HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Interpretive Summary Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Nydam, D. V. Articles by Mohammed, H. O. Search for Related Content PubMed PubMed Citation Articles by Nydam, D. V. Articles by Mohammed, H. O.

Quantitative Risk Assessment of Cryptosporidium Species Infection in Dairy Calves

Department of Population Medicine and Diagnostic Science, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853

Corresponding author: Hussni O. Mohammed; e-mail: hom1{at}cornell.edu.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 
Cryptosporidium parvum is a zoonotic protozoan that infects many different mammals including cattle and humans. Cryptosporidiosis has become a concern for dairy producers because of the direct losses due to calves not performing well and the potential for environmental contamination with C. parvum. Identifying modifiable control points in the dynamics of infection in dairy herds will help identify management strategies that mitigate its risk. The quantitative risk assessment approach provides estimates of the risk associated with these factors so that cost-effective strategies can be implemented. Using published data from epidemiologic studies and a stochastic approach, we modeled the risk that C. parvum presents to dairy calves in 2 geographic areas: 1) the New York City Watershed (NYCW) in southeastern New York, and 2) the entire United States. The approach focused on 2 possible areas of exposure—the rearing environment and the maternity environment. In addition, we evaluated the contribution of many risk factors (e.g., age, housing, flies) to the end-state (i.e., total) risk to identify areas of intervention to decrease the risk to dairy calves. Expected risks from C. parvum in US dairy herds in rearing and maternity environments were 41.7 and 33.9%, respectively. In the NYCW, the expected risks from C. parvum in the rearing and maternity environments were 0.36 and 0.33%, respectively. In the US scenarios, the immediate environment contributed most of the risk to calves, whereas in the NYCW scenario, it was new calf infection. Therefore, within the NYCW, risk management activities may be focused on preventing new calf infections, whereas in the general US population, cleaning of calf housing would be a good choice for resource allocation. Despite the many assumptions inherent with modeling techniques, its usefulness to quantify the likelihood of risk and identify risk management areas is illustrated.

Key Words: cattle • risk assessment • Cryptosporidium • watershed

Abbreviation key: NYCW = New York City Watershed, P1, P2, P3 = probabilities of sources of C. parvum from infected calves (P1), oocysts in the calf rearing environment (P2), or oocysts on fomites (P3)

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 
Cryptosporidium parvum is a zoonotic protozoan recognized as one of the primary pathogens causing diarrhea in neonatal calves (de la Fuente et al., 1999; Naciri et al., 1999). The organism, depending on genotype (Morgan et al., 1997; Peng et al., 1997), causes usually self-limiting diarrhea in immunocompetent human patients, but causes life-threatening disease in those with immunodeficiencies (O’Donoghue, 1995). It has emerged as one of the most recognized causes of water-borne outbreaks of gastrointestinal illness (MacKenzie et al., 1994) as well as being associated with foodborne outbreaks (Quiroz et al., 2000) and sporadic cases (McLauchlin et al., 2000). By their location in watersheds, cattle have been implicated as a source of C. parvum associated with these examples (Smith and Rose, 1990), despite lack of direct evidence. The possible losses to a dairy operation from C. parvum are many. One is the direct loss in profitability from calves that do not perform as well due to morbidity. Other losses include less directly quantifiable losses that may arise from government restrictions on animal agriculture because of the zoonotic potential of the organism, as well as decreased consumer confidence in animal agricultural products.

It is necessary to find management approaches to prevent the risk of cryptosporidiosis. There are multiple reasons to take a preventive approach to this disease, not least among them is that there are no consistently effective and approved chemotherapeutics for cattle or humans (Woods et al., 1996; Blagburn and Soave, 1997). Furthermore, at recommended concentrations, most commercial disinfectants are not effective at killing C. parvum oocysts (Campbell et al., 1982; Ares-Mazas et al., 1997). Because of their small size, oocysts evade most water treatment facilities attempts to remove them and thus remain in finished water. To date, no vaccines have been marketed that have proven efficacious in preventing cryptosporidiosis, although there is current experimental work in this area (Perryman et al., 1999). Another reason to prevent the disease is that the life cycle of C. parvum is such that oocysts are immediately infective upon excretion and are transmitted via the fecal-oral route.

To design cost-effective strategies to minimize the associated risk of Cryptosporidium in dairy herds, a systematic approach of examining the dynamics of infection in the population and its environment is necessary. Quantitative risk assessment is one approach that provides a means to assess the risk and its impact (Vose, 2000), including microbial risk assessments for potentially zoonotic pathogens such as Escherichia coli O157:H7 and Listeria monocytogenes (Bemrah et al., 1998; Ebel et al., 2004). Two common risk assessment models are currently used of which an adaptation to animal health has been proposed by the Office Internationale des Epizooties (OIE, 2001). Carrying out a comprehensive risk assessment for cryptosporidiosis in watersheds is a complex task. A reasonable approach is the parsimonious one in which the complex risk scenario is partitioned into complementary units that can be integrated through multiple studies toward a comprehensive risk assessment.

