M IGAN

 

 

 

 

 

 

 

 

 

 

 

I!Iii/Iizflijifliflfl‘fliiiiii
Z4750 LIBRARY 9013
Michigan State
University

AN EPIDEMIOLOGICAL STUDY OF ANTIMICROBIAL
RESIDUES DETECTED IN MICHIGAN COWS' MILK

presented by
Suzanne Noel Gibbons-Burgener, DVM

has been accepted towards fulfillment
of the requirements for

Ph.D. degreein Large Animal Clinical

Sciences

y Marat professor

Date 2/2410‘9
VI 1

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

f“- fl-

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

 

DATE DUE DATE DUE DATE DUE

 

oecoT80 322002:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

woo c/crnaoatmpes-pn

AN EPIDEMIOLOGICAL STUDY OF ANTIMICROBIAL RESIDUES DETECTED
IN MICI-HGAN COWS’ MILK

By

Suzanne Noel Gibbons-Burgener, DVM

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Large Animal Clinical Sciences
(Epidemiology)

2000

ABSTRACT

AN EPIDEMIOLOGICAL STUDY OF ANTIMICROBIAL RESIDUES DETECTED
IN MICHIGAN COWS’ MILK

By

Suzanne Noel Gibbons-Burgener, DVM

Following widely publicized accounts of undetected antimicrobial residues in
milk making it to market in the late 19808, the Milk and Dairy Beef Quality Assurance
Program (QAP) was developed. Though designed to prevent drug residues, the impact of
the QAP on residue occurrence has yet to be determined. A commonly promoted and
adopted preventative practice has been the unapproved use of residue detection assays to
test for antimicrobial residues in milk from treated individual cows. The reliability of
these assays in testing individual cow milk has yet to be established.

The epidemiological research presented here was achieved through two main
studies that addressed six objectives. The first study was a retrospective study of
Michigan dairy farms evaluating the Milk and Dairy Beef Quality Assurance Program
(QAP) and its role in the prevention of violative antimicrobial residues in milk and the
adoption of prudent drug management practices. The first and second objectives were to
determine if QAP certification and specific management factors were associated with a
reduced risk of having antimicrobial residues in milk. Certification in the QAP was
associated with a tendency toward reduced risk (OR=O.3 [0.07-1.32]) of having
experienced a violative residue in bulk-tank milk. The risk of having had a residue was
reduced on farms treating >10% of their herd for metritis, and having their milk processor

perform residue testing. However, on-farm residue testing and maintaining written

identification records of treated cows was associated with an increased risk of having had
a residue. In a separate set of analyses the associations between QAP certification and the
use of prudent drug management practices were evaluated (Objective 3). Involuntary
certification was associated with maintenance of good written treatment records and
performance of on-farm residue testing. Voluntary certification was weakly associated
with use of refrigerated drug storage. These results suggest that farms adopted specific
management practices, irrespective of certification.

The second study was a longitudinal experimental study evaluating the reliability
of 3 on-farm assays when used to test individual cow milk for antimicrobial residues
following treatment for mild clinical mastitis. Methods were developed (Objective 4) to
improve the high performance liquid chromatography (HPLC) analyses for detection of
ampicillin and pirlimycin in milk. The reliability of the assays was expressed as
sensitivity, specificity, and positive and negative predictive values (Objective 5). Ranging
from 32.14 to 73.68%, the positive predictive values were poor for all three assays when
using the assays’ detection limits. Additional statistical analyses were used to determine
whether somatic cell count, Ing, bacterial isolates or specific antimicrobial treatments
were associated with false-positive results (Objective 6). Milk IgGl concentrations were
positively associated with false positive results from the all 3 assays.

The tendency of the QAP to prevent violative residues provides encouraging
information for the continued promotion and implementation of the Program. Dairy
producers and veterinarians can use the findings to target their residue prevention efforts.
Producers should reconsider their reliance on screening assays for testing individual

cows’ milk on-farm as a primary tool for residue prevention.

ACKNOWLEDGMENTS

Many people have made the completion of this dissertation possible. I am forever
grateful to my family and friends for their encouragement, support and love. Without
them, “real life” may have passed me by.

I would like to thank my advisory committee; Drs. Ron Erskine, Greg Fink, Joe
Leykam, Jim Lloyd and Scott McEwen for their guidance and patience. They’ve
provided me with a diverse set of skills and steered me toward the “big picture”. I am
lucky to have worked with such a wonderful group of people.

I am most indebted to my major advisor, Dr. John B. Kaneene. As an advisor, he
tailor-made a program that allowed me to experience the many facets of academic
pursuit. As a colleague, he respected me and the contribution I could make to the
veterinary profession. As a fiend, John is genuinely caring and compassionate about my
happiness and success beyond the University.

Additional gratitude goes to all of the Population Medicine Center and
Macromolecular Structure Facility faculty, staff and graduate students. I enjoyed being
part of both teams.

The success of my research was dependent on the cooperation of hundreds of
Michigan Grade A dairy farms and the MDA — Dairy Division. I especially thank the
eight farms that participated in the residue detection assay study.

Above all, I thank God for his unending love and all the blessings bestowed upon

me.

iv

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................... ix

LIST OF FIGURES .......................................................................................................... xi

INTRODUCTION
RATIONALE ......................................................................................................... l
PROBLEM STATEMENT ..................................................................................... 2
OBJECTIVES ........................................................................................................ 2
HYPOTHESES ....................................................................................................... 3
OVERVIEW .......................................................................................................... 3

CHAPTER 1

ANTIMICROBIAL RESIDUES IN MILK: A REVIEW
Introduction ............................................................................................................ 5
Concerns regarding residues in meat and milk ...................................................... 5
Risk factors for antimicrobial residues in milk ...................................................... 7
Preventing antimicrobial residues .......................................................................... 8
Antimicrobial residue testing ............................................................................... 1]
Areas for future study ........................................................................................... 14
REFERENCES .................................................................................................... 16

CHAPTER 2

EVALUATION OF CERTIFICATION IN THE MILK AND DAIRY BEEF
QUALITY ASSURANCE PROGRAM AND RELATED FACTORS ON THE
RISK OF HAVING VIOLATIVE ANTIBIOTIC RESIDUES IN MILK FROM

DAIRIES IN MICHIGAN
ABSTRACT ......................................................................................................... 21
INTRODUCTION ................................................................................................ 23
MATERIALS AND METHODS
Design ...................................................................................................... 24
Sample selection ...................................................................................... 24
Data collection ......................................................................................... 25
Statistical analysis .................................................................................... 25
RESULTS ............................................................................................................ 26
DISCUSSION ...................................................................................................... 29
FOOTNOTES ...................................................................................................... 34
REFERENCES .................................................................................................... 35
FIGURES ............................................................................................................. 37
TABLES ............................................................................................................... 38

CHAPTER 3

INFLUENCE OF THE MILK AND DAIRY BEEF QUALITY ASSURANCE
PROGRAM ON MICHIGAN DAIRY FARM DRUG MANAGEMENT
PRACTICES

ABSTRACT ......................................................................................................... 43
INTRODUCTION ................................................................................................ 45
MATERIALS AND METHODS
Study design ............................................................................................. 46
Study population ...................................................................................... 46
Data collection ......................................................................................... 47
Statistical analyses ................................................................................... 47
RESULTS ............................................................................................................ 48
DISCUSSION ...................................................................................................... 50
FOOTNOTES ...................................................................................................... 54
REFERENCES .................................................................................................... 55
TABLES ............................................................................................................... 57
CHAPTER 4

IDENTIFICATION AND QUANTIFICATION OF AMPICILLIN,
CEPHAPIRIN AND PIRLIMYCIN IN COWS’ MILK USING HIGH
PERFORMANCE LIQUID CHROMATOGRAPHY AND FLUORESCENCE
DETECTION

ABSTRACT ......................................................................................................... 63
INTRODUCTION ................................................................................................ 64
MATERIALS AND METHODS
Equipment used ........................................................................................ 65
Reagents and solutions used .................................................................... 66
Extraction and derivatization of ampicillin .............................................. 68
Extraction of cephapirin and pirlimycin .................................................. 68
Derivatization for pirlimycin detection .................................................... 69
Detection of ampicillin residue ................................................................ 69
Detection of pirlimycin ............................................................................ 70
Detection of cephapirin ............................................................................ 70
Quantification of antimicrobials .............................................................. 70
RESULTS AND DISCUSSION .......................................................................... 71
FOOTNOTES ...................................................................................................... 74
REFERENCES .................................................................................................... 75
FIGURES ............................................................................................................. 77

vi

CHAPTER 5

AN EPIDEMIOLOGICAL EVALUATION OF THE RELIABILITY OF
BULK-TANK RESIDUE DETECTION ASSAYS USED TO TEST
INDIVIDUAL COW MILK

ABSTRACT ......................................................................................................... 79
INTRODUCTION ................................................................................................ 81
MATERIALS AND METHODS
Study design ............................................................................................. 82
Case definition ......................................................................................... 82
Sample size .............................................................................................. 82
Treatment groups ..................................................................................... 82
Sample collection ..................................................................................... 83
Testing for antimicrobials ........................................................................ 84
Statistical analysis .................................................................................... 84
RESULTS ............................................................................................................ 85
DISCUSSION ...................................................................................................... 86
REFERENCES .................................................................................................... 90
FIGURES ............................................................................................................. 92
TABLES ............................................................................................................... 96
CHAPTER 6

RISK FACTORS ASSOCIATED WITH FALSE-POSITIVE RESULTS
USING THREE ON-F ARM BULK TANK RESIDUE DETECTION ASSAYS

ON INDIVIDUAL COW MILK
ABSTRACT ......................................................................................................... 99
INTRODUCTION .............................................................................................. 10]
MATERIALS AND METHODS
Study design ........................................................................................... 102
Case definition ....................................................................................... 102
Sample size ............................................................................................ 102
Treatment groups ................................................................................... 102
Sample collection ................................................................................... 103
Testing for somatic cell count ................................................................ 103
Quantification of bovine IgG1 ................................................................ 103
Bacteriologic culture of milk ................................................................. 104
Testing for antimicrobial residues .......................................................... 104
Statistical analysis .................................................................................. 104
RESULTS .......................................................................................................... 105
DISCUSSION .................................................................................................... 106
FOOTNOTES .................................................................................................... 108
REFERENCES .................................................................................................. 109
TABLES ............................................................................................................. 1]]

vii

SUMMARY AND CONCLUSIONS

SUMMARY ....................................................................................................... 116
CONCLUSIONS ................................................................................................ 1 17
APPENDICES
APPENDIX 1
Dairy Production Management Study Survey ........................................ 121
APPENDIX 2
Sample Size Calculations ....................................................................... 125
REFERENCES .............................................................................................................. 127

viii

Table 1-1

Table 2-1

Table 2-2

Table 2-3

Table 2-4

Table 3-1

Table 3-2

Table 3-3

Table 3-4

Table 5-1

Table 5-2

Table 5-3

LIST OF TABLES

Ten critical control points of the QAP Residue Prevention
Protocol.

Herd size categories and the number of case operations from each
category.

Distribution of categorical risk factors among case and control
operations.

Distribution of continuous risk factors among cases and controls.

Results of the conditional multivariable logistic regression
analysis of associations among chemical residue occurrence in
milk and herd-level risk factors.

Percentages of Michigan dairy farms that used various drug-use
management practices.

Results of multivariable logistic regression analysis of
associations among use of refrigerated drug storage and farm
management factors.

Results of multivariable logistic regression analysis of
associations between use of on-farm drug testing and farm
management factors.

Results of multivariable logistic regression analysis of
associations between maintenance of good records and farm
management factors.

FDA tolerance levels for specific antimicrobials in marketable
milk and the visual minimal detection limits of three on-farm
residue detection assays.

Sensitivity, specificity and predictive values for the detection of
antimicrobials at or above the reported assay detection limit by
three on-farm residue detection assays evaluated in a field trial.

Sensitivity, specificity and predictive values for the detection of

antimicrobials above the F DA-established tolerance levels by
three on-farm residue detection assays evaluated in a field trial.

ix

Page

38

39

41

42

57

60

61

62

96

97

98

Table 6-1

Table 6-2

Table 6-3

Table 6-4

Table 6-5

Means and standard deviations of somatic cell count (SCC) and
milk IgG1 concentration for false-positive residue detection assay
results.

Frequency distribution of antimicrobial therapy as a risk factor
for false-positive residue detection assay results.

Frequency distribution of bacteriologic culture results as risk
factors for false-positive residue detection assays results.

Univariate logistic regression results used to estimate the relative
risk of possible milk components producing a false-positive
result.

Multivariable logistic model of risk factors associated with false-
positive SNAP B-lactam assay results.

Page
111

112

113

114

115

Figure 2-1

Figure 4-1

Figure 4-2

Figure 5-1

Figure 5—2

Figure 5-3

Figure 5-4

LIST OF FIGURES

Geographic distribution of Michigan agricultural statistics
districts defined by the National and Michigan Agricultural
Statistics Services.

Chromatogram of the fluorescent detection of 20 ppb ampicillin
extracted from raw milk.

Chromatogram of the fluorescent detection of 200 ppb pirlimycin
extracted from raw milk.

Assignment and treatment options of cases in the antimicrobial
and control treatment groups.

An example of how sampling frequency was determined for the
collection of the pre and post-treatment samples using the most
prior antimicrobial treated cow’s sampling interval for the control
group cow.

2 x 2 charts for the distribution of assay results using the reported
antimicrobial detection limits for each assay.

2 x 2 charts for the distribution of assay results using the FDA-
established antimicrobial tolerance levels.

xi

Page
37

77

78

92

93

94

95

INTRODUCTION

RATIONALE

Since the introduction of penicillin nearly half a century ago, the use of
antimicrobials to prevent and treat diseases in cattle has increased tremendously. With
the use of antimicrobials in livestock came the potential risk of residues these
medications could pose. Harmful effects ascribed to antimicrobial residues found in meat
and milk include: manufacturing difficulties in products requiring fermentation or live
cultures, allergic reactions following consumption of residues in food, and potential
contributions to the development of antimicrobial resistant bacterial strains.
Additionally, consumer perceptions can have adverse effects on the demand for
implicated foods. The dairy industry has long recognized the need to be proactive in
requiring quality raw milk from farms. Following widely publicized accounts of
undetected antimicrobial residues in milk making it to market in the late 19805, the
National Milk Producers Federation and the American Veterinary Medical Association
developed the Milk and Dairy Beef Quality Assurance Program (QAP). The QAP
provides 10 critical control points meant to reduce the incidence of individual farms
experiencing violative residues. There have been limited epidemiological studies
determining risk factors for farms experiencing residues. Many of the critical control
points address risk factors identified by these studies, however much of the QAP appears
to be based more on common sense and intuition than on scientific studies. The eighth
critical control point emphasizes the use of residue screening assays in the prevention of

residues in the bulk tank. Many in the industry, including veterinarians, producers, and

dairy co-operatives, have interpreted this recommendation to imply testing individual
cows’ milk is the best prevention. Several studies have indicated that when testing
individual cow milk, false—positive and false-violative results were obtained using the
currently available assays that were validated for testing commingled milk. Therefore,
there is a need for field trials to adequately evaluate the use of residue detection assays

and other components of the QAP in the pursuit of on-farm residue prevention.

PROBLEM STATEMENT

Consumer, industry and dairy producer concerns about potential drug residues in
milk continues to drive the dairy industry toward investing more resources into the
monitoring and prevention of residues. Though designed to prevent drug residues, the
impact of the QAP on residue occurrence has yet to be determined. Specifically, it is
important to evaluate the Program’s influence in changing producers’ drug use
management practices. One of the most promoted and commonly adopted preventative
practice has been the use of residue detection assays to test for antimicrobial residues in
milk from treated individual cows prior to keeping and marketing its milk. However, the
assays are only approved for use in testing commingled (tanker) milk. The reliability of

these same assays in testing individual cow milk has yet to be established.

OBJECTIVES
This dissertation focuses on the epidemiological evaluation of the prevention of

antimicrobial residues in milk. The objectives of the two main studies were to:

1. Determine if QAP certification was associated with a reduced risk of having
antimicrobial residues in milk.

2. Define specific management factors that may have predisposed dairy farms to having
violative antimicrobial residues in milk.

3. Determine if QAP certification was associated with the use of prudent drug
management practices.

4. Develop robust gold standard methods for use in determining the reliability of the
Delvo-SP, Penzyme Milk Test and SNAP B-lactam assays in the detection of
ampicillin, cephapirin and pirlimycin in raw milk.

5. Determine the reliability of the Delvo-SP, Penzyme Farm Milk Test and SNAP B-
lactam residue test assays when used to test individual cow milk.

6. Identify risk factors that may be associated with false assay results.

HYPOTHESES

Chapters 2, 3, 5 and 6 each contain specific hypotheses.

OVERVIEW

Chapter 1 is a literature review of the epidemiology and detection of antimicrobial
residues in milk. The remaining chapters have been written in a format suitable for
independent publication. Two main studies were conducted to accomplish the six
objectives. The second and third chapters present findings from a retrospective study
evaluating the effect of the QAP, while chapters 4-6 contain the findings from a

prospective study evaluating the reliability of residue detection assays used to test

individual cow milk. Chapter 2 evaluates the impact of QAP certification in the
prevention of antimicrobial residues. Chapter 3 evaluates the potential associations
between the use of specific drug management practices and a dairy farm’s QAP
certification status (non-, involuntarily and voluntarily QAP-certified). Chapter 4
describes the development of two high pressure liquid chromatography methods used as
gold standards in the detection of trace amounts of ampicillin, cephapirin and pirlimycin.
Chapter 5 evaluates the reliability of the Delvo-SP, Penzyme Farm Milk Test and SNAP
B-lactam assays when used to test milk from individual cows diagnosed and treated for
mild clinical mastitis. Chapter 6 then identifies risk factors that may contribute to the
occurrence of false-results when using the residue test assays on individual cow milk.
The contribution of both main studies to the understanding of the epidemiology of residue

prevention is presented in the overall summary.

CHAPTER 1

ANTIMICROBIAL RESIDUES IN MILK - A REVIEW

Introduction

Residues in food products of animal origin are the presence of foreign substances
that can be parent compounds, their metabolites or other substances produced as a
consequence of administering the parent compound. There are a variety of chemicals that
can result in residues in meat and milk, including antimicrobials, insecticides,
antihelrninthics, hormones, heavy metals and pesticides. Most chemicals are excreted in
the urine, feces and milk of the exposed animal. Veterinary pharmaceuticals approved for
use in food producing animals include meat, and possibly milk, withholding periods on
their labels. These withholding periods are based on the pharmacokinetics of the drug
and provide a timeline within which the animal will excrete enough of the drug and its
metabolites to allow any remaining residue to fall below the established tolerance or safe
level. A violative residue occurs when the chemical is detected in milk or tissues at a
level exceeding the tolerance limit. Condemnation of the carcass or load of milk is the

usual course of action.

Concerns regarding residues in meat and milk
Processors of dairy products were some of the first to note the adverse effect of
antimicrobial residues on the production of products requiring live cultures and

fermentation (Stoltz and Hankinson, 1953; Albright et a1, 1961). Public health concerns

regarding residues in food products have evolved as more pharmaceuticals have been
utilized in animal production and the testing methods have become more sensitive (Engel,
1980). Often the difficulty is deciphering whether the concerns are based on reality or
perception. Consumers are particularly concerned when the apparent hazard isn’t visible
and avoidance is, therefore, perceived as out of their control. Research has shown that
consumers rarely list antimicrobials or hormones as a major concern. Yet, when directly
asked, half the people believed that those chemicals could pose a serious hazard (Bruhn,
1996). Does the consumer have reason to worry? The most significant hazard centers on
the link between antimicrobial use in food-producing animals and the development of
microorganisms that are resistant to those antimicrobials (Franco, et al., 1990; Brady, et
al., 1993). Though it’s the actual use of antimicrobials that have come under scrutiny,
residues may be seen as evidence of possible misuse of the drugs. In addition, there is the
remote possibility of an allergic reaction to some drugs, especially B-lactams (Huber,
1986; Kindred and Hubbert, 1993).

Acknowledging the potential negative impact of press articles (Ingersoll, 1989,
1990) publicizing alleged antimicrobial residues in retail dairy products, the National
Conference on Interstate Milk Shippers strengthened the grade A Pasteurized Milk
Ordinance by mandating increased testing of marketed milk (Center for Veterinary
Medicine, 1996). With active surveillance of milk at the creamery in place, it quickly
became apparent that prevention of antimicrobial residues should begin at the farm level.
Though not essential, knowing the causes of violative residues could aid the development

of preventative practices. Because total elimination of pharmaceutical use in the

treatment of diseased cows isn’t a realistic option, other preventative measures should be

explored.

Risk factors for antimicrobial residues in milk

Few studies have scientifically identified risk factors for residues in milk. A case-
control study by Kaneene and Ahl reported larger herd size, more hired employees, and
use of pre-medicated feeds were associated with an increased risk of a farm having
experienced a residue during the prior 5 years (Kaneene and Ahl, 1987). Those farms
with residues were also more likely to acknowledge the importance of adhering to
withdrawal periods and have residue testing equipment available. Another case-control
study (McEwen et al, 1991 a) of farms with violative residues in their milk concluded that
the employment of part-time labor to milk cows was associated with an increased risk of
residues. Management factors that reduced the risk were pipeline milking in tie stalls, use
of antimicrobial test assays, use of separate equipment when milking treated cows and the
belief that increasing the withholding period was necessary when increasing the drug
dose. With the addition of increased frequency of intramammary antibiotic treatments
associated with an increased risk of having had a residue, an earlier dairy farm survey by
the same group had similar findings (McEwen et a1, 1991 b). To date, the research
evaluating potential causes of residues has been restricted to retrospective studies,
because violative residues are rare occurrences. A study (Kaneene and Willeberg, 1989)
cautioned that results of these retrospective studies may exhibit significant recall and

information biases.

Preventing antimicrobial residues

In an effort to reaffirm the commitment of the US dairy industry to maintaining
the quality and safety of the nation's milk supply, the National Milk Producers Federation
and American Veterinary Medical Association completed the development of the Milk
and Dairy Beef Quality Assurance Program (QAP) in 1992. Official certification is given
when the producer espouses the principles of providing a high quality product by
preventing residues in milk and dairy beef. Certification can be voluntarily pursued by
producers, or it can be involuntarily implemented (i.e., required) in instances of a residue
violation in milk.

The QAP is based on the hazard analysis critical control point (HACCP)
principles. The residue prevention protocol comprises 10 critical control points (Table 1-
1), or good management practices, focusing on drug use protocol, herd health
management practices, record-keeping and employee education (Boeckman and Carlson,
1997). Many of the critical control points address risk factors that were identified in
studies for the increased risk of drug residue occurrence (Kaneene and Ahl, 1987;
McEwen et al, 1991 a & b). Other components of the critical control points, such as
maintaining treatment records and properly storing medications, appear to have been
incorporated into the QAP because of anecdotal and intuitive reasons (Day, 1993). Does
the QAP accomplish its goal of reducing the incidence of violative drug residues? That

question has yet to be answered.

Table 1-1 Ten critical control points of the QAP Residue Prevention Protocol.

