella
The cross-contamination and survival of
Salmonella enteritidis PT4 on sterile and
non-sterile foodstuff
Salmonella
Development of a model for evaluation
of microbial cross-contamination in the
kitchen
Enterobacter
Quantification and variability analysis
Enterobacter
118

meat, hand
cutting board,
raw chicken,
hands, beef

No curve

36

108

2 No curve

cutting board,
meat
laminate, cloth,
stainless steel,
fingers, work
surface
fingers, egg,
work surfaces

8

82

1

Beta,
50

452

1

egg, melon, meat
cutting board,
vegetables, hand,
meat
chicken, hand,

Table 6.16 (cont’d)

Gill and
2002 Jones

of bacterial cross-contamination rates in
common food service tasks
Glove barriers to bacterial crosscontamination between hands to food
Transfer of bacteria from fabrics to
hands and other fabrics: development
and application of a quantitative method
using Staphylococcus aureus as a model
Effects of wearing knitted or rubber
gloves on the transfer of Escherichia
coli between hands and meat

2002 Gorman

A study of cross-contamination of foodborne pathogens in the domestic kitchen
in the Republic of Ireland

Montville
2000 et al.

2001 Sattar et al.

Midelet
and
2002 Carpentier
Harrison et
2003 al.
Kusumanin
2003 grum et al.

Transfer of microorganisms, including
Listeria monocytogenes, from various
materials to beef
Bacteria transfer and crosscontamination potential associated with
paper-towel dispensing
Survival of foodborne pathogens on
stainless steel surfaces and cross-

Weibull,
Gamma
Enterobacter

91

480

1 No curve

spigot, lettuce,
cutting board
chicken, lettuce,
glove

Staphylococc
us

32

64

1 No curve

finger pad, fabric

44

88

4 No curve

meat, glove,
hand

1 No curve

meat, board,
dishcloth, hands

stainless steel,
PU, PVC, meat

E. coli
Escherichia,
Salmonella,
Staphylococc
us,
Campylobact
er
Listeria,
Staphylococc
us,
Pseudomona
s,
Comomonas
Micrococcus,
Serratia
Salmonella,
Staphylococc

119

13

90

12

144

2 No curve

2

1000

1 No curve

169

882

1 No curve

paper towel
dispenser, hand
Stainless steel,
vegetables, meat,

Table 6.16 (cont’d)
contamination to foods

Mattick et
2002 al.
Montville
and
2003 Schaffner

2003

2004

2004

2006
2005

The survival of foodborne pathogens
during domestic washing-up and
subsequent transfer onto washing-up
sponges, kitchen surfaces and food

Inoculum size influences bacterial cross
contamination between surfaces
Transfer of Salmonella and
Moore et
Campylobacter from Stainless Steel to
al.
Romaine Lettuce
A Quantitative Analysis of Crosscontamination of Salmonella and
Kusumanin Campylobacter spp via Domestic
grum
Kitchen surfaces
Phyicochemical Factors Influencing
Bacterial Transfer from Contact Lenses
Vermeltfoo to Surfaces with Different Roughness
rt et al.
and Wettability
Transfer Coefficient Models for
Escherichia coli O157:H7 on Contacts
between Beef Tissue and High-Density
Flores et al. Polyethylene Surfaces
Cross contamination between
Lin et al.
processing equipment and deli meats by

us,
Campylobact
er
Salmonella,
E.coli,
Campylobact
er

sponges

17

82

1 No curve

1

1053

1 No curve

meat, sponge,
kitchen surface
meat, cutting
board, lettuce,
hand, spigot,
glove

20

80

7

150

3 No curve
Monte
carlo,
beta1 poisson

stainless steel to
vegetables
raw meat,
stainless steel,
vegetables,
cutting board

1

36

3 No curve

contact lens,
glass, silicone

E. coli

28

152

Listeria

3

61

Enterobacter
Salmonella,
Campylobact
er
Salmonella,
Campylobact
er
Pseudomona
s,
Staphylococc
us