The many factors associated with cryptosporidiosis and their complex interaction necessitates a systematic investigation of a multistep process that incorporates the inherent uncertainty involved at each step. Risk assessment, one of the components of risk analysis, offers a means to a systematic investigation. The present investigation focused on the risk C. parvum poses to dairy calves, on-farm, using data from 2 different geographic areas—the New York City Watershed (NYCW) and the United States as a whole. The process also identified some areas for potential risk mitigation activities that could decrease the likelihood of this protozoan on dairy farms.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 
We adopted a risk assessment approach to address the stated objectives. This involves organizing and analyzing the information to arrive at an estimate of the probability of the risk occurring. The model we are following is based on the model of Covello and Merkhofer (1993) and adopted by the Office Internationale des Epizooties. The framework of the model consists of 4 steps: release assessment, exposure assessment, consequence assessment, and risk characterization. We used the scenario pathway analysis and event-tree methods to describe the dynamics of infection of Cryptosporidium in dairy calves. A conceptual framework for the event tree is presented in Figure 1. The end-state or final probability represents the risk (likelihood) of infection to susceptible calves in the dairy environment (Figure 1). Four different risk scenarios were investigated in this study: rearing environment on dairies in the NYCW (1a; Figure 2); maternity environment on dairies in the NYCW (1b); rearing areas on dairies in the United States at large (2a); and maternity environment on dairies in the United States (2b); each generating an end-state probability.

View larger version (18K): [in this window] [in a new window]   Figure 1. The conceptual framework for the event tree used in the model to describe the dynamics of Cryptosporidium parvum in dairy calves. P1, P2, and P3 represent the probability of C. parvum at each stage. Downward arrows indicate a route of transmission. No = Absence of risk.

View larger version (25K): [in this window] [in a new window]   Figure 2. Scenario pathway model showing the different factors playing a role on the risk of Cryptosporidium parvum at each level of the event tree for the New York City Watershed rearing environment model (Scenario 1a).

The second event in the pathway was that Cryptosporidium would be in the existing rearing or maternity areas (P2). Cryptosporidium does not multiply in the external environment. If the probability of oocysts in the calves’ environment was nonzero then there would be a risk equal to P2. Several factors that play a role in modifying this risk are explained below.

The risk of infection with Cryptosporidium was exacerbated by transmission through inanimate objects and fomites. Calves may be exposed to oocysts via fomites and the likelihood of that risk equals P3. The probability of fomites contributing to the likelihood of Cryptosporidium was calculated as described below and, again, if there were no oocysts on fomites, then there would be no risk to calf. This probability was modified by several factors that are explained below. The end-state risk was computed as the probability of the risk at the source, environment, and the transmission by fomites, P1, P2, and P3, respectively.

Factors Operating at the Source Level (P1)
Calves are likely the primary source of Cryptosporidium in dairy herds. Although some studies have shown that adult cows and wildlife might contribute to the risk posed by this organism, we have not accounted for these 2 sources because either their role is not clear in the literature or was assumed to be negligible. Several factors modify the likelihood of release and perpetuation of C. parvum from a calf. Data for this level were obtained primarily from 2 sources (Garber et al., 1994; Mohammed et al., 1999). The study of Mohammed et al. (1999) was conducted in a watershed area that included the Catskill/Delaware portion of the NYCW in southeastern New York State, and was designated scenario 1. The study of Garber et al. (1999) was conducted over the entire United States and is designated as scenario 2.

Inputs to P1 in the NYCW (Scenario 1)
Estimates of the likelihood of Cryptosporidium at the source for the NYCW scenario were computed based on the findings reported by Mohammed et al. (1999). This likelihood was a function of 13 variables explained below and shown in Table 1 and Figure 2.

General Management
General management was significantly associated with the risk of infection of C. parvum in dairy herds in the NYCW and consisted of 4 risk factors—change of herd size, number of milking cows, presence of animals other than cattle on the farm, and the distance of the barn from the house septic system. Again, an indexing approach was used to determine the value of the risk factor (Z_i) that represents general management. The index value for general management was estimated at 8.992.

Age and Season
Age of the calf and the season in which the sample was taken were also significant in the final model and thus contributed to the likelihood of calf infection (P1) in the scenario path. Summer was assigned a value of 1 in the model and the age of the calf was assigned a value of 15 d, mid-age of the highest risk.

Inputs to P1 in the US (Scenario 2)
Estimates of the likelihood of Cryptosporidium at the source level in the United States were computed based on the findings of Garber et al. (1994). It is a function of 3 variables at the farm level that were significantly associated with the likelihood of the oocyst and are presented in Table 1. These factors were herd size, the summer season, and grouping of calving cows in the maternity facility. A herd size of 100 to 200 cows was assigned a value of 1 in the computation, as were the summer season and multiple cow maternity facilities.