 

Critical
control point Description of critical control point

1 Practice healthy herd management

2 Establish and maintain a valid veterinarian/client/patient
relationship

3 Use only approved drugs (Rx and OTC)

4 Ensure all drugs used on the farm have labels that comply
with state and/or federal labeling requirements

5 Store all drugs correctly

6 Administer all drugs prOperly and visibly identify all
treated animals

7 Maintain and use proper treatment records on all treated
animals

8 Use drug residue screening assays to test animals
receiving drugs in an extra-label manner.

9 Implement employee/family awareness (education) of

proper drug use to avoid marketing adulterated products

10 Review farm plan for residue prevention annually and
recertify every 2 years.

 

 

 

In a study that evaluated the use of an on—farm risk assessment tool (Sischo et al,
1997), the authors expressed concern that although the QAP does a good job of
articulating the hazards of residues, the program is deficient in three necessary
components of any HACCP program. Specifically, the Program doesn’t provide adequate
motivation and tools to allow farm owners to assess their own risk of illegal residues,
develop a plan to reduce their risk, or monitor their progress toward residue prevention.
In their study, the treatment and control herds received a copy of the QAP booklet and
were evaluated by use of the risk assessment tool. The treatment group received

additional information and guidance that led to a farm plan to reduce their risk of

residues. Although the overall risk of antibiotic residues was reduced by approximately
19%, there was no significant difference between the groups. It is difficult to ascertain
whether the risk assessment tool, the QAP booklet, or the combination of these two
factors had the greatest impact on risk reduction. Nevertheless, their study is one of the
few that have addressed the challenge of evaluating the QAP.

Ideally, the adoption of all the good management practices would significantly
reduce a farm’s risk of residue occurrence. Some farms may be more interested in
making only some management changes and want to know which of the critical control
points are most beneficial. Even before the QAP was instituted, dairy co-operatives,
veterinarians and extension personnel were promoting the use of on-farm residue
screening assays, such as the Delvotest, Penzyme Milk Test, Charm Farm, SNAP, Cite
Probe and LacTek assays. The assays were seen as “insurance” that a farm’s bulk-tank
was clear of antimicrobials prior to marketing (Jones and Seymour, 1988; Adams, 1993).
The assays available for on-farm testing are relatively simple to use and give a qualitative
result (yes, maybe or no drug is present). Some of the same screening assays have
recently been validated for regulatory use in testing commingled milk at creameries
(Center for Veterinary Medicine, 1996).

There are several reasons why the use of on-farrn residue screening assays may
not be the panacea they appear to be. Producers are encouraged to either submit samples
or test milk on-farm from individual treated cows prior to including the cow’s milk in the
bulk-tank. An important point to remember is that the screening assays are approved and
labeled for use in testing commingled milk only. Their reliability in testing individual

cow milk has not been established. A recent study (Slenning and Gardner, 1997)

10

evaluating the economic risk of using on-farm residue testing programs, indicated that not
testing was slightly less costly to a farm. Though small changes in milk price or assay
costs will alter the dynamics of the economic models, it is evident that indiscriminate
testing of treated animals is not a cost-effective recommendation as “insurance” against
residues.

Since the changes to the grade A Pasteurized Milk Ordinance in 1991 mandated
the screening of every tanker-load of raw milk for at least B-lactam antimicrobial
residues, there has been an increased emphasis on the preharvest prevention of residues at
the farm level (Adams, 1994). Unfortunately, the only tests available are approved for
screening commingled milk, and have been reported to produce false-positive results
when used to test individual cow milk for antimicrobial residues (Sischo and Burns,
1993; Cullor et al, 1994; Andrew et a1, 1997). Regardless of the potential drawbacks of
residue testing, many producers feel the benefits outweigh the negatives and opt for either
on-farm or milk handler testing of individual cow milk in an efi’ort to avoid illegal

residues.

Antimicrobial Residue Testing

Although the specificity and sensitivity of the variety of assays have been
established for commingled milk in controlled laboratory conditions, concern over the
accuracy under field conditions exists, because to date, few studies have been conducted
that validate these assays in a field setting (Cullor, 1996; Gardner et al, 1996). Approval
of residue assays is dependent on the testing of milk spiked with known quantities of

specific antimicrobials. In an effort to increase analytical sensitivity, many of the assays

11

test for antimicrobials below the US Food and Drug (FDA) established tolerance or safe
levels (Mitchell et al, 1998). This creates a dilemma when discussing false-positive
results. Researchers have struggled with whether to use the term false-violative when an
assay result is positive, but the quantity is below the established tolerance level (Gardner
et al, 1996; Mitchell et al, 1998). The picture becomes even more grey when we consider
testing individual cow milk which would be diluted if included in a farm’s bulk tank.
The regulatory tolerance levels have been established for commingled milk being sold on
the market. It is apparent that the analytical sensitivity of the current on-farm assays
probably results in needless disposal of individual cow milk. Unless quantitative
methods, such as high performance liquid chromatography (HPLC) are used, there is no
accurate way of determining whether a sample contains a violative level of an
antimicrobial.

A new European database describes many biochemistry methods developed to
identify and quantitate antimicrobials in milk (Van Eeckhout et al, 1998). Mass
spectrometry has been used to identify b-lactam residues at their tolerance levels (Heller
and Ngoh, 1998). Beta-lactam antimicrobials comprise the majority of antimicrobials
approved for use in lactating cattle and the majority of residues detected in milk. HPLC is
the most commonly employed method of quantifying antimicrobials. It has been used to
identify and quantify penicillins and cephalosporins (Briguglio and Lau-Cam, 1984;
Dasenbrock and LaCourse, 1998; Hong et al, 1995; Moats, 1993; Moats, 1994; Moats
and Harik-Khan, 1995; Moats and Romanowski, 1998), macrolides (Heller, 1997),
sulfonamides (Smedley, 1994; Schwartz and Lightfield, 1995) and tetracyclines (White et

al, 1993). A combination of mass spectrometry and HPLC has also been explored

12

(Heller, 1996; Homish et al, 1995; Straub et al, 1994; Tyczkowska et al, 1994). Several
studies have compared screening assay and liquid chromatography results (Anderson et
al, 1998; Ang et al, 1997; Hank-Khan and Moats, 1995). Unlike most of the studies,
Anderson et a1. and Ang et al. used milk with incurred instead of spiked residues. One
study used milk from only two cows (Ang et al., 1997) and the other (Anderson et al.,
1998) utilized six residue screening assays to determine the qualitative status of specific
HPLC fractions. None of these studies were specifically designed to determine the
reliability of on-farm residue screening assays. The technical expertise, required
equipment and reagents, and lengthy analysis time make the HPLC methods impractical
and cost prohibitive for routine residue testing, but should be considered when gold
standard detection methods are necessary.

Because violative residues in milk are a relatively rare occurrence (<0.1% of bulk
tanks), the likelihood of false-negative results is negligible. Consequently, research
evaluating the accuracy of on-farm residue detection assays has focused on testing
individual cow milk and the possible sources of false-positive results. Possible causes of
false-positive results include elevated somatic cell count (Sischo and Burns, 1993; Van
Eenennaam et al, 1993); increased lactoferrin and lysozyme concentrations (Carlsson and
Bjorck, 1989); lower milk production, increased parity and increased coliforrn counts
(Andrew et al, 1997); and other inhibitory substances in the milk (Tyler et al, 1992;
Cullor et a1, 1994). The screening assays represent an array of detection methods, such
as microbial inhibition, microbial receptor, enzymatic colourimetric, and receptor binding
assays (Mitchell et al., 1998). The different detection methods may be affected

differently by the potential causes of false-positive results.

13

Producers are still left wondering whether and how to use on-farm residue
detection assays. The use of on-farm assays may have its place in investigating the source
of a violative residue on a farm (Musser and Anderson, 1999). The case study by Musser
and Anderson describes how a combination of detective work and cautious use of a series
of assays was used to identify the probable source of a violative residue on a farm. Again
the authors discouraged indiscriminate use of on-farrn assays to test all cattle. Additional
studies are needed to determine the reliability and usefirlness of on-farm residue detection

assays in the prevention of violative residues.

Areas for future study

There are few epidemiological studies of the prevention of residues in milk.
Introduction of the QAP was a major industry initiative to be pro-active in the prevention
of antimicrobial residues. The tremendous amount of financial and human resources
invested in the QAP nationwide warrants scientific evaluation of the efficacy of the
program in attaining its goals. Its impact and success in changing dairy management
practices to those considered prudent in the prevention of drug residues have not been
reported. By scientifically determining strengths and weaknesses of the QAP, the
evaluation of the Program will elicit dialogue regarding potential improvements.

Evaluating preventative practices for antimicrobial residues hinges on the correct
classification of residue occurrence. We need to know what we’re preventing. Many of
the practices focus on individual animal management. Because there are no residue
screening assays approved or validated for use in testing individual cow milk, and there

are numerous reports of false-positive results when assays approved for commingled milk

14

testing are utilized for individual cow milk testing, it is extremely important to determine
whether on-farm residue screening assays provide reliable results for both producers and
researchers. Misclassification (false positive and negative) of results can lead to
unnecessary disposal of milk on farms, misleading research findings and perhaps the
exposure of people to unnecessary residues. Field-based epidemiological studies will
provide much needed information regarding the usefulness of on-farrn assays and their

appropriateness as research tools.

15

REFERENCES

Adams JB. Assuring a residue-free food supply: Milk. J Am Vet Med Assn
1993;202:1723-1725.

Albright JL, Tuckey SL, Woods GT. Antibiotics in milk - a review. J Dairy Sci
1961;44:779-807.

Andrew SM, F robish RA, Paape MJ, Maturin LJ. Evaluation of selected antibiotic
residue screening tests for milk from individual cows and examination of factors that
affect the probability of false-positive outcomes. J Dairy Sci 1997;80:3050—3057.

Ang CYW, Luo W, Call VL, Righter HF. Comparison of liquid chromatography with
microbial inhibition assay for detection of incurred amoxicillin and ampicillin residues in
milk. J Agric Food Chem 1997;45:4351-4356.

Boeckman S, Carlson KR Milk and dairy beef quality assurance program: milk and
dairy beef residue prevention protocol. Stratford, IA: Agri-Education, Inc, 1997;1-70.

Brady MS, White N, Katz SE. Resistance development potential of

antibiotic/antimicrobial residue levels designated as "safe levels.” J Food Prot
1993;56:229-233.

Briguglio GT, Lau-Cam CA. Separation and identification of nine penicillins by reverse
phase liquid chromatography. J AOAC Int 1984;67:228-231.

Bruhn CM. Consumer Perceptions and concerns about veterinary drug residues. In:
Moats WA and Medina MB, ed. Veterinary Drug Residues: Food Safety. Washington,
DC, American Chemical Society, 1996, pp 18-21.

Carlsson A, Bjorck L. Lactoferrin and lysozyme in milk during acute mastitis and their
inhibitory effect in Delvotest P. J Dairy Sci 1989;72:3166-3175.

Center for Veterinary Medicine. Appendix N, Pasteurized Milk Ordinance. In C VM
Update: Milk monitor with antimicro drug screening test. 1996, pp 5-6.

Cullor J S. Dilemmas associated with antibiotic residue testing in milk. In: Moats WA
and Medina MB, ed. Veterinary Drug Residues: Food Safety. Washington, DC,
American Chemical Society, 1996, pp 44-57.

Cullor J S, van Eenennaam A, Gardner 1, et al.. Performance of various tests used to

screen antibiotic residues in milk samples from individual animals. J AOAC Int
1994;77:862-870.

16

Dasenbrock CO, LaCourse WR. Assay for cephapirin and ampicillin in raw milk by
high-performance liquid chromatography - integrated pulsed amperometric detection.
Anal Chem, 1998;70:2415-2420.

Day J. A subcommittee review of the quality assurance initiative: implementation issues
from the Implementation and Communication Subcommittee of the Drug Residue
Committee, in Proceedings. 32nd Annu Meet National Mastitis Council 1993;144-146.

Engel RE. Current food safety and quality service residue control program. J Am Vet
Med Assn 1980;176:1145-1147.

Franco DA, Webb J, Taylor CE. Antibiotic and sulfonamide residues in meat:
implications for human health. J Food Prot 1990;53:178-185.

Gardener IA, Cullor J S, Galey FD, et al.. Alternatives for validation of diagnostic assays
used to detect antibiotic residues in milk. J Am Vet Med Assn 1996;209:46-52.

Harik-Khan R, Moats WA. Identification and measurement of b-lactam antibiotic
residues in milk: integration of screening kits with liquid chromatography. J AOAC Int
1995;78:978-986.

Heller, DN. Determination and confirmation of pirlimycin residue in bovine milk and
liver by liquid chromatography/thermospray mass spectrometry: interlaboratory study. J
AOAC Int 1996;79:1054-1061.

Heller, DN. Determination of Pirlimycin residue in milk by liquid chromatographic
analysis of the 9-fluorenylmethyl chloroformate derivative. J AOA C Int 1997;80:975-
981.

Heller DN, Ngoh MA. Electrospray ionization and tandem ion trap mass spectrometry
for the confirmation of seven B-lactam antibiotics in bovine milk. Rapid Commun Mass
Spectrom 1998;12:2031-2040.

Hong CC, Lin CL, Tsai CE, Kondo F. Simultaneous identification and determination of
residual penicillins by use of high-performance liquid chromatography with
spectrophotometric or fluorometric detectors. Am J Vet Res 1995;56:297-303.

Homish RE, Cazars AR, Chester ST, Roof RD. Identification and determination of
pirlimycin residue in bovine milk and liver by high-performance liquid chromatography-

thermospray mass spectrometry. J Chromatogr B 1995;674:219-235.

Huber WG. Allergenicity of antibacterial drug residues. In: Rico AG, ed., Drug residues
in animals. Academic Press, Toronto, ONT 1986, pp 33-49.

17

Ingersoll B. Milk is found tainted with a range of drugs farmers give cattle. The Wall
Street Journal, December 29, 1989; A1, A2.

Ingersoll B. New York milk supply highly tainted, TV station says, based on own survey.
The Wall Street Journal, February 8, 1990; A4.

Jones GM and Seymour EH. Cowside antibiotic residue testing. J Dairy Sci,
1988;71:1691-1699.

Kaneene J B and Ahl AS. Drug residues in dairy cattle industry: Epidemiological
evaluation of factors influencing their occurrence. J Dairy Sci, 1987;70:2176-2180.

Kaneene JB, Willeberg P. Influence of management factors in the occurrence of antibiotic
residues in milk: a case-control study of Michigan dairy herds, with examples of
suspected information bias. Acta Vet Scand, 1989;84:473-476.

Kindred TP, Hubbert WT. Residue prevention strategies in the United States. J Am Vet
Med Assn, 1993 ;202:46-49.

McEwen SA, Black WD, Meek AH. Antibiotic residue prevention methods, farm
management, and occurrence of antibiotic residues in milk. J Dairy Sci 1991;74:2128-
2137 (a).

McEwen SA, Meek AH, Black WD. A dairy farm survey of antibiotic treatment
practices, residue control methods and associations with inhibitors in milk. J Food
Protect, 1991;54:454-459 (b).

Mitchell JM, Griffiths MW, McEwen SA, et al.. Antimicrobial drug residues in milk and
meat: causes, concerns, prevalence, regulations, tests, and test performance. J Food Prot,
1998;61:742-756.

Moats, WA. Determination of cephapirin and desacetylcephapirin in milk using
automated liquid chromatographic cleanup and ion-pairing liquid chromatography. J
AOAC Int, 1993;76:535-539.

Moats, WA. Determination of ampicillin and amoxicillin in milk with an automated
liquid chromatographic cleanup. J AOAC Int, 1994;77:41-45.

Moats WA, Harik-Khan. Liquid chromatographic determination of B-lactam antibiotics
in milk: a multiresidue approach. J AOAC Int, 1995;78:49-54.

Moats, WA, Romanowski, RD. Multiresidue determination of b-lactarn antibiotics in

milk and tissues with the aid of high-performance liquid chromatographic fractionation
for clean up. J Chromatogr A, 1998;812:237-247.

18

Musser JMB, Anderson KL. Using drug residue screening tests to investigate antibiotic
contamination of milk. Vet Med, 1999;94:474-479.

Schwartz DP, Lightfield AR. Practical screening procedures for sulfarnethazine and N4-
acetylsulfamethazine in milk at low parts-per-billion levels. J AOAC Int, 1995;78:967-
970.

Sischo WM. Quality milk and tests for antibiotic residues. J Dairy Sci, 1996;79:1065-
1073.

Sischo WM, Burns CM. Field trial of four cowside antibiotic-residue screening tests. J
Am Vet Med Assn, 1993;202:1249-1254.

Sischo WM, Kieman NE, Burns CM, et a1. Implementing a quality assurance program
using a risk assessment tool on dairy operations. J Dairy Sci, 1997;80:777-787.

Slenning BD, Gardner IA. Economic evaluation of risks to producers who use milk
residue testing programs. J Am Vet Med Assn, 1997;211:419-427.

Smedley MD. Liquid chromatographic determination of multiple sulfonamide residues in
bovine milk: collaborative study. J AOAC Int, 1994;77:1112-1122.

Stoltz E1, Hankinson DJ. Antibiotics and lactic acid starter cultures. Appl Microbiol,
1953;1221-29.

Straub R, Linder M, Voyksner RD. Determination of B-lactam residues in milk using
perfusive-particle liquid chromatography combined with ultrasonic nebulization
electrospray mass spectrometry. Anal Chem, 1994;66:3651-3658.

Tyczkowska KL, Voyksner RD, Straub RF, Aronson AL. Simultaneous multiresidue
analysis of b-lactam antibiotics in bovine milk by liquid chromatography with ultraviolet
detection and confirmation by electrospray mass spectrometry. J AOAC Int,
1994;77:1122-1131.

Tyler J W, Cullor J S, Erskine RJ, et al. Milk antimicrobial residue assay results in cattle
with experimental, endotoxin-induced mastitis. J Am Vet Med Assn, 1992;201:1378-
1384.

Van Eeckhout NJ, Van Peteghem CH, Helbo VC, et al.. New database on hormone and
veterinary drug residue determination in animal products. Analyst, 1998;123:2423-2427.

Van Eenennaam AL, Cullor J S, Perani L, et al.. Evaluation of milk antibiotic residue

screening tests in cattle with naturally occurring mastitis. J Dairy Sci 1993;76:3041-
3053.

19

White CR, Moats WA, Kotula KL. Optimization of a liquid chromatographic method for

determination of oxytetracycline, tetracycline and chlortetracycline in milk. J AOAC Int,
1993;76:549-554.

20

CHAPTER 2

EVALUATION OF CERTIFICATION IN THE MILK AND DAIRY BEEF
QUALITY ASSURANCE PROGRAM AND RELATED FACTORS ON THE
RISK OF HAVING VIOLATIVE ANTIBIOTIC RESIDUES IN MILK
FROM DAIRIES IN MICHIGAN

ABSTRACT

Objectives—To determine if certification in a Milk and Dairy Beef Quality Assurance
Program (QAP) was associated with a reduced risk of having antibiotic residues in milk
and to define specific management factors that may have predisposed dairy farms to
having violative antibiotic residues in milk.

Sample Population—124 dairy farms in Michigan that had at least 1 violative residue in
milk during 1993 and 248 randomly selected control farms in Michigan that did not have
violative residues in milk during 1993.

Procedure—A pretested structured questionnaire was mailed to case and control farms.
A conditional multivariable logistic regression model was developed to determine risk
factors associated with having a violative antibiotic residue in milk.
Results—Certification in the QAP tended to reduce the risk of having a violative
antibiotic residue. Annual treatment of > 10% of a herd for metritis was associated with a
reduced risk of having a violative residue. Evidence suggested that a routine request for a
milk processor to perform residue testing was associated with a decreased risk of having
had a violative antibiotic residue, but routine on-farm residue testing was associated with
an increased risk of having had a residue.

Conclusion and Clinical Relevance— QAP certification was associated with a tendency

21

toward reduced risk of having had a violative antibiotic residue in milk. Other risk factors
associated with violative antibiotic residues are addressed by various critical control

points in the QAP and may be indicators for strengths and weaknesses of a QAP.

22

INTRODUCTION

On Dec 29, 1989 and Feb 8, 1990, The Wall Street Journal (Ingersoll, 1989 &
1990) publicized alleged antimicrobial residues in retail dairy products. The negative
impact that these reports had on consumer perceptions about food safety were
substantiated by subsequent national surveys (Tillison, 1991). In an effort to reaffirm the
commitment of the US dairy industry to maintaining the quality and safety of the nation's
milk supply, the National Milk Producers Federation and American Veterinary Medical
Association completed the development of the Milk and Dairy Beef Quality Assurance

Program (QAP) in 1992.

The QAP is based on Hazard Analysis Critical-Control Point (HACCP) principles
and is probably one of the most ambitious programs that the industry has undertaken. The
QAP has 10 critical control points that producers, in conjunction with their veterinarian,
should regularly monitor and evaluate in an effort to formulate their own unique plan of
action to minimize the risk of violative drug residues. The 10 critical control points center
around a drug-use protocol, managerial practices, personnel policies, and management
strategies for maintaining cattle health (AVMA and NMPF, 1991). Many of these critical
control points are associated with violative antibiotic residues in milk (Kaneene et al,
1986; Kaneene and Ahl, 1987; Kaneene and Willeberg, 1989; McEwen et al, 1991 a & b).

The National Milk Producers Federation and AVMA have conducted an extensive
national campaign to encourage voluntary adoption of the QAP by dairy operations. As an
additional incentive, several states have enacted legislation that reduces the penalty for

having a violative residue in milk if the operation has been certified in the QAP prior to

23

the residue violation (Fluid Milk Act, 1996; Adulterated dairy products, 1998). The
tremendous amount of financial and human resources invested in the QAP warrants
scientific evaluation of the efficacy of the program in attaining its goals. Therefore, the
objectives for the study reported here were to determine if QAP certification was
associated with a reduced risk of having antibiotic residues in milk and to define specific
management factors that may have predisposed dairy farms to having violative residues in
milk. Specifically, the following hypothesis was tested: dairy operations that have had a

violative drug residue in milk were less likely to have participated in the QAP.

MATERIALS AND METHODS

Design—A case-control study of Grade-A dairy herds in Michigan was conducted
in 1994 to evaluate risk factors hypothesized to be associated with antibiotic residues in
milk. A case was defined as a Grade-A dairy farm that had at least 1 violative antibiotic
residue in milk shipped to market during 1993. A control farm was defined as a Grade-A
dairy farm that did not have a violative antibiotic residue in milk shipped to market
during 1993.

Sample selection—A list of 124 farms with violative antibiotic residues in milk
during 1993 was obtained from the Michigan Department of Agriculture, Dairy Division,
and those farms were included as case farms in the study. To control for varying degrees
of general herd management and labor requirements as well as geographic differences in
available veterinary services and milk marketing, case farms were stratified on the basis
of geographic region (agricultural statistics district; Fig 2-1) and herd size. Two control

farms were randomly selected from the records of the Michigan Department of

24

Agriculture, Dairy Division for each case farm located in their respective strata.
Therefore, proportionately equivalent sample distributions of case and control farms
across strata were generated.

