J. Dairy Sci. 89:4229–4236
© American Dairy Science Association, 2006.
Risk Factors Associated with Cryptosporidium Infection on Dairy Farms
in a New York State Watershed
S. R. Starkey, K. R. Kimber, S. E. Wade, S. L. Schaaf, M. E. White, and H. O. Mohammed1
Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853
ABSTRACT
A cross-sectional study was carried out to determine
the prevalence of Cryptosporidium parvum-like oocyst
shedding on dairy farms in a watershed in New York
State and to identify the factors that put animals at
risk. A proportional sample of dairy herds in the targeted area was obtained, and animals were selected
using a stratified sampling design to ensure representation of the population at risk. Fecal samples were collected per rectum and analyzed for the presence of C.
parvum-like oocysts using the quantitative centrifugation concentration flotation technique and a proprietary
enzyme-linked immunoassay. Additionally, isolates of
Cryptosporidium were examined via bidirectional DNA
sequencing. Data on putative risk factors were collected
at the time of sampling and analyzed for association
using logistic regression. The herd prevalence was 42%
and the overall animal prevalence was 3.2%. The prevalence among animals less than 60 d of age was 20%.
The likelihood of shedding Cryptosporidium decreased
with the age of the animal and varied with the type of
barn water source. Both the number of unweaned calves
present at the time of the study, and whether the calves
were tied vs. not tied increased the risk of infection.
There was significant agreement between the flotation and PCR techniques. Sequencing revealed that
50% of the isolates were Cryptosporidium bovis, an isolate thought to be nonzoonotic.
Key words: prevalence, cross-sectional, Cryptosporidium, risk factor
INTRODUCTION
Since members of the genus Cryptosporidium were
first found to infect humans and animals over 30 yr
ago, extensive research has been conducted on these
protozoa (Panciera et al., 1971; Navin and Juranek,
1984). Epidemiologic studies have shown that the majority of human disease attributable to members of the
Received January 27, 2006.
Accepted May 8, 2006.
1
Corresponding author: hom1@cornell.edu
genus results from the species Cryptosporidium parvum and Cryptosporidium hominis (Xiao and Ryan,
2004). Although C. parvum has been isolated from multiple mammalian species, including cattle, C. hominis
is believed to be an exclusively human pathogen
(Hunter and Thompson, 2005). Infections in humans
typically cause a self-limiting disease of 7 to 10 d duration, characterized by diarrhea with associated vomiting and abdominal discomfort. However, among
young, elderly, and immunocompromised individuals,
the condition can become severe and potentially life
threatening (Current et al., 1983; Nannini and Okhuysen 2002; Tumwine et al., 2005). The introduction of
antiretroviral therapy has reduced the impact of Cryptosporidium among many HIV patients in developed
nations (Maggi et al., 2000).
Transmission of Cryptosporidium species takes place
via the fecal–oral route, and susceptible hosts contract
the infection by ingesting the infective form of the parasite, the sporulated oocyst. Human cryptosporidiosis is
most often associated with the consumption of contaminated drinking water or exposure to contaminated recreational water, such as public swimming pools. A small
number of cases have also been attributed to contaminated foods such as salads (Fayer et al., 2000).
In cattle, C. parvum typically causes disease in calves
and has been identified as one of the primary etiologic
agents of neonatal calf diarrhea (Naciri et al., 1999).
Herd-level prevalence of C. parvum has been reported
to range from 13 to 100% (Wade et al., 2000; Santin et
al., 2004), although a portion of this variability is likely
attributable to study design and the geographical region under investigation. A new species, Cryptosporidium bovis, has recently been described based on both
genomic and host range differences (Fayer et al., 2005).
To date, this organism has been identified only in cattle
and is thus considered unlikely to be zoonotic. Cryptosporidium bovis is indistinguishable from C. parvum
using traditional diagnostic methods such as flotation
and ELISA techniques; thus, the term “C. parvum-like”
has been used to describe infections identified using
such methods (Starkey et al., 2005).