120

10 curve fit
1 No curve

beef, HDPE
meat, slicer

Table 6.16 (cont’d)
Listeria monocytogenes

2005 Luber et al.

2005 Vorst et al.

2006 Vorst et al.
Aarnisalo
2007 et al.

Dawson et
2006 al

Knobben et
2006 al
Moore et
2007 al.
2007 Perez-

Quantification of Campylobacter
Species Cross-Contamination during
Handling of Contaminated Fresh
Chicken Parts in Kitchens
Transfer of Listeria monocyotgenes
during Mechanical Slicing of Turkey
Breast, Bologna, and Salami
Transfer of Listeria monocytogenes
during slicing of turkey breast, bologna,
and salami with simulated kitchen
knives
Modelling transfer of Listeria
monocytogenes during slicing of
'gravad' salmon
Residence time and food contact time
effects on transfer of Salmonella
Typhimurium from tile, wood, and
carpet: testing the five-second rule
Transfer of bacteria between
biomaterials surfaces in the operating
room-an experimental study
Recovery and Transfer of Salmonella
Typhimurium from Four Different
Domestic Food Contact Surfaces
Modeling transfer of Escherichia coli

Campylobact
er

27

168

meat, hands,
plate, sausage,
cutting board,
knife, cucumber,
1 No Curve bread

Listeria

14

256

3 Curve fit

blade, meat

Listeria

13

255

3 Curve fit

meat, knife

Listeria

7

198

Exponent
3 ial

slicer, fish

22

163

13

92

2 Curve Fit
glove,
broach,
gown,
3 light

Wood, tile,
carpet, meat
hydrophobicity
and roughness,
wet vs. dry,
friction

2 curve fit
3 Log-

chicken, hand,
gloves, lettuce,
slicer blade, meat

Salmonella
Staphylococc
us,
Propionibact
erium

Salmonella
E.coli,

121

8
6

49
115

Table 6.16 (cont’d)
Rodriguez
et al.

Rodriguez
2007 et al.

2007 Rodriguez

2006 Rodriguez

Keskinen
2008 et al.

Keskinen
2007 et al.

2008 S. Sheen
2008 Sheen and

O157:H7 and Staphylococcus aureus
during slicing of a cooked meat product
Effect of Biofilm Dryness on the
Transfer of Listeria monocytogenes
Biofilms Grown on Stainless Steel to
Bologna and Hard Salami
Effects of inoculation level, material
hydration, and stainless steel surface
roughness on the transfer of Listeria
monocytogenes from inoculated
bologna to stainless steel and highdensity polyethylene
Evaluation of the Transfer of Listeria
monocytogenes from Stainless Steel and
High Density Polyethylene to Bologna
and American Cheese
Imapct of bacterial stress and biofilm
forming ability on transfer of surfacedried Listeria monocytogenes during
slicing of deli meats
Transfer of Surface-Dried Listeria
monocytogenes from Stainless Steel
Knife Blades to Roast Turkey Breast
Modeling Surface Transfer of Listeria
monocytogenes on Salami during
Slicing
Modeling Surface Transfer of Listeria

Staphylococc
us

Listeria

Listeria

linear and
Weibull

1

17

8

64

1 No curve

food composition
and water
activity

5

inoc. Level,
topography, dry,
wet
stainless steel,
HDPE, meat,
cheese

Listeria

1

64

3 No curve

Listeria

9

130

3 Curve fit

Listeria

8

133

Listeria
Listeria

4
7

192
280

122

meat,
3 knife
Curve,
changing
3 variable
3 Curve fit

slicer, meat
characterized n
blade and cutting
speed , time, RH,
temp.

slicer, meat
meat, slicing

Table 6.16 (cont’d)
Hwang

VerhoeffBakkenes
2007 et al
Fravalo et
2009 al
Jimene et
2009 al.