Data collection—A pretested self-administered questionnaire8 was developed to
confidentially obtain data from the 372 farms in the study (124 case and 248 control
farms). The questionnaire consisted of questions focusing on herd health management,
drug use, record keeping, personnel management, and descriptive characteristics of each
farm during 1993. Six dairy farms not included in the study were used to test the
questionnaire. On the basis of that preliminary test, substantial changes were not required
in the questionnaire. The questionnaire was sent to the farms in an initial mailing, and, if
necessary, 2 reminder mailings were sent approximately 1 month apart. Data from
completed questionnaires were recorded in a relational database program.b

Statistical analysis — A x2 goodness-of-fit test was performed to determine
whether the geographic distributions of farms was comparable to those of respondents, on
the basis of agricultural district as well as the overall population of Grade-A dairy farms
in Michigan. Distributions of questionnaire respondents and farms included in the study
were similarly evaluated on the basis of herd size. Those variables deemed to be
biologically plausible risk factors for having antibiotic residues in milk were identified.
Descriptive statistics were determined, including frequencies for categoric risk factors
and mean i SD for continuous risk factors. Potential correlations among risk factors were
evaluated; using Pearson's and Spearman rank correlation coefficients.

The outcome of interest was the binary variable of whether a farm had a violative

residue in milk during 1993. Univariate analyses between each risk factor and the

25

outcome variable were performed to aid the development of a conditional multivariate
logistic regression model. A conditional multivariate logistic regression model,
controlling for agricultural district and herd size, was developed by using a backward
stepwise technique (Hosmer and Lemeshow, 1989). The initial model included those
variables that had a univariate parameter estimate with a significance of P 5 0.30 and
those variables that were forced into the model. Reasons for forcing a variable into the
model included the ability to evaluate the risk factor of greatest biological interest, the
variable was a member of a categorical variable with multiple levels, or the variable was
significantly correlated with a variable otherwise included in the model. Potential
confounding by specific risk factors was evaluated, using methods described by
Kleinbaum (Kleinbaum et a1, 1988). Deviance and degrees of freedom for the initial and
final regression models were compared, using the likelihood ratio statistic, to ensure that

the 2 models did not differ significantly (P > 0.05).

RESULTS

After 3 separate mailings of the questionnaire, 45 (36%) case farms and 121
(49%) control farms responded, resulting in an overall response rate of 45%. Of the 166
returned questionnaires, 158 (95%) had relatively complete, useable responses. Analysis
of x2 goodness-of-fit tests indicated that the 372 farms included in the study did not
significantly deviate in geographic distribution from all Grade-A dairy farms in Michigan
(12 = 6.24; P = 0.62) or from those who responded to the questionnaire ()8 = 5.96; P =
0.65). Distribution of herd size (Table 2-1) did not differ between all 372 farms and the

case-farm respondents (x2 = 6.38; P = 0.27) and control-farm respondents (x2 = 5.93; P =

26

0.31). The distribution of Grade-A dairy farms in Michigan on the basis of herd size was
not available; hence, comparison to farms used in the study was not performed. The
proportion of responding case farms with QAP certification was the same as that of the
nonresponding case farms (12 = 0.009; P = 0.92). The same comparison was not
performed for control farms, because certification status of nonresponding control farms
could not be ascertained.

Frequency distributions of categoric risk factors among responding case and
control farms were tabulated (Table 2-2). For example, 40 case and 114 control farms
provided an answer to the question regarding whether they routinely milked treated cows
last, and 22 (55%) case and 54 (47%) control farms responded that they did milk treated
cows last. Distributions of continuous risk factors for case and control farms was
determined (Table 2-3). Significant correlations were between routine use of on-farm test
kits and a routine request that a milk processor perform antibiotic residue testing (R =
-0.49; P < 0.001), use of another bucket when milking treated cows and use of another
milking claw when milking treated cows (R = 0.51; P < 0.001), and number of full-time
farm workers and mean herd size (R = 0.70; P < 0.001).

Using results of univariate analyses, 11 risk factors were identified for inclusion in
the initial logistic regression model. Those risk factors included use of another milking
claw when milking treated cows, use of another bucket when milking treated cows,
written identification records of cows treated, routine performance of on-farm residue
testing, routine request that milk processor perform antibiotic residue testing, purchase of
over-the-counter drugs from nonveterinarian sources, purchase of prescription drugs from

a veterinarian, annual treatment of 5 10% of herd for mastitis, annual treatment of > 40%

27

of herd for mastitis, annual treatment of >10% of herd for metritis, and mean size of
milking herd in 1993 (Tables 2-2 and 2-3). In addition, certification in the QAP prior to
an antibiotic residue in milk was forced into the model because it was the risk factor of
primary importance (main effect) in this study. Because of the strong correlation with
mean herd size, number of full-time workers was also forced into the model. After the
stepwise procedures (Hosmer and Lemeshow, 1989), the model was reduced to 8 risk
factors (certification in the QAP, annual treatment of >10% of herd for metritis, annual
treatment of 5 10% of herd for mastitis, annual treatment of >40% of herd for mastitis,
routine request for milk processor to perform residue testing, routine performance of
on-farm residue testing, written identification records of treated cows, and purchase of
prescription drugs from a veterinarian; Table 2-4). None of the modifiers investigated
made a significant contribution to the model. The final model was significantly (x2 =
21.54; P = 0.006) correlated with a herd having a violative antibiotic residue in milk,
suggesting that the model would be sufficient to explain the odds of having an antibiotic
residue in milk. The x2 of the log-likelihood statistic (x2 = 6.06; P = 0.30) between initial
and final models was not significant, suggesting that the predictive ability of the model
was not significantly diminished by its reduction.

Written identification records of treated cows was the only risk factor significantly
associated (odds ratio [OR] = 4.78; P = 0.03) with an increased risk of having a violative
antibiotic residue. Annual treatment of >10% of herd for metritis (OR = 0.20; P = 0.02)
was significantly associated with a decreased risk of having had a violative antibiotic
residue was. Certification in the QAP prior to having had a violative antibiotic residue

tended to be associated (OR = 0.28; P = 0.11) with a reduced risk of having had a

28

violative antibiotic residue in milk.

During the model building process, models including on-farm residue testing or
residue testing performed by milk processors indicated that these 2 variables were each
significantly associated with explaining antibiotic residues. However, these variables
were negatively correlated and, therefore, must be retained together in the model to avoid
biased estimates. Their independent contribution to the model appears not significant, but
their elimination resulted in a likelihood ratio statistic of 5.12 (P = 0.08) and evidence
that they were significant confounders. Although separately not significant risk factors,
variables for the 2 categories of mastitis treatment and purchase of prescription drugs
from a veterinarian were significant confounders in the model and, consequently, were

retained.

DISCUSSION

The main objective of the QAP is to reduce the incidence and risk of drug
residues in beef and milk (Adams, 1993; AVMA and NMPF, 1991). To test our
hypothesis that certification in the QAP reduced the risk of having had antibiotic residues
in milk we forced the risk factor for prior certification into the model. Analysis of this
factor indicated a tendency (OR = 0.28; P = 0.11) for a QAP-certified farm to have a 70%
reduction in risk of having a violative antibiotic residue in milk. This association may
have been weakened by measurement error regarding the degree to which a producer and
their veterinarian actually participated in the QAP. It cannot be assumed that all the
certified farms fully embraced the program and made major management changes.

Although the analysis was restricted to farms with voluntary certification, the binary

29

variable for QAP certification did not discern conscientious from less-conscientious
producers.

Control farms (35/109, 32.1%) were more likely to have participated in the QAP
than case farms (10/41, 24.4%; Table 2-2). The study attempted to determine the duration
of participation in the QAP by inquiring about the dates of initial certification and
recertification. Date of certification for case farms were confirmed through analysis of
records of the Michigan Department of Agriculture, Dairy Division, but information
pertaining to some of the control farms participating in the program was deficient. This
deficiency of information was probably the result of a lack of compliance with the request
that a copy of the signed certificate be sent to the office of the Dairy Division. Five of 35
( 14.3%) of certified control farms did not report the date of certification. An additional 18
of 35 (51.4%) of certified control farms reported only the year but not the day or month of
initial certification. A simple comparison of operations that reported at least the year of
certification did not indicate a difference in the percentage of participants in the QAP
from the case or control groups (5/10, [50%] of participating case farms and 16/30
[53.3%] of participating control farms) that were certified before Jan 1, 1993 (i.e., prior to
the onset of the study). Because the campaign for statewide adaptation of the QAP was
initiated in 1992, it would have been critical to know the month, in addition to the year, of
certification to enable us to include duration of participation in the analyses.
Consequently, the effect of duration was not evaluated.

The rate for overall voluntary participation of respondents in this study in the
QAP (30%) far exceeded the 1993 national participation rate and the estimated

participation rate for the state of Michigan (4 and 10%, respectivelyc). On the basis of the

30

x2 analysis for comparison of certification for responding and nonresponding case farms,
there was minimal chance that nonresponders biased our estimation of the rate of
certification. This discrepancy might have been a result of national underreporting of
certification and may indicate the need for a more reliable census of farms involved in
QAP certification and participation. The issues of underreporting participation and
duration need to be addressed by studies evaluating the efficacy of the QAP.

Increased drug use is believed to increase the risk of having violative residues in
milk and meat. Mastitis is the most common disease of lactating dairy cows and is
frequently treated with antimicrobial agents (Gardner et al, 1990; Cullor, 1993; Hady et
al, 1993). Using the second mastitis category as a reference, which included the mean
annual incidence of mastitis in Michigan, a tendency toward a reduced risk of residues
when treating less than the mean number of cows in a herd and a tendency toward an
increased risk of residues when treating more than the mean number of cows in herd were
consistent with other studies (Kaneene and Ahl, 1987; McEwen et al, 1991 b).
Unfortunately, results of the study reported here did not indicate a significant association
among various categories of mastitis treatment and having violative residues.

In Michigan, metritis was the second most commonly reported disease of dairy
cows between 1986 and 1989 (Kaneene and Hurd, 1988; Kaneene et al, 1990). Farms
having treated >10% of the herd annually for metritis were associated with 80% less risk
of having had a residue in milk, compared with farms having treated 310% of the herd.
In Michigan, veterinarians (Kaneene and Miller, 1994) most often treat metritis. We
interpreted this finding to indicate that increased presence of a veterinarian for metritis

treatment possibly provides increased guidance to producers for drug use and withdrawal

31

periods (Kaneene and Miller, 1992). In addition, visible identification or segregation of
sick and treated cows might be enhanced when treatments extend beyond otherwise-
routine mastitis treatments.

Critical control point No. 8 of the QAP emphasizes the availability and use of
drug residue tests. The majority of case and control respondents in this study indicated
that they used residue testing during 1993. When on-farm and milk processor testing were
evaluated separately, farms that routinely requested that the milk processor perform
residue testing had a reduced risk of having violative residues, while on-farm residue
testing was associated with increased risk of having had a violative residue. The
difference between the risk associated with on-farm residue testing and that conducted by
milk processors has 2 possible explanations. First, on-farm testing might have been
initiated after the farm had a violative residue, as was suggested by other studies
(Kaneene and Ahl, 1987; McEwen et al, 1991 b). Secondly, milk processors were more
likely to have had extensive experience in performing the tests.

Milk processors performing the requested tests had more knowledge and
information regarding the use of antibiotic residue test kits. It was important to have used
a test that was designed to detect the compound for which the sample was being tested
and that was able to detect the compound at or below the legal tolerance level.
Furthermore, milk processors should have ensured that any tests were performed correctly
and consistently. These criteria might not have been met when performing tests on-farm.
Realizing that on-farm and milk processor testing were not mutually exclusive (a farm
could perform none, one, or both) and that a small, but significant, negative correlation

was detected between the 2 testing variables, the possibility of erroneous on-farm testing

32

should be evaluated more thoroughly in future studies. Most importantly, it would be
prudent for dairies that want to reduce the risk of a violative residue in milk to entrust
antibiotic testing needs with their milk processor.

Unexpectedly, keeping written identification records of treated cows was
associated with a fivefold increase in the risk of having had a violative antibiotic residue.
Maintenance of complete and accurate records for cattle treated on a farm is the basis of
critical control point No. 7. A similar finding was previously reported (McEwen et al,
1991 b). In that retrospective study, case farms (violative residue detected) were more
likely to maintain records of treatment, and significantly more farmers on case farms held -
the opinion that insufficient record keeping of treated cattle increased the risk for a
violative residue. This paradoxic phenomenon might be the result of case farms having
adopted the good management practice of recording the identification of treated cows
afier notification of a violative residue and mandatory completion of the QAP. As the
responsibility of proving proper residue prevention practices shifts toward dairy farms, it
is important that complete and accurate records be maintained prior to having a violative
antibiotic residue.

Certification in the QAP had a tendency to reduce the risk of having a violative
antibiotic residue in milk. Dairy operations that treated > 10% of the herd for metritis had
a decreased risk of having violative antibiotic residues in milk. A positive association
between maintaining written identification records of treated cows and having a violative
residue was identified, but this finding probably indicates a change in management
implemented after notification of having had a violative residue. Although the routine

request that a milk processor perform residue testing was associated with a decreased risk

33

of having a violative residue, routine on-farm testing was associated with an increased
risk of having had a violative residue in milk. Specific risk factors associated with having
a violative residue are addressed by various critical control points in the QAP and may be

indicators for some of the program's strengths and weaknesses.

FOOTNOTES
aQuestionnaire available on request and as Appendix 1.
bRbase for DOS, Microrim Inc, Redmond, Wash.

°McCarthy W. Michigan Department of Agriculture, Dairy Division, Lansing, MI:
Personal communication, 1994.

34

REFERENCES

Adams JB. Assuring a residue-free food supply: milk. J Am Vet Med Assoc
1993;202:1723-1725.

Adulterated dairy products, Minn. Stat. 32.21, Subd. 1-4. Minnesota statues vol. 1. St
Paul, Minn: Revisor of Statues, State of Minnesota, 1998;920-922.

American Veterinary Medical Association and National Milk Producers Federation. Milk
and Dairy Beef Residue Prevention: a quality assurance program. J Am Vet Med Assoc
1991; 199 (Suppl):1-23.

Cullor J S. The control, treatment, and prevention of the various types of bovine mastitis.
Vet Med 1993;88:571-579.

Fluid Milk Act of 1965, PA. 1965, No. 233, as amended by RA. 1993, No.5. Michigan
Compiled Laws Annotated 1996 Sections 286.1 to 288. St Paul, Minn: West Publishing
Co, 1996; 417-419.

Gardner IA, Hird DW, Guterback WW, et a1. Mortality, morbidity, case-fatality, and
culling rates for California dairy cattle as evaluated by the National Animal Health
Monitoring System, 1986-87. Prev Vet Med 1990;82157-170.

Hady PJ, Lloyd J W, Kaneene J B. Antibacterial use in lactating dairy cattle. J Am Vet Med
Assoc 1993;203:210-220.

Hosmer DW, Lemeshow S. Chapters 4-7. In: Barnett V, Bradley RA, Hunter J S, et al,
eds. Applied logistic regression. New York: John Wiley & Sons, 1989;82-213.

Ingersoll B. Milk is found tainted with a range of drugs farmers give cattle. The Wall
Street Journal, December 29, 1989; Al, A2.

Ingersoll B. New York milk supply highly tainted, TV station says, based on own survey.
The Wall Street Journal, February 8, 1990; A4.

Kaneene JB, Ahl AS. Drug residues in dairy cattle industry: Epidemiological evaluation
of factors influencing their occurrence. J Dairy Sci 1987;70:2176-2180.

Kaneene J B, Coe PH, Smith JH, et a1. Drug residues in milk after intrauterine injection of
oxytetracycline, lincomycin-spectinomycin, and povidone-iodine in cows with metritis.

Am J Vet Res 1986;47:1363-1365.

Kaneene JB, Hurd HS. National Animal Health Monitoring System Round l—Final
Report. East Lansing, Mich: Michigan State University, 1988;63-74.

35

Kaneene JB, Hurd HS, Miller R, et a1. National Animal Health Monitoring System Round
2—Final Report. East Lansing, Mich: Michigan State University, 1990;26-34.

Kaneene JB, Miller R. Epidemiological study of metritis in Michigan dairy cattle. Vet Res
1994;25:253-257.

Kaneene JB, Miller RA. Description and evaluation of the influence of veterinary
presence on the use of antibiotics and sulfonamides in dairy herds. J Am Vet Med Assoc
1992;201:68-76.

Kaneene JB, Willeberg P. Influence of management factors in the occurrence of antibiotic
residues in milk: a case-control study of Michigan dairy herds, with examples of
suspected information bias. Acta Vet Scand 1989;84:473-476.

Kleinbaum DG, Kupper LL, Muller KE. Confounding and interaction in regression. In:
Payne M, ed. Applied regression analysis and other multivariable methods. Belmont,
Calif: Duxbury Press, 1988;163-180.

McEwen SA, Black WD, Meek AH. Antibiotic residue prevention methods, farm
management, and occurrence of antibiotic residues in milk. J Dairy Sci 1991
(a);74:2128-2137.

McEwen SA, Meek AH, Black WD. A dairy farm survey of antibiotic treatment practices,
residue control methods and associations with inhibitors in milk. J. Food Prot 1991
(b);54:454-459.

Tillison J. Consumer research review, edited by the National Dairy Promotion and

Research Board. Animal Health & Milk Quality - A Situation Analysis. Jefferson, Wis:
Morgan & Myers, 1991;10-11.

36

Figure 2-1 - Geographic distribution of Michigan agricultural statistics districts defined
by the National and Michigan Agricultural Statistics Services. The numbers of case farms
in the sample (denominator) and number that responded (numerator) in each district are
indicated.

 

 

 

 

 

 

 

37

Table 2-1- Herd size categories and the number of case operations from each category.

 

Case farms

 

 

 

Herd size category No. of lactatingows No. in sample No. respondinL

A 10-39 24 11
B 40-79 45 13
C 80-119 22 5
D 120-159 17 1 1
E 160-249 8 2
F >250 8 3

 

 

38

Table 2-2. Distribution of categorical risk factors among case and control operations.

 

Responding operations

 

 

 

Case Control
Risk factor No.l %2 No.l %2
Certification in QAP prior to residue3 41 24.4 109 32.1
Use of a different milking claw for treated cows3 39 35.9 112 48.2
Treated cows milked last 40 55.0 1 14 47.4
Diverted pipeline when milking treated cows 35 34.3 105 35.2
Use qf a different bucket when milking treated 39 64.1 115 74.8
cows
Every cow received nonlactating treatment for 41 82.9 117 84.6
mastitis
All cows were routinely deworrned 41 31.7 117 35.0
Visible identification placed on treated cows 41 87.8 1 17 89.7
No treatment records kept 41 17.1 117 20.5
Record of treated cow identification" 41 73.2 116 62.9
Milkers had access to treatment records 35 94.3 93 94.6
Use of a drug test to determine milk withholding 41 70.7 117 72.6
period
Use of the drug label to determine milk withholding 41 78.0 117 71.8
period
Routinely perform on-farm chemical residue testing3 41 80.5 1 17 50.4
Routinely request milk processor perform residue 41 29.3 117 53.0
testing3
Purchase over-the-counter drugs from a veterinarian 41 39.0 117 34.2
Purchase over-the-counter drugs from a 41 82.9 117 70.1
nonveterinarian source3
Purchase prescription drugs from a veterinarian3 41 92.7 117 99.1
Purchase prescription drugs from a nonveterinarian 41 9.8 117 6.8
source
Annual treatment of > 10% of herd for metritis3 41 24.4 117 43.6

 

39

 

Table 2-2 continued

 

Annual treatment 5 10% of herd for mastitis3 41

Annual treatment of 10.1-39.9% of herd for mastitis 41
(reference category)

Annual treatment of > 40% of herd for mastitis3 41
Treated at least one cow in an extra-label manner 41
during 1993

22.0
53.6

24.4
36.6

117
117

117
117

38.5
48.7

12.8
35.9

 

 

QAP = Milk and Dairy Beef Quality Assurance Program.

1Number of operations providing a response. 2Percentage of responding operations
indicating "yes." 3Risk factors included in the initial logistic regression model.

 

40

 

Table 2-3. Distribution of continuous risk factors among cases and controls.

 

Responding operations

 

 

Case Control

Risk factor No.l Mean SD No.l Mean SD
No. of full-time workers 40 1.425 2.04 115 1.609 3.21
No. of part-time workers 39 1.333 1.53 115 1.400 1.73
Milking herd (No. of cows.)2 41 124.95 159.84 117 108.61 90.98
Rolling herd average (lbs of 36 18,836 3,350 110 19,249 3,328
milk)

Annual percentage of herd 38 41.55 193.92 109 9.54 34.28

treated in an extra-label manner

 

1No. of operations providing a response. 2Risk factors included in the initial logistic
regression model. To conve1t lb of milk to kg, divide value by 2.2.

 

 

 

41

Table 2-4 - Results of the conditional multivariable logistic regression analysis of
associations among chemical residue occurrence in milk and herd-level risk factors.

 

 

Risk Factor b' SEgb) P’ Odds 95% (:1
Ratio (odds ratio)

Certification in QAP prior to -1.19 0.75 0.1 l 0.30 0.07-1.32

residue

Annual treatment of 110% of -l .61 0.71 0.02 0.20 0.05-0.80

herd for metritis

Record of identification of treated 1.56 0.70 0.03 4.78 1.21-18.77
cows

Routinely request that milk -0.88 0.67 0.18 0.41 0.11-1.54
processor perform residue testing

Routinely perform on-farm residue 0.79 0.72 0.28 2.20 0.54-9.04
testing

Annual treatment of _<_10% of -1.08 0.72 0.13 0.34 0.08-1.39
herd for mastitis
Annual treatment of 3 40% of 0.13 0.70 0.85 1.14 0.29-4.49

herd for mastitis

Purchase prescription drugs from a -2.01 1.68 0.23 0.13 0.01-3.61
veterinarian

 

lMultivariate logistic regression coefficient. 2Standard error of b. 3P-value for Wald test
statistic. QAP = Milk and Dairy Beef Quality Assurance Program. 95% CI = 95%
confidence interval.

Likelihood Ratio Statistic = 6.06; 5 df; P > 0.05.

 

 

 

42

CHAPTER 3

INFLUENCE OF THE MILK AND DAIRY BEEF QUALITY
ASSURANCE PROGRAM ON MICHIGAN DAIRY FARM DRUG
MANAGEMENT PRACTICES

ABSTRACT

Objective - To test the hypothesis that dairy farms certified in the Milk and Dairy Beef
Quality Assurance Program (QAP) were more likely to use prudent drug management
practices than farms that were not certified.

Design — Cross-sectional study.

Sample Population — 141 Michigan dairy farms, of which 74 were not certified in the
QAP, 30 were involuntarily certified, and 37 were voluntarily certified.

Procedure — Dairy producers completed a self-administered questionnaire that focused
on herd health management, drug use, record-keeping, personnel management, and
descriptive characteristics of their farm during 1993. Separate multivariable logistic
regression models were developed to determine the association of QAP certification with
each of the management practices.

Results — Results suggested that farms adopted specific management practices,
irrespective of certification. Large percentages of farms used visible identification and
non-emergency veterinary services and discussed residue prevention with employees.

Involuntary certification was associated with maintenance of good written treatment

43

records and performance of on-farm drug residue testing. Voluntary certification was
weakly associated with use of refrigerated drug storage.

Conclusions and Clinical Relevance — QAP certification appeared to have been
associated with the adoption of only a few prudent drug use practices, although QAP
materials and framework were developed to assist veterinarians in the promotion of
disease prevention, client communication, and residue prevention practices on farms.
Veterinary care would benefit from the development and encouragement of better record

keeping on farms.

44

INTRODUCTION

The Milk and Dairy Beef Quality Assurance Program (QAP) was jointly
developed by the National Milk Producers Federation and the AVMA in 1991. The QAP
is based on the hazard analysis critical control point (HACCP) principles. The residue
prevention protocol comprises 10 critical control points focusing on drug use protocol,
herd health management practices, record-keeping and employee education (Boeckman
and Carlson, 1997). Many of the critical control points address risk factors that were
identified in studies by Kaneene et a1 and McEwen et al of the increased risk of drug
residue occurrence (Kaneene and Ahl, 1987; McEwen et a1, 1991 a). Dairy farms become
certified in the QAP by using the 10 critical control points to review, with their
veterinarian, their farms’ residue prevention practices, and define areas for improvement.
Official certification is given when the producer espouses the principles of providing a
high quality product by preventing residues in milk and dairy beef. Certification can be
voluntarily pursued by producers, or it can be involuntarily implemented (i.e., required) in
instances of a residue violation in milk.

In the report of a study that evaluated the use of an on-farm risk assessment tool
(Sischo et al, 1997), the authors expressed concern that although the QAP does a good
job of articulating the hazards of residues, the program is deficient in 3 necessary
components of any HACCP program. Specifically, these authors pointed out that the
program doesn’t provide adequate motivation and tools to allow farm owners to assess
their own risk of illegal residues, develop a plan to reduce their risk, or monitor their
progress toward residue prevention. In their study, the treatment and control groups

received a copy of the QAP booklet and were evaluated by use of the risk assessment

45

tool. The treatment group received additional information and guidance that led to a farm
plan to reduce their risk of residues. Although the overall risk of antibiotic residues was
reduced by approximately 19%, there was no significant difference between the groups.
It is difficult to ascertain whether the risk assessment tool, the QAP booklet, or the
combination of these 2 factors had the greatest impact on risk reduction. Nevertheless,
their study is one of the few that have addressed the challenge of evaluating the QAP.
Intuitively, the QAP seems to focus on important drug residue hazards. Its impact
and success in changing dairy management practices to those considered prudent in the
prevention of drug residues has not been reported. The purpose of the study reported here
was to test the hypothesis that dairy farms certified in the QAP were more likely to use

prudent drug management practices than those farms that were not certified.

MATERIALS AND METHODS

Study design - As part of a larger study (Gibbons-Burgener et al, 1999) of
Michigan dairy farms that had a violative drug residue in milk, a cross-sectional study
was undertaken to test the hypothesis that QAP certification was associated with

implementation of certain drug management practices.

Study population - Sample selection for the original study has been described
(Gibbons-Burgener et al, 1999). Briefly, 166 of 372 (45%) farms that received the
questionnaire, returned it. Results of goodness-of-fit analyses indicated that the
respondents were geographically representative of dairy farms in Michigan. One hundred

forty-one responding farms provided complete information regarding various drug

46

management practices and were included in the cross-sectional study sample. Of the 141
farms, 74 were not certified in the QAP, 30 were involuntarily certified, and 37 were
voluntarily certified. Approximately a third of the study population had a residue

violation in 1993. Of these farms, 73% were in involuntarily certified.

Data collection - The pretested, self-administered questionnairea focused on herd
health management, drug use, record keeping, personnel management, and descriptive

characteristics of the farm during 1993.

Statistical analyses - Outcome variables that represented management practices
for 7 of the 10 critical control points of the QAP were identified in the questionnaire.
Only those management practices believed to have biologically plausible associations
with QAP participation were considered. Because we suspected that the various
outcomes were related, Spearman rank correlation analysis was performed with all
outcome variables to test the degrees of correlation. In addition to univariate analyses,
separate multivariable logistic regression models were developed to determine the
association of QAP certification with each of the management practices. Because few
independent variables were considered for inclusion in each model, reduction techniques
were not used. Three categories were used to designate QAP certification: non-certified,
involuntarily certified as a result of a drug residue violation, or voluntarily certified.
Farms with a violative drug residue after becoming QAP certified were included in the
voluntarily certified group. Additional variables were believed to potentially have

biological relationships (primary or confounding) with each management practice. For

47

example, mean milking herd size and whether the producer was a college or technical
school graduate were considered potential variables associated with all practices.
Independent variables considered for inclusion in specific models included whether the
farm utilized a milking parlor, whether the farm utilized routine, nonemergency
veterinary services, and whether cows were treated with drugs in an off-label manner.
Each model was estimated twice to facilitate the 3-way comparison between the
categories of non-certified, involuntarily certified, and voluntarily certified. The
comparison group was switched from non-certified to involuntarily certified to evaluate a
difference between voluntary and involuntary certification. Seemingly unrelated 2-
equation probit regression modeling was used to determine whether significant
correlations between outcome variables had a substantial influence on results of logistic

regression modeling (Hardin, 1996). Significance was set at P 5 0.05.

RESULTS

Percentages of responding farm owners that indicated that they performed specific
drug-use managerial practices were tabulated (Table 3-1). Spearman rank correlation
analysis revealed significant correlations between numerous outcome variables. The
correlations of most practical importance included: recorded reason for treatment and
recorded type of drug used (r = 0.51), recorded type of drug used and recorded dose of
drug used (r = 0.63), recorded type of drug used and recorded date(s) of treatment ( r =
0.50), recorded identification of treated cow and recorded date(s) of treatment (r = 0.66),

withdrawal period determined by label and withdrawal period determined by a asking a

48

veterinarian (r = 0.50), and routinely requested off-farm testing for residues and on-farm
testing used routinely (r = -0.47).

Most of the models revealed minimal association between QAP certification and
the various drug management practices; results of 3 multivariable models provided the
most significant results. The first model considered the use of refrigerated drug storage
as its outcome variable (Table 3-2); voluntarily certified farms were almost 3 times more
likely than noncertified farms to use refiigerated drug storage, and herd size was
significantly associated with refrigeration.

The second model used on-farm drug testing as the outcome variable (Table 3-3).
To emphasize differences that seemed inherent in instances of involuntary certification,
analysis was performed with involuntarily certified farms as the base comparison group.
Compared with involuntarily certified farms, noncertified and voluntarily certified farms
were less likely to have used on-farm residue testing.

The third model used good treatment records as the outcome variable (Table 3-4).
The term “good” was defined as records that included the treated cow’s identification,
date of treatment, and drug or dose used. In this model, involuntarily certified farms were
2.5 times more likely than noncertified farms to maintain good treatment records.

Results of the seemingly unrelated 2-equation probit regression modeling
indicated that correlations detected among some outcome variables did not significantly

bias results of multivariable logistic regression models.

49

DISCUSSION

Recommendations for proper drug storage often include maintaining the drug
within a certain temperature range, as well as avoiding exposure to environmental factors
such as sunlight and excessive humidity. Three of the most commonly used drugs
approved for use in lactating cows (procaine penicillin G, ceftiofirr sodiumb and oxytocin)
are supplied with labels that recommend refrigeration (Arrioja-Dechert, 1997). Herd size
may influence the type and quantity of these drugs on a given farm, which could explain
its strong positive association with refrigerated storage. Independent of herd size, there
was weak evidence that voluntarily certified farms were more likely (by an approximate
factor of 3) than noncertified farms to use refrigerated storage. This association was not
significant (P = 0.086) but is worthy of discussion. Of particular interest is the reduced
overall effect of QAP on storage method when involuntarily and voluntarily certified
farms were used together as an index for QAP certification. A true difference between
involuntarily certified farms and voluntarily certified farms in the use of refrigerated
storage may be related to conscientiousness or other unmeasured factors of producers
voluntarily seeking certification.

Forty-five percent of the study farms requested off-farm residue testing, whereas
60% used on-farm drug tests. Although not mutually exclusive, substantial contrast
among types of QAP certification was detected only for on-farm testing. Involuntarily
certified farms were 5 times more likely than noncertified farms and 3.5 times more likely
than voluntarily certified farms to perform on-farm residue testing. Of particular interest
is the difference between involuntarily and voluntarily certified farms. This may be an

indication that having a violative residue compels a farm to adopt on-farm residue testing

50

in an attempt to avoid additional violations. This finding is consistent with results of
previous studies (Kaneene and Ahl, 1987; McEwen et al, 1991 b).

The eighth critical control point in the QAP encourages the use of antibiotic
screening tests when drugs are used in an off-label manner. The use of drug screening
assays for milk of individual cow’s has yet to be approved; nevertheless, it is a commonly
accepted practice (Sischo, 1996). There is evidence that assays designed to test
commingled milk may produce false-positive or false-violative results when used on milk
from an individual cow (Sischo and Burns, 1993; Tyler et al, 1992; Van Eenannaam et al,
1993), as well as pose variable economic risk (Slenning and Gardner, 1997). The QAP
directs the producer to the drug’s label to determine the correct withdrawal period and
states that drug testing is unnecessary when using drugs according to the label (Boeckman
and Carlson, 1997). Voluntarily certified farms were more likely to indicate that some of
their cows were treated in an off-label manner (Table 3-1). If the QAP was indeed
causing the producers to adopt drug-testing technology, we would have expected the
voluntarily certified farms to have more need and, consequently, be more inclined to use
individual cow testing. However, farms involuntarily certified in the QAP (all of which
had a violative residue in 1993) were more likely to adopt on-farm drug testing, and
noncertified farms were more likely to request off-farm drug testing. Perhaps the eighth
critical control point is perceived as the most simple and reliable management change a
farm can make. A broader objective of the QAP is the promotion of preventative health
practices that reduce disease and the need for treatment; however, many veterinarians and

producers are unsure how to determine practical and meaningful indices for herd health

51

improvement. As stated, the program is probably lacking the necessary tools for farm
operators to evaluate their progress toward implementing the QAP (Sischo et al, 1997).

To take into account the multiple correlations among the different record types,
we performed additional statistical methods, with mixed success. We decided to define a
level of record keeping that wasn’t necessarily optimal, but that could be considered
“good.” Consistent with 1996 National Animal Health Monitoring System data (USDA-
APHIS-VS, 1996), 21% of the producers reported keeping no written treatment records.
Maintaining complete treatment records is often thought to be one of the most important
residue prevention practices (McEwen et al, 1991 b; Day, 1993). In the study reported
here, farms with involuntary certification were 2.5 times more likely to keep good written
records than farms that had never been certified. Although not statistically different from
either of the other groups, a higher percentage of voluntarily certified farms maintained
good treatment records than did noncertified farms, whereas a lower percentage of
voluntarily certified farms maintained good treatment records than involuntarily certified
farms.

The QAP introduces the concept of complete record keeping and provides a
template for a daily treatment record (Boeckman and Carlson, 1997). Written treatment
records have the potential to be used not only in a residue prevention capacity, but also
epidemiologically. The incidence of treatable diseases and the efficacy of treatment may
be evaluated on a herd basis. As reported (McEwen et al, 1991 b) farms that have had a
violative residue may be more attuned to deficiencies in residue prevention and more
motivated to alter selected practices. In the study reported here, producers who were

forced to review the protocol after a violative residue was detected may have discovered

52

that their records were lacking and that this hindered an explanation or defense of the
violation. Perhaps, it was not until the need arose that the necessity for good records
became apparent. Our results do not clearly indicate that QAP alone had an impact on
record-keeping practices.

We expected QAP certification to be associated with more prudent drug
management practices. The use of a cross-sectional study design hindered the assessment
of temporality in the adoption of management practices. Consistent with results reported
by another study (Sischo et a1, 1997), we found that farms may adopt management
practices irrespective of certification. Large percentages of farms in each of the 3 QAP
groups used visible identification and nonemergency veterinary services and chose to
discuss residue prevention with their employees. Three possible explanations for these
findings are that herd size may have been a better indicator for levels of herd
management, farms with prior experience with residues may have already altered their
drug use practices, and education level may play an important role in adopting certain
management practices. With a low rate of voluntary participation in the program, the
influence of involuntary certification following a residue violation needs to be addressed.
The administration of the QAP alone may be insufficient to prompt producers to adopt
specific drug management practices.

Involuntary certification was associated with the maintenance of good written
treatment records and the performance of on-farm drug residue testing. Voluntary
certification was weakly associated with the use of refiigerated drug storage. Although
QAP certification appeared to have been influential in the adoption of only a few prudent

drug use practices, veterinarians and producers may be using the general concepts of the

53

QAP to improve many of their drug residue prevention practices without formalizing
QAP certification. The QAP materials and framework were developed to assist
veterinarians in the promotion of disease prevention, client communication, and residue
prevention practices on client farms. However, results of the study reported here could
not clearly indicate the efficacy of the QAP with regard to residue prevention practices,
possibly as a result of the small number of farms in the study and the cross-sectional

design. A prospective study with a large sample size may overcome these shortcomings.

FOOTNOTES

“Questionnaire available from authors by request and in Appendix 1.

bNaxcel®, Pharmacia & Upjohn, Kalamazoo, MI.

54

REFERENCES

Arrioja-Dechert A. Compendium of veterinary products. 4th ed. Port Huron, MI: North
American Compendiums Inc, 1997;109-13 10.

Boeckman S, Carlson KR Milk and dairy beef quality assurance program: milk and
dairy beef residue prevention protocol. Stratford, IA: Agri-Education, Inc, 1997;1-70.

Day J. A subcommittee review of the quality assurance initiative: implementation issues
from the Implementation and Communication Subcommittee of the Drug Residue
Committee, in Proceedings. 32nd Annu Meet National Mastitis Council 1993;144-146.

Gibbons-Burgener SN, Kaneene J B, Lloyd J W, et al. Evaluation of certification in the
Milk and Dairy Beef Quality Assurance Program and associated factors on the risk of
having violative antibiotic residues in milk from dairy farms in Michigan. Am J Vet Res

1999:60;1312-1316.
Hardin JW. Bivariate probit models. Stata Technical Bulletin 1996;33:15-20.

Kaneene J B, Ahl AS. Drug residues in dairy cattle industry: Epidemiological evaluation
of factors influencing their occurrence. J Dairy Sci 1987;70:2176-2180.

McEwen SA, Black WD, Meek AH. Antibiotic residue prevention methods, farm
management, and occurrence of antibiotic residues in milk. J Dairy Sci 1991(a);74:2128-
2137.

McEwen SA, Meek AH, Black WD. A dairy farm survey of antibiotic treatment
practices, residue control methods and associations with inhibitors in milk. J Food
Protect 1991(b);54:454-459.

Sischo WM. Quality milk and tests for antibiotic residues. J Dairy Sci 1996;79:1065-
1073.

Sischo WM, Burns CM. Field trial of four cowside antibiotic-residue screening tests. J
Am Vet Med Assoc 1993;202:1249-1254.

Sischo WM, Kieman NE, Burns CM, et al. Implementing a quality assurance program
using a risk assessment tool on dairy operations. J Dairy Sci 1997;80:777-787.

Slenning BD, Gardner IA. Economic evaluation of risks to producers who use milk
residue testing programs. J Am Vet Med Assoc 1997;211:419-427.

Tyler J W, Cullor J S, Erskine RJ, et a1. Milk antimicrobial residue assay results in cattle

with experimental, endotoxin-induced mastitis. J Am Vet Med Assoc 1992;201:1378-
1384.

55

USDA-APHIS-VS. Part 111: Reference of 1996 diary health and health management
National Animal Health Monitoring System. In: NAHMS Dairy '96. Washington, DC:
USDA, 1996;17.

Van Eenannaam AL, Cullor J S, Perani L, et al. Evaluation of milk antibiotic residue
screening tests in cattle with naturally occurring mastitis. J Dairy Sci 1993;76:3041-
3053.

56

Table 3-1. Percentages of Michigan dairy farms that used various drug-use management
practices.

 

No Involuntary Voluntary
Management practice QAP ' QAP b QAP °

 

 

Valid veterinarian-client-patient relationshipd

Nonemergency veterinary care used 74 70 81

Veterinarian provided Rx drugs 100 97 92

Veterinarian provided OTC drugs 35 43 27
Use of OTC and Rx drugsd

Nonveterinarian source for Rx drugs 3 7 14

Nonveterinarian source for OTC drugs 70 87 73

Cows treated in off-label manner 34 33 49

Drug storaged

Drugs stored in cabinet 61 70 62
Drugs stored on an open shelf or table 36 33 32
Drugs stored in refrigerator 69 70 89
Drugs locked in storage 5 3 3

Proper drug administration and identification of
treated cowsd

Used visible identification on treated 88 87 89
cows

Milked treated cows last 50 52 47
Used different milking claw on treated 44 39 43
cows

Diverted milk pipeline when milking 37 32 29
treated cows

Used a special bucket when milking 71 71 69

treated cows

 

57

 

Table 3-1 — continued

 

 

Maintain treatment recordsd

Recorded reason for treatment
Recorded type of drug used

Recorded dose of drug used

Recorded identification of treated cow
Recorded date(s) of treatment
Recorded udder quarter treated

No written records

Kept records at milking location

Kept records where drugs are stored
Keeps records in cattle housing area

Use of drug residue screening testsd
On-farm testing used routinely

Routinely requested off-farm (milk handler)
to test for residues
No optional testing performed on milk

Withdrawal period determined by label
Withdrawal period determined by milk test

Withdrawal period determined by asking
veterinarian

Withdrawal period determined by past
experience

Withdrawal period determined by
information from other producers

34
42
28
59
66
34
26
38
15

49

50

12

74

65

57

18

47
6O
33
70
77
40
20
53
17
10

83

37

80

73

57

33

53
56

78
78
42
1 1
49
22

62

43

70

81

54

27

 

58

 

Table 3-1 continued

 

Employee educationd

Discussed residue avoidance with employees 68
Discussed avoidance with new employees 24
Routinely discussed avoidance throughout year 32
Discussed avoidance when problems occurred 26

76
43
37
27

76
32
35
30

 

 

aF arms (11 = 74) were not enrolled in the Milk & Dairy Beef Quality Assurance Program
(QAP). bFarms (n = 30) were involuntarily enrolled in the QAP. °Farms (n = 37) were

voluntarily enrolled in the QAP.

dCritical control point as defined in a drug residue prevention protocol.

Rx = Prescription. OTC = Over-the-counter.

 

59

 

Table 3-2. Results of multivariable logistic regression analysis of associations among
use of refiigerated drug storage and farm management factors.

 

 

95% Confidence

Variable p P Odds ratio interval
QAP certification

None NA NA 1.0 NA

Involuntary 0.089 0.860 1.09 0.41 — 2.93

Voluntary 1.057 0.086 2.88 0.86 — 9.63
Herd size 0.016 0.005 1.016 1.005 — 1.027
College graduate" 0.513 0.237 1.67 0.71 — 3.90
Parlor milkingb 0.055 0.910 1.06 0.41 — 2.72
Intercept -0.690 0. 144 NA NA

 

 

Model -2 log likelihood: x2 = 24.76 (5 degrees of freedom; P < 0.001)

NA = Not applicable. aProducer was a college or technical school graduate. bF arm used
a milking parlor.

 

60

 

Table 3-3. Results of multivariable logistic regression analysis of associations between
use of on-farm drug testing and farm management factors.

 

 

95% Confidence

Variable [3 P Odds ratio interval
QAP certification

None -1.668 0.002 0.19 0.06 - 0.55

Involuntary NA NA 1.0 NA

Voluntary -1.249 0.041 0.29 0.09 — 0.95
Herd size 0.003 0.160 1.003 0.999 — 1.006
College graduateal -0.263 0.473 0.77 0.37 — 1.58
Off-label drug useb -0105 0.786 0.90 0.42 — 1.92
Non-emergency -0.019 0.965 0.98 0.42 — 2.27
veterinary carec
Intercept 1.53 1 0.01 1 NA NA

 

 

Model -2 log likelihood: x2 = 14.40 (6 degrees of freedom; P = 0.026).

a|Producer was a college or technical school graduate. bDrugs were used in an off-label
manner. cNon-emergency veterinary care was used on farm.

 

61

 

Table 3-4. Results of multivariable logistic regression analysis of associations between
maintenance of good“ records and farm management factors.

 

 

95% Confidence

Variable [3 P Odds ratio interval
QAP certification

None NA NA 1.0 NA

Involuntary 0.895 0.046 2.45 1.01 — 5.91

Voluntary 0.567 0.187 1.76 0.76 - 4.09
Herd size 0.003 0.147 1.003 0.999 — 1.006
College graduatea 0.548 0.124 1.76 0.86 — 3.48
Intercept -1 . l 83 0.001 NA NA

 

 

Model -2 log likelihood: x2 = 9.70 (4 degrees of freedom; P = 0.046).
*Good records defined as records that included treated cow’s identification, date(s) of
treatment and drug or dose used. “Producer was a college or technical school graduate.

 

62

 

CHAPTER 4

IDENTIFICATION AND QUANTIFICATION OF AMPICILLIN, CEPHAPIRIN
AND PIRLIMYCIN IN COWS’ MILK USING HIGH PERFORMANCE LIQUID
CHROMATOGRAPHY AND FLUORESCENCE DETECTION

ABSTRACT

To determine the reliability of results from three antimicrobial assays used to test
individual cows’ milk; it was essential to quantify the antimicrobials that were potentially
present. Previously described biochemistry methods were modified to better
accommodate the blinded evaluation of milk samples collected as part of a field trial.
Two extraction techniques were necessary for the recovery of ampicillin, cephapirin and
pirlimycin. Additionally, two derivatizations were used to elicit fluorescent products
from ampicillin and pirlimycin. Naturally occurring proteins, lipids and their breakdown
products hindered the clean extraction of the antimicrobials from milk. Three separate
HPLC methods were necessary to obtain adequate detection sensitivity for each
antimicrobial. Reversed phase HPLC with a C-18, 5m, 4.6 X 220mm column was used
in all methods. Sensitivity was substantially enhanced by the use of fluorometric
detection, in place of ultraviolet, to detect Fmoced pirlimycin and derivatized ampicillin.
Identification and separation of cephapirin were best accomplished by premixing the B
buffer (1 :1 0.01M KH2P04zCH3CN) to achieve the desired gradient. The methods
presented should improve the reliability of HPLC analyses used to detect ampicillin and

pirlimycin in milk at or below the established FDA tolerance levels.

63

INTRODUCTION

Antimicrobial residues in milk present several public health and manufacturing
problems (Brady et al., 1993; Mitchell et al., 1998; Waltner-Toews and McEwen, 1994).
In an effort to prevent milk with residues from being marketed, the dairy creamery tests
each tanker for antimicrobial residues prior to accepting the milk. The detection of
antimicrobials in raw milk has been made easier with the use of various residue-screening
assays. Depending on the assay, a qualitative positive or negative result is produced by
an enzymatic reactiona, receptor bindingb, or growth inhibitionc. There has been concern
that other components of raw milk, such as somatic cells (Sischo and Burns, 1993; Van
Eenennaam et al., 1993) or lactoferrin (Carlsson eta1.,1989) may produce false assay
results. Unapproved use of the screening assays to test individual cow milk on farms has
been promoted and widely adopted (Gibbons-Burgener et al., 2000). However, the
reliability of the assays when used to test individual cow milk have yet to be determined.

As part of a longitudinal experimental study to determine the reliability of the
SNAP B-lactam, Penzyme Milk Test and Delvo-SP assays for testing individual cow
milk, gold standard methods for identification and quantification of antimicrobials used
in the treatment of mastitis on the study farms were essential. Specifically, cows in the
antimicrobial treatment group were treated at the producer’s or veterinarian’s discretion
with a FDA approved intramammary preparation containing cephapirin, hetacillin or
pirlimycin. Hetacillin is readily metabolized into ampicillin. Chromatographic methods
are considered the most sensitive and reliable gold standard methods for evaluating the
presence of antimicrobials. Limited studies have been published describing the

identification and quantification of pirlimycin (Hornish et al., 1992; Hornish et al., 1995;

64

Heller, 1996; Heller, 1997). The detection of cephapirin and ampicillin in milk have been
more widely studied (Moats and Romanowski, 1998; Moats, 1993; Moats, 1994;
Dasenbrock and LaCourse, 1998; Tyczkowska et al., 1994).

Preliminary studies in our laboratory indicated that the published extraction and
detection methods were inadequate for blinded screening of the large number of field
samples to be tested. The objective of this part of the overall study was to determine
robust gold standard methods for the identification and quantification of ampicillin,
cephapirin and pirlimycin in order to evaluate the reliability of three on-farm residue

detection assays when used to test individual cows’ milk.

MATERIALS AND METHODS

Equipment Used

Waters 717+ autosampler

Waters Model 510 millipore pump

Waters millipore automated gradient controller

Waters 474 Scanning Fluorescence Detector

Waters 486 tunable (U.V.) absorbance detector
Perkin-Elmer, Supercosil C-18, 5m, 4.6 X 220mm column

Beckman System Gold® analog interface module 406

65

Reagents & Solutions Used

Reference standards for ampicillin“, cephapirind and pirlimycine were used. Stock

solutions of 1 mg/ml of active drug were made for each of the antimicrobials by the

addition of filtered milli-Q water (MQW). Additional standard concentrations were made

by diluting the stock solutions with MQW. Standards were maintained in a —20° C

freezer when not in use. All acetonitrile and methanol used in solutions were either

synthesis or HPLC grade.

Extraction & Derivatization Solutions

a)

b)

d)

26. 7% T richloroacetic (T CA) acid: 500 g of trichloroacetic acid crystalf was
reconstituted to 100% with 500 ml of MQW. Additional dilution with MQW made
26.7% TCA solution.

2 M Sodium hydroxide (2 M NaOH): Combined 20 g of sodium hydroxide pelletsf
(FW 40.0) with 250 m1 of MQW.

2 N Hydrochloric acid (HCI): Diluted 1 part 12 M HCl with 5 parts MQW.
Sore'nson citrate buffer: Dissolved 21 g of granular citric acid monohydratef in 200
ml of 1 M NaOH solution. The solution was diluted to 1 liter with MQW. The
addition of HCl acid is used to adjust the solution pH to 2.5.

0.1% Mercury bichloride (HgC [2): Combined 0.2 mg of dry HgClzd in 200 ml of
Sorénson citrate buffer.

0.1 M Tetraethylammonium chloride (EtWCL): 8.3 g of tetraethylarnmonium

chloride hydrateg was mixed with 500 ml of MQW.

66

g) pH 6.0 KH2P04sNa2HP04 buffer: Combined 50 ml of 0.01 M monobasic potassium

h)

j)

k)

phosphate and 10 ml of 0.01 M dibasic sodium phosphate. The pH was adjusted to

6.0 using glacial acetic acid.

0.25 mM Sodium hydroxide (NaOH): Combined 2.5 ml of 10 mM sodium hydroxide
(made by diluting 1 ml 1 M NaOHf with 99 ml MQW) with 97.5 ml of MQW..

F moc 100 ppm solution: Combined 10 i 0.5 mg 9-Fluorenylmethyl chloroformate
(Fmoc chloride)h with 10 ml of acetonitrile in a 20 ml vial. Transferred the solution
to a 150-250 ml bottle using an additional 90 ml of acetonitrile to rinse the vial
(Heller, 1997). The Fmoc solution was sealed and stored at 4°C. (Weigh powders on
a balance in a chamber and seal the vial until adding acetonitrile in a hood).

F moc 10 ppm solution: Combined 10 ml of 100 ppm Fmoc and 90 ml of acetonitrile.

Sealed and stored the solution at 4°C .

2 M Disodium hydrogen phosphate (2 'M NazHP04): Combined 56.8 gm of

disodium hydrogen phosphate anhydrous powderf with 200 ml of MQW.

Bufler solutions

a) 55:45 CH 30H:H20 : For each liter of buffer, 550 m1 of methanol was combined with

450 ml of MQW. Buffer was sparged with helium for 30 minutes.

b) 0.01 M KH2P04: 1.36 g of granular potassium phosphatef was mixed in one liter of

MQW and then filtered.

c) 1:1 0.01 M KH2P041CH3CH : 500 ml of premixed/filtered 0.01M KH2P04 was

mixed with 500 ml acetonitrile.

67

d) 4:3 :3 1% CH 3C00H.'CH 30H:CH 3CN: For each liter of buffer we combined 4 ml of
glacial acetic acid, 396 ml of MQW, 300 ml of methanol and 300 m1 of acetonitrile. The

solution was mixed for 5 minutes and sparged with helium for 30 minutes.

Extraction and Derivatization of Ampicillin

Milk sample was thawed in an ice bath and briefly vortexed prior to use. One ml
of milk was combined with 4 m1 of MQW and 3 ml 26.7% TCA in a 15 ml centrifuge
tube. Tube contents were vortexed for 10 seconds and then centrifuged 5 minutes at 1000
g. 4.5 ml of supernatant was pipetted and filtered through glass wool into a second 15 ml
centrifuge tube, avoiding inclusion of the top fat layer. We added 0.5 ml of 2 M NaOH to
the filtered solution and vortexed it 3 seconds and incubated for 5 minutes at room
temperature. That was followed by the addition of 0.5 ml of 2 N HCl and 1 ml of 0.1%
HgC 12 solution. Again the solution was vortexed 3 seconds and incubated for 5 minutes
at room temperature. The pH of the solution was adjusted to 6.2 by adding pre-warmed
2 M NazHPO4. The solution was incubated at 38-40° C for 25 minutes. Six ml of ethyl
acetate was added and the solution was vigorously shaken for 5 minutes. The solution
was centrifuged for 5 minutes at 1000 g. The top, organic layer (approximately 5 ml) was

decanted and retained . Liquid was evaporated in speed vac with no heat.

Extraction of Cephapirin and Pirlimycin
Milk samples were thawed in an ice bath and briefly vortexed prior to use.

Combined 4 ml of milk and 0.8 ml of Et4NCl in small beaker. Slowly added 16 ml of

CH3CH while continuously vortexing the solution. The solution was incubated for 10

68

minutes at 4°C. Supernatant was filtered through glass wool. 0.8 ml of 6.0 pH buffer
was added and thoroughly mixed. Solution was transferred to glass tubes and the sample
was dried in a speed-vac with no heat. We used 4 tubes initially and later combined 3
tubes into one resulting in one tube with residue from 1 ml of milk and one tube with

residue from 3 ml of milk.

Derivatization for Pirlimycin Detection

Following the extraction and drying of the tube with residue from 1 m1 of milk, a
derivatization process may be undertaken 3-12 hours prior to analysis of sample in the
HPLC system. One hundred m1 of standard in water may also be derivatized in this
manner. To the dry residue we added 0.5 ml of 0.67 NaOH (0.4 m1 added to 100 pl of
standard in water) and vortexed 10 seconds. Added 0.5 m1 of 100 ppm Fmoc solution
(use 10 ppm Fmoc with standard in water) and vortexed for 10 seconds. The mixture was
incubated at room temperature for 1 hour and then vortexed 10 seconds. A minimum of

two additional hours of incubation is required prior to analysis on the HPLC system.

Detection of Ampicillin Residue

The dry HgClz derivatized residue was brought up in 100 pl of 100% methanol
and vortexed for 5 seconds. The entire 100 pl was transferred to a 250 pl autosampler
tube. Using an isocratic 55:45 CH3OHzMQW buffer system with a flow rate of 0.8
ml/min, 20 pl of a sample could be injected every 30 minutes with no detected carryover
from previous sample. An excitation A = 345 nm and emission it = 420 nm on the

fluorescence detector was used for optimal detection of ampicillin residues.

69

Detection of Pirlimycin

Fmoced samples and standards are suspended in a 1 ml solution during the
derivatization process. We transferred 200 p1 of each sample to a 250 pl autosampler
tube. An isocratic program using 423:3 1% acetic acidzCH3OH2CH3CN was used with an
excitation it = 260 nm and emission l. = 315 nm on the fluorescent detector to detect

pirlimycin residues. 50 p1 of each sample was injected into the system every 35 minutes.

Detection of Cephapirin

The dried sample containing 3 ml milk extract is eluted in 200 pl of 0.01 M
KH2P04. The solution is transferred to an autosampler tube and centrifirged for 4
minutes. The supernatant is decanted to another tube. A gradient program using 0.01 M

KH2P04 and 1:1 0.01 M KH2P042CH3CH buffers is used with UV. detection at A = 290

nm. 50 pl of each sample was injected into the HPLC system every 65 minutes.

Quantification of Antimicrobials

Standard linear regression curves were developed for each drug based on the areas
under the curve/peak at the specific retention time for the known blank and spiked milk
samples. The areas at the same retention times measured on blinded study samples were
then placed in their respective regression model to produce the estimated concentration of

each drug in the sample.

70

RESULTS AND DISCUSSION

The ideal method would have involved one extraction protocol and one HPLC
system. However, it became apparent that that was not possible and the most robust
methods were sought. Several studies have used residue screening assays to test fractions
collected as part of a liquid chromatography cleanup method (Harik-Khan and Moats,
1995; Moats and Romanowski, 1998; Anderson et al., 1998). Since we were determining
the reliability of three of the assays it was deemed inappropriate to use the same assays in
the development of the gold standards. Fraction collection and cleanup were to be
avoided if possible due to time constraints.

Moats indicated difficulties in effectively clearing a LC column of ampicillin
(Moats, 1994) and our preliminary investigations bore the same finding. Reports of
derivatization and fluorometric detection methods used to identify <50 ppb of ampicillin
in plasma, serum, kidneys and liver provided a new route to explore in detecting
ampicillin in milk (Miyazaki etal., 1983; Hong et al., 1995). Slight modifications in the
extraction methods used with serum and plasma samples were made to accommodate the
use of milk samples. By increasing the initial TCA concentration to 26.7% we more
effectively deproteinated the milk products. Another modification was the need to add
greater quantities of NazHPO4 at a higher molarity (2 M instead of 0.67 M) to bring the
pH up to 6.2. The extraction method produced an unimpressive 11.5 — 12.3% recovery
rate of ampicillin fiom milk. Fluorescence detection provided a peak that was clearly
evident at 5.15 - 5.25 minutes in the presence of ampicillin below the lowest reported
assay detection level of 7.7 ppb (Figure 4-1). By increasing the amount injected into the

HPLC system from 20 pl to 40 or 50 pl one could improve the sensitivity two-fold.

71

Additional testing that included amoxicillin, cephapirin and pirlimycin produced no
unique detectable fluorescent products. This was advantageous in verifying the
specificity of this method for ampicillin identification. Conversely, it was a detriment in
the pursuit of a single extraction method for the three drugs being studied.

The extraction method commonly used with LC detection of B-lactams (Moats
and Romanowski, 1998) was found to be adequate for the extraction of the macrolide,
pirlimycin. Recovery rates for pirlimycin were 85 — 102% down to 62 ppb. Analytical
sensitivity of the HPLC method was approximately 30 ppb which was well below the
400 ppb FDA tolerance level and the 50-200 ppb detection level of the Delvo-SP assay.
Identification and quantification of pirlimycin were greatly improved with the use of
F moc derivatization (Heller, 1997). Fluorescent products from only pirlimycin were
detected using an ultra-violet detector with k = 264 nm. Detection at smaller
concentrations was further enhanced by the use of a fluorescence detector (Figure 4-2). A
previous study suggested that the excitation wavelength would be 275 nm and the
emission wavelength would be 315 run when detecting F moc bound products (Chou et
al., 1989). The scanning feature on the fluorescence detector allowed us to further hone
the excitation wavelength to 260 nm and maintain the 315 nm emission wavelength.
Again we were able to detect only one of the drugs of interest using this derivatization
method.

Cephapirin proved to be the most difficult of the three drugs to identify a clean
peak using any of the proposed methods. Altering the wavelengths for fluorescence
detection using either derivatization method produced negative results. Coelution of milk

by-products at the retention times for desacetylcephapirin and cephapirin (20.9 and 25.32

72

minutes respectively) reduced the sensitivity of cephapirin detection. Without the
potential benefit of fraction cleanup, the detection limit for cephapirin was 42 ppb.
Unfortunately the sensitivity was inadequate for detecting cephapirin below its .20 ppb
tolerance level and 5 ppb assay detection level.

In conclusion, a single method for the simultaneous identification and
quantification of ampicillin, cephapirin and pirlimycin could not be found or developed.
For the evaluation of 200 blinded milk samples, the use of derivatization methods and
fluorescence detection provided greater detection sensitivity and specificity for ampicillin

and pirlimycin than methods currently used to evaluate isolated samples.

73

FOOTNOTES

aPenzyme Milk Test, Cultor Food Science Group, New York, NY.
bSNAP B-lactam, IDEXX Laboratories Inc, Westbrook, ME.
cDelvo-SP, Gist Brocades Food Ingredients Inc, Menomonee Falls, WI.
d Sigma Chemical Co., St. Louis, MO.

6 Pharmacia & Upjohn, Kalamazoo, MI.
fJ.T. Baker, Phillipsburg, NJ.

3 Aldrich Chemical Co., Milwaukee, WI.

h Pierce, Rockford, IL.

74

REFERENCES

Anderson KL, Moats WA, Rushing J E, O’Carroll JM. Detection of milk antibiotic
residues by use of screening tests and liquid chromatography after intramammary

administration of amoxicillin or penicillin G in cows with clinical mastitis. Am J Vet Res
1998;59:1096-1100.

Brady MS, White N, Katz SE. Resistance development potential of

antibiotic/antimicrobial residue levels designated as “safe levels”. J Food Prot
1993;56:229-233.

Carlsson A, Bjorck L, Persson K. Lactoferrin and lysozyme in milk during acute mastitis
and their inhibitory effect in Delvotest P. J Dairy Sci 1989;72:3166-3175.

Chou TY, Gao CX, Grinberg N, Krull IS. Chiral polymeric reagents for off-line and on-
line derivatizations of enantiomers in high-performance liquid chromatography with

ultraviolet and fluorescence detection: an enantiomer recognition approach Anal Chem
1989;61:1548-1558.

Dasenbrock CO, LaCourse WR. Assay for cephapirin and ampicillin in raw milk by
high-performance liquid chromatography - integrated pulsed amperometric detection.
Anal Chem 1998;70:2415-2420.

Gibbons-Burgener SN, Kaneene J B, Lloyd J W, Erskine RJ. Influence of the Milk and
Dairy Beef Quality Assurance Program on dairy farm drug management practices. J Am
Vet Med Assn 2000;216:1960-1964.

Harik-Khan R, Moats WA. Identification and measurement of beta-lactam antibiotic
residues in milk: integration of screening kits with liquid chromatography. J AOAC Int
1995;78:978-986.

Heller, DN. Determination and confirmation of pirlimycin residue in bovine milk and
liver by liquid chromatography/thermospray mass spectrometry: interlaboratory study. J
AOAC Int 1996;79:1054-1061.

Heller, DN. Determination of Pirlimycin residue in milk by liquid chromatographic
analysis of the 9-fluorenylmethyl chloroformate derivative. J AOAC Int 1997;80:975-
981.

Hong CC, Lin CL, Tsai CE, Kondo F. Simultaneous identification and determination of

residual penicillins by use of high-performance liquid chromatography with
spectrophotometric or fluorometric detectors. Am J Vet Res 1995;56:297-303.

75

Hornish RE, Arnold TS, Bacynskyj L, et al.. Pirlimycin in the dairy cow. In: Hutson
DH, ed. Xenobiotics in Food producing animals — metabolism and residues, ACS

symposium series 5 03 . American Chemical Society, Washington, DC, 1992;132-147.

Hornish RE, Cazars AR, Chester ST, Roof RD. Identification and determination of
pirlimycin residue in bovine milk and liver by high-performance liquid chromatography-
thermospray mass spectrometry. J Chromatogr B 1995;674:219-235.

Mitchell JM, Griffiths MW, McEwen, SA, et al.. Antimicrobial drug residues in milk
and meat: causes, concerns, prevalence, regulations, tests, and test performance. J Food
Prot 1998;61:742-756.

Miyazaki K, Ohtani K, Sunada K, Arita T. Determination of ampicillin, amoxicillin,
cephalexin, and cephradine in plasma by high-performance liquid chromatography using
fluorometric detection. J Chromatography 1983;276:478-482.

Moats, WA. Determination of cephapirin and desacetylcephapirin in milk using
automated liquid chromatographic cleanup and ion-pairing liquid chromatography. J
AOAC Int 1993;76:535-539.

Moats, WA. Determination of ampicillin and amoxicillin in milk with an automated
liquid chromatographic cleanup. J AOAC Int 1994;77:41-45.

Moats, WA, Romanowski, RD. Multiresidue determination of B-lactam antibiotics in
milk and tissues with the aid of high-performance liquid chromatographic fractionation
for clean up. J Chromatogr A 1998;812:237-247.

Sischo WM, Burns CM. Field trial of four cowside antibiotic-residue screening test. J
Am Vet Med Assn 1993;202:1249-1254.

Tyczkowska KL, Voyksner RD, Straub RF, Aronson AL. Simultaneous multiresidue
analysis of b-lactam antibiotics in bovine milk by liquid chromatography with ultraviolet
detection and confirmation by electrospray mass spectrometry. J AOAC Int
1994;77:1122-1131.

Van Eenennaam AL, Cullor J S, Perani L, eta1.. Evaluation of milk antibiotic residue
screening tests in cattle with naturally occurring mastitis. J Dairy Sci 1993;76:3041-
3053.

Waltner-Toews D, McEwen SA. Residues of antibacterial and antiparasitic drugs in
foods of animal origin: a risk assessment. Prev Vet Med 1994;20:219-234.

76

Figure 4-1 — Chromatogram ofthe fluorescent detection (cm A. = 345, em A = 420) of 20
ppb ampicillin extracted from raw milk.

0.100

Absorbance (AU)

0.050

- Ampicillin
5.15 min.

 

 

 

0.000

 

 

 

r r I r I r 1 I r l “‘T‘
8 O o
O. o
Minutes N

77

Figure 4-2 — Chromatogram ofthe fluorescent detection (ex A = 260. cm A = 315) of200
ppb pirlimycin extracted from raw milk.

 

 

 

 

 

 

 

 

 

 

 

 

 

o
o
O.
N -
A
I)
<1
v
o d
‘8
E
0 ..
U)
.0
<2
o
8
.r “ '1 i"
.4
Pirlimycin
- . 9.64 min.
'.
d I
d
1
I
c, 1
O
o _ J
o
T j fl I I j I I 1 l I
o o
O. o 8
Minutes

78

CHAPTER 5
AN EPIDEMIOLOGICAL EVALUATION OF THE RELIABILITY OF

BULK-TANK RESIDUE DETECTION ASSAYS USED TO TEST
INDIVIDUAL COW MILK

ABSTRACT

Objectives— to determine the likelihood of false assay results when using the SNAP B-
lactam, Penzyme Milk Test and Delvo-SP assays to test for antimicrobial residues in
individual cows’ milk.

Sample Population—1 1 1 cows diagnosed with mild clinical mastitis on one of eight
participating dairy farms.

Procedure—Cows were randomly assigned to either the antimicrobial or control
treatment group. Pretreatment and post-treatrnent milk samples were collected. Post-
treatment samples were randomly tested twice using each of the 3 on-farm residue
detection assays and once using high performance liquid chromatography methods. The
reliability of each of the assays was determined using sensitivity, specificity, positive and
negative predictive values and the kappa statistic.

Results—The Delvo-SP and SNAP B-lactam assays displayed >90% sensitivity,
while the sensitivity of the Penzyme Milk Test was only 60%. Ranging from 39.29 to
73.68%, the positive predictive values were poor for all three assays. The kappa statistics
for the SNAP B-lactam, Penzyme Milk Test and Delvo-SP were 0.846, 0.545 and 0.813
respectively.

Conclusion and Clinical Relevance— The kappa statistics provided strong evidence

that all three assays produced good to excellent repeatability. The poor positive predictive

79

value of the SNAP B-lactam assay was most likely due to an undocumented cross-
reactivity with pirlimycin residues. With such low positive predictive values and
incidence of violative antimicrobial levels, the usefulness of the three residue detection
assays in deciding the fate of milk from cows receiving treatment for mastitis is highly

questionable.

80

INTRODUCTION

Dairy farmers, veterinarians, dairy manufacturers and researchers believe it is
highly desirable to have at least one quick and reliable test for the detection of
antimicrobials in milk. Prior to the approval of new drugs for use in lactating cattle,
pharmaceutical companies must demonstrate that an assay or method exists that detects
their drug in marketable milk. The ability to test individual cow milk for residues is
essential in the determination of labeled withholding periods. However, earlier studies
(McEwen, et al., 1991; Gibbons-Burgener, etal., 1999) have found that farmers depend
more on residue testing than labels when deciding to withhold milk from a treated cow.
The practice of testing individual cow milk off or on farms is widespread and promoted.
Since there are no rapid assays labeled for use in testing individual cows’ milk, it has
become acceptable to use approved commingled milk testing assays.

Numerous reports have demonstrated that current on-farm testing of milk from
individual cows for drug residues often yields false-positive results and that caution is
warranted in their use (Andrew, et al., 1997; Cullor, 1992; Cullor, et al., 1994; Sischo and
Burns, 1993; Van Eenennaam, et al., 1993). Although the specificity and sensitivity of
the assays have been established for commingled milk in controlled laboratory conditions,
concern over the accuracy under field conditions exists, because to date, few studies have
been conducted that validate these assays in a field setting. In fact, no screening assay has
been recognized by the FDA for use on milk from individual cows. The objective of this
study, therefore, was to determine the likelihood of false assay results when using the
SNAP B-lactam, Penzyme Milk Test and Delvo-SP assays to test for antimicrobial

residues in individual cows’ milk.

81

MATERIALS AND METHODS

Study Design - A longitudinal experimental study of cows developing mild clinical
mastitis was conducted to evaluate the reliability of the SNAP B-lactarn, Penzyme Milk
Test and Delvo-SP assays when used to test individual cow milk for antimicrobial
residues.

Case Definition - A mild clinical mastitis case was defined as l) visibly abnormal milk
stripped from the quarter, 2) quarter may be enlarged or reddened and 3) conventional
therapy (intramammary antimicrobial infirsions or other non-antimicrobial therapy) was
believed to be an appropriate treatment. Cows were specifically excluded from the study
if they 1) received antimicrobial treatment for any reason within the last 30 days, 2) were
previously included in this study (a repeat case), 3) had a concurrent illness requiring
antimicrobial treatment or 4) had a severe case of mastitis that required systemic (IV, IM
or SQ) antimicrobial therapy.

Sample Size - Sample sizes for estimating a single proportion (i.e. F alse-positive
results/all results) were much less than those required for evaluating potential
associations with risk factors. By estimating that 10% of the population would have a
positive assay result and allowing a 5% margin of error (or = 0.05), the required sample
size was 138 tests (Appendix 2).

Treatment Groups - After diagnosing a case of mild clinical mastitis and collecting the
initial milk sample, cows were randomly assigned to one of two treatment groups (Figure
5-1). Cows assigned to the Antimicrobial treatment group were treated as directed on the

label with a FDA approved IMM antimicrobial therapy selected by either the producer or

82

veterinarian. Label dose and number of treatments were followed unless the cow was
removed from the study. Cows assigned to the control treatment group received
appropriate treatment that did not include any form of antimicrobial therapy. Other
treatments included anti-inflammatory drugs, oxytocin, saline infusions or no drug

therapy.

Sample Collection - Two samples (pre and post-treatment) from each mastitis case were
collected. Prior to collection of the samples, foremilk from each quarter was discarded.
An aseptic collection of at least 5 ml of milk from the affected quarter was obtained for
bacterial culture using standard methods (Harmon, et al., 1990). An 80-ml composite
milk sample, comprised of approximately 20 ml of milk from each quarter was collected.
The composite milk was then divided as follows: 1) 30 ml in a plastic vial with
preservative tablet for infrared somatic cell count, 2) 10 ml in a plastic vial for IgGl
analysis, and 3) four 5-8 ml aliquots in plastic vials for antimicrobial residue analyses.
The aliquots were frozen at -70°C until needed for residue analyses.

The initial sample was collected following the diagnosis of mild clinical mastitis
and before initiating treatment (Figure 5-2). The second sample was collected at the
milking following the completion of the labeled milk-withholding period for cows in the
antimicrobial treatment group. To simulate similar potential recovery times from
diagnosis to second sample, the timing for the collection of the second sample from a cow
in the control treatment group was determined by using the same withholding period as
the last (most recently) antimicrobial treated cow. Treatment with other drugs may have

required a variety of actual withholding periods prior to shipment of milk from the farm.

83

The label’s recommended withholding period and instructions were to be observed prior
to including milk in the bulk tank.

Testing for antimicrobials - Milk samples were thawed in an ice water bath and
vortexed briefly to thoroughly mix. Pre-treatrnent samples were randomized and tested
once with each assay. Each post-treatment sample was randomized twice and tested
twice. Except for the use of individual cow and thawed samples, the three assays being
evaluated (Delvo-SP, SNAP B-lactam and Penzyme Milk Test) were run according to
each assay’s directions. The same person throughout the study performed visual
interpretation of the assay results.

High performance liquid chromatography (HPLC) was used to identify and
quantify the presence of potential antimicrobial residues in each of the samples. The
specific extraction and detection methods have been previously described (Gibbons-
Burgener, et al., 2000). By comparing the quantity of each antimicrobial found on HPLC
with its FDA tolerance level and assay detection limits (Table 5-1) we were able to
determine if the residue should have been detected by each assay and whether a given

residue was considered violative (above the FDA tolerance level).

Statistical analysis - The reliability of each of the residue detection assays was expressed
in terms of sensitivity, specificity, positive predictive value and negative predictive value
using equations 1, 2, 3 and 4 respectively. The reliability statistics were calculated first
using the specific assay’s detection limits and second using the FDA-established

tolerance levels for each of the antimicrobials.

84

Equation 1

Sensitivity = No. tests with positive results on HPLC & assay X 100
No. all tests with positive results on HPLC (+/- assay)

Equation 2

Specificity = No. tests with negative results on HPLC & assay X 100
No. all tests with negative results on HPLC (+/- assay)

Equation 3

Positive Predictive = No. tests with positive results on HPLC & assay X 100
Value No. all tests with positive results on assay (+/- HPLC)

Equation 4
Negative Predictive = No. tests with negative results on HPLC & assay X 100
Value No. all tests with negative results on assay (+/- HPLC)
Calculating the kappa statistic for each of the three commercial assays compared
the concordance of the first and second tests run on the post-treatment samples (Equation
5).
Equation 5 (Rosner, 1995)
Kappa = (P0 - Pc)/(1- P6)
P0 = observed probability of concordance between 2 testings

P¢ = expected probability of concordance between 2 testings

RESULTS

Eight farms participated in the study and collected data. Of the 111 cows that
developed clinical mastitis and were enrolled as cases, 92 remained in the study through
their post-treatment sample collection (83% case retention rate). About half (45/92) of

the cows were in the antimicrobial treatment group and received either pirlimycin (26/45,

85

57.8%), hetacillin (9/45, 20%) or cephapirin (IO/45, 22.2%) IMM therapy. We were
unable to interpret the assay results from 3 of the cows. Occasionally a milk sample
would produce no visual result (dud) on a given assay. Consequently, of the potential
178 post-treatment tests run on each assay, the evaluation of the SNAP B-lactam,
Penzyme Milk Test and Delvo-SP assays included 168, 175 and 177 tests respectively.

The frequency distributions of the assay results (Figure 5-3 and 5-4) indicate
relatively low numbers of positive results. Sensitivity, specificity and predictive values
for the 3 assays are in Tables 5-2 and 5-3. Twenty-three cows had levels of an
antimicrobial that were detectable by at least one of the commercial assays. Only 6 of
those 23 animals had violative levels. The statistical sensitivities of the SNAP B-lactam,
Penzyme Milk Test and Delvo-SP assays to detect the violative samples were 83.33, 62.5
and 91.67% respectively (Table 5-3).

The kappa statistics for the pairs of tests run on the SNAP B-lactam (K =

0.846) and Delvo-SP (1c = 0.813) were similar. The Penzyme Milk Test had a lower

statistic (K = 0.545).

DISCUSSION

Unlike some earlier studies (Harik-Khan and Moats, 1995; Halbert, et al., 1996;
Andrew, et al., 1997) this study was an experimental field trial designed to simulate the
collection and testing of milk as commonly performed on farms. This meant samples
were collected from diseased cows requiring their milk to be withheld from the bulk-tank,
the samples weren’t spiked with known quantities of antimicrobials and withholding

periods were observed to best simulate the return of milk to the bulk tank. Another

86

consequence of the study design was that the enrollment of farms willing and capable of
following the study protocol included farms with relatively low incidence rates of mild
clinical mastitis. These study requirements limited the number of accessible cases.
However, the obtained sample sizes provided adequate precision in the calculation of the
reliability statistics.

The testing done on the pre-treatment samples was performed only to ensure the
absence of pre-existing residues or abnormalities that could have interfered with the
interpretation of the post-treatment samples. Because all the study cows were initially
diagnosed with clinical mastitis and abnormal looking milk is a manufacturer
contraindication for the use of the Penzyme and Delvo assays, it was inappropriate to use
the pre-treatment test results to evaluate the reliability of the tests. In addition, testing for
residues prior to treatment on a farm is a rare request. By blindly testing the post-
treatrnent samples twice and comparing the results with the kappa statistic we found
strong evidence that all three assays produced good to excellent outcome repeatability.
This is consistent with the report of a discussion fortun which also recommended the use
of two assays in series instead of simply repeating the same assay on presumptive positive
samples (Gardner, et al., 1996). However, in this study the tests couldn’t be evaluated in
series. Different assays may test for different antimicrobials at different detection limits,
which can make interpreting series tests more difficult. This was particularly evident in
this study where only the Delvo-SP assay was reported to be able to detect pirlimycin. As
recommended by the discussion forum, it would be most beneficial to have a sensitive

initial assay and a more specific second assay. The three assays we evaluated had similar

87

detection limits for ampicillin and cephapirin, and provided little improvement of

specificity to distinguish false positive results.

The SNAP [3-lactam and Delvo-SP had comparable statistical sensitivities. The
Penzyme Milk Test had a few more false-negative results, which (combined with a low
prevalence of detectable residues) had a profound effect on lowering the sensitivity
reported by this study. Specificities for all three assays were comparable. The predictive
values for the assays are better indicators of what the dairy farmer encounters in deciding
to discard or sell the milk from a tested cow. The positive predictive value is the
likelihood that a positive assay result has truly detected an antimicrobial residue within its
detection limits or above the F DA-established tolerance level. Each of the assays had
lower positive predictive values than one might have expected with fairly good
specificities. This phenomenon is particularly evident when the disease (detectable
residue) prevalence is low. The SNAP B-lactarn assay had a poor positive predictive
value, which is most likely due to previously undocumented cross-reactivity with
pirlimycin residues. Six of the 17 false-positive results were recorded for samples with
HPLC-detected pirlimycin residues. If the assay had been approved for pirlimycin
detection the positive predictive value would have improved to almost 61%. With only a
rare false-negative result, each of the assays exhibited an excellent negative predictive
value. Assay users should feel very confident in a negative result if they run an

appropriate assay developed to detect the administered antimicrobial.

88

 

This study found that 6 of the 45 cows treated according to the label with an
antimicrobial approved for use in lactating cattle had violative levels in their milk after
observing the labeled milk-withholding period. It is difficult to discuss violative levels in
terms of individual cow milk, because the standards are set for commingled or bulk-tank
milk where an individual cow’s milk is usually diluted. As at the creamery, the farmer
can not quantify the amount of antimicrobial in the milk following a positive assay result.
The conservative approach to a positive assay test on an individual cow’s milk is to
discard the milk and retest at a later milking. If this practice had been followed with the
study cows, milk from 17 of the cows may have been unnecessarily discarded. With such
low positive predictive values and prevalence of violative antimicrobial levels, the
usefulness of the three residue detection assays in deciding the fate of milk from cows
receiving treatment for mastitis is highly questionable if farms want to minimize the

quantity of unnecessarily discarded milk.

89

REFERENCES

Andrew SM, Frobish RA, Paape MJ, Maturin LJ. Evaluation of selected antibiotic
residue screening tests for milk from individual cows and examination of factors that
affect the probability of false-positive outcomes. J Dairy Sci 1997;80:3050—3057.

Cullor J S. Tests for identifying antibiotic residues in milk: how well do they work? Vet
Med 1992;87:1235-1241.

Cullor J S, van Eenennaam A, Gardner 1, et al.. Performance of various tests used to
screen antibiotic residues in milk samples from individual animals. J AOAC
1994;77:862-870.

Gardener IA, Cullor J S, Galey FD, et al.. Alternatives for validation of diagnostic assays
used to detect antibiotic residues in milk. J Am Vet Med Assn 1996;209:46-52.

Gibbons-Burgener SN, Kaneene JB, Lloyd J W, Erskine RJ. Evaluation of certification in
the Milk and Dairy Beef Quality Assurance Program and associated factors on the risk of
having violative antibiotic residues in milk from dairy farms in Michigan. Am J Vet Res
1999;60:1312-1316.

Gibbons-Burgener SN, Leykam J F , Kaneene J B. Identification and quantification of
ampicillin, cephapirin and pirlimycin in cows’ milk using high performance liquid
chromatography and fluorescence detection. J Vet Diagn Invest, 2000(manuscript
submitted).

Halbert LW, Erskine RJ, Bartlett PC, Johnson GL. Incidence of false-positive results for
assays used to detect antibiotics in milk. J Food Prot 1996;59:886-888.

Harmon RJ, Eberhart RJ, Jasper DE, et al.. Microbiological procedures for the diagnosis
of bovine udder infections, ed 3. Arlington, VA, National Mastitis Council, 1990.

Harik-Khan R, Moats WA. Identification and measurement of B-lactam antibiotic
residues in milk: integration of screening kits with liquid chromatography. J AOAC Intl
1995;78:978-986.

McEwen SA, Meek AH, Black WD. A dairy farm survey of antibiotic treatment
practices, residue control methods and associations with inhibitors in milk. J Food Prot

1991;54:454-459.

Rosner BA. Chapter 10 Hypothesis testing: Categorical data. In Fundamentals of
Biostatistics, 4‘11 ed. London, England, International Thomson Publishing, 1995;423-426.

Sischo WM and Burns CM. Field trial of four cowside antibiotic-residue screening tests.
J Am Vet Med Assn 1993;202:1249-1254.

90

Van Eenennaam AL, Cullor J S, Perani L, et al.. Evaluation of milk antibiotic residue
screening tests in cattle with naturally occurring mastitis. J Dairy Sci 1993;76:3041-
3053.

91

 

  

 

 

72:85. _8_=__o 2:: E 38

 

Agnew E2585 .888 15 1.380355 05 E momma mo 98on Banach vac Sufi—H9934 Tm 0.53..—

92

Eootnxo ”x...

 

 

 

 

 

 

 

v m N _.
Ill-#llllgllll 35:00
So... as am we on VN N_. o
I I I I I I xm<
acted 29:23 5:).
N osmtd ”xh F
29:3 2953 “BE
ocooom mic—.85

.38 95% 35:8 05 com BEBE mam—98m M38 @885 Cam/xv 35828?“ SE ~88 05
m5? 838% “8:58:88 can can 05 mo 880:8 05 com 855.8% 83 8888a 3:958 30: Co 29:88 :< I «in 0.53,.—

93

Figure 5-3 — 2 x 2 charts for the distribution of assay results using the reported

antimicrobial detection limits for each assay.

+
Delvo-SP Assay
+
Penzyme Milk
Test Assay -
+

SNAP B—lactam
Assay -

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

HPLC
+ _
28 10 38
3 136 139
31 146 177
HPLC
+ -
6 4 10
4 161 165
10 165 175
HPLC
+ ..
ll 17 28
1 139 140
12 156 168

94

 

Figure 5-4 — 2 x 2 charts for the distribution of assay results using FDA-established
antimicrobial tolerance levels.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

HPLC
+ -
+ l 1 25 36
Delvo-SP Assay
- 1 138 139
12 163 175
HPLC
+ -
+ 5 4 9
Penzyme Milk
Test Assay - 3 161 164
8 165 173
HPLC
+ -
+ 5 19 24
SNAP B—lactam
Assay - 1 139 140
6 158 164

 

95

Table 5-1. FDA tolerance levels for specific antimicrobials in marketable milk and the
visual minimal detection limits of 3 on-farm residue detection assays.

 

FDA tolerance SNAP b-lactam Penzyme Milk

 

level d.l.' Test d.l.‘l Delvo-SP d.l.'
Antimicrobial (ppb) (ppb) (ppb) (ppb)
Ampicillinb 10 4-6 4-6 4
Cephapirin 20 2 4-8 5
Pirlimycin 400 NAc NAc 50-200

 

 

ad.l.=minimal detection limits / analytical sensitivity, bAmpicillin is the immediate
product of hetacillin metabolism, NA = no information available regarding ability to

detect residue or the potential for cross-reactivity.

 

96

 

 

88“ mo 8983:": a 62.3 38:85 n .>.m ._ . BEBE cocoumcoo £63 05 mo £8: comm: v8 .833 a

 

 

 

R.— Gmdo 1 8.8V vwfio 80.3 1 86% «9mm $.98 1 8.53 m _ .mg 836 1 3.2» dem mmtoZoQ

“8%

m: 3.8 .. 3.8V ”08 8.; 1 age o8 $28 1 3.3 mm? 3.5 1 433 o8 0.52 688

m3 338 1 8.08 omda Amvdm 1 on. _ NV omdm 308 1 ~— 53 o—dw 858 1 ~23 Nb. 5 gum—tn m<Zm
.x. as as as

.a .8 same ..>.._ «>582 .8 some at; 358.— .8 same £65625 .8 same £5588 8.2

 

 

.35 20¢ a E 88538 £88 .8888 2662 Sammie 8.5g 3 :8:
8888 >88 8:82 05 gene .8 “a mRBEomEEB no 8828 05 co.“ 82.3 38:85 can 36¢?on .bmfixmcom .Ntm 033—.

97

 

83“ mo 8985:": o dig o>fio€8q n .>.m a . BEBE 8:09.38 $3 05 .«o £8: H25: 98 832 a

 

 

 

mt 30.8 1 863 undo 2 :3 1 m3: whom and» 1 owns 83 868 1 ~28 S. 5 “51¢on

“mob

m: A8,? 1 $5 \1 _ .3 88m 1 8. _ 8 3.3 $25 1 35$ 3.3 at; 1 $63 0% x52 085$

«2 30.8 1 8.8V amda 5.9 1 25 8.3 80.2. 1 3.5V 8.? 308 1 3.3 2.8 83%; m<Zm
.x. .x. .x. .x.

9.. be same .3... 2::qu “no :3 .3: 358.— Eu came £95925 «so :me Ezzaom 5.2

 

 

.35 20¢ a E 38390 @988 .838ro 0352 83.38 025 3 £05.
8:828 uosmznfimmv1<am 2: 959m maneuogucm mo 582% 05 8w 8:13 0386on can 56$”.on .ESEmcom .m1m «Bah

98

CHAPTER 6

RISK FACTORS ASSOCIATED WITH F ALSE-POSITIVE RESULTS USING
THREE ON-FARM BULK TANK RESIDUE DETECTION ASSAYS ON
INDIVIDUAL COW MILK

ABSTRACT

Objectives— To determine whether SCC, IgGl, microbiologic isolates or specific
antimicrobial treatments were associated with false-positive results when using the SNAP
B-lactam, Penzyme Milk Test and Delvo-SP assays to test milk from cows treated for
naturally occurring mastitis.

Sample Population—1 11 cows diagnosed with mild clinical mastitis on one of eight
participating dairy farms.

Procedure—Cows were randomly assigned to either the antimicrobial or control
treatment group. Post-treatment samples were randomly tested twice using each of the 3
on-farm residue detection assays and once using HPLC methods. Microbiologic culture
was performed using milk from the affected quarter. Composite milk samples were also
tested for somatic cell count and bovine lg61 concentration. Logistic regression was used
to determine associations between risk factors and false-positive assay results.
Results—Inflammatory-related milk proteins, represented by bovine IgG1 concentrations,
were positively associated with false-positive results from the SNAP B-lactam, Penzyme
Milk Test and Delvo-SP assays. The SNAP B-lactam assay was less likely to have a false-

positive result when the initial bacteriologic culture result was negative. Low numbers of

99

false-positive results precluded multivariable statistical analysis and warrants caution in
over-interpreting the apparent associations.

Conclusion and Clinical Relevance— This study demonstrated that components in milk
other than antimicrobials were associated with positive assay results, which sheds more
doubt on the use of on-farm antimicrobial residue screening assays to test milk from

individual cows.

100

INTRODUCTION

Having reliable testing methods for antimicrobials in milk is essential to the dairy
industry. Since the changes to the grade A Pasteurized Milk Ordinance in 1991 mandated
the screening of every tanker-load of raw milk for at least B-lactam antimicrobial
residues, there has been an increased emphasis on the preharvest prevention of residues at
the farm level (Adams, 1994). Producers are encouraged to either submit samples or test
milk on—farm from individual treated cows prior to including the cow’s milk in the bulk-
tank. Unfortunately, the only tests available are approved for screening commingled
milk, and have been reported to produce false-positive results when used to test
individual cow milk for antimicrobial residues (Cullor et al, 1994; Sischo and Burns,
1993; Andrew et al, 1997; Gibbons-Burgener et al, 2000).

Because violative residues in milk are a relatively rare occurrence (<0. 1% of bulk
tanks), the likelihood of false-negative results is negligible. Consequently, research
evaluating the accuracy of on-farm residue detection assays has focused on the possible
sources of false-positive results. Possible causes of false-positive results include elevated
somatic cell count (Sischo and Burns, 1993; Van Eenennaam et al, 1993); increased
lactoferrin and lysozyme concentrations (Carlsson and Bjorck, 1989); lower milk
production, increased parity and increased coliform counts (Andrew et al, 1997); and
other inhibitory substances in the milk (Tyler et al, 1992; Cullor et al, 1994). The
objective of this study was to determine whether SCC, IgGl, microbiologic isolates or
specific antimicrobial treatments were associated with false-positive results when using
the SNAP B-lactam, Penzyme Milk Test and Delvo-SP assays to test milk from cows

treated for naturally occurring mastitis.

101

MATERIALS AND METHODS

Study Design - A longitudinal experimental study of cows developing mild clinical
mastitis was conducted to evaluate the reliability of the SNAP B-lactam, Penzyme Milk
Test and Delvo-SP assays when used to test individual cow milk for antimicrobial
residues.

Case Definition - A mild clinical mastitis case was defined as 1) visibly abnormal milk
stripped from the quarter, 2) quarter may be enlarged or reddened and 3) conventional
therapy (intramammary antimicrobial infusions or other non-antimicrobial therapy) was
believed to be an appropriate treatment. Cows were specifically excluded from the study
if they 1) received antimicrobial treatment for any reason within the last 30 days, 2) were
previously included in this study (a repeat case), 3) had a concurrent illness requiring
antimicrobial treatment or 4) had a severe case of mastitis that required systemic (IV, IM
or SQ) antimicrobial therapy.

Sample Size - The initial sample size calculations (Appendix 2) indicated that 199 cows
were needed in each treatment group (400 total) to evaluate the most conservative
estimate for risk factor exposure (antimicrobial treatment group).

Treatment Groups - After diagnosing a case of mild clinical mastitis and collecting the
initial milk sample, cows were randomly assigned to one of two treatment groups. Cows
assigned to the Antimicrobial treatment group were treated as directed on the label with a
FDA approved IMM antimicrobial therapy selected by either the producer or veterinarian.
Label dose and number of treatments were followed unless the cow was removed from

the study. Cows assigned to the control treatment group received appropriate treatment

102

that did not include any form of antimicrobial therapy. Other treatments included anti-

inflammatory drugs, oxytocin, saline infusions or no drug therapy.

Sample Collection - As previously described (Gibbons-Burgener et a1, 2000 b),
pretreatment and post-treatment samples from each mastitis case were collected. Prior to
collection of the samples, foremilk from each quarter was discarded. An aseptic
collection of at least 5 ml of milk from the affected quarter was obtained for bacterial
culture using standard methods (Harmon et al, 1990). An 80-ml composite milk sample
was divided as follows: 1) 30 ml in a plastic vial with preservative tablet for infrared
somatic cell count (SCC), 2) 10 ml in a plastic vial for IgG; analysis, and 3) four 5-8 ml
aliquots in plastic vials for antimicrobial residue analyses. The aliquots were frozen at -
70°C until needed for residue analyses.

Testing for somatic cell count - Raw milk with the preservative was stored in the
refrigerator 4-48 hours prior to delivering the sample to the Michigan Dairy Health
Improvement Association laboratory. The laboratory utilizes infrared instrumentation to
determine SCC.

Quantification of bovine IgG; - A technique described by Guidry, et al. as modified
(Erskine et al, 1989) was used to analyze the milk for bovine IgGl. Proteins in the 10 ml
sample of composite milk were stabilized by the addition of 10 pl of 1 M benzamidine
hydrochloric acid and maintained at -10°C. Samples were thawed in an ice bath and
vortexed in preparation for the various steps of centrifugation and dilution (Guidry et al,

1980) that precede the application of a commercial radioimmunodiffusion assay“.

103

Diffusion diameters were measured and interpreted according to the manufacturer’s
directions.

Bacteriologic culture of milk - Frozen aseptically-collected samples were thawed and
vortexed. Sheep blood agar was inoculated with 10 pl of milk and aerobically incubated
at 37°C for up to 48 hours. When necessary, additional testing was performed to
determine the catalase status of streptococcus species and coagulase status of
staphylococcus species. Specific culturing techniques for the identification of
Mycoplasma infection were not performed.

Testing for antimicrobial residues - Performance of the SNAP b-lactam, Penzyme Milk
Test and Delvo-SP assays and high performance liquid chromatography (HPLC) have
been previously described by the authors (Gibbons-Burgener et al, 2000 a & b). Briefly,
post-treatment samples were randomly tested twice using each of the assays. Assay
results were compared to the HPLC identification and quantification of potential residues
to determine the true positive or negative status of the sample. A positive assay result
was considered false-positive when HPLC results indicated that antimicrobials were not
present at or above the assay detection limit. A false-negative was the product of a
negative assay result and the HPLC detection of an antimicrobial at or above the reported
assay detection limit.

Statistical analysis - Descriptive statistics were calculated for the false-positive, true-
positive and negative assay results. Logistic regression modeling was used to evaluate
potential associations between risk factors and false-positive results obtained with each of
the three residue assays. A false-positive assay result on one or both of the repeated tests

for each assay was the outcome for all of the models. Independent variables considered

104

in the models were somatic cell count (x 106 cells/ml), IgGl concentration (mg/ml),
negative initial bacteriologic culture (1, 0), and antimicrobial mastitis therapy (none,
pirlimycin, hetacillin or cephapirin). When multivariable modeling was possible
backward reduction techniques that considered confounding and effect modification

(Hosmer and Lemeshow, 1989) were utilized.

RESULTS

Eighty-nine cases of mild clinical mastitis were included in the analyses.
Because some assay results were uninterpretable, the SNAP B-lactam and Penzyme Milk
Test had 86 and 88 cases respectively available for analysis. Means and standard
deviations are presented for the continuous variables and frequencies for the categorical
variables in tables 6-1, 6-2 and 6-3. Specific bacteria that were isolated from the samples
were E. coli, Streptococcus spp. (catalase negative), Staphylococcus spp. (coagulase
negative), Klebsiella spp., S. aureus, C. bovis and A. pyogenes.

The univariate logistic regression results (Table 6-4) excluded those cases that had
true positive results and computed the relative risk of having a false positive result in a
sample population of known negatives. Since there were only 3 false-positive Penzyme
results, the univariate results are of questionable value and multivariable modeling was
impossible. The 6 false-positive Delvo-SP results also restricted the possibility of
multivariable modeling with data for that assay. The reduced logistic model for the

SNAP B-lactam assay retained two variables that confounded each other (Table 6-5).

105

DISCUSSION

The achieved sample size (n=89) was substantially less than the predetermined
size of approximately 400 cases. Statistical analyses were restricted by the limited
number of cows that had false-positive assay results and the relatively small sample size.
However, there was evidence that inflammatory-related milk proteins, represented by
bovine IgGl concentrations, were positively associated (Tables 6-4 and 6-5) with false-
positive results by all three residue detection assays. It is important to remember that a
cow with mastitis will experience an increased influx of many different proteins including
serum albumin, IgGl, Ing, lactoferrin and lysozyme into her milk. This study only
quantified the milk Ig61 levels, because its levels are correlated with those of serum
albumin and Ing (Erskine et al, 1989; Li-Chan and Kummer, 1997). Because of the
close association of the various inflammatory proteins and their increased presence in
milk from mastitis cases we feel confident to conclude that an inflammatory-related
protein in the milk is associated with the occurrence of false-positive results, but cannot
definitively ascribe the results to IgGl concentrations.

Each of the assays we evaluated had a different mechanism of antimicrobial
detection. The SNAP B-lactam is a receptor binding assay. The Penzyme Milk Test is an
enzymatic colourimetric assay and involves the inactivation of an enzyme by
antimicrobials. The Delvo-SP is a traditional microbial inhibition assay with a color
indicator. It is unlikely that a single inflammatory protein profoundly interferes with all 3
of the assays. Additional laboratory studies controlling the levels of each protein would

be beneficial in resolving which of the proteins are culpable in altering the results of the

106

screening assays. Until such studies are conducted it is advisable to consider the
diagnostic specificity of the assays to be compromised by inflammatory proteins.

The SNAP B-lactam assay was less likely to have a false-positive result when the
initial bacteriologic culture result was negative. This finding can also be interpreted as an
increased risk of false-positive results when bacteria have been isolated prior to treatment
of a cow with mastitis. Other studies have reported similar rates of negative cultures in
the diagnosis of clinical mastitis (Anderson et al, 1982; Smith et al, 1985; White and
Montgomery, 1987). In addition to not being present in the milk, bacteria may be present
in very low numbers not detected by standard culturing methods (Dinsmore et al, 1992).
The culture results from this study were from pre-treatment samples and the antimicrobial
detection assay results are from post-treatment samples, which raises speculation as to
whether the live bacteria were present in the post-treatment samples. The presence of
bacteria usually precedes the elevation of SCC. We found that initial culturing of bacteria
was significantly associated (p=0.005) with pre-treatment SCC, but only marginally
(p=0.087) with post-treatment SCC. The lack of association between SCC and false-
positive results indicates that another aspect of bacterial presence may need to be
considered. A possibility could be the bacterial production of an antibiotic-like substance
that is bound by the conjugated enzyme in the SNAP assay. Again, the actual mechanism
of the association is unknown, but should be considered when testing individual cow
samples for antimicrobial residues.

The authors previously suggested that the SNAP B-lactam assay may
experience an undocumented cross-reactivity with pirlimycin (Gibbons-Burgener et al,

2000 b). It was surprising that having received pirlimycin IMM therapy was not a

107

significant risk factor for false-positive results. We were also unable to confirm the
association of SCC with false-positive results reported by other studies (Sischo and
Burns, 1993; Van Eenennaam et al, 1993). There appeared to be a trend toward higher
SCCs in samples testing positive on the SNAP and Delvo assays (Table 6-1), yet the
insignificant regression model findings may be the result of large standard errors and
insufficient statistical power.

The smaller than desired sample size that was achieved and the relatively low
occurrence of false-positive results by the 3 residue detection assays prevented the
application of more extensive modeling techniques and compromised the statistical
precision and power of the simpler models. However, this study was able to demonstrate
that components in milk other than antimicrobials were associated with positive assay
results, which sheds more doubt on the use of on—farm antimicrobial residue screening

assays to test milk from individual cows.

FOOTNOTES

3 VMRD Inc., Pullman, Wash.

108

REFERENCES

Adams J B. Results of drug screening from a producer’s view. J Dairy Sci,
1994;77:1933-1935.

Anderson KL, Smith AR, Gustafsson BK, et al.. Diagnosis and treatment of acute
mastitis in a large dairy herd. J Am Vet Med Assn, 1982;181 1690-693.

Andrew SM, F robish RA, Paape MJ, Maturin LJ. Evaluation of selected antibiotic
residue screening tests for milk from individual cows and examination of factors that
affect the probability of false-positive outcomes. J Dairy Sci 1997;80:3050-3057.

Carlsson A, Bjorck L. Lactoferrin and lysozyme in milk during acute mastitis and their
inhibitory effect in Delvotest P. J Dairy Sci, 1989;72:3166-3175.

Cullor J S. Tests for identifying antibiotic residues in milk: how well do they work? Vet
Med 1992;87:1235-1241.

Cullor J S, van Eenennaam A, Gardner 1, et al.. Performance of various tests used to
screen antibiotic residues in milk samples from individual animals. J AOAC,
1994;77:862-870.

Dinsmore RP, English PB, Gonzalez RN, Sears PM. Use of augmented cultural
techniques in the diagnosis of the bacterial cause of clinical bovine mastitis. J Dairy Sci,
1992;75:2706-2712.

Erskine RJ, Eberhart RJ, Grasso PJ, Scholz RW. Induction of Escherichia coli mastitis
in cows fed selenium-deficient or selenium-supplemented diets. Am J Vet Res,
1989;50:2093-2100.

Gardener IA, Cullor J S, Galey FD, et al.. Alternatives for validation of diagnostic assays
used to detect antibiotic residues in milk. J Am Vet Med Assn 1996;209:46-52.

Gibbons-Burgener SN, Leykam J F , Kaneene JB. Identification and quantification of
ampicillin, cephapirin and pirlimycin in cows’ milk using high performance liquid
chromatography and fluorescence detection. J Vet Diagn Invest, (manuscript submitted
2000). (a)

Gibbons-Burgener SN, Kaneene J B, Leykam J F, et al.. An epidemiological evaluation of
the reliability of bulk-tank residue detection assays used to test individual cow milk. Am
J Vet Res, (manuscript submitted 2000). (b)

Guidry AJ, Paape MJ, Pearson RE. Effect of udder inflammation on milk
immunoglobulins and phagocytosis. Am J Vet Res, 1980;41:751-753.

109

Harmon RJ, Eberhart RJ, Jasper DE, et al., Microbiological procedures for the diagnosis
of bovine udder infections, ed. 3. Arlington, VA, National Mastitis Council, 1990.

Hosmer DW, Lemeshow 8. Chapters 4-7. In: Barnett V, Bradley RA, Hunter .1 S, et al,
eds. Applied logistic regression. New York: John Wiley & Sons, 1989;82-213.

Li-Chan ECY, Kummer A. Influence of standards and antibodies in immunochemical
assays for quantitation of immunoglobulin G in bovine milk. J Dairy Sci, 1997;80:1038-
1046.

Sischo WM, Burns CM. Field trial of four cowside antibiotic-residue screening tests. J
Am Vet Med Assn, 1993;202:1249-1254.

Smith KL, Todhunter DA, Shoenberger PS. Environmental mastitis: cause, prevalence,
prevention. J Dairy Sci, 1985;68:1531-1553.

Tyler J W, Cullor J S, Erskine RJ, et al.. Milk antimicrobial drug residue assay results in
cattle with experimental, endotoxin-induced mastitis. J Am Vet Med Assn,
1992;201:13781384.

Van Eenennaam AL, Cullor J S, Perani L, et al.. Evaluation of milk antibiotic residue
screening tests in cattle with naturally occurring mastitis. J Dairy Sci, 1993;76:3041-
3053.

White ME, Montgomery ME. The resemblance of clinical attributes between mastitis

cows with no growth on bacterial milk cultures and those with gram-positive bacteria
cultured. Can J Vet Res, 1987;51:181-184.

110

Table 6-1 - Means and standard deviations of somatic cell count (SCC) and milk

IgG. concentration for false-positive residue detection assay results.

 

 

Assay Result' 11 Mean std. dev.
Overall SCC 91 3396.81 2948.57
Ioglo SCC 91 7.490 1.495
IgG1 89 122.0 107.49
SNAP b-lactam SCC TN 68 2928.78 2679.28
FF 1] 4235.82 3593.65
TP 6 3735.17 3370.04
10310 SCC TN 68 7.30 1.556
FF 1 1 7.828 1.272
TP 6 7.722 1.282
IgGl TN 68 103.28 74.88
FP 10 169.70 126.17
TP 6 202.83 253.66
Penzyme Milk
Test SCC TN 79 3222.53 2903.65
FF 3 2904.0 2137.78
TP 4 3463.5 2122.42
Iogm SCC TN 79 7.411 1.530
FF 3 7.612 1.222
TP 4 7.967 0.745
IgGl TN 78 104.96 59.48
FF 3 278.67 317.26
TP 4 181.75 144.88
Delvo-SP SCC TN 66 3007.73 2938.88
FF 6 3853.0 2069.95
TP 16 4111.25 2995.62
logm SCC TN 66 7.282 1.582
FF 6 7.964 1.083
TP 16 7.907 1.150
IgG1 TN 66 96.23 38.45
FF 5 252.0 225.49
TP 15 189.93 188.22

 

 

a Results for the assay when compared to the HPLC identification and quantification
of antimicrobials. TN = true negative, FP = false positive, TP = true positive

 

111

 

Table 6-2 - Frequency distribution of antimicrobial therapy as a risk factor for
false-positive residue detection assay results.

 

 

Antimicrobial Therapy

Assay None Pirlimycin Hetacillin Cephapirin

Overall 45 26 9 9

SNAP B-lactam TN3 39 19 6 5
FF 4 5 l 1
TP 2 1 0 3

Penzyme Milk

Test TN 41 25 7 7
FP 3 0 O 0
TP 1 1 1 1

Delvo-SP TN 40 14 6 7
FF 2 3 1 0
TP 3 9 2 2

 

 

’ Results for the assay when compared to the HPLC identification and quantification of
antimicrobials. TN = true negative, FP = false positive, TP = true positive

 

112

 

Table 6-3 - Frequency distribution of bacteriologic culture results as risk factors
for false-positive residue detection assay results.

 

Bacteriologic Culture

 

Gram +
Noncontagious pleomorphic

Assay Neg. Coliform gram + cocci rods S. aureus

Overall 33 21 21 3 2

SNAP B-lactam TNa 29 l 1 16 2 2
PP 1 5 4 l 0
TP 2 3 1 0 0

Penzyme Milk

Test TN 31 15 20 3 2
FF 2 1 0 0 0
TP 0 4 O 0 0

Delvo-SP TN 26 15 14 3 1
FF 3 1 2 0 0
TP 4 5 5 0 1

 

 

a Results for the assay when compared to the HPLC identification and quantification of
antimicrobials. TN = true negative, FP = falsgositive, TP = true positive

 

113

 

Table 6-4 - Univariate logistic regression results used to estimate the relative risk

of possible milk components producing a false-positive result. All samples

included were HPLC confirmed negative for antimicrobials the specific assay was

 

 

designed to detect.
Assay Risk Factor [3" S.E.b pc ORd
SNAP B-lactam
Neg. bacterial culture ~2.269 1.081 0.036 0.103
SSC (106 cells/m1) 0.0001 0.0001 0.162 1.0
logio SCC (106 cells/m1) 0.298 0.277 0.283 1.347
Milk bovine IgG. (mg/m1) 0.0058 0.003 0.066 1.006
Antimicrobial IMM
therapy
None reference NA NA 1 .0
Pirlimycin 0.942 0.727 0.195 2.566
Hetacillin 0.486 1.201 0.686 1.625
Cephapirin 0.668 1.215 0.583 1.950
Penzyme Milk Test
Neg. bacterial culture 0.948 1.248 0.447 2.581
SSC (106 cells/ml) -0.00004 0.0002 0.866 1.0
logio SCC (106 cells/ml) 0.111 0.439 0.80 1.117
Milk bovine IgG1 (mg/m1) 0.0055 0.0027 0.0412 1.006
Delvo-SP
Neg. bacterial culture 0.208 0.858 0.809 1.231
SSC (106 cells/m1) 0.0001 0.0001 0.491 1.0
loglo SCC (10‘5 cells/ml) 0.414 0.394 0.294 1.513
Milk bovine IgGl (mg/ml) 0.021 0.009 0.021 1.021

 

 

8 Parameter estimate; b standard error of B; c p—value for the Wald x2; d Odds ratio

 

114

 

Table 6-5- Multivariable logistic model of risk factors associated with false-
positive SNAP B-lactam assay results (n=78).

 

 

Risk Factor 13" 5.18." If ou‘I (95% CI)
Intercept -2 .495 0.721 0.0005
Neg. bacteriologic culture -4.314 2.413 0.074 0.013 (0.000-1.515)

Milk bovine IgG1 (mg/ml) 0.012 0.005 0.026 1.012 0.002-1.022)

 

 

-2 Log likelihood x2 = 13.342, 2 df, p=0.0013

3 Parameter estimate; b standard error of B; c p-value for the Wald x2; d Odds ratio and
the 95% confidence interval for OR.

 

115

 

SUMMARY AND CONCLUSIONS

SUMMARY

The epidemiological research presented in this dissertation was achieved through
two main studies. The first study was a retrospective study of Michigan dairy farms
evaluating the Milk and Dairy Beef Quality Assurance Program (QAP) and its role in the
prevention of violative antimicrobial residues in milk and the adoption of prudent drug
management practices. The initial study (case-control) population consisted of case
farms having experienced a violative residue in their milk during 1993, and control farms
not having experienced a residue during the same period. In the second part (cross-
sectional) of the first study, the distinction between voluntary and involuntary
participation in the Program was included in the analyses. 1993 was the first full year of
QAP participation in Michigan and offered a unique opportunity to evaluate the initial
impact of the Program. Though the temporal relationships between some management
practices and residue occurrence and QAP participation were unclear, the overall study
represents one of the first scientific evaluations of the QAP.

The use of on-farm antimicrobial screening assays has consistently emerged as a
management tool believed to aid the prevention of residues. The second study, therefore,
was a longitudinal experimental study evaluating the reliability of 3 on—farm assays when
used to test individual cow milk. The development of novel biochemistry methods for
gold standard antimicrobial detection was necessary to the completion of the second
study. The study population for the experimental study consisted of cows diagnosed with

mild clinical mastitis. The cases were randomly assigned to either the antimicrobial or

116

control treatment group. The study was designed to simulate the collection and testing of
milk samples as commonly performed on farms. The reliability of 3 on-farm residue
screening assays was expressed as sensitivity, specificity, positive predictive value and
negative predictive value. Additional statistical analyses were used to determine whether
somatic cell count, IgG 1, bacterial isolates or specific antimicrobial treatments were

associated with false-positive results when using the 3 assays to test individual cow milk.

CONCLUSIONS

There were six objectives addressed by the two main studies of this dissertation.
The first objective was to determine if QAP certification was associated with a reduced
risk of having antimicrobial residues in milk. Certification in the QAP was associated
with a tendency toward reduced risk (OR=O.3 [0.07-1.32]) of having experienced a
violative residue in bulk-tank milk. Though it was not statistically significant, the
potential protective effect of the QAP was encouraging. Additional research may
demonstrate a stronger association if the distinction between conscientious and less
conscientious QAP participants is made.

The second objective was closely related to the first and was to define specific
management factors that may have predisposed dairy farms to having violative
antimicrobial residues in milk. The risk of having had a violative residue was reduced on
farms treating >10% of their herd for metritis, and having their milk processor perform
residue testing. However, on-farm residue testing was associated with an increased risk
of having had a residue. A positive association between maintaining written

identification records of treated cows and having a violative residue was identified, but

117

this finding probably indicates a change in management implemented afier notification of
having a violative residue. Specific risk factors associated with having a violative residue
are addressed by various critical control points in the QAP and may be indicators for
some of the program’s strengths and weaknesses.

In a separate set of analyses the associations between QAP certification and the
use of prudent drug management practices were evaluated (Objective 3). Results
suggested that farms adopted specific management practices, irrespective of certification.
Large percentages of farms used visible identification and non-emergency veterinary
services and discussed residue prevention with employees. Involuntary certification was
associated with maintenance of good written treatment records and performance of on-
farm drug residue testing. Voluntary certification was weakly associated with use of
refrigerated drug storage. QAP certification appeared to have been associated with the
adoption of only a few prudent drug use practices, although QAP materials and
framework were developed to assist veterinarians in the promotion of disease prevention,
client communication, and residue prevention practices on farms.

Results from the first main study evaluating the QAP indicated strong
associations between the use of on-farm residue testing and violative residue occurrence
and involuntary QAP certification. These were strong indications that farms were
adopting on-farm individual cow testing to avoid additional violative residues. This
unapproved and unvalidated use of antimicrobial screening assays led to the second main
study designed to determine the reliability of three commonly used on—farm commercial

assays when testing individual cow milk.

118

The residue assay study was dependent on the gold standard methods for
detection and quantification of antimicrobials. The fourth objective was to develop
robust gold standard methods for use in determining the reliability of the Delvo-SP,
Penzyme Milk Test and SNAP B-lactam assays in the detection of ampicillin, cephapirin
and pirlimycin in raw milk. Previously described biochemistry methods were modified
to better accommodate the blinded evaluation of milk samples collected as part of the
field trial. The methods developed should improve the high performance liquid
chromatography (HPLC) analyses used to detect ampicillin and pirlimycin in milk.

The fifth objective was to determine the reliability of the Delvo-SP, Penzyme
Farm Milk Test and SNAP B-lactam residue test assays when used to test individual cow
milk. Comparing the post-treatment assay results to the HPLC results, the Delvo-SP and
SNAP B-lactam assays displayed >90% sensitivity, while the sensitivity of the Penzyme
Milk Test was only 60%. Ranging from 32.14 to 73.68%, the positive predictive values
were poor for all three assays when using the assays’ detection limits. The positive
predictive values were even less when the FDA tolerance levels were used as the cut-offs.
The poor positive predictive value of the SNAP B-lactam assay was most likely due to an
undocumented cross-reactivity with pirlimycin residues. The kappa statistics provided
strong evidence that all three assays produced good to excellent repeatability. With such
low positive predictive values and incidence of violative antimicrobial levels, the
usefulness of the three residue detection assays in deciding the fate of milk from cows
receiving treatment for mastitis is highly questionable.

Another objective of the residue assay study (overall Objective 6) was to ascertain

possible associations between specific milk components and false-positive assay results.

119

Inflammatory-related milk proteins, represented by bovine IgG1 concentrations, were
positively associated with false-positive results from the SNAP B-lactam, Penzyme Milk
Test and Delvo-SP assays. The SNAP B-lactam assay was less likely to have a false-
positive result when the initial bacteriologic culture result was negative. Low numbers of
false-positive results and a relatively small sample size compromised the statistical power
of the study and warrants caution in over-interpreting the apparent associations.

Even with the possible limitations, both studies make significant contributions to
the epidemiology of antimicrobial residues in milk. The tendency of the QAP to prevent
violative residues provides encouraging information for the continued promotion and
implementation of the Program. Dairy producers and veterinarians can use the findings
to target their residue prevention efforts. To reduce the likelihood of recall bias and allow
for the better establishment of temporal relationships, questions regarding QAP
participation and management practices should be administered at the initiation of a
residue inquiry. Producers may reconsider their reliance on screening assays for testing
individual cows’ milk on-farm as a primary tool for residue prevention. Researchers and
regulatory personnel may use these findings to improve the accurate detection and
investigation of violative residues. Future studies evaluating the reliability of on-farm
residue assays should include further assessment of inflammatory-related milk proteins

that may be interfering or cross-reacting with the assays.

120

APPENDIX 1

Dairy Production Management Study

If you would prefer to answer the questions during a telephone conversation, please mark the box below
and provide your phone number so we can contact you. Return the entire questionnaire in the enclosed
envelope and we will contact you soon.

I would prefer to be contacted by telephone
My telephone number is
The best time to call is

 

 

1. Please provide the following information on your farm's principal operator:

a. Age _ (yrs.)

b. Education level

Did not complete high school
High school diploma

Some college

Technical school graduate
Bachelor's degree

Some graduate school
Graduate degree

 

lllll

 

2. Have you completed the Milk and Dairy Beef Quality Assurance Program? Yes or No

If yes. when were you initially certified?
If you have been recertilied, when?

 

 

Please use the period January 1993 - December 1993 to answer the following
quesfions.

3. How many workers did you employ full-time?
How many workers did you employ part-time?

 

4. How many people on your farm were allowed to administer drugs with withdrawal time
requirements to your dairy herd?
5. Which of the following management tool(s) did your farm utilize in 1993? (check all that apply)
DHlA

 

Dairy Comp 305 or other individual farm computer package
Daily milk weights for each cow

 

 

 

Other (specify)
None
6. What was your average herd size (lactating plus dry cows) during the period from January 1.
1993 through December 31. 1993? cows
7. What was the predominant breed of dairy cows on your farm?
8. What was your herd's rolling average (average milk production per cow per year) from January
1993 - December 1993? lbs per cow per year

 

121

10.

11.

12.

13.

14.

15.

Which of the following best describes the milking operation of your farm (check only one)?

Milking Parlor
Stanchion/Comfort Stalls with Milk Pipeline

Stanchion/Comfort Stalls with Bucket Milker
Other (describe):

 

 

Please describe the protocol used to milk treated cows. (circle Yes 9_r No after every part, a-d.)

a Were treated cows milked last? Yes or No

b. Was a different milker claw used when milking treated cows? Yes or No

c. Was a milk pipeline (with milk diverted at end) utilized when milking treated cows? Yes or No
d. Was a special Bucket used when milking treated cows? Yes or No

What veterinary services did you use during 1993. Please check the one best answer:

Emergency, problem. and sick cow work only
Emergencies plus regularly scheduled visits
Regularly scheduled visits only

Other - please specify:

a.
b.
c.
d
If you had regularly scheduled veterinary visits, how often did they occur? Please check the one
best answer:

 

 

a Every other month

b. Once per month

c. Twice per month

d. Weekly

9. Other - please specify:

 

 

Approximately how many cases of the following diseases were treated each month (or year) in
your adult herd?

Mastitis cases/month or _ cases/yr

Retained placenta _ cases/month or _ cases/yr

Metritis (uterine hfection) _ cases/month or _ cases/yr

Respiratory disease _ cases/month or __ cases/yr

Lameness _ cases/month or _ cases/yr

Digestive problems (eg. Hardware _ cases/month or __ cases/yr
dianhea, DA)

Did the above case numbers come from your farm records? Yes or No

Please indicate which of the following preventative procedures were included in your herd health
management (check all that apply):

Dry-cow intramammary treatment for evegy cow ‘
Dry-cow intramammary treatment only for selected individual cows
Vaccination program [other than Brucellosis (BANGS) vaccination]
All cows routinely given a magnet

Use of afoot bath by all cows

All cows routinely dewormed

122

16.

17.

18.

19.

21.

How did you identify lactating cows treated with antibiotics? Please check all that apply and circle
the primary type of identification used.

Leg band

Paint

Tail tape

Special tags

Other - please specify:

 

Which of the following records did you maintain for treatments administered to cows? (check all
that apply)

Reason for treatment (disease or condition)

Type of drug used

Dosage given

ID or name of cow(s) treated

Date the treatment was given

Quarter(s) treated (if appropriate)

No records were kept

 

 

If you kept records, where were they kept? (check all that apply)

In milking parlor __ In StanchionlComfort Stall Area
In drug storage area _ In bam where cows are housed
Other location:

 

Did the people that milk your cows have access to these records? Yes or No

How did you determine how long to withhold a cow’s milk after treatment? Please check all of
the following that apply and circle your primary source of information.

Past experience
Drug residue testing
Information from other producers
Read the drug's label
Ask the veterinarian
Other - please specify:

 

Which of the following best describes your use of available residue testing for milk in 1993?
(check all that apply)

Bulk Tank Testing was available and used routinely on the farm

Individual Cow Testing was available and used on the farm

Testing was performed by milk handler or off the farm

No optional residue testing was done on milk

Other - please specify:

 

 

 

Which of the following describe how you stored drugs (check all that apply)?

In locked cabinets

In unlocked closed cabinets

On non~enclosed shelves or table
In a locked refrigerator

In an unlocked refrigerator

Other (describe):

 

123

 

 

“Please use the following definitions when answering question 23.

Over the Counter (OTC) Drugs are those drugs that can be purchased anywhere without a
veterinarian's prescription or supervision.

Prescription (Rx) Drugs are those drugs that require a veterinarian's prescription and supervision.
(For example: Lutalyse, Naxcel, Oxytocin, Gentocin, Quartermaster, Dari-clox and others.)

 

23. What was your m‘mary source of OTC and Rx drugs for your dairy herd? Please check the one
best answer in each column:

OTC Drugs Rx Drugs
a. Veterinarian

b. Local feed or livestock supply store
c. Mail-order catalog
(1 Other - please specify:

 

 

 

“Please use the following definition when answering question 24.

Extra-label or Off-label Use is the use of a drug in a manner that is different than what the
manufacturer's label specifies. [For example: administering a higher dose of penicillin than what the
manufacturer recommends, or administering a drug to a lactating cow that the manufacturer states is only

approved for non—lactating cattle]

 

24. In your herd, approximately how many cows were treated with drugs used in an off-label manner
per month? cows/month

25. Did you discuss ways of avoiding drug residues in milk with your employees during 1993?
Yes or No

If you did, when did you discuss residue avoidance with employees?

When employees were new
Routinely - times a year
When problems occurred

Other - please describe:

 

Please add any comments you may have regarding dairy management or the Dairy Quality Assurance
Program:

 

 

 

 

 

124

 

 

APPENDIX 2

Sample Size Calculations

A. Sample size required per group when using the 2 statistic to compare proportions of
dichotomous variables.

 

(Z. 220—2)+Zilp.(1-P.)+p.(1-p.))’

(p. - p.12

 

Assumptions:
(1 = 0.05, two-tailed test
B = 0.20 (80% power)

pc = 0.20 (proportion of samples from non-exposed cows that produce false-
positive results)
pi = 0.10 (proportion of samples from exposed (antibiotic treatment) cows

that produce false-positive results)
1’) = 0.15

 

(1.96./2(.15)(.85) + .84.[2(.8) + .1(.9))2 _198 7 _
(.2-.1)’ — ' 7

199 each group

Adjustments to sample size:

The sample size will be increased by approximately 15% to adjust for the possibilities of
l) a cow suffering a second disease problem between the pre-treatment and post-
withholding period sample collection, or 2) a cow dying or being culled prior to
collection of post-withholding period sample.

Adjusted sample size = 460 230 in antimicrobial treatment group
230 in non-antimicrobial treatment group

125

Sample Size Calculations
B. Total sample size required when estimating a single proportion (Kelsey et al, 1996).
n=Z2 p (1-p)/L2
Assumptions:
Z = 1.96 (or = 0.05)

p = estimated proportion of population having a particular exposure
L = margin of error of estimated proportion

 

 

 

 

p L n
0.01 0.01 380
0.02 95
0.05 15
0.02 0.01 753
0.02 188
0.05 30
0.05 0.01 1824
0.02 456
0.05 73
0.1 0.01 3456
0.02 864
0.05 138
0.9 0.05 138
0.95 0.05 73

 

126

REFERENCES

Adams JB. Assuring a residue-free food supply: Milk. J Am Vet Med Assn
1993;202:1723-1725.

Adams J B. Results of drug screening from a producer’s view. J Dairy Sci,
1994;77:1933-1935.

Adulterated dairy products, Minn. Stat. 32.21, Subd. 14. Minnesota statues v01. 1. St
Paul, Minn: Revisor of Statues, State of Minnesota, 1998;920-922.

Albright J L, Tuckey SL, Woods GT. Antibiotics in milk - a review. J Dairy Sci,
1961;44:779-807.

American Veterinary Medical Association and National Milk Producers Federation. Milk
and Dairy Beef Residue Prevention: a quality assurance program. J Am Vet Med Assoc
1991;199 (Suppl):1-23.

Anderson KL, Moats WA, Rushing J E, O’Carroll J M. Detection of milk antibiotic
residues by use of screening tests and liquid chromatography after antramammary
administration of amoxicillin or penicillin G in cows with clinical mastitis. Am J Vet Res
1998;59:1096-1100.

Anderson KL, Smith AR, Gustafsson BK, et al.. Diagnosis and treatment of acute
mastitis in a large dairy herd. J Am Vet Med Assn, 1982;181:690-693.

Andrew SM, Frobish RA, Paape MJ, Maturin LJ. Evaluation of selected antibiotic
residue screening tests for milk from individual cows and examination of factors that
affect the probability of false-positive outcomes. J Dairy Sci 1997;80:3050-3057.

Ang CYW, Luo W, Call VL, Righter HF. Comparison of liquid chromatography with
microbial inhibition assay for detection of incurred amoxicillin and ampicillin residues in
milk. J Agric Food Chem, 1997;45:4351-4356.

Arrioj a-Dechert A. Compendium of veterinary products. 4th ed. Port Huron, MI: North
American Compendiums Inc, 1997;109-1310.

Boeckman S, Carlson KR. Milk and dairy beef quality assurance program: milk and
dairy beef residue prevention protocol. Stratford, IA: Agri-Education, Inc, 1997;1-70.

Brady MS, White N, Katz SE. Resistance development potential of
antibiotic/antimicrobial residue levels designated as "safe levels.” J Food Prot,
1993;56:229-233.

Briguglio GT, Lau-Cam CA. Separation and identification of nine penicillins by reverse
phase liquid chromatography. J AOAC Int 1984;67:228-231.

127

Bruhn CM. Consumer Perceptions and concerns about veterinary drug residues. In:
Moats WA and Medina MB, ed. Veterinary Drug Residues: Food Safety. Washington,
DC, American Chemical Society, 1996, pp 18-21.

Carlsson A, Bjorck L. Lactoferrin and lysozyme in milk during acute mastitis and their
inhibitory effect in Delvotest P. J Dairy Sci, 1989;72:3166-3175.

Center for Veterinary Medicine. Appendix N, Pasteurized Milk Ordinance. In C VM
Update: Milk monitor with antimicro drug screening test. 1996, pp 5-6.

Chou TY, Gao CX, Grinberg N, Krull IS. Chiral polymeric reagents for off-line and on-
line derivatizations of enantiomers in high-performance liquid chromatography with
ultraviolet and fluorescence detection: an enantiomer recognition approach. Anal Chem
1989;61:1548-1558.

Cullor J S. Tests for identifying antibiotic residues in milk: how well do they work? Vet
Med 1992;87:1235-1241.

Cullor J S. The control, treatment, and prevention of the various types of bovine mastitis.
Vet Med 1993;88:571-579.

Cullor J S. Dilemmas associated with antibiotic residue testing in milk. In: Moats WA
and Medina MB, ed. Veterinary Drug Residues: Food Safety. Washington, DC,
American Chemical Society, 1996, pp 44-57.

Cullor J S, Van Eenennaam A, Gardner 1, et al.. Performance of various tests used to
screen antibiotic residues in milk samples from individual animals. J AOAC Int
1994;77:862—870.

Dasenbrock CO, LaCourse WR. Assay for cephapirin and ampicillin in raw milk by
high-performance liquid chromatography - integrated pulsed amperometric detection.
Anal Chem, 1998;70:2415-2420.

Day J. A subcommittee review of the quality assurance initiative: implementation issues
from the Implementation and Communication Subcommittee of the Drug Residue
Committee, in Proceedings. 32nd Annu Meet National Mastitis Council 1993;144-146.

Dinsmore RP, English PB, Gonzalez RN, Sears PM. Use of augmented cultural
techniques in the diagnosis of the bacterial cause of clinical bovine mastitis. J Dairy Sci,
1992;75:2706-2712.

Engel RE. Current food safety and quality service residue control program. J Am Vet
Med Assn 1980;176:1145-1147.

128

Erskine RJ, Eberhart RJ, Grasso PJ, Scholz RW. Induction of Escherichia coli mastitis in
cows fed selenium-deficient or selenium-supplemented diets. Am J Vet Res
1989;50:2093-2100.

Fluid Milk Act of 1965, RA. 1965, No. 233, as amended by RA. 1993, No.5. Michigan
Compiled Laws Annotated 1996 Sections 286.1 to 288. St Paul, Minn: West Publishing
Co, 1996; 417-419.

Franco DA, Webb J, Taylor CE. Antibiotic and sulfonamide residues in meat:
implications for human health. J Food Prot, 1990;53:178-185.

Gardener IA, Cullor J S, Galey FD, et al.. Alternatives for validation of diagnostic assays
used to detect antibiotic residues in milk. J Am Vet Med Assn 1996;209:46-52.

Gardner IA, Hird DW, Guterback WW, et a1. Mortality, morbidity, case-fatality, and
culling rates for California dairy cattle as evaluated by the National Animal Health
Monitoring System, 1986-87. Prev Vet Med 1990;8zlS7-170.

Gibbons-Burgener SN, Kaneene J B, Lloyd J W, Erskine RJ. Influence of the Milk and
Dairy Beef Quality Assurance Program on dairy farm drug management practices. J Am
Vet Med Assn 2000;216:1960-1964.

Gibbons-Burgener SN, Kaneene J B, Leykam JF, et al.. An epidemiological evaluation of
the reliability of bulk-tank residue detection assays used to test individual cow milk. Am
J Vet Res, (manuscript submitted 2000). (b)

Gibbons-Burgener SN, Kaneene J B, Lloyd J W, et a1. Associations of milk and dairy beef
quality assurance program certification and related risk factors with the occurrence of
antibiotic residues in Michigan milk. Am J Vet Res 1999:60;1312-1316.

Gibbons-Burgener SN, Leykam J F , Kaneene J B. Identification and quantification of
ampicillin, cephapirin and pirlimycin in cows’ milk using high performance liquid
chromatography and fluorescence detection. J Vet Diagn Invest, (manuscript submitted
2000). (a)

Guidry AJ, Paape MJ, Pearson RE. Effect of udder inflammation on milk
immunoglobulins and phagocytosis. Am J Vet Res, 1980;41:751-753.

Hady PJ, Lloyd J W, Kaneene JB. Antibacterial use in lactating dairy cattle. J Am Vet
Med Assoc 1993;203:210-220.

Halbert LW, Erskine RJ, Bartlett PC, Johnson GL. Incidence of false-positive results for
assays used to detect antibiotics in milk. J Food Prot 1996;59:886-888.

Hardin JW. Bivariate probit models. Stata Technical Bulletin 1996;33:15-20.

129

Harik-Khan R, Moats WA. Identification and measurement of B-lactam antibiotic

residues in milk: integration of screening kits with liquid chromatography. J AOAC Int.
1995;78:978-986.

Harmon RJ, Eberhart RJ, Jasper DE, et al.. Microbiological procedures for the diagnosis
of bovine udder infections, ed 3. Arlington, VA, National Mastitis Council, 1990.

Heller, DN. Determination and confirmation of pirlimycin residue in bovine milk and
liver by liquid chromatography/thermospray mass spectrometry: interlaboratory study. J
AOAC Int, 1996;79:1054-1061.

Heller, DN. Determination of Pirlimycin residue in milk by liquid chromatographic
analysis of the 9-fluorenylmethyl chloroformate derivative. J A OAC Int, 1997;80:975-
981.

Heller DN, Ngoh MA. Electrospray ionization and tandem ion trap mass spectrometry

for the confirmation of seven B-lactam antibiotics in bovine milk. Rapid Commun Mass
Spectrom, 1998;12:2031-2040.

Hong CC, Lin CL, Tsai CE, Kondo F. Simultaneous identification and determination of
residual penicillins by use of hi gh-perforrnance liquid chromatography with
spectrophotometric or fluorometric detectors. Am J Vet Res, 1995;56:297-303.

Homish RE, Arnold TS, Bacynskyj L, et al.. Pirlimycin in the dairy cow. In: Hudson
DH, ed. Xenobiotics in Food producing animals - metabolism and residues, ACS
symposium series 503, American Chemical Society, Washington, DC, 1992; pp. 132-
147.

Homish RE, Cazars AR, Chester ST, Roof RD. Identification and determination of
pirlimycin residue in bovine milk and liver by high-performance liquid chromatography-
thermospray mass spectrometry. J Chromatogr B, 1995;674:219-235.

Hosmer DW, Lemeshow S. Chapters 4-7. In: Barnett V, Bradley RA, Hunter J S, et al,
eds. Applied logistic regression. New York: John Wiley & Sons, 1989;82-213.

Huber WG. Allergenicity of antibacterial drug residues. In: Rico AG, ed., Drug residues
in animals. Academic Press, Toronto, ONT 1986, pp 33-49.

Ingersoll B. Milk is found tainted with a range of drugs farmers give cattle. The Wall
Street Journal, December 29, 1989; Al, A2.

Ingersoll B. New York milk supply highly tainted, TV station says, based on own survey.
The Wall Street Journal, February 8, 1990; A4.

Jones GM and Seymour EH. Cowside antibiotic residue testing. J Dairy Sci,
1988;71:1691-1699.

130

Kaneene J B and Ahl AS. Drug residues in dairy cattle industry: Epidemiological
evaluation of factors influencing their occurrence. J Dairy Sci, 1987;70:2176-2180.

Kaneene J B, Coe PH, Smith J H, et a1. Drug residues in milk afier intrauterine injection of
oxytetracycline, lincomycin-spectinomycin, and povidone-iodine in cows with metritis.
Am J Vet Res 1986;47:1363-1365.

Kaneene J B, Hurd HS. National Animal Health Monitoring System Round l—Final
Report. East Lansing, Mich: Michigan State University, 1988;63-74.

Kaneene J B, Hurd HS, Miller R, et a1. National Animal Health Monitoring System
Round 2—Final Report. East Lansing, Mich: Michigan State University, 1990;26-34.

Kaneene J B, Miller R. Epidemiological study of metritis in Michigan dairy cattle. Vet Res
1994;25:253-257.

Kaneene J B, Miller RA. Description and evaluation of the influence of veterinary
presence on the use of antibiotics and sulfonamides in dairy herds. J Am Vet Med Assn
1992;201:68-76.

Kaneene J B, Willeberg P. Influence of management factors in the occurrence of
antibiotic residues in milk: a case-control study of Michigan dairy herds, with examples
of suspected information bias. Acta Vet Scand, 1989;84:473-476.

Kelsey J L, Whittemore AS, Evans AS, Thompson WD. Methods of sampling and
estimation of sample size. Ianethods in observational epidemiology ed. New York,
NY: Oxford University Press, 1996:337-338.

Kindred TP, Hubbert WT. Residue prevention strategies in the United States. J Am Vet
Med Assn, 1993 ;202:46-49.

Kleinbaum DG, Kupper LL, Muller KE. Confounding and interaction inregression. In:
Payne M, ed. Applied regression analysis and other multivariable methods. Belmont,
Calif: Duxbury Press, 1988;163-180.

Li-Chan ECY, Kummer A. Influence of standards and antibodies in immunochemical
assays for quantitation of immunoglobulin G in bovine milk. J Dairy Sci, 1997;80:1038-
1046.

McEwen SA, Black WD, Meek AH. Antibiotic residue prevention methods, farm

management, and occurrence of antibiotic residues in milk. J Dairy Sci, 1991;74:2128-
2137. (a)

131

McEwen SA, Meek AH, Black WD. A dairy farm survey of antibiotic treatment
practices, residue control methods and associations with inhibitors in milk. J Food
Protect, 1991;54:454-459. (b)

Mitchell JM, Griffiths MW, McEwen SA, et al.. Antimicrobial drug residues in milk and
meat: causes, concerns, prevalence, regulations, tests, and test performance. J Food Prot,
1998;61:742-7568.

Miyazaki K, Ohtani K, Sunada K, Arita T. Determination of ampicillin, amoxicillin,
cephalexin, and cephradine in plasma by high-performance liquid chromatography using
fluorometric detection J Chromatography 1983;276:478-482.

Moats, WA. Determination of cephapirin and desacetylcephapirin in milk using
automated liquid chromatographic cleanup and ion-pairing liquid chromatography. J
AOAC Int, 1993;76:535-539.

Moats, WA. Determination of ampicillin and amoxicillin in milk with an automated
liquid chromatographic cleanup. J AOAC Int, 1994;77:41-45.

Moats WA, Harik-Khan. Liquid chromatographic determination of B-lactam antibiotics
in milk: a multiresidue approach. J AOAC Int, 1995;78:49-54.

Moats, WA, Romanowski, RD. Multiresidue determination of b-lactam antibiotics in
milk and tissues with the aid of high-performance liquid chromatographic fiactionation
for clean up. J Chromatogr A, 1998;812:237-247.

Musser J MB, Anderson KL. Using drug residue screening tests to investigate antibiotic
contamination of milk. Vet Med, 1999;94:474-479.

Schwartz DP, Lightfield AR. Practical screening procedures for sulfarnethazine and N4-
acetylsulfamethazine in milk at low parts-per-billion levels. J A OAC Int, 1995;78:967-
970.

Sischo WM. Quality milk and tests for antibiotic residues. J Dairy Sci, 1996;79:1065-
1073.

Sischo WM, Burns CM. Field trial of four cowside antibiotic-residue screening tests. J
Am Vet Med Assn, 1993;202:1249-1254.

Sischo WM, Kieman NE, Burns CM, et a1. Implementing a quality assurance program
using a risk assessment tool on dairy operations. J Dairy Sci, 1996;80:777-787.

Slenning BD, Gardner IA. Economic evaluation of risks to producers who use milk
residue testing programs. J Am Vet Med Assn, 1997;211:419-427.

132

Smedley MD. Liquid chromatographic determination of multiple sulfonamide residues
in bovine milk: collaborative study. J AOAC Int, 1994;77:1112-1122.

Smith KL, Todhunter DA, Shoenberger PS. Environmental mastitis: cause, prevalence,
prevention. J Dairy Sci, 1985;68:1531-1553.

Stoltz EI, Hankinson DJ. Antibiotics and lactic acid starter cultures. Appl Microbiol,
1953;] :21—29.

Straub R, Linder M, Voyksner RD. Determination of B-lactam residues in milk using
perfusive-particle liquid chromatography combined with ultrasonic nebulization
electrospray mass spectrometry. Anal Chem, 1994;66:3651-3658.

Tillison J. Consumer research review, edited by the National Dairy Promotion and
Research Board. Animal Health & Milk Quality - A Situation Analysis. Jefferson, Wis:
Morgan & Myers, 1991;10-11.

Tyczkowska KL, Voyksner RD, Straub RF, Aronson AL. Simultaneous multiresidue
analysis of B-lactam antibiotics in bovine milk by liquid chromatography with ultraviolet

detection and confirmation by electrospray mass spectrometry. J AOAC Int,
1994;77:1122-1131.

Tyler J W, Cullor J S, Erskine RJ, et al. Milk antimicrobial residue assay results in cattle
with experimental, endotoxin-induced mastitis. J Am Vet Med Assn, 1992;201:1378-
1384.

USDA-APHIS-VS. Part 111: Reference of 1996 diary health and health management
National Animal Health Monitoring System. In: NAHMS Dairy ’96. Washington, DC:
USDA, 1996;17.

Van Eeckhout NJ, Van Peteghem CH, Helbo VC, et al.. New database on hormone and
veterinary drug residue determination in animal products. Analyst, 1998;123:2423-2427.

Van Eenennaam AL, Cullor J S, Perani L, et al.. Evaluation of milk antibiotic residue
screening tests in cattle with naturally occurring mastitis. J Dairy Sci 1993;76:3041-
3053.

Waltner-Toews D, McEwen SA. Residues of antibacterial and antiparasitic drugs in
foods of animal origin: a risk assessment. Prev Vet Med 1994;20:219-234.

White CR, Moats WA, Kotula KL. Optimization of a liquid chromatographic method for

determination of oxytetracycline, tetracycline and chlortetracycline in milk. J AOAC Int,
1993;76:549-554.

133

White ME, Montgomery ME. The resemblance of clinical attributes between mastitis

cows with no growth on bacterial milk cultures and those with gram-positive bacteria
cultured. CanJ Vet Res, 1987;51:181-184.

134

"llllllllllllllllllll