Research efforts were increased after Cryptosporidium species were found to be responsible for several
4229
4230
STARKEY ET AL.
large waterborne disease outbreaks in North America,
the United Kingdom, and Europe (Patel et al., 1998;
Ong et al., 1999; Glaberman et al., 2002). The largest
such outbreak occurred in Milwaukee in 1993 (Mac
Kenzie et al., 1994). In their 1994 paper, Mac Kenzie
et al. speculated that the source of Cryptosporidium
oocysts may have been contamination of the body of
water from which a failed water treatment plant drew
its intake—Milwaukee harbor—by cattle herds or
slaughterhouses located along 2 rivers leading to the
harbor. In addition to animal sources, the authors also
commented on the possible role of human sewage and
mixed-source runoff attributable to spring rains and
thawing snow. Subsequent molecular work indicated
that C. hominis was responsible for the Milwaukee outbreak (Sulaiman et al., 2001). This finding essentially
removed cattle from consideration as the source of the
outbreak because C. hominis has never been isolated
from cattle, and it is considered an anthroponotic pathogen (Morgan-Ryan et al., 2002). However, C. parvum
has been identified at the molecular level in a number
of smaller waterborne outbreaks (Ong et al., 1999; Glaberman et al., 2002). This finding, along with the fact
that cattle are quite common within watersheds in the
United States, has led to further speculation about their
role in the contamination of drinking water supplies
(Xiao and Ryan, 2004). Furthermore, human dose–response studies have shown that C. parvum of bovine
origin can have an infectious dose 50 (ID50) as low as
87 oocysts (Okhuysen et al., 1999). Therefore, there is
little doubt that cattle pose a potential, albeit imperfectly quantified, risk to human health with regard to
C. parvum. An additional source of this imperfect risk
quantification stems from the relative lack of information about the prevalence of the nonzoonotic C. bovis,
especially among cattle populations within watersheds.
Dairy farms represent a significant industry within
New York State, with cash receipts from milk totaling
over $1.56 billion in 2002, more than half of all agricultural receipts collected that year (New York State Department of Agriculture and Markets, 2006). Such operations are often located within the prime agricultural
lands that form watersheds. Academic institutions as
well as state and local government agencies have conducted extensive research on the dynamics of infection
of Cryptosporidium in dairy cattle as a means to better
understand the level of risk these animals pose to the
environment. By way of continuation of such research
activities, the present study aimed to determine the
prevalence of C. parvum in the Upper Susquehanna
Watershed and investigate the potential risk factors
associated with the prevalence of C. parvum (as identified by traditional methods) in the target population.
Additionally, the potential role of cattle in the risk to
Journal of Dairy Science Vol. 89 No. 11, 2006
water quality was examined by differentiating between
zoonotic and nonzoonotic isolates using DNA sequencing.
MATERIALS AND METHODS
Study Design
The target population consisted of all cattle on dairy
farms within the approximately 275-square-mile portion of the Upper Susquehanna River Watershed located in Delaware County, New York. This region spans
the northern border of the county and covers approximately 20% of the county’s 1,460 square miles. A crosssectional study was designed and conducted over a 3mo period (June to August). Sampling was conducted
over this relatively short time period to reduce any
potential confounding effects of season on the prevalence of Cryptosporidium.
The study population consisted of 19 dairy farms selected from the target population of approximately 100
farms. A block sampling design was adopted to ensure
representation of all geographic areas covered by the
target population. In this design, farms were grouped
into clusters according to their geographic location
within the county (i.e., towns). Within this framework,
a proportional sampling scheme was also used. Thus,
although all areas were represented among the selected
farms, more study farms were included from those areas with a relatively higher proportion of dairy operations. While following the above study design, we used a
participatory approach for the selection of study farms.
This approach involved collaboration with the Delaware County Soil and Water Conservation District and
representatives of the local farmers’ association. The
collaborators provided planning and logistical support
and facilitated study farm selection by granting access
to their ongoing, countywide farm census data. In addition to providing census data, the collaborators also
provided detailed local knowledge during the study
farm selection phase. We believe that the participatory
approach, in conjunction with the proportional block
design, led to the selection of a representative study
population.
Sample and Data Collection
An age-stratified sampling design was used to collect
fecal samples on the study farms (Wade et al., 2000;
Santin et al., 2004). Animals less than 6 mo of age
were differentially targeted to improve the chances of
detecting those shedding C. parvum-like oocysts. In the
present study, the fecal sampling protocol called for the
sampling on study farms of all calves less than 6 mo of
age, up to a maximum of 15 animals. If more than 15
CRYPTOSPORIDIAL PREVALENCE IN A NEW YORK WATERSHED
4231
Table 1. Summary of variables from dairy farm interviews used in data analysis
Variable (risk factor)
General farm management
Source of barn water
Barn type
Quarantine pen
Herd size increased by 10% or more in past year
Herd size decreased by 10% or more in past year
Milkers allowed access to pasture in summer
Bred heifer housing (summer)
Bred heifer housing (summer) outside
Rodent control on farm
Maternity management
Dedicated calving area
Do calvings occur at pasture during the summer?
Calf management
Preweaning
Total number of preweaned calves at the time of the study
Are calves allowed to suckle dam?
When are calves separated from dam?
Where are calves moved immediately after separation?
Type of preweaned calf housing:
Bedding added daily?