Sheen and
2010 Hwang

monocytogenes from slicer to deli meat
during mechanical Slicing
Quantification of Campylobacter jejuni
Cross-Contamination via Hands,
Cutlery, and Cutting Board During
Preparation of a Chicken Fruit Salad
Campylobacter transfer from naturally
contaminated chicken thighs to cutting
boards in inversly related to initial load
Survival of Salmonella on refrigerated
chicken carcasses and subsequent
transfer to cutting board
Mathematical modeling the cross
contamination of E. coli O157:H7 on
the surface of RTE meat product while
slicing

Campylobact
er
Campylobact
er

Salonella

E.coli

123

31

150

3 No curve

meat, cutlery,
cutting board,
hand s
HDPE, cutting
board, raw meat

4

109

Curve Fit

10

40

2 Curve fit

meat, cutting
board

280

Curve fit,
changing
3 variables

meat, slicer

7

REFERENCES

124

REFERENCES

Aarnisalo K, Sheen S, Raaska L, Tamplin M. 2007. Modelling transfer of Listeria
monocytogenes during slicing of 'gravad' salmon. International Journal of Food
Microbiology 118(1):69-78.
Abbott A, Rutter PR, Berkeley RCW. 1983. The influence of ionic-strength, ph and a protein
Layer On The Interaction Between Streptococcus-mutans And Glass Surfaces. Journal of
General Microbiology 129(FEB):439-45.
Adams BW, Mead GC. 1983. Incidence and properties of Staphylococcus-aureus associated with
turkeys during processing and further-processing operations. Journal of hygiene
91(3):479-90.
An YH, Friedman RJ. 1996. Prevention of sepsis in total joint arthroplasty. Journal of Hospital
Infection 33(2):93-108.
An YH, Friedman RJ. 1998. Concise review of mechanisms of bacterial adhesion to biomaterial
surfaces. Journal of Biomedical Materials Research 43(3):338-48.
An YH, Friedman RJ, Draughn RA, Smith EA, Nicholson JH, John JF. 1995. Rapid
quantification of Staphylococci adhered to titanium surfaces using image analyzed
epifluorescence microscopy. Journal of Microbiological Methods 24(1):29-40.
Arnold JW, Bailey GW. 2000. Surface finishes on stainless steel reduce bacterial attachment and
early biofilm formation: Scanning electron and atomic force microscopy study. Poultry
Science 79(12):1839-45.
Arnold JW, Niesel DW, Annable CR, Hess CB, Asuncion M, Cho YJ, Peterson JW, klimpel gr.
1993. Tumor-necrosis-factor-alpha mediates the early pathology in Salmonella infection
of the gastrointestinal-tract. Microbial pathogenesis 14(3):217-27.
Arnold JW, Silvers S. 2000. Comparison of poultry processing equipment surfaces for
susceptibility to bacterial attachment and biofilm formation. Poultry Science 79(8):121521.

125

Baker AS, Greenham LW. 1988. Release of gentamicin from acrylic bone-cement - elution and
diffusion studies. Journal of bone and joint surgery-american volume 70a(10):1551-7.
Barth E, Myrvik QM, Wagner W, Gristina AG. 1989. Invitro and invivo comparative
colonization of Staphylococcus-aureus and Staphylococcus-epidermidis on orthopedic
implant materials. Biomaterials 10(5):325-8.
Beumer RR, Tegiffel MC, Deboer E, Rombouts FM. 1996. Growth of Listeria monocytogenes
on sliced cooked meat products. Food Microbiology 13(4):333-40.
Bridgett MJ, Davies MC, Denyer SP. 1992. Control of Staphylococcal adhesion to polystyrene
surfaces by polymer surface modification with surfactants. Biomaterials 13(7):411-6.
Buchholz AL, Davidson GR, Marks BP, Todd ECD, Ryser ET. 2012. Quantitative transfer of
Escherichia coli o157:h7 to equipment during small-scale production of fresh-cut leafy
greens. Journal of Food Protection 75(7):1184-97.
Campos D, Hall NO, Ortra-Ramirez A, Marks BP, Booren AM, Ryser ET. 2007. Quantitative
transfer of Eschericia coli O157:H7 between beef and equipment contact surfaces. 2007
IFT Abstract . July 28-Aug 1. Chicago, Il.
CDC. 2010. Listeriosis technical information. Center for Disease Control. Atlanta, Georgia.
Available at: http://www.cdc.gov/listeria/. Assessed: 11 Novmeber 2010.
Chen YH, Jackson KM, Chea FP, Schaffner DW. 2001. Quantification and variability analysis of
bacterial cross-contamination rates in common food service tasks. Journal of Food
Protection 64(1):72-80.
Chen YH, Ross EH, Scott VN, Gombas DE. 2003. Listeria monocytogenes: low levels equal low
risk. Journal of Food Protection 66(4):570-7.
Cliver DO. 2006. Cutting boards in Salmonella cross-contamination. J. AOAC Int. 89(2):538-42.
Costerton JW. 1995. Overview of microbial biofilms. Journal of Industrial Microbiology
15(3):137-40.