Calculation of P1
Probability P1 for both the NYCW and US scenarios was computed using the logistic regression model as follows (Kleinbaum et al., 1982; Hosmer and Lemeshow, 1989):

([1])

where P (Cp) is the probability a calf shedding Cryptosporidium, is the constant at standard conditions, and ß is the change in the probability of infection due to the respective factor Z_i.

Exposure Assessment
Although the shedding of C. parvum has been observed in different age groups, we believe that the population at highest risk is calves between 1 and 30 d of age (O’Handley et al., 1999; Wade et al., 2000). Calves become exposed to the oocyst immediately after birth where the oocyst is prevalent in the surrounding environment (Sischo et al., 2000). Although no dose-response studies have been carried out in calves, it has been shown that all calves less than 30 d are at high risk of exposure to Cryptosporidium on farms where the infection is endemic (Moore et al., 2003). All calves that were exposed experimentally to C. parvum using different dosages (1 x 10⁴, 1.0 x 10⁶, 5 x 10⁶ oocysts) developed infection and shed the organism in their feces (Naciri et al., 1993; Harp and Goff, 1995; Fayer et al., 1998). Because infected animals shed significantly higher numbers of oocysts than the doses that were used in experimental infection, we believe that susceptible calves in endemic farms are at constant risk of exposure and that they will contract the infection if they are exposed (Fayer et al., 1998; Nydam et al., 2001).

Risk of Cryptosporidium at Existing Rearing Area (P2)
The calf rearing environment and the maternity environment risk factor categories were assumed to be the same for both the NYCW and US scenarios and were thus modeled similarly, but using input values specific to each population. The respective scenarios were designated 1a for the NYCW rearing environment, 2a for the US rearing environment, 1b for the NYCW maternity environment, and 2b for the US maternity environment (Table 1).

Rearing Environment Scenario
Data on the rearing environment were primarily obtained from Faubert and Litvinsky (2000). The probability that an infected calf contributes to the likelihood of the hazard in the existing environment was accounted for. Sites that also contributed to P2 included soil from the area around the calves, the calf hutches, and the walls of the calf area (Table 1 and Figure 2).

Maternity Environment Scenario
Data on the maternity environment were again primarily obtained from Faubert and Litvinsky (2000). Sites that contributed to P2 in the maternity environment included the calving pen floor, the alleys near the calving pen, the calf area, and soil from the exercise area (Table 1). The probability that a calf was shedding also contributed to the likelihood of the hazard in the maternity environment and was accounted for.

It was assumed that each of the 4 factors in the rearing environment and the 5 factors in the maternity environment contributed separately to the likelihood of Cryptosporidium at the respective P2 values. Thus, P2 was calculated as the joint probability of the independent events in both cases. The estimation of the true probability of each of the various sites in the environment contributing to P2 was assumed to follow the Beta distributions of the form ₁ = r + 1 and ₂ = n – r + 1, where r = the number of times C. parvum oocysts were found, and n = the number of trials in which they were looked for (Table 1; Vose, 2000).

Role of Fomites in the Risk of Cryptosporidium (P3)
For all 4 scenarios modeled, P3 is comprised of the contribution of fomites [feed buckets (Faubert and Litvinsky, 2000) and flies (Graczyk et al., 1999)], the immediate environment, and existing C. parvum (Figure 1). The probability of existing C. parvum was obtained from the data of Garber et al. (1994) and Mohammed et al. (1999) less a factor of 50% that accounted for (approximately) routine cleaning of some surfaces where C. parvum may exist (Faubert and Litvinsky, 2000). The P3 was calculated as a joint probability of independent events (feed buckets, flies, and P2) and the disjoint probability of the existing C. parvum. The probability that the feed buckets contributed to P3 was modeled using a Beta distribution as above (Table 1).

Consequence Assessment
Infected calves excrete oocysts in their feces for 3 to 13 d (average 7 d) postinfection (Fayer et al., 1998; Nydam et al., 2001). The number of oocysts excreted by an infected calf in the immediate environment varies between 4 x 10⁶ and 4.15 x 10⁷ oocysts/g per day (Fayer et al., 1998; Nydam et al., 2001). This amount is much higher than the reported infective dose used in experimental studies (Fayer et al., 1998; Moore et al., 2003). Because all calves less than 30 d old are at risk, irrespective of whether they were fed colostrum (Harp et al., 1989), we assumed that raising calves on farms with a history of C. parvum would put them at risk for infection. This risk was dependent on the presence of the infection on the farm. Experimental studies on the survival of C. parvum oocysts in the environment illustrated that changes in temperature and animal waste pile management affected its viability (Jenkins et al., 1999). An observational study showed that the likelihood of the oocyst being in environment samples is affected by pH and temperature of the soil (Barwick et al., 2003). In our analysis, we assumed that the oocysts would survive in the immediate environment of the calf for at least 30 d, the period during which the animal is at risk.

Risk Characterization
The risk presented to a calf, relative to the presence of oocysts, was estimated using the results of the previous steps as inputs (Table 1). The risk for each of the 4 scenarios is calculated as follows:

([2])

where P (end-state) is the final probability outcome and Pi is the estimate of the probability at each event in Figure 1 or 1 of the other 3 corresponding scenarios (Table 1). This calculation assumes that the probability of P1 was independent from P2 and P3. However, this assumption is likely not true in this scenario. For example, the presence of oocysts in the environment is likely correlated with the probability of oocysts shed by calves. The degree of dependency is adjusted for in the final simulation using the RiskCorrmat function in @Risk (version 4; Palisade Corp., Newfield, NY), which produces a matrix of rank correlation coefficients between each of the input values.

To simulate the distribution of the end-state risk of Cryptosporidium to calves in the dairy environment, P1, P2, and P3 were assigned probability distributions (Table 1). For P1, a Pert distribution was assumed with values of (0.01, 0.042, 0.07) for scenario 1 and (0.7, 0.765, 0.82) for scenario 2. The Pert distribution is a version of the 4-parameter Beta and it allows the 4 parameters to be defined by inputs of minimum, most likely, and maximum values. Both P2 and P3 were assumed normally distributed and their distribution was approximated by the Poisson distribution (Vose, 2000). The Poisson distribution is often used to model events of microorganism concentrations, especially when the number of events is low (Bemrah et al., 1998). As the mean and variance get larger, the Poisson distribution approaches normal (Casella and Berger, 1990). The values of the mean and standard deviation for the risk at both P2 and P3 are given for each of the scenarios in Table 1.

The final probability distribution estimates the risk to a calf posed by C. parvum in each of the 4 scenarios. The distribution was generated using Monte Carlo simulation in @Risk software (version 4, Palisade Corp.) carried out with 10,000 iterations of the respective distributions for each probability with Latin Hypercube sampling. Monte Carlo simulation briefly involves using probability distributions, sampling from these distributions, repeatedly running the model (iterations), and then analyzing the sets of outputs. Latin Hypercube sampling uses stratified sampling without replacement and thus has advantages over Monte Carlo sampling when sample sizes are small. One such advantage is that it does not over- or undersample some parts of the input distribution and therefore better reproduces the shape of it, leading to statistics (e.g., mean and standard deviation) that are nearer to the theoretical values (Vose, 2000). When sample sizes are large, the results provided are similar.

Sensitivity Analyses
Sensitivity analyses were performed to identify the impact of the inputs on the outcomes. This was accomplished by Spearman’s rank order correlation analysis because it makes no assumptions about the relationship of the composite of the distributions used.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 
Calf Infection (P1)
A simulated distribution of the risk of calf infection with C. parvum given the inputs above was obtained (Table 2). The risk for the calves in the NYCW was 1.1 to 7% with a mean of 4.2%. The model predicted that 90% of the probabilities of infection would be less than 5.7%. For the calves in the US scenario, the probability ranged from 70.3 to 82% with a mean of 76.5%. The 90% interval estimate for this probability was 73 to 79%.

Fomites and Existing C. parvum (P3)
The probability of fomites contributing oocysts and of oocysts previously existing despite cleaning had a mean of 6.9% and a range of 6.4 to 7.3% in the NYCW rearing environment model (Table 2). The corresponding risk in the United States as a whole was 28.1% (18 to 35.4%). In the maternity environment scenarios, the estimated value of P3 in the NYCW was 6.7% with a range of 6.3 to 7.0%. The estimated value of P3 in the United States was 24.4% (range = 16.7 to 30.2%).

End-State Hazard (1a, 2a, 1b, 2b)
Our analysis showed that the risk to a calf in the NYCW in the rearing environment (1a) had a mean of 0.357% and the distribution of this probability ranged from 0.076 to 0.768% (Table 3 and Figure 3). The corresponding risk to a calf in the US scenario ranged from 14.7 to 88.9% with a mean of 41.7%. Despite the wide range in scenario 2a, 90% of the values were between 31 and 55% (Figure 4). The end-state range of risk for calves in the maternity environment of the NYCW (1b) was 0.080 to 0.660% and the mean was 0.339% (Figure 3). For calves in the US maternity environment, the model predicted a range of 9.68 to 62.4% and a mean of 33.9% for the risk that C. parvum presents. For scenario 2b, 90% of the values were between 24.7 and 43.6% (Figure 4).

View larger version (13K): [in this window] [in a new window]   Figure 3. Cumulative probability for the hazard of Cryptosporidium parvum as determined for the New York City Watershed rearing (scenario 1a, – – –) and maternity (scenario 1b, —) environments.

View larger version (12K): [in this window] [in a new window]   Figure 4. Cumulative probability for the hazard of Cryptosporidium parvum as determined for the US rearing (scenario 2a, – – –) and maternity (scenario 2b, —) environments.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

Risk analysis is a methodological tool for supporting rational and systematic decision making in the face of uncertainty about the likelihood of various hazards occurring. The method used here combines a deterministic approach, which is based on actual observations collected on the study units, with stochastic Monte Carlo simulation. The stochastic approach complements the deterministic approach by providing insight to unobserved events. A limitation of an exclusively deterministic approach is that it only reports the average expected outcome. This limitation can be overcome by performing worst- and best-case scenarios, but this approach overemphasizes extreme situations because they are not as likely to occur as the expected scenario. An advantage of including deterministic inputs is that it allows for greater validity when generalizing the results to similar populations.

The 4 scenarios investigated represent different populations and different areas where the hazard, C. parvum, may present a risk to calves. The probability of the hazard was much higher in the broader US population than in the restricted NYCW population. Furthermore, the risk was higher in the rearing environment scenarios than in the maternity environment. Of the 3 broad probability inputs (P1, P2, and P3) that determined the end-state risk, the immediate environment (P2) contributed most to the final probability in the US population, whereas calf infection (P1) was the most important input in the NYCW scenario. This may indicate that if the endemic probability of shedding is already quite high, it is most important to focus risk management activities at the environment level to limit transmission and exposure of C. parvum. In contrast, if the level of calf infection is relatively low, it might be advantageous to take actions to prevent new calves from shedding oocysts. The initial prevalences of the C. parvum in animals in these 2 geographic areas, NYCW and the United States, were obtained from the published literature (Garber et al., 1994; Wade et al., 2000) and were different to begin with. In addition to the difference in the target populations (initial prevalence, transmission parameters, animal density, etc.), there was a difference in the methods used to determine the prevalence of C. parvum. Each of these might have contributed to the observed differences in the computed risk in our current study.

Within the various factors that comprised the contribution of P2 (immediate environment) in the rearing environment, calf housing played the largest role in the end-state hazard. Thus, the logical risk management activity that would decrease the risk the most would be cleaning hutches between calves and placing hutches or pens on a surface that could be easily cleaned and which drains well. Examples of this include concrete, large gravel, and geotextile fabric. It is then necessary to provide clean, dry bedding over that surface (Harp and Goff, 1998). Washing the hutches or pens should be done with the hottest (~71°C) high-pressure water available followed by thorough drying (Harp and Goff, 1998).

In the maternity environments for both the NYCW and US scenarios, the area reserved for calves after birth and the calving pen floor were the most significant contributors to the P2 input. This is logical in that these areas provide more exposure to newborn calves than do the aisles around the maternity pens and the corresponding soil.

Many assumptions are necessarily made in this approach and therefore, the results should be interpreted with this in mind. The data used in this investigation were taken solely from peer-reviewed publications and thus, are as reliable as the scientific community allows. It cannot be assumed, however, that these data are equally applicable to all external populations. Further assumptions include the season of sampling (summer), the size of the herd (100 to 200 cows), and the approximate age of the calf (15 d).

To facilitate understanding the complex mathematical approach and explain the objectives to the stakeholders, including farmers, we opted to select simplified assumptions. In reality, the dynamics of this infection are more complex. There are more factors that contribute to the risk of exposure and infection that C. parvum poses to a calf, other than the ones modeled here—calves shedding oocysts, oocysts in the immediate environment, and fomites. For example, the role of other sources of this organism, for example, the contribution of wildlife in exposure of calves or environmental contamination, is not accurately known and thus is not included in the model, but could be a factor.

The probability distributions used constitute another group of assumptions. We used a Beta distribution, where the graphic plotting of the raw data was suggestive, to obtain the mean and standard deviation for the probabilities of P2 and P3. The Beta distribution returns an estimate of the true probability of an event occurring given the number of times that event occurs in a given number of trials (Vose, 2000). The Pert distribution was chosen to model P1 because we arrived at a valid most likely value via the logistic model, but wanted to incorporate some uncertainty around this value. The Pert distribution is 4 times more sensitive to the most likely value than to the minimum and maximum values when compared with the somewhat analogous Triangular distribution. In addition, choosing the Triangular distribution in this case would overemphasize the tails of the distribution or require knowing the absolute minimum and maximum values for calf infection. The Normal approximation to the Poisson distribution was chosen to account for our uncertainty and the variability in the data for P2 and P3. As the mean of the Poisson distribution gets larger and the coefficient of variability (standard deviation/mean) approaches 0, the Normal approximates the Poisson. This is the case in this analysis. A Poisson process is a good choice to model P2 and P3 because, in this process, the likelihood of occurrence of the event has a continuum of opportunity that ranges between zero and infinity and there is a probability of it occurring no matter how small the unit of exposure might be (Vose, 2000). Of the Poisson processes, the Normal approximation to the Poisson distribution is a good choice for this analysis because it is a continuous distribution that will return more than discrete probabilities.

Modeling this scenario is somewhat easier than a similar one might be for a bacterial pathogen that can readily multiply in the environment, changing the probability of its occurrence from one stage to another in a scenario path. In contrast, C. parvum cannot grow or multiply outside of the host; thus, the probability of their existence should not increase from one stage to another. Furthermore, because the oocysts are so environmentally hardy, one would not expect a large loss in the viability of the oocysts that could take place if one were modeling the path from the calves, through the environment, and to a stream. Although the loss might not be great in the current scenario, some oocysts do die or are not viable for some other reason. Loss of viability data for each step could be added if it was more readily available for the environments modeled here.

The relatively large differences in outcomes between the NYCW and US scenarios reflect the large amount of variability that can occur between different populations. This further illustrates the danger of generalizing the results of this, or any other, study beyond the target population without taking great care. This also shows perhaps a more dangerous situation that can occur in risk assessment and that is trying to arrive at a viable solution using only mathematical constructs and not basing the studies on observed data from the population to which one wishes to make inference. This illustrates the utility of our approach, showing that different risks are computed and different risk management strategies are highlighted depending on factors specific to the population from which one draws data and makes inferences.

The scenario in this model was limited to the risk that presence of C. parvum oocysts presents to calves and did not include formal modeling of a dose-response assessment. This was because there were not sufficient data to validly estimate the number of oocysts it would take to cause clinical disease in an average calf on a dairy farm. Furthermore, using the Beta-Poisson model to approximate it, as has been done in similar situations, may not be the correct approach (Teunis and Hovelaar, 2000). When such data become available, they can easily be incorporated into the model to extend the assessment. In addition, further risk assessments on this topic could take into account the number of oocysts at each stage and their genotype.

This model is just one step in a number of approaches or actions that can be used to find the best solutions for controlling C. parvum at the farm, environment, and watershed levels. It is logical to first attempt to model this risk at the only place where the pathogen can live and multiply—in the animal. Further models that could be considered include the transport of oocysts from the host to agricultural run-off and from run-off to consumers of water. Ideally, then, all models would be combined to yield a complete risk assessment of zoonotic cryptosporidiosis.

Even with the restrictions we have noted, this is, to our knowledge, the first attempt to model the risk that C. parvum presents to calves in the dairy environment. It shows some of the advantages and disadvantages of combining deterministic and stochastic approaches to risk assessment as well as some further avenues of investigation where data are lacking. Further, and most importantly, it illustrates some areas for possible actions to decrease the risk in dairy calves.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 
Potential risk of C. parvum to dairy calves in a specific watershed was compared with risk for the general calf population in the US. Youngstock is known to be at high risk for this zoonotic pathogen; therefore, our analysis focused on areas of probably highest exposure—maternity and rearing environments. Our analysis showed that calves are at a higher risk of infection with this organism in the general US population compared with animals in NYCW. This difference in the risk between the 2 regions was attributed to the expected endemic shedding rate of this water and foodborne pathogen among these cattle populations. If risk mitigation strategies were designed, they should be directed toward the hygiene of calf housing in the general US calf population, but toward preventing entry of new infected calves in the lower animal density NYCW population.

	ACKNOWLEDGEMENTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 
This project was supported in part by USDA grant no. 98-38420-5803 and by funds from the Water Resources Institute at Cornell University.

Received for publication January 13, 2005. Accepted for publication July 13, 2005.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES

 


Ahl, A. S., J. A. Acree, P. S. Gipson, R. M. McDowell, L. Miller, and M. D. McElvaine. 1993. Standardization of nomenclature for animal health risk analysis. Rev. Sci. Tech. Off. Int. Epiz. 12:1045–1053.

Ares-Mazas, E., M. J. Lorenzo, J. A. Casal, B. Fernandez da Ponte, J. A. Castro, and F. Freire. 1997. Effect of a commercial disinfectant ("Virkon") on mouse experimental infection by Cryptosporidium parvum. J. Hosp. Infect. 36:141–145.[Medline]

Barwick, R. S., H. O. Mohammed, M. E. White, and R. B. Bryant. 2003. Factors associated with the likelihood of Giardia spp. and Cryptosporidium spp. in soil from dairy farms. J. Dairy Sci. 86:784–791.[Abstract/Free Full Text]

Bemrah, N., M. Sanaa, M. H. Cassin, M. W. Griffiths, and O. Cerf. 1998. Quantitative risk assessment of human listeriosis from consumption of soft cheese made from raw milk. Prev. Vet. Med. 37:129–145.[Medline]

Blagburn, B. L., and R. Soave. 1997. Prophylaxis and chemotherapy: Human and animal. Pages 111–128 in Cryptosporidium and cryptosporidiosis. R. Fayer, ed. CRC Press, Boca Raton, FL.

Campbell, I., A. S. Tzipori, G. Hutchison, and K. W. Angus. 1982. Effect of disinfectants on survival of cryptosporidium oocysts. Vet. Rec. 111:414–415.[Medline]

Casella, G., and R. L. Berger. 1990. Statistical Inference. Duxbury Press, Belmont, CA.

Covello, V. T., and M. W. Merkhofer. 1993. Risk Assessment Methods. Approaches for Assessing Health and Environmental Risks. Plenum Press, New York, NY.

de la Fuente, R., M. Luzon, J. A. Ruiz-Santa-Quiteria, A. Garcia, D. Cid, J. A. Orden, S. Garcia, R. Sanz, and M. Gomez-Bautista. 1999. Cryptosporidium and concurrent infections with other major enteropathogens in 1 to 30-day-old diarrheic dairy calves in central Spain. Vet. Parasitol. 80:179–185.[Medline]

Ebel, E., W. Schlosser, J. Kause, K. Orloski, T. Roberts, C. Narrod, S. Malcolm, M. Coleman, and M. Powell. 2004. Draft risk assessment of the public health impact of Escherichia coli O157:H7 in ground beef. J. Food Prot. 67:1991–1999.[Medline]

Faubert, G. M., and Y. Litvinsky. 2000. Natural transmission of Cryptosporidium parvum between dams and calves on a dairy farm. J. Parasitol. 86:495–500.[Medline]

Fayer, R., L. Gasbarre, P. Pasquali, A. Canals, S. Almeria, and D. Zarlenga. 1998. Cryptosporidium parvum infection in bovine neonates: Dynamic clinical, parasitic and immunologic patterns. Int. J. Parasitol. 28:49–56.[Medline]

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. JAVMA 205:86–91.

Graczyk, T. K., R. Fayer, M. R. Cranfield, B. Mhangami-Ruwende, R. Knight, J. M. Trout, and H. Bixler. 1999. Filth flies are transport hosts of Cryptosporidium parvum. Emerg. Infect. Dis. 5:726–727.[Medline]

Harp, J. A., and J. P. Goff. 1995. Protection of calves with a vaccine against Cryptosporidium parvum. J. Parasitol. 81:54–57.[Medline]

Harp, J. A., and J. P. Goff. 1998. Strategies for the control of Cryptosporidium parvum infection in calves. J. Dairy Sci. 81:289–294.[Abstract]

Harp, J. A., D. B. Woodmansee, and H. W. Moon. 1989. Effects of colostral antibody on susceptibility of calves to Cryptosporidium parvum infection. Am. J. Vet. Res. 50:2117–2119.[Medline]

Hosmer, D. W., and S. W. Lemeshow. 1989. Applied Logistic Regression. Wiley, New York, NY.

Jenkins, M. B., M. J. Walker, D. D. Bowman, L. C. Anthony, and W. C. Ghiorse. 1999. Use of a sentinel system for field measurements of Cryptosporidium parvum oocyst inactivation in soil and animal waste. Appl. Environ. Microbiol. 65:1998–2005.[Abstract/Free Full Text]

Kleinbaum, D. G., L. L. Kupper, and H. Morganstern. 1982. Epidemiologic Research: Principles and Quantitative Methods. Van Norstrand Reinhold, New York, NY.

MacKenzie, W. R., N. J. Hoxie, M. E. Proctor, M. G. 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. 21; 331:161–167.

McLauchlin, J., C. Amar, S. Pedraza-Diaz, and G. L. Nichols. 2000. Molecular epidemiological analysis of Cryptosporidium spp. in the United Kingdom: Results of genotyping Cryptosporidium spp. in 1,705 fecal samples from humans and 105 fecal samples from livestock animals. J. Clin. Microbiol. 38:3984–3990.[Abstract/Free Full Text]

Mohammed, H. O. 1991. A multivariate indexing system for hygiene in relation to the risk of Mycoplasma gallisepticum infection in chickens. Prev. Vet. Med. 9:75–83.

Mohammed, H. O., and T. E. Carpenter. 1991. Use of a multivariable indexing score for hygiene in relation to egg production. Am. J. Vet. Res. 52:970–973.[Medline]

Mohammed, H. O., S. E. Wade, and S. Schaaf. 1999. Risk factors associated with Cryptosporidium parvum infection in dairy cattle in southeastern New York State. Vet. Parasitol. 83:1–13.[Medline]

Moore, D. A., E. R. Atwill, J. H. Kirk, D. Brahmbhatt, L. H. Alonso, L. Hou, M. D. Singer, and T. D. Miller. 2003. Prophylactic use of decoquinate for infections with Cryptosporidium parvum in experimentally challenged neonatal calves. JAVMA 223:839–845.

Morgan, U. M., C. C. Constantine, D. A. Forbes, and R. C. A. Thompson. 1997. Differentiation between human and animal isolates of Cryptosporidium parvum using rDNA sequencing and direct PCR analysis. J. Parasitol. 83:825–830.[Medline]

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.[Medline]

Naciri, M., R. Mancassola, P. Yvore, and J. E. Peeters. 1993. The effect of halofuginone lactate on experimental Cryptosporidium parvum infections in calves. Vet. Parasitol. 45:199–207.[Medline]

Nydam, D. V., S. E. Wade, S. L. Schaaf, and H. O. Mohammed. 2001. Number of Cryptosporidium parvum oocysts or Giardia spp. cysts shed by dairy calves after natural infection. Am. J. Vet. Res. 62:1612–1615.[Medline]

O’Donoghue, P. J. 1995. Cryptosporidium and cryptosporidiosis in man and animals. Int. J. Parasitol. 25:139–195.[Medline]

Office Internationale des Epizooties (OIE). 2001. International Animal Health Code. OIE, Paris, France.

O’Handley, R. M., C. Cockwill, T. A. McAllister, M. Jelinski, D. W. Morck, and M. E. Olson. 1999. Duration of naturally acquired giardiosis and cryptosporidiosis in dairy calves and their association with diarrhea. JAVMA 214:391–396.

Peng, M. M., L. Xiao, A. R. Freeman, M. J. Arrowood, A. A. Escalante, A. C. Weltman, C. S. Ong, W. R. MacKenzie, A. A. Lal, and C. B. Beard. 1997. Genetic polymorphism among Cryptosporidium parvum isolates: Evidence of two distinct human transmission cycles. Emerg. Infect. Dis. 3:567–573.[Medline]

Perryman, L. E., S. J. Kapil, M. L. Jones, and E. L. Hunt. 1999. Protection of calves against cryptosporidiosis with immune bovine colostrum induced by a Cryptosporidium parvum recombinant protein. Vaccine 23:2142–2149.

Quiroz, E. S., C. Bern, J. R. MacArthur, L. Xiao, M. Fletcher, M. J. Arrowood, D. K. Shay, M. E. Levy, R. I. Glass, and A. Lal. 2000. An outbreak of cryptosporidiosis linked to a foodhandler. J. Infect. Dis. 181:695–700.[Medline]

Sischo, W. M., E. R. Atwill, L. E. Lanyon, and J. George. 2000. Cryptosporidia on dairy farms and the role these farms may have in contaminating surface water supplies in the northeastern United States. Prev. Vet. Med. 29:43:253–267.

Smith, H. V., and J. B. Rose. 1990. Waterborne cryptosporidiosis. Parasitol. Today 6:8–12.[Medline]

Teunis, P. F. M., and A. H. Hovelaar. 2000. The beta-Poisson dose-response model is not a single-hit model. Risk Anal. 4:513–520.

Vose, D. 2000. Risk Analysis: A quantitative guide. 2nd ed. Wiley and Sons, West Sussex, UK.

Wade, S. E., H. O. Mohammed, and S. L. Schaaf. 2000. Prevalence of Giardia spp. Cryptosporidium parvum and Cryptosporidium andersoni (syn. C. muris) [correction of Cryptosporidium parvum and Cryptosporidium muris (C. andersoni)] in 109 dairy herds in five counties of southeastern New York. Vet. Parasitol. 93:1–11.[Medline]

Woods, K. M., M. V. Nesterenko, and S. J. Upton. 1996. Efficacy of 101 antimicrobials and other agents on the development of Cryptosporidium parvum in vitro. Ann. Trop. Med. Parasitol. 90:603–615.[Medline]

This article has been cited by other articles:

				O. Sunnotel, W. J. Snelling, N. McDonough, L. Browne, J. E. Moore, J. S. G. Dooley, and C. J. Lowery Effectiveness of Standard UV Depuration at Inactivating Cryptosporidium parvum Recovered from Spiked Pacific Oysters (Crassostrea gigas) Appl. Envir. Microbiol., August 15, 2007; 73(16): 5083 - 5087. [Abstract] [Full Text] [PDF]

				O. Sunnotel, W. J. Snelling, L. Xiao, K. Moule, J. E. Moore, B. C. Millar, J. S. G. Dooley, and C. J. Lowery Rapid and sensitive detection of single cryptosporidium oocysts from archived glass slides. J. Clin. Microbiol., September 1, 2006; 44(9): 3285 - 3291. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Interpretive Summary Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Nydam, D. V. Articles by Mohammed, H. O. Search for Related Content PubMed PubMed Citation Articles by Nydam, D. V. Articles by Mohammed, H. O.