Frequency of total bedding change or cleaning in unweaned
calf housing
Primary method for total bedding change or cleaning in
unweaned calf housing
Primary feed for unweaned calves
Calf feeding hygiene
Cleaning of feeding utensils between feedings
Calf contacts after separation from dam
Postweaning
Number of calves from weaning to 6 mo of age
Summer provision of water to calves
Preweaning calf bedding type
Postweaning contacts
Frequency of cleaning weaned (to 6 mo) calf housing
Description
Categorical:
Categorical:
Categorical:
Categorical:
Categorical:
Categorical:
Categorical:
inside
Categorical:
Categorical:
Well, other
Free stall, tie stall
Yes, no
Yes, no
Yes, no
Yes, no
Inside: Free stall, individual pen, group pen, tie stall, not
Free stall, individual pen, group pen, tie stall, not outside
Present or absent
Categorical: Yes, no
Categorical: Yes, no
Continuous
Categorical:
Categorical:
Categorical:
Categorical:
Categorical:
Yes, no
Immediately, not immediately
Tied near dam, moved to final preweaned calf housing
Tied, group pens, hutches, individual pens
Yes, no
Categorical: Never, daily, every other day, weekly or less frequently
Categorical: Not cleaned, soiled bedding removed, all bedding removed
and washed with water, all bedding removed and washed with
detergent
Categorical: Milk replacer, other
Categorical: Utensils not shared, utensils shared and not cleaned,
utensils shared and cleaned between calves
Categorical: Cleaned, not cleaned
Categorical: Other unweaned animals, weaned animals to 6 mo of age,
6 mo old to bred, bred heifers, adult cattle, no physical contact
Continuous
Categorical: None, periodic, continuous
Categorical: Stray or hay, sawdust, other
Categorical: Other unweaned animals, weaned animals to 6 mo of age,
6 mo old to bred, bred heifers, adult cattle, no physical contact
Categorical: Never, daily, every other day, weekly or less frequently
such animals were present, 80% of all calves less than
6 mo of age were to be sampled. Fecal samples were
also collected from 5 heifers from 6 mo of age until first
freshening, 5 lactating cows, and 5 dry cows at each
study farm.
At the time of sampling, study personnel collected
detailed farm management and demographic data
through the administration of a comprehensive management questionnaire. The questionnaire was conducted in the form of an in-person interview with the
farm owner or manager. A single member of the study
team administered the questionnaire on all farms to
eliminate interobserver error. The questionnaire covered a large range of physical and management factors
pertaining to each farm (Table 1). Emphasis was placed
on factors hypothesized to relate to the prevalence of
this parasite, such as colostrum management, nature
and frequency of bedding changes, nature of manure
management, open or closed herd status, and existence
of rodent or wild bird problems.
Sample Processing
Fecal samples were collected from the rectum of each
animal and immediately placed in uniquely labeled
screw-topped specimen containers and stored on frozen
cold packs while being transported to the laboratory.
For the flotation and immunogenic techniques, samples
were stored at 4°C in the laboratory and processed
within 2 wk of collection. A standard quantitative centrifugation concentration flotation technique was initially used on all samples. For each sample, 1 g of feces
was processed using sugar (sg 1.33) as the flotation
medium to recover C. parvum-like oocysts. Microscopic
examination was carried out using bright-field and
phase-contrast microscopy. An animal was considered
Journal of Dairy Science Vol. 89 No. 11, 2006
4232
STARKEY ET AL.
positive if an oocyst with the correct morphology (i.e.,
optical properties, internal structure, size, and shape)
was detected in the sample. Criteria used to identify
C. parvum-like oocysts included the following: measuring 4 to 6 mm, being spherical with a residuum and
sporozoites, refracting pink in sugar, and having a halo
in phase. A commercially available ELISA (ProSpecT
Cryptosporidium Polyclonal Microplate Assay; AlexonTrend, Inc., Lenexa, KS) was performed on all samples
collected from animals 45 d of age or less.
DNA Sequencing
Polymerase chain reaction products were purified using Exonuclease I/Shrimp Alkaline Phosphatase (ExoSAP-IT; USB Corporation, Cleveland, OH). Purified
products were sequenced using the internal primers
described above in 9-L reactions using an automated
sequencer (3730 DNA Analyzer; Applied Biosystems,
Foster City, CA). Samples were sequenced in both directions and sequence chromatograms from each strand
were aligned and inspected using MEGA 3.1 (Kumar
et al., 2004).