126

Dankert J, Hogt AH, Feijen J. 1986. Biomedical polymers - bacterial adhesion, colonization, and
infection. Crc Critical Reviews in Biocompatibility 2(3):219-301.
De Reu K, Debeuckelaere W, Botteldoorn N, De Block J, Herman L. 2002. Hygienic parameters,
toxins and pathogen occurrence in raw milk cheeses. Journal of Food Safety 22(3):18396.
Den Aantrekker ED, Boom RM, Zwietering MH, van Schothorst M. 2003. Quantifying
recontamination through factory environments - a review. International Journal of Food
Microbiology 80(2):117-30.
Dhir VK, Dodd CER. 1995. Susceptibility of suspended and surface-attached salmonellaenteritidis to biocides and elevated-temperatures. Applied and Environmental
Microbiology 61(5):1731-8.
Dickson JS, Daniels EK. 1991. Attachment of Salmonella-Typhimurium and Listeriamonocytogenes to glass as affected by surface-film thickness, cell-density, and bacterial
motility. Journal of Industrial Microbiology 8(4):281-3.
Elson R, Burgess F, Little CL, Mitchell RT, Local Authorities C. 2004. Microbiological
examination of ready-to-eat cold sliced meats and pate from catering and retail premises
in the UK. Journal of Applied Microbiology 96(3):499-509.
Farber BF, Wolff AG. 1992. The use of nonsteroidal antiinflammatory drugs to prevent
adherence of Staphylococcus-epidermidis to medical polymers. Journal of Infectious
Diseases 166(4):861-5.
Fletcher M, Loeb GI. 1979. Influence of substratum characteristics on the attachment of a marine
Pseudomonad to solid-surfaces. Applied and Environmental Microbiology 37(1):67-72.
Fletcher M, Marshall KC. 1982. Bubble contact-angle method for evaluating substratum
interfacial characteristics and its relevance to bacterial attachment. Applied and
Environmental Microbiology 44(1):184-92.
Franco CM, Quinto EJ, Fente C, rodriguezotero JL, Dominguez L, Cepeda A. 1995.
Determination of the principal sources of Listeria spp contamination in poultry meat and
a poultry processing plant. Journal of Food Protection 58(12):1320-5.

127

Fratamico PM, Schultz FJ, Benedict RC, Buchanan RL, Cooke PH. 1996. Factors influencing
attachment of Escherichia coli O157:H7 to beef tissues and removal using selected
sanitizing rinses. Journal of Food Protection 59(5):453-9.
FSIS. 2009. Verification procedures for enforcement, Investigations and Analysis Officers
(EIAOS) for the Listeria monocytogenes regulation and routine risk-based Listeria
monocytogenes sampling program.
FSIS. 2010a. Food Saftey and Inspection Service. FSIS Comparative Risk Assessment for
Listeria monocytogenes in ready-to-eat meat and poultry deli meats report. Available at:
http://www.fsis.usda.gov/PDF/Comparative_RA_Lm_Exec_Summ_May2010.pdf Accessed 1
November 2010.
FSIS. 2010b. Food Saftey and Inspection Service. FSIS Comparative Risk Assessment for
Listeria monocytogenes in ready-to-eat meat and poultry deli meats report. Available at:
http://www.fsis.usda.gov/PDF/Comparative_RA_Lm_Exec_Summ_May2010.pdf Accessed 1
November 2010.
Gibbs PA, Patterson JT, Thompson JK. 1978. Distribution of staphylococcus-aureus in a poultryprocessing plant. Journal of Applied Bacteriology 44(3):401-10.
Gillespie I, Little C, Mitchell R. 2000. Microbiological examination of cold ready-to-eat sliced
meats from catering establishments in the United Kingdom. Journal of Applied
Microbiology 88(3):467-74.
Glass KA, Doyle MP. 1989. Fate of Listeria-monocytogenes in processed meat-products during
refrigerated storage. Applied and Environmental Microbiology 55(6):1565-9.
Grau FH, Vanderlinde PB. 1992. Occurrence, numbers, and growth of Listeria-monocytogenes
on some vacuum-packaged processed meats. Journal of Food Protection 55(1):4-7.
Gray ED, Verstegen M, Peters G, Regelmann WE. 1984. Effect of extracellular slime substance
from Staphylococcus epidermidis on the human cellular immune-response. Lancet
1(8373):365-7.
Gristina AG, Hobgood CD, Webb LX, Myrvik QN. 1987. Adhesive colonization of biomaterials
and antibiotic-resistance. Biomaterials 8(6):423-6.

128

Gristina AG, Jennings RA, Naylor PT, Myrvik QN, Webb LX. 1989. Comparative invitro
antibiotic-resistance of surface-colonizing coagulase-negative Staphylococci.
Antimicrobial Agents and Chemotherapy 33(6):813-6.
Harber MJ, Mackenzie R, Asscher AW. 1983. A rapid bioluminescence method for quantifying
bacterial adhesion to polystyrene. Journal of General Microbiology 129(MAR):621-32.
Harrison WA, Griffith CJ, Ayers T, Michaels B. 2003. Bacterial transfer and crosscontamination potential associated with paper-towel dispensing. American Journal of
Infection Control 31(7):387-91.
Hilbert LR, Bagge-Ravn D, Kold J, Gram L. 2003. Influence of surface roughness of stainless
steel on microbial adhesion and corrosion resistance. International Biodeterioration &
Biodegradation 52(3):175-85.
Hoelzer K, Pouillot R, Gallagher D, Silverman MB, Kause J, Dennis S. 2012. Estimation of
Listeria monocytogenes transfer coefficients and efficacy of bacterial removal through
cleaning and sanitation. International Journal of Food Microbiology 157(2):267-77.
Hoffman AD, Wiedmann M. 2001. Comparative evaluation of culture- and BAX polymerase
chain reaction-based detection methods for Listeria spp. And Listeria monocytogenes in
environmental and raw fish samples. Journal of Food Protection 64(10):1521-6.
Hogt AH, Dankert J, Devries JA, Feijen J. 1983. Adhesion of coagulase-negative Staphylococci
to biomaterials. Journal of General Microbiology 129(SEP):2959-68.
Hogt AH, Dankert J, Feijen J, Devries JA. 1982. Cell-surface hydrophobicity of Staphylococcus
species and adhesion onto biomaterials. Antonie Van Leeuwenhoek Journal of
Microbiology 48(5):496-8.
Ivanek R, Grohn YT, Wiedmann M, Wells MT. 2004. Mathematical model of Listeria
monocytogenes cross-contamination in a fish processing plant. Journal of Food Protection
67(12):2688-97.
Jackson TC, Acuff GR, Lucia LM, Prasai RK, Benner RA, Terry CT. 1993. Survey of residential
refrigerators for the presence of Listeria-monocytogenes. Journal of food protection
56(10):874-5.

129

Johnson JL, Doyle MP, Cassens RG. 1990. Listeria-monocytogenes and other listeria spp in meat
and meat-products - a review. Journal of Food Protection 53(1):81-91.
Kang D, Eifert JD, Williams RC, Pao S. 2007. Evaluation of quantitative recovery methods for
Listeria monocytogenes applied to stainless steel. J. AOAC Int. 90(3):810-6.
Keskinen LA, Todd ECD, Ryser ET. 2008. Impact of bacterial stress and biofllm-forming ability
on transfer of surface-dried Listeria monocytogenes during slicing of delicatessen meats.
International Journal of Food Microbiology 127(3):298-304.
Krekeler C, Ziehr H, Klein J. 1989. Physical methods for characterization of microbial cellsurfaces. Experientia 45(11-12):1047-55.
Krysinski EP, Brown LJ, Marchisello TJ. 1992. Effect of cleaners and sanitizers on Listeriamonocytogenes attached to product contact surfaces. Journal of Food Protection
55(4):246-51.
Kusumaningrum HD, Riboldi G, Hazeleger WC, Beumer RR. 2003. Survival of foodborne
pathogens on stainless steel surfaces and cross-contamination to foods. International
Journal of Food Microbiology 85(3):227-36.
Lin CM, Takeuchi K, Zhang L, Dohm CB, Meyer JD, Hall PA, Doyle MP. 2006. Crosscontamination between processing equipment and deli meats by Listeria monocytogenes.
Journal of Food Protection 69(1):71-9.
Locci R, Peters G, Pulverer G. 1981. Microbial colonization of prosthetic devices .1.
Microtopographical characteristics of intravenous catheters as detected by scanning
electron-microscopy. Zentralblatt Fur Bakteriologie Mikrobiologie Und Hygiene Serie BUmwelthygiene Krankenhaushygiene Arbeitshygiene Praventive Medizin 173(5):285-92.
Ludwicka A, Jansen B, Wadstrom T, Pulverer G. 1984. Attachment of Staphylococci to various
synthetic-polymers. Zentralblatt Fur Bakteriologie Mikrobiologie Und Hygiene Series aMedical Microbiology Infectious Diseases Virology Parasitology 256(4):479-89.
Lunden JM, Autio TJ, Korkeala HJ. 2002. Transfer of persistent Listeria monocytogenes
contamination between food-processing plants associated with a dicing machine. Journal
of Food Protection 65(7):1129-33.

130

Mafu AA, Roy D, Goulet J, Magny P. 1990a. Attachment of Listeria-monocytogenes to stainlesssteel, glass, polypropylene, and rubber surfaces after short contact times. Journal of Food
Protection 53(9):742-6.
Mafu AA, Roy D, Goulet J, Savoie L. 1991. Characterization of physicochemical forces
involved in adhesion of Listeria-monocytogenes to surfaces. Applied and Environmental
Microbiology 57(7):1969-73.
Mafu AA, Roy D, Goulet J, Savoie L, Roy R. 1990b. Efficiency of sanitizing agents for
destroying Listeria-monocytogenes on contaminated surfaces. Journal of Dairy Science
73(12):3428-32.
Marshall KC, Stout R, Mitchell R. 1971. Mechanism of initial events in sorption of marine
bacteria to surfaces. Journal of General Microbiology 68(NOV):337-&.
Mcallister EW, Carey LC, Brady PG, Heller R, Kovacs SG. 1993. The role of polymeric surface
smoothness of biliary stents in bacterial adherence, biofilm deposition, and stent
occlusion. Gastrointestinal Endoscopy 39(3):422-5.
Mceldowney S, Fletcher M. 1988. Bacterial desorption from food container and food-processing
surfaces. Microbial Ecology 15(2):229-37.
Merritt K, Shafer JW, Brown SA. 1979. Implant site infection-rates with porous and dense
materials. Journal of Biomedical Materials Research 13(1):101-8.
Midelet G, Carpentier B. 2002. Transfer of microorganisms, including Listeria monocytogenes,
from various materials to beef. Applied and Environmental Microbiology 68(8):4015-24.
Montville R, Schaffner DW. 2003. Inoculum size influences bacterial cross contamination
between surfaces. Applied and Environmental Microbiology 69(12):7188-93.
Moore G, Blair IS, mcdowell DA. 2007. Recovery and transfer of Salmonella Typhimurium
from four different domestic food contact surfaces. Journal of Food Protection
70(10):2273-80.

131

Mosteller TM, Bishop JR. 1993. Sanitizer efficacy against attached bacteria in a milk biofilm.
Journal of Food Protection 56(1):34-41.
Naylor PT, Myrvik QN, Gristina A. 1990. Antibiotic-resistance of biomaterial-adherent
coagulase-negative and coagulase-positive Staphylococci. Clinical Orthopaedics and
Related Research (261):126-33.
Norrung B, Andersen JK, Schlundt J. 1999. Incidence and control of Listeria monocytogenes in
foods in Denmark. International Journal of Food Microbiology 53(2-3):195-203.
Oh DH, Marshall DL. 1995. Destruction of Listeria-monocytogenes biofilms on stainless-steel
using monolaurin and heat. Journal of Food Protection 58(3):251-5.
Okabe S, Hirata K, Watanabe Y. 1995. Dynamic changes in spatial microbial distribution in
mixed-population biofilms: Experimental results and model simulation. Water Science
and Technology 32(8):67-74.
Ong YL, Razatos A, Georgiou G, Sharma MM. 1999. Adhesion forces between E-coli bacteria
and biomaterial surfaces. Langmuir 15(8):2719-25.
Pascual A, Dearellano ER, Martinez LM, Perea EJ. 1993. Effect of polyurethane catheters and
bacterial biofilms on the in-vitro activity of antimicrobials against Staphylococcusepidermidis. Journal of Hospital Infection 24(3):211-8.
Perez-Rodriguez F, Valero A, Carrasco E, Garcia RM, Zurera G. 2008. Understanding and
modelling bacterial transfer to foods: a review. Trends in Food Science & Technology
19(3):131-44.
Pether JVS, Gilbert RJ. 1971. Survival of Salmonellas on finger-tips and transfer of organisms to
foods. Journal of Hygiene-Cambridge 69(4):673-&.
Reynolds EC, Wong A. 1983. Effect of adsorbed protein on hydroxyapatite zeta potential and
Streptococcus-mutans adherence. Infection and Immunity 39(3):1285-90.
Richard J, Piton C. 1986. Semi-log model for interpreting the results of swabbing surfaces
naturally contaminated. Journal of Applied Bacteriology 60(3):243-9.

132

Rodriguez A, Autio WR, mclandsborough LA. 2007a. Effect of biofilm dryness on the transfer
of Listeria monocytogenes biofilms grown on stainless steel to Bologna and hard salami.
Journal of Food Protection 70(11):2480-4.
Rodriguez A, Autio WR, mclandsborough LA. 2007b. Effects of inoculation level, material
hydration, and stainless steel surface roughness on the transfer of Listeria monocytogenes
from inoculated bologna to stainless steel and high-density polyethylene. Journal of Food
Protection 70(6):1423-8.
Rodriguez A, mclandsborough LA. 2007. Evaluation of the transfer of Listeria monocytogenes
from stainless steel and high-density polyethylene to bologna and American cheese.
Journal of Food Protection 70(3):600-6.
Satou N, Satou J, Shintani H, Okuda K. 1988. Adherence of Streptococci to surface-modified
glass. Journal of General Microbiology 134:1299-305.
Schaffner DW. 2004. Mathematical frameworks for modelling Listeria cross-contamination in
food-processing plants. Journal of Food Science 69(6):R155-R9.
Scott E, Bloomfield SF. 1990. The survival and transfer of microbial-contamination via cloths,
hands and utensils. Journal of Applied Bacteriology 68(3):271-8.
Sheen S. 2008. Modeling surface transfer of Listeria monocytogenes on salami during slicing.
Journal of Food Science 73(6):E304-E11.
Sheen S, Costa S, Cooke P. 2010. Impact of mechanical shear on the survival of Listeria
monocytogenes on surfaces. Journal of Food Science 75(6):E387-E93.
Sheen S, Hwang CA. 2010. Mathematical modeling the cross-contamination of Escherichia coli
O157:H7 on the surface of ready-to-eat meat product while slicing. Food Microbiology
27(1):37-43.
Sheth NK, Franson TR, Sohnle PG. 1985. Influence of bacterial adherence to intravascular
catheters on invitro antibiotic susceptibility. Lancet 2(8467):1266-8.

133

Smith JL, Fratamico PM. 1995. Factors involved in the emergence and persistence of food-borne
diseases. Journal of Food Protection 58(6):696-708.
Sofos JN, Geornaras I. 2010. Overview of current meat hygiene and safety risks and summary of
recent studies on biofilms, and control of Escherichia coli O157:H7 in nonintact, and
Listeria monocytogenes in ready-to-eat, meat products. Meat Science 86(1):2-14.
Somers EB, Schoeni JL, Wong ACL. 1994. Effect of trisodium phosphate on biofilm and
planktonic cells of Campylobacter-jejuni, Escherichia-coli O157-H7, Listeriamonocytogenes and Salmonella-Typhimurium. International Journal of Food
Microbiology 22(4):269-76.
Speier JL, Malek JR. 1982. Destruction of microorganisms by contact with solid-surfaces.
Journal of Colloid and Interface Science 89(1):68-76.
Sugarman B, Musher D. 1981. Adherence of bacteria to suture materials. Proceedings of the
Society for Experimental Biology and Medicine 167(2):156-60.
Swaminathan B, Cabanes D, Zhang W, Cossart P. 2007. Listeria monocytogenes. Food
Microbiology: Fundamentals and Frontiers, Third Edition:457-+.
Takahashi H, Kuramoto S, Miya S, Kimura B. 2010. Desiccation survival of Listeria
monocytogenes and other potential foodborne pathogens on stainless steel surfaces is
affected by different food soils Food Control
Takeuchi K, Frank JF. 2000. Penetration of Escherichia coli O157 : H7 into lettuce tissues as
affected by inoculum size and temperature and the effect of chlorine treatment on cell
viability. Journal of Food Protection 63(4):434-40.
USDA-NIFSI. 2010. National Institute for Food and Agriculture. Priority areas for Listeria
monocytognes. Http://www.csrees.usda.gov/foodsafetybiosecurity.cfm. Accessed November
2012.
Van Oss CJ. 2003. Long-range and short-range mechanisms of hydrophobic attraction and
hydrophilic repulsion in specific and aspecific interactions. Journal of Molecular
Recognition 16(4):177-90.

134

Vanloosdrecht MCM, Lyklema J, Norde W, Schraa G, Zehnder AJB. 1987. The role of bacterialcell wall hydrophobicity in adhesion. Applied and Environmental Microbiology
53(8):1893-7.
Vorst KL, Todd ECD, Ryser ET. 2006a. Transfer of Listeria monocytogenes during mechanical
slicing of turkey breast, bologna, and salami. Journal of Food Protection 69(3):619-26.
Vorst KL, Todd ECD, Ryser ET. 2006b. Transfer of Listeria monocytogenes during slicing of
turkey breast, bologna, and salami with simulated kitchen knives. Journal of Food
Protection 69(12):2939-46.
Wang CH, Muriana PM. 1994. Incidence of Listeria-monocytogenes in packages of retail franks.
Journal of Food Protection 57(5):382-6.
Westergren G, Olsson J. 1983. Hydrophobicity and adherence of oral Streptococci after repeated
subculture invitro. Infection and Immunity 40(1):432-5.
Whitehead KA, Rogers D, Colligon J, Wright C, Verran J. 2006. Use of the atomic force
microscope to determine the effect of substratum surface topography on the ease of
bacterial removal. Colloids and Surfaces B-Biointerfaces 51(1):44-53.
Wirtanen G, Husmark U, mattilasandholm T. 1996. Microbial evaluation of the biotransfer
potential from surfaces with Bacillus biofilms after rinsing and cleaning procedures in
closed food-processing systems. Journal of Food Protection 59(7):727-33.
Zhinong Y, Todd ECD, Ryser ET. 2006. Quantitative Transfer of Listeria monocytognes
Between Conveyor Belts and Deli Meats. IAFP Oral Presentation. August 2006. .
Zobell CE. 1943. The effect of solid surfaces upon bacterial activity. Journal of Bacteriology
46(1):39-56.
Zottola EA. 1994. Microbial attachment and biofilm formation - a new problem for the foodindustry. Food Technology 48(7):107-14.

135