This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright
Transcript
This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies areencouraged to visit:
j ourna l ho me pag e: ww w.elsev i er .com/ locate /prev etmed
Environmental contamination with Mycobacterium avium subsp.paratuberculosis in endemically infected dairy herds
R.L. Smitha,∗, Y.H. Schukkenb, A.K. Pradhanb,c, J.M. Smithd, R.H. Whitlocke, J.S. Van Kessel f,D.R. Wolfgangg, Y.T. Grohna
a Section of Epidemiology, Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853,United Statesb Quality Milk Production Services, Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca,NY 14853, United Statesc Department of Nutrition and Food Science & Center for Food Safety and Security Systems, University of Maryland, College Park, MD 20742, United Statesd Department of Animal Science, University of Vermont, Burlington, VT 05405, United Statese Department of Clinical Studies, New Bolton Center, School of Veterinary Medicine, University of Pennsylvania, Kennett Square, PA 19348, United Statesf Environmental Microbial and Food Safety Laboratory, ANRI, USDA-ARS, Beltsville, MD 20705, United Statesg Department of Veterinary and Biomedical Science, Penn State University, University Park, PA 16802, United States
a r t i c l e i n f o
Article history:Received 28 January 2011Received in revised form 21 June 2011Accepted 22 June 2011
Environmental contamination with Mycobacterium avium subsp. paratuberculosis (MAP) isthought to be one of the primary sources of infection for dairy cattle. The exact link betweenfecal shedding of MAP by individual cows and environmental contamination levels at theherd level was explored with a cross-sectional analysis of longitudinally collected sampleson 3 dairy farms. Composite samples from multiple environmental sites in 3 commercialdairy herds in the Northeast US were cultured quarterly for MAP, providing 1131 samples(133 (11.8%) were culture-positive), and all adult animals in the herds were tested biannu-ally by fecal culture (FC), for 6 years. Of the environmental sites sampled, manure storageareas and shared alleyways were most likely to be culture-positive. Environmental sampleresults were compared to FC results from either the concurrent or previous sampling dateat both the herd and the pen level. At the herd level, a 1 log unit increase in average fecalshedding increased the odds of a positive non-pen environmental sample by a factor of 6and increased the average amount of MAP in non-pen samples by 2.9 cfu/g. At the pen level,a 1 log unit increase in average fecal shedding in the pen increased the odds of a positiveenvironment by a factor of 2.4 and the average amount of MAP was increased by 3.5 cfu/g.We were not able to model the relationship between non-pen environmental sample sta-tus and the distance between shedding animals and the sample’s location, and neighboringpens did not significantly affect the results of the pen-level analysis. The amount of MAP inpen-level samples and the probability of a pen testing positive for MAP were both positivelybut non-significantly correlated with the number of animals in the pen shedding >30 cfu/gof MAP. At least 6 environmental samples met the criteria for the U.S. Voluntary BovineJohne’s Disease Control Program on 47 of the 72 sampling dates; of these, 19 of the 47FC-positive sampling dates were positive by the 6-sample environmental testing method,resulting in a herd sensitivity of 0.40 (95% CI: 0.26–0.54). None of the 3 FC-negative samplingdates produced positive environmental samples. Although environmental sampling can be
2 R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9
used as a tool in understanding the level of MAP infection in a herd or pen, it did not appearto be a sensitive diagnostic method for herd positivity in these low prevalence herds, andits use may require caution.
Johne’s disease in cattle is caused by a chronic intestinalinfection with Mycobacterium avium subsp. paratuberculo-sis (MAP), due to ingestion of the organism. After a longlatent period, animals infected with MAP begin to shed theorganism in their feces (Benedictus et al., 2008), therebycontaminating the farm environment. This environmentalcontamination with MAP is thought to be one of the pri-mary sources of infection for dairy cattle (Nielsen and Toft,2009).