Nested PCR
Data Management and Statistical Analysis
A nested PCR targeting the 18S rRNA gene was used
to allow for the differentiation between C. parvum and
C. bovis, allowing elucidation of the zoonotic risk posed
by cattle within watersheds. All samples testing positive by flotation, ELISA, or both were subjected to the
PCR protocol. Additionally, 100 flotation-negative samples were analyzed with this method to determine the
degree of agreement between the techniques, as measured by the kappa statistic.
Immediately on return to the laboratory, a 1-mL subsample of feces of approximately 50% solids was obtained from each individual fecal sample and stored at
−20°C in a sealed tube prior to DNA extraction. Upon
thawing the feces, we used a modified mechanical disruption method for DNA extraction (Zhu et al., 1998).
Samples were examined at the 18S rRNA gene loci
using a nested PCR protocol. Two sets of primers amplifying a final fragment of approximately 830 bp were
used. The primary reaction consisted of 1 L of 1:10
diluted DNA solution obtained from the extraction procedure, added to a mixture consisting of 10.8 L of
reverse osmosis water, 2 L of 1× PCR buffer, 4.8 L
of MgCl2 (50 mM), 0.4 L of deoxyribonucleoside triphosphates, 0.4 L of the forward primer (SSU 1) 5′GAT AAC CGT GGT AAT TCT AGA GCTA-3′ (10 M),
0.4 L of the reverse primer (SSU 2) 5′- TAA GGT GCT
GAA GGA GTA AGG -3′ (10 M), and 0.2 L of Taq
DNA polymerase. The secondary reaction consisted of
1 L of the product from the primary reaction added
to a mixture consisting of 13.2 L of water, 2 L of
1× PCR buffer, 2.4 L of MgCl2 (50 mM), 0.4 L of
deoxyribonucleoside triphosphates, 0.4 L of the forward nested primer (SSU 3) 5′-GAA RGG TYG TAT
TTA TTA GAT AAA GGAAC -3′ (10 M), 0.4 L of the
reverse nested primer (SSU 4) 5′-AAG GAG TAA GGA
ACA ACC TCC A-3′ (10 M), and 0.2 L of Taq DNA
polymerase. Both the primary and secondary reactions
were run under the same conditions: 35 cycles of 96°C
for 45 s, 55°C for 45 s, and 72°C for 1 min.
Journal of Dairy Science Vol. 89 No. 11, 2006
Data management, coalition, and validation were
performed using Microsoft Access 2000 and Microsoft
Excel 2000 (Microsoft, Redbank, CA). The SAS statistical program (Version 9.1 for Windows; SAS Institute,
Cary, NC) was used to generate descriptive statistics
and manage data. Prevalence was determined as the
proportion of samples testing positive for C. parvumlike oocysts relative to those testing negative. Because
all samples were analyzed with the flotation method,
prevalence is reported based on the results yielded by
this method. Prevalence was examined in the population as a whole, at the herd level, and by age.
Fecal flotation results were used in the statistical
analyses to identify risk factors associated with the
prevalence of C. parvum-like oocyst shedding. These
analyses were restricted to animals less than 60 d of
age. Such a restriction was used to focus on the population perceived to be at greatest risk, and to increase
the power of the statistical analyses. This cut-off was
based on the results of a study conducted in an adjacent
watershed, which identified an age range of 3 to 60 d
among C. parvum-like-positive animals (mean 15 d,
standard deviation 6.6 d; Starkey et al., 2005). To this
end, a logistic regression model was used to identify
and examine the significance of various risk factors
when examined simultaneously. During this process,
appropriate transformations, including polynomial
transformations, were assessed for continuous variables in an attempt to improve model fit. Individual
animal and herd management factors obtained from
the management questionnaire were first examined for
association with the prevalence using appropriate bivariate techniques. Discrete factors were examined by
χ2 test of independence or Fisher’s exact test when expected cell counts were less than 5. Continuous variables were examined by unconditional logistic regression. The model-building step continued with the investigation of correlation among significant variables (P ≤
0.2) grouped under each of the management categories
(general farm management, maternity management,
CRYPTOSPORIDIAL PREVALENCE IN A NEW YORK WATERSHED
preweaning calf management, postweaning calf management). Pearson correlation coefficients were examined, and when a correlation was found between variables, the decision regarding which to retain was based
on biological plausibility. When 3 or more variables
were retained in a management category after the
above steps, a best subset selection procedure was run,
using Mallow’s Cp statistic, to determine which to advance to the final model-building step. Age was included
as an independent variable in these steps to control
for this important risk factor throughout the modelbuilding process. Ultimately, all the significant variables chosen at each management level were submitted,
along with age, to a final best subset selection procedure
for the selection of the best model of association between
parameters of interest and the prevalence of C. parvumlike oocyst shedding.
Because the sampling units (the animals) in this
study are clustered into herds, we assumed that this
clustering would lead to a correlation in the likelihood of
infection within the study population. This correlation
between responses occurs because they are dependent
on exogenous factors that are associated with these
responses (i.e., infection with the organism). Conditioning on an observed set of these factors by controlling
for their effect in the analysis and including them as
covariates in the logistic regression analysis will sometimes achieve approximate conditional independence.
However, more often this correlation in the response
arises from both observed and unobserved risk factors.
We assumed that the unobserved risk factors were randomly distributed among farms, and the overall significance of this assumption was evaluated by using a
mixed-effect logistic regression model (Rosner, 1989).
The mixed-effect logistic regression model was specified
as follows:
P(CP/α, βi, σ) =
1
[1 +
exp−(α + Σ βiZi + iσ)]
where P(CP/α, β, σ) is the probability that an animal
within a herd level i would have been shedding C.
parvum-like oocysts given a set of fixed factors Zi with
an effect of βi. The likelihood ratio test was used to
evaluate the significance of the farm random effect parameter in the mixed model. The mixed-effect logistic
regression analysis was performed using the EGRET
statistical software (Cytel Statistical Software, Cambridge, MA). The effect of each factor on the likelihood
of infection with the organism was quantified by the
odds ratio (OR), which was computed as the exponent
of the respective regression coefficient.
4233
RESULTS
Descriptive
A total of 453 samples were collected from animals
on the 19 study farms. These samples were obtained
from 184 animals of 6 mo of age or less (including 81
animals of 60 d of age or less), 104 heifers older than
6 mo, and 165 adult animals (lactating and dry cows).
There was 92% agreement between the flotation and
the ELISA techniques used in this study; thus, results
are reported for the flotation technique because animals
of all ages were subjected to this test. Of the 453 samples, 16 were determined to be positive by flotation for
C. parvum-like oocysts, giving an overall prevalence of
3.53%. This figure is higher than that of 0.9% reported
in a cross-section study previously performed in an adjacent watershed (Wade et al., 2000). When examining
herd-level prevalence, C. parvum-like oocysts were evident in at least one animal on 8 of the 19 study farms,
yielding a herd-level prevalence of 42%. Prevalence
among animals sampled on the study farms ranged
from 3.1 to 21.4%.
The average age of animals shedding oocysts was 16
d, with a standard deviation of 9.2 d and a range of 3
to 31 d. When examining animals less than or equal to
60 d of age, prevalence was found to be 20% across the
study subjects. The prevalence was higher still when
animals of 31 d of age or younger were examined, with
32% of such animals testing positive at the time of
sampling.
Risk Factors
After variables within in each management category
were examined using bivariate techniques and significant variables were investigated with Pearson correlation coefficients, several variables remained and were
subsequently analyzed using a best subset selection
procedure (Table 2). The variables under consideration
during the development of the final best subset model
were age, type of water supply for the barn, type of
housing for milking cows, nature of calf housing (tied
vs. not tied), total number of unweaned calves on a
farm, and contact of postweaned calves with adult
cows (binary).
The factors that were ultimately chosen for the multivariate investigation of association with the prevalence
of C. parvum-like oocyst shedding were age, source of
barn water, calf housing (tied vs. not tied), and number
of unweaned stock on the farm. The results of the logistic regression based on these parameters, including parameter estimates, OR, and confidence intervals, can be
seen in Table 3. None of the polynomial transformations
were significant. The likelihood of shedding oocysts deJournal of Dairy Science Vol. 89 No. 11, 2006
4234
STARKEY ET AL.
Table 2. Significant variables1 as identified by bivariate analysis, after accounting for correlation patterns,
grouped by management categories
Management category
Variable
P-value
General farm management
Age
Barn water source
Barn type
Milking cows allowed access to pasture during summer
0.0137
0.133
0.0033
0.185
Maternity management
Not significant
Preweaning calf management
Number of unweaned calves at the time of the study
Calf housing (tied vs. not tied)
Calf housing (greenhouse vs. not greenhouse)
Contact with older calves (weaned to 6 mo)
Frequency of total bedding change
Summer provision of water to calves
0.0134
0.165
0.0836
0.0807
0.0077
0.0818
Postweaning calf management
Contact with adult cows
0.0180
1
Significant at P ≤ 0.2.
creased as the source of barn water changed from nonwell to well sources (OR = 0.02) and as the age of the
animal increased (OR = 0.9). Calves housed in tie stalls
were at greater risk of shedding the protozoa than those
in other forms of housing, such as greenhouses, individual pens, or group pens. The likelihood of shedding
C. parvum-like oocysts increased with the number of
unweaned calves in the herd (Table 3). Investigation of
the hierarchical nature of the data set was performed
using a mixed-effect logistic regression model. This
analysis indicated that there was no significant correlation of risk by group among the study subjects, as indicated by a nonsignificant random effect parameter (data
not shown).
PCR and Sequencing Results
Of the 16 flotation-positive samples, 12 were successfully amplified using the nested PCR. An additional
100 flotation-negative samples were analyzed with the
PCR protocol to determine the degree of agreement between the 2 techniques. Statistical analysis of all samples examined by PCR and flotation indicated a kappa
statistic of 0.71, a level considered to be associated with
a high degree of association above that expected by
chance alone (Fleiss, 1981). Subsequent DNA sequence
analysis of the PCR-positive samples indicated that
50% of the isolates were C. bovis, a nonzoonotic Cryptosporidium species (Fayer et al., 2005).
DISCUSSION
The study area forms part of the headwaters of the
Susquehanna River, the source of water for several
downstream communities, including the City of Binghamton. Additionally, the river provides a large proportion of Chesapeake Bay’s fresh water, an area of increasing environmental concern and home to extensive
fishing and to crustacean and bivalve industries (Fayer
et al., 2002). Dairy cattle have been implicated as the
source of drinking water contamination with C. parvum, although evidence to this effect is not entirely
conclusive (Xiao and Ryan, 2004). Therefore, quantification of the prevalence of C. parvum infection and
identification of associated risk factors among dairy cattle in watersheds remains an important undertaking.
Table 3. Results of the logistic regression analysis for the association between putative risk factors and
odds of infection with Cryptosporidium parvum
Factor
Age
Barn water source
Well
Non-well
Calf housing
Tied
Not tied
Number of unweaned calves
Intercept
Journal of Dairy Science Vol. 89 No. 11, 2006
Regression
coefficient
Standard
error
−0.105
0.0353
−3.90
0
1.335
3.77
0
0.383
−3.96
1.26
0.125
1.70
Estimated
odds
ratio
0.020
1
43.4
1
95%
confidence
interval
0.001, 0.285
3.66, 515
CRYPTOSPORIDIAL PREVALENCE IN A NEW YORK WATERSHED
It is also important that future studies consider the
role of C. bovis. As reported in this study and elsewhere
in the literature (Santin et al., 2004), up to 50% of
isolates that would have been diagnosed as C. parvum
via traditional methods (i.e., C. parvum-like) are in fact
C. bovis, a nonzoonotic species of Cryptosporidium.
The overall prevalence of 3.9% observed in this study
via the flotation method is higher than the 0.9% reported in a cross-sectional study conducted in an adjacent watershed (Wade et al., 2000). Although similar
age stratification was used in both study designs, additional attention was paid to the youngest animals
within the less than 6-mo-old category in the present
study, potentially explaining the increase in observed
prevalence. The prevalence of C. parvum-like oocyst
shedding among unweaned animals in this study was
20%, a level significantly higher than the overall prevalence and in keeping with other reports in the literature
that this age group has the highest prevalence of the
organism (Santin et al., 2004). The present study identified a relatively high herd-level prevalence of the organism of 42%. This finding is also in keeping with previous
reports in the literature (Garber et al., 1994; Wade et
al., 2000).
As indicated by the kappa statistic of 0.71, there was
a high degree of agreement between the flotation and
the PCR methods used in this study. For the purposes
of reporting prevalence and investigating risk factors
in the present study, we chose to use the results of
flotation. This decision was based on the fact that all
samples were subjected to this test and thus all had an
equal chance of becoming positive. Additionally, this
test remains in common use among practitioners and in
diagnostic laboratories throughout the United States.
This decision subsequently precluded the investigation
of factors that would have led to a differential prevalence of C. parvum and C. bovis in the present study.
Such work would require a larger number of positive
animals to draw conclusions, and could initially be performed as a case-control study or a larger retrospective
cross-sectional study.
For the purpose of model building, we focused our
analysis on animals that were less than 60 d of age.
The rationale for this strategy is 2-fold. First, from our
previous studies we learned that among the populations
of cattle in New York State, the risk of infection is
limited to younger calves that are typically less than
60 d of age (Wade et al., 2000; Starkey et al., 2005).
Second, other researchers found similar results in other
cattle populations outside New York State (Santin et
al., 2004). Given these findings and the potential for
increased statistical power, we opted to restrict the
analysis to cattle below 60 d of age.
4235
Among the findings of interest in the final logistic
regression model was the estimated 37% increase in
the odds of infection among those animals whose barn
water source was not a well. In the study population,
non-well sources of water were either springs or
streams. It is possible that these water sources are
associated with an increased prevalence because they
may have a greater risk of fecal contamination relative
to well water. In the final model, there was also a significant increase in the risk of infection in calves that
were housed in a tied manner, with a point estimate
indicating a 434% increase in the odds of infection. Most
other unweaned calves in the study population were
housed in greenhouses. One possible explanation for
this increased risk among tied animals is the relative
ease of contact with immediate neighbors in most tied
environments, thereby increasing the risk of fecal–oral
contact. However, it is important to note that even moderate physical separation may not eliminate the risk of
transmission of this organism owing to the frequently
propulsive nature of the diarrhea associated with the
infection. The number of unweaned animals on a property was significantly associated with an increase in
the risk of C. parvum-like oocyst shedding. This finding
is logical, because an increased number of at-risk (i.e.,
unweaned) animals would be expected to increase both
the chances of detecting infection and the likelihood of
an uninfected animal encountering an infected animal
or infectious oocyst in its environment.
The study was designed to increase both internal and
external validity; however, a large-scale longitudinal
study needs to be conducted to confirm the identity of
the identified risk factors and to allow causal inferences
to be made. Considering increasing evidence regarding
the relatively high prevalence of C. bovis, further work
is also needed to better quantify the risk cattle pose to
water supplies and downstream water users. An integrative approach combining stochastic and deterministic reasoning with data in the literature may aid in
identifying a useful determination of risk.
ACKNOWLEDGMENTS
Many organizations and individuals contributed to
the success of this project, including the Water Resource
Institute at Cornell University, the Delaware County
Soil and Water Conservation district, and the students
and staff of the Parasitology Section at Cornell University’s College of Veterinary Medicine. This work was
partially supported by a grant from the federal Environmental Protection Agency (grant #403) and from the
USDA-Cooperative State Research, Education, and Extension Service program (grant #2002-35212-12317).
Journal of Dairy Science Vol. 89 No. 11, 2006
4236
STARKEY ET AL.
REFERENCES
Current, W. L., N. C. Reese, J. V. Ernst, W. S. Bailey, M. B. Heyman,
and W. M. Weinstein. 1983. Human cryptosporidiosis in immunocompetent and immunodeficient persons. Studies of an outbreak
and experimental transmission. N. Engl. J. Med. 308:1252–1257.
Fayer, R., U. Morgan, and S. J. Upton. 2000. Epidemiology of Cryptosporidium: Transmission, detection and identification. Int. J.
Parasitol. 30:1305–1322.
Fayer, R., J. M. Trout, E. J. Lewis, L. Xiao, A. Lal, M. C. Jenkins,
and T. K. Graczyk. 2002. Temporal variability of Cryptosporidium
in the Chesapeake Bay. Parasitol. Res. 88:998–1003.
Fayer, R., M. Santin, and L. Xiao. 2005. Cryptosporidium bovis n.
sp. (Apicomplexa: Cryptosporidiidae) in cattle (Bos taurus). J.
Parasitol. 91:624–629.
Fleiss, J. L. 1981. Statistical Methods for Rates and Proportions. 2nd
ed. John Wiley & Sons, New York, NY.
Garber, L. P., M. D. Salman, H. S. Hurd, T. Keefe, and J. L. Schlater.
1994. Potential risk factors for Cryptosporidium infection in dairy
calves. J. Am. Vet. Med. Assoc. 205:86–91.
Glaberman, S., J. E. Moore, C. J. Lowery, R. M. Chalmers, I. Sulaiman, K. Elwin, P. J. Rooney, B. C. Millar, J. S. G. Dooley, A. A.
Lal, and L. Xiao. 2002. Three drinking-water-associated cryptosporidiosis outbreaks, Northern Ireland. Emerg. Infect. Dis.
8:631–633.
Hunter, P. R., and R. C. Thompson. 2005. The zoonotic transmission
of Giardia and Cryptosporidium. Int. J. Parasitol. 35:1181–1190.
Kumar, S., K. Tamura, and M. Nei. 2004. MEGA3: Integrated software for molecular evolutionary genetics analysis and sequence
alignment. Brief. Bioinform. 5:150–163.
Mac Kenzie, W. R., N. J. Hoxie, M. E. Proctor, M. S. Gradus, K. A.
Blair, D. E. Peterson, J. J. Kazmierczak, D. G. Addiss, K. R.
Fox, J. B. Rose, and J. P. Davis. 1994. A massive outbreak in
Milwaukee of Cryptosporidium infection transmitted through the
public water supply. N. Engl. J. Med. 331:161–167.
Maggi, P., A. Larocca, M. Quarto, G. Serio, O. Brandonisio, and G.
Angarano. 2000. Effect of antiretroviral therapy on cryptosporidiosis and microsporidiosis in patients infected with human immunodeficiency virus type 1. Eur. J. Clin. Microbiol. Infect. Dis.
19:213–217.
Morgan-Ryan, U. M., A. Fall, L. A. Ward, N. Hijjawi, I. Sulaiman,
R. Fayer, R. C. Thompson, M. Olson, A. Lal, and L. Xiao. 2002.
Cryptosporidium hominis n.sp. (Apicomplexa: Cryptosporidiidae)
from Homo sapiens. J. Eukary. Micro. 49:433–440.
Naciri, M., M. P. Lefay, R. Mancassola, P. Poirier, and R. Chermette.
1999. Role of Cryptosporidium parvum as a pathogen in neonatal
diarrhoea complex in suckling and dairy calves in France. Vet.
Parasitol. 85:245–257.
Nannini, E. C., and P. C. Okhuysen. 2002. HIV1 and the gut in the
era of highly active antiretroviral therapy. Curr. Gastroenterol.
Rep. 4:392–398.
Journal of Dairy Science Vol. 89 No. 11, 2006
Navin, T. R., and D. D. Juranek. 1984. Cryptosporidiosis: Clinical,
epidemiologic, and parasitologic review. Rev. Infect. Dis.
6:313–327.
New York State Department of Agriculture and Markets. 2006. Subject: Ag Facts. http://www.agmkt.state.ny.us/agfacts.html Accessed Apr. 11, 2006.
Okhuysen, P. C., C. L. Chappell, J. H. Crabb, C. R. Sterling, and H.
L. DuPont. 1999. Virulence of three distinct Cryptosporidium
parvum isolates for healthy adults. J. Infect. Dis. 180:1275–1281.
Ong, C. S. L., D. L. Eisler, S. H. Goh, J. Tomblin, F. M. Awad-ElKariem, C. B. Beard, L. Xiao, I. Sulaiman, A. Lal, M. Fyfe, A.
King, W. R. Bowie, and J. L. Isaac-Renton. 1999. Molecular epidemiology of cryptosporidiosis outbreaks and transmission in British Columbia, Canada. Am. J. Trop. Med. Hyg. 61:63–69.
Panciera, R. J., R. W. Thomassen, and F. M. Garner. 1971. Cryptosporidial infection in a calf. Vet. Pathol. 8:479–484.
Patel, S., S. Pedraza-Diaz, J. McLauchlin, and D. P. Casemore. 1998.
Molecular characterisation of Cryptosporidium parvum from two
large suspected waterborne outbreaks. Outbreak Control Team
South and West Devon 1995, Incident Management Team and
Further Epidemiological and Microbiological Studies Subgroup
North Thames 1997. Commun. Dis. Public Health 1:231–233.
Rosner, B. 1989. Multivariate methods for clustered binary data with
more than one level of nesting. J. Am. Stat. Assoc. 84:373–380.
Santin, M., J. M. Trout, L. Xiao, L. Zhou, E. Greiner, and R. Fayer.
2004. Prevalence and age-related variation of Cryptosporidium
species and genotypes in dairy calves. Vet. Parasitol. 122:103–
117.
Starkey, S. R., S. E. Wade, S. Schaaf, and H. O. Mohammed. 2005.
Incidence of Cryptosporidium parvum in the dairy cattle population in a New York City watershed. Vet. Parasitol. 131:197–205.
Sulaiman, I. M., A. A. Lal, and L. Xiao. 2001. A population genetic
study of the Cryptosporidium parvum human genotype parasites.
J. Eukaryot. Microbiol. (Suppl.):24S–27S.
Tumwine, J. K., A. Kekitiinwa, S. Bakeera-Kitaka, G. Ndeezi, R.
Downing, X. Feng, D. E. Akiyoshi, and S. Tzipori. 2005. Cryptosporidiosis and microsporidiosis in Ugandan children with persistent diarrhea with and without concurrent infection with the
human immunodeficiency virus. Am. J. Trop. Med. Hyg.
73:921–925.
Wade, S. E., H. O. Mohammed, and S. L. Schaaf. 2000. Prevalence
of Giardia sp., Cryptosporidium parvum and Cryptosporidium
muris (C. andersoni) in 109 dairy herds in five counties of southeastern New York. Vet. Parasitol. 93:1–11.
Xiao, L., and U. M. Ryan. 2004. Cryptosporidiosis: An update in
molecular epidemiology. Curr. Opin. Infect. Dis. 17:483–490.
Zhu, G., M. J. Marchewka, J. G. Ennis, and J. S. Keithly. 1998. Direct
isolation of DNA from patient stools for polymerase chain reaction
detection of Cryptosporidium parvum. J. Infect. Dis. 177:1443–
1446.