Environmental sampling has been evaluated for identi-fication of herd MAP status (Lombard et al., 2006; Raizmanet al., 2004). The number of MAP-positive environmentalcultures (EC) on a dairy farm has been found to be pro-portionate to the seroprevalence in the herd (Berghauset al., 2006) as well as the fecal culture (FC) prevalence(Pillars et al., 2009). In assessing the reliability of repeatenvironmental samples, the majority of variation in MAPconcentration has appeared to come from the source ofthe sample, both the dairy herd and the pen within theherd (Aly et al., 2009; Pillars et al., 2009). The NationalAnimal Health Monitoring System (NAHMS) has used stan-dardized environmental sampling (USDA:APHIS:VS, 2010)to determine the apparent herd-level prevalence of MAPnationally, currently estimated in dairy herds at 68%(USDA:APHIS:VS, 2008). However, no studies have exam-ined the relationship between the concentration of MAP inenvironmental samples and in the feces of individual cows.
The objective of this study was to longitudinallydescribe environmental MAP contamination in endemi-cally infected dairy herds, and to correlate that contami-nation to fecal shedding by individual animals.
2. Materials and methods
Sample collection and isolation of MAP from samplesused in this study have been previously described (Pradhanet al., 2009). Briefly, 1 dairy herd in each of 3 states (herdA in New York, herd B in Pennsylvania, and herd C in Ver-mont) was visited quarterly by members of the RegionalDairy Quality Management Alliance (RDQMA) from 2004through 2009. A number of environmental sites were con-sistently sampled on each quarterly visit from mappedlocations using the same method. Source water, water fromdrinking troughs of cows and heifers, and standing wateror manure slurry in freestall barns were collected by a tubescooped through the water/slurry. Feed from pens of cowsand calves, calf bedding, corn silage, and manure compos-ites from milking cows, nonlactating cows, periparturientcows, calf pens, and milking alleyways were collected man-ually by a freshly gloved hand; at each of these sites (i.e.,an alleyway), material from 4–6 locations (i.e., 4–6 pointsalong the alleyway) was combined into one composite
sample. On some sampling dates, bird droppings, flies andinsects were also collected. The samples were classifiedas Alleyway, Pen (consisting only of adult cow pens), Pit(consisting of manure storage areas), and Other. Biannu-ally, fecal samples were collected rectally from every adultcow present in the herds. All fecal samples were tested byfour-tube fecal culture (Pradhan et al., 2009), and the totalsum of cfu across 4 tubes was multiplied by the conversionfactor, 5.3, to determine cfu/g (Pradhan et al., 2011).
Environments sampled included the 6 minimum sitesfor herd-level MAP testing recommended by the Volun-tary Bovine Johne’s Disease Control Program (VBJDCP)(USDA:APHIS:VS, 2010): 2 each of cow housing alleywaysor gutters, manure storage areas, and another manure con-centration area. The culture results of these 6 qualifyingsamples were combined to create a single variable, theStandard 6 (S6). When more than 6 samples qualifying forthe S6 were collected, 6 were chosen randomly from thequalifying samples to include in the S6 results.
The vector of EC results can be defined as Y, where theYi are independent Bernoulli random variables where 1indicates the presence of MAP in an EC, E[Yi] = �(Xi), andVar(Yi) = �(Xi) (1 − �(Xi)) and Xi is the vector of explana-tory variables for sample i, then
�(Xi) = exp(�′Xi + �zi)
1 + exp(�′Xi + �zi)(1)
where � is the vector of the effects of explanatory vari-ables, zi is the sampling date, and � is normally distributedwith a mean of 0 and variance of �2
�. A random effects
model was used as ECs from the same sampling date wereassumed to not be independent. For this logistic model,several configurations of Xi were considered: all containedsample type (alley, pen, pit, or other) and herd (A, B, orC), the prevalence-based model contained the natural logof the proportion of MAP-positive animals in the herd, theamount-based model contained the base-10 log of the aver-age shedding level (cfu/g) of all individual adults, and thehigh-shedder-based model contained the number of ani-mals shedding ≥30 cfu/g in the herd. In addition, the results
R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9 3
of EC may be modeled as a continuous variable ϕi, theamount of MAP (cfu/g) in the sample. In this case,
log(ϕi) = �′Xi + � ′zi + ei (2)
with ei being the random error associated with sample iand all other variables defined as above for Eq. (1).
The logistic model in Eq. (1) was also fit to the databasewith pen samples excluded, and to data limited to pen sam-ples in which the FC results were limited to animals in thepen from which the sample was taken. In the model of pensamples only, sample type was removed from Xi. The linearmodel in Eq. (2) was also used for the separate databases(pen and non-pen samples), adding an additional term toXi for the distance-corrected average shedding level in thepens,
∑ilog(fi)/di, where fi is the shedding level in pen i
and di is the distance between pen i and the center of thesampled area (in meters); a similar analysis was also per-formed using the squared distance between pens, d2 (m2),as the effect of distance is not necessarily linear.
2.2. VBJDCP samples
The primary analysis of the S6 evaluated whether therewas a relationship between the presence of MAP in FC andthe presence of MAP in the S6. This evaluation was based ona logistic regression model (Eq. (1)), where �(Xi) was theprobability that any FC were MAP positive, zi was unity,and Xi consisted of the herd (A, B, or C) and either the over-all result of VBJDCP sampling (positive if ≥1/6 S6 cultureswas positive), the number of S6 cultures testing positive,or the base-10 log of the average amount of MAP in the 6S6 samples (cfu/g). In any of these cases, exp(ˇn) was theodds ratio for herd shedding status with a 1-unit increasein the S6 results, where ˇn was the parameter estimateassociated with that result. The relationship between theresults of the S6 and the MAP infection level within theherd was also examined, using a linear model with the nat-ural log of the herd’s fecal culture prevalence, ln(mi(�)), orthe base-10 log of the average MAP shedding level for theherd, log(fi(�)), as the response variable; herd was includedas a fixed variable. The same three previously defined S6predictor variables were examined separately.
All models were fit with the lmer function in thelme4 package (Bates and Maechler, 2010) for R 2.11.1(R Development Core Team, 2010), which was accessed
Table 1Distribution of 1131 environmental samples by type for 3 commercialdairy herds in the Northeast US; samples were collected quarterly over a5 year period (2004–2009) and cultured for Mycobacterium avium subsp.paratuberculosis.
A total of 1131 EC results were recorded on the 3 farmsduring the study period, 133 (11.8%) of which were positivefor MAP. Of these samples, 545 (125 or 22.9% positive) hadconcurrent FC results and 383 (47 or 12.3% positive) wereassociated with FC results from the previous quarter. Thedistribution of results across environmental sample type inthe 3 herds is shown in Table 1. Fig. 1 shows a time seriesfor the average amount of MAP in FCs and the number andresults of ECs in the 3 study herds. There is variation in thenumber of samples included, as on some sampling dates agreater number of ECs was collected for reasons associatedwith other research questions, while on others some of thesamples were contaminated and could not be included inthe results.
The models using the full database were found to bepoorly fitted, with normality assumptions failing. Thus,analysis was focused on the separate datasets of pensamples and non-pen samples, which were able to meetstatistical assumptions and produce well-fitted models.
There were 250 ECs from adult animal pens with fulldata for analysis, 42 of which were positive for MAP (16.8%).The average number of high shedders in a pen at any time-point was 0.27 for herd A (median = 0, range 0–4), 0 for herdB, and 0.25 for herd C (median = 0, range 0–1). Results of thelogistic regression (Eq. (1)) for pen sample data are shownin Table 2; linearity assumptions for explanatory variables
4 R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9
Fig. 1. Timeline of Mycobacterium avium subsp. paratuberculosis (MAP) sampling intensity and results for each of 3 commercial US dairy herds between2004 and 2009. Bars represent the total number of environmental samples collected in the herd at that sampling date, with positive samples representedin dark gray and negative samples in light gray. The base-10 log of the average amount of MAP (cfu/g) in individual fecal samples is represented by theblack line, with error bars representing the base-10 log of the standard deviation of MAP (in cfu/g) in individual fecal samples as a rough indicator of thevariability in samples.
were met for all models. A 1-log increase in average shed-ding level or an increase of 0.027 in the prevalence of fecalshedders in a pen resulted in a 2.4-fold increase in the oddsof a positive sample, and 1 additional high shedder resultedin a 2-fold increase. Herd effects were not significant for anymodel.
Median distance between pens was 43 m for herd A(range 16–108), 39 m for herd B (range 36–54), and 31 mfor herd C (range 23–48). The linear models based on preva-lence and number of high-shedders in the pen did not meetnormality assumptions for explanatory variables and ran-dom errors and produced poorly fitted models; the results
Table 2Results of logistic regression for the probability of a Mycobacterium avium subsp. paratuberculosis (MAP)-positive environmental sample culture from anadult cow pen by average MAP shedding level (log(average cfu/g), above), the proportion of animals in the pen shedding MAP, or the number of animalsin the pen shedding ≥30 cfu/g MAP in their feces, in 3 commercial US dairy herds, either on the date of sampling or in the previous quarter, between 2004and 2009. Sampling date is included as a random variable.a
Variable Estimate (OR) Standard Error Pr(|z|)Logistic regression based on fecal shedding(Intercept) −1.98 0.26 <0.01Average fecal shedding (log cfu/g) 0.88 (2.41) 0.29 <0.01
Logistic regression based on fecal prevalence(Intercept) 0.40 0.93 0.67ln(fecal prevalenceb) 0.86 (2.36) 0.34 0.01
Logistic regression based on high-shedders(Intercept) −2.05 0.26 <0.01High sheddersc 0.63 (1.88) 0.32 0. 05
a Sampling date had a variance of 0.83 for the shedding model, 0.72 for the prevalence model, and 0.68 for the high-shedders model.b Expressed as a proportion, natural log (base e).c Number of animals shedding ≥30 cfu/g MAP in their feces.
Author's personal copy
R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9 5
Table 3Results of a linear regression for Mycobacterium avium subsp. paratuberculosis (MAP) concentration (log(cfu/g)) in environmental samples from adult cowpens in 3 commercial US dairy herds, either on the date of sampling or in the previous quarter, between 2004 and 2009. Sampling date is included as arandom variable.a
a Sampling date had a variance of 0.09.b Average shedding level for animals in the pen at the time of sampling.c Sum of average shedding level for animals in other pens divided by squared distance (in m) between the center of the sampled pen and the center of
the other pens.
of the linear model based on average fecal shedding in thepens is shown in Table 3. A 2 log unit increase in the averageshedding level of animals in a pen was required to raise thecontamination of samples in that pen by 1 log. The effect ofMAP shedding by cows in other pens was not significantlyrelated to the amount of MAP in a pen’s environment. Herdeffects were not significant for any model.
There were 542 samples available for analyzing theresults of EC from non-pen sources, of which 55 (10.1%)
were positive for MAP. Results of the logistic and linearregressions based on whole-herd prediction variables areshown in Tables 4 and 5, respectively. A 1-log increase inaverage shedding increased the odds of finding MAP 6-foldand the average amount of MAP by 0.5-log in alleyway orpit samples. Fecal prevalence had a positive relationshipwith both the probability of finding MAP and the averageamount of MAP in environmental samples, but this rela-tionship was strongest in alleyway samples. The number
Table 4Results of logistic regression for the probability of a Mycobacterium avium subsp. paratuberculosis (MAP)-positive environmental sample culture from asource other than adult cow pens by average MAP shedding level in the herd (log(average cfu/g)), the proportion of animals in the herd shedding MAP, orthe number of animals in the herd shedding ≥30 cfu/g MAP in their feces, in 3 commercial US dairy herds, either on the date of sampling or in the previousquarter, between 2004 and 2009. Sampling date is included as a random variable.a
Variable Estimate (OR) Standard Error Pr(>|z|) �2 (p > �2)
Logistic regression based on fecal shedding(Intercept) −3.34 0.49 <0.01Average fecal shedding 1.78 0.44 <0.01(log cfu/g) (5.93)Type (Alley is base) Other −3.83 0.94 <0.01
(0.02) 78.32Pit 0.84 0.44 0.06 (<0.01)
(2.32)
Logistic regression based on fecal prevalence(Intercept) 8.8 3.1 <0.01ln(prevalence) 4.5 1.2 <0.01
(90.02)Type (Alley is base) Other −398.8 73426 1.00
(6.4 × 10−174) 21.39Pit −1.8 3.1 0.56 (<0.01)
(0.17)ln(prevalence):Type Other −133.4 24510 1.00
(1.16 × 10−58) 15.77Pit −1.1 1.3 0.41 (<0.01)
(0.33)
Logistic regression based on high-shedders(Intercept) −3.68 0.76 <0.01high sheddersb 0.64 0.40 0.11
(1.90)Type (Alley is base) Other −3.24 0.99 <0.01
(0.04) 54.12Pit 1.28 0.57 0.02 (<0.01)
(3.6)
a Sampling date had a variance of 2.90 for the shedding model, 3.43 for the prevalence model, and 4.70 for the high-shedders model.b Number of animals shedding ≥30 cfu/g MAP in their feces.
Author's personal copy
6 R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9
Table 5Results of a linear regression for Mycobacterium avium subsp. paratuberculosis (MAP) concentration (log(cfu/g)) in environmental samples from a sourceother than adult cow pens by average MAP shedding level in the herd (log(average cfu/g)), the proportion of animals in the herd shedding MAP, or thenumber of animals in the herd shedding ≥30 cfu/g MAP in their feces, in 3 commercial US dairy herds, either on the date of sampling or in the previousquarter, between 2004 and 2009. Sampling date is included as a random variable.a
Linear regression based on fecal sheddingIntercept −0.13 0.07 0.06Average fecal sheddingb, log(cfu/g) 0.46 0.09 <0.01Type (Alley is base) Other −0.11 0.06 0.06 12.06
Pit 0.15 0.07 0.06 (<0.01)shedding:Type (Alley is base) Other −0.43 0.08 <0.01 34.69
Pit 0.03 0.09 0.35 (<0.01)
Linear regression based on fecal prevalenceIntercept 2.21 0.49ln(prevalence) 0.83 0.18Type (Alley is base) Other −2.80 0.44 <0.01 63.85
Pit −0.02 0.48 0.39 (<0.01)ln(prevalence):Type (Alley is base) Other −0.97 0.16 <0.01 49.43
Pit −0.07 0.18 0.36 (<0.01)
Linear regression based on high sheddersIntercept −0.11 0.12Number of high shedders 0.14 0.08Type Other −0.05 0.10 0.34 5.12(Alley is base) Pit 0.21 0.12 0.10 (0.08)high shedders:Type Other −0.17 0.07 0.03 12.13(Alley is base) Pit 0.04 0.08 0.34 (<0.01)
a Sampling date had a variance of 0.13 for the shedding model, 0.18 for the prevalence model, and 0.19 for the high-shedder model.b Average shedding level for animals in the herd at the time of sampling.
of high-shedding animals was not significantly related tothe odds of finding MAP or the amount of MAP in a sample.Herd effects were not significant for any model.
For the distance-corrected linear regression with non-pen samples, pit samples were excluded; 249 samplesremained, of which 13 samples from alleyways and 0samples from other sites were positive for MAP (5.2%).Median distance between non-pen sampling sites and anypen was 78 m for herd A (range 19:174), 33 m for herd B(range 11:93), and 50 m for herd C (range 11:131). Linear-ity assumptions were not met for these models (results notshown).
3.2. VBJDCP samples
Of the 483 collected environmental samples meetingthe criteria for the S6, 92 were positive for MAP (19.0%).There were 50 sampling dates with both FC results and atleast 6 S6 samples from the same or previous quarter.
The results of the logistic model for VBJDCP results(Table 6) showed that none of the S6 results (dichotomous,count, or average contamination) were significant predic-tors of MAP presence in the herd (p = 0.998 for each variable,data not shown). Likewise, none of the S6 predictor vari-ables were significantly related to the FC prevalence in theherd (p = 0.4 for each variable, data not shown), nor washerd a significant variable in any of the models. However,all S6 predictor variables were positively correlated withthe average fecal shedding in the herd (Table 4).
The S6 results correctly identified as positive 19 of47 dates (40%) in which animals were shedding MAP(and correctly identified as negative 3 of 3 dates with nopositive MAP FC samples). This indicates a relative sen-
sitivity of 0.40 (95% CI: 0.26–0.54) compared to wholeherd FC, and supported our assumption of a specificity of1. The NAHMS environmental survey, using similar sam-pling methods to our study, found a MAP prevalence of0.681 among dairy operations. Assuming the sensitivity ofenvironmental sampling observed in our study, the trueherd prevalence of MAP would therefore be 0.977 (95% CI:0.921–0.983).
4. Discussion
Our study showed that environmental contaminationwith MAP was significantly correlated with MAP sheddinglevels in individual animals. Other studies have consideredthe relationship between environmental contaminationand fecal shedding in dairy cattle, with similar findings.Contamination of bedding with Klebsiella pneumoniae hasbeen associated with fecal shedding in animals using thebedding, but only a subset of animals were sampled andthere was insufficient variation in Klebsiella concentra-tion in bedding to allow further analysis (Munoz et al.,2006). Environmental sample positivity in an endemicallyinfected herd has appeared to correspond temporally withactive Salmonella shedding (Wray et al., 1989). The currentstudy shows that average shedding level, as well as shed-ding prevalence, is important in understanding the amountof environmental MAP contamination via feces. Pillars et al.(2009) likewise found a relationship between prevalenceand environmental MAP contamination, but did not con-sider either the amount of fecal shedding or the amount ofenvironmental contamination. The results presented hereusefully extend their findings; the presence of one or two
Author's personal copy
R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9 7
Table 6Results of linear regression for average fecal shedding, log(cfu/g), of Mycobacterium avium subsp. paratuberculosis (MAP) in individual animals in commercialUS dairy herds between 2004 and 2009 by results of 6 VBJDCP-standard environmental samples (the S6), by the dichotomous results (S6 result, top), thetotal number of positive samples (Positive count, middle) or the average MAP concentration in the samples (mean log(cfu/g), bottom).
Dichotomous Model(Intercept) 0.79 0.14 <0.01 14.02S6 result 0.48 0.20 0.02Farm B −0.83 0.19 <0.01(A is base) C −0.47 0.22 0.03
Count Model(Intercept) 0.74 0.13 <0.01 12.69Positive count 0.17 0.05 <0.01Farm B −0.92 0.18 <0.01(A is base) C −0.46 0.19 0.02
Concentration Model(Intercept) 0.76 0.13 <0.01 12.71log(mean cfu) a 0.27 0.08 <0.01Farm B −0.83 0.18 <0.01(A is base) C −0.37 0.19 0.05
a Log (base-10) transformation.
animals with extremely high shedding rates (super shed-ders) could bias a model based only on prevalence.
This study confirmed that the environment of knownMAP positive dairy farms can be contaminated by animalsshedding MAP, though the proportion of positive sampleswas lower than in previous studies (Lombard et al., 2006;Pillars et al., 2009); the difference is likely a combinationof herd prevalence (which was low for 2 of the 3 herdsin this study) and sample types (which were more var-ied in the current study than in previous studies). Certainenvironments on the farm were more likely to be contam-inated with MAP and had higher average contaminationlevels. These high risk areas included manure storage areasand shared alleyways, parts of the farm in which themanure of adult cows are mixed. High MAP concentrationin fecal cultures increased the amount of MAP found inmanure storage areas, which seems a logical and expectedresult; previous studies have found that MAP can survivein manure storage areas for >200 days (Jorgensen, 1977;Lovell et al., 1944). Fig. 2 shows that there were a large num-ber of negative environmental samples even when highcontemporary fecal shedding levels were measured; thesenegative environmental samples were primarily from areaslabeled ‘Other’ (Table 1), which were not frequented byadult cattle or in contact with their manure and which wereoften not sampled in previous studies of environmentalcontamination. However, negative EC samples from alleysand manure collection areas were also observed in somecases with high contemporary average FC shedding levels(Fig. 2).
Our data lacked quarterly fecal sampling, limiting ouranalysis to the effects of either concurrent samples orsamples from the previous quarters. A recent longitudinalstudy found that the number of MAP-positive EC samplesincreased with increasing FC prevalence, but the conversewas not consistently true, indicating that some envi-ronmental contamination may remain despite successfulreduction of incidence (Pillars et al., 2009). Therefore, wechose to include fecal sampling results from the previousquarter as a proxy for the current shedding level. As these
herds were not yet provided with FC results from the pre-vious quarter during the quarterly EC sampling, little or notest-based culling would be present to bias this assump-tion.
We were interested in the effect of proximity of shed-ding animals on MAP contamination in farm environments,and we found that samples taken from freestall pens, inwhich adult animals directly shared an environment, weresignificantly related to the results of individual FC insidethose pens. The results of pen sampling were correlatedwith the concurrent presence of high-shedding animals. Asthese animals have lower milk production (Nielsen et al.,2008; Smith et al., 2009) and longer calving intervals (Smithet al., 2010), it may be economically desirable to detect theirpresence and remove them. Additionally, high sheddinganimals pose an important risk for other MAP susceptibleanimals in the herd (Lu et al., 2010; Pradhan et al., 2011).Our study observed a correlation between the probabilityof finding MAP in adult cow pens and the presence of highshedders, suggesting that regular sampling of adult cowpens may be a good method for detecting the presenceof highly infectious animals, in agreement with Aly et al.(2009).
We were also interested in the ability of sheddinganimals in the adult cow pens to contaminate other envi-ronments on the farm. Specifically, we wanted to identifya link between EC results for non-pen/non-pit samplesand the distance-corrected shedding level in the herd.Such a link could not be identified in the current data, asstatistical assumptions were not met, although a relation-ship was observed between all non-pen samples and theherd-level results of FC; there were very few positive non-pen/non-pit environmental samples, so the power of thedistance-corrected analysis was low. We also were lim-ited to straight-line distance analyses, ignoring walls andfences, which does not necessarily reflect traffic patternsand other contamination methods. In addition, on one farm(Herd C), alleyways frequently passed through adult cowpens, making distance calculations inappropriate. How-ever, the low number of positive samples suggests that the
Author's personal copy
8 R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9
Fig. 2. Predicted (lines) and observed (symbols) amounts of environmental contamination by Mycobacterium avium subsp. paratuberculosis (MAP) (inlog(cfu/g)) in 3 commercial US dairy herds between 2004 and 2009 based on the sample type (shared alleys; manure pits; all other) and the average fecalsampling of MAP in all adult animals in the herd in the current or previous quarters (in log(cfu/g)). Shedding levels were log-transformed (base-10) afteradding 0.5 to all values.
level of MAP contamination in feed bunks, water sources,and other such locations was negligible; this in itself sug-gests that the optimal use of hygiene program resourcesis to focus on MAP contamination of maternity pens andother known transmission methods.
Our study showed that a single S6 sampling had a rel-ative sensitivity of only 0.40 compared to whole-herd FC;there were only 3 sampling dates with no positive fecalcultures, and the herds were known to be endemicallyinfected, so specificity calculations would be inappropriate.These results agree with the simulations of Tavornpanichet al. (2008), which estimated the herd sensitivity of thistesting method to be 0.38 with a within-herd prevalence of0.05; the 3 herds in this study had a prevalence of 0.05 forthe majority of sampling dates. This has important impli-cations for the interpretation of results from this samplingmethod; a negative result by this method does not neces-sarily mean that the sampled herd is MAP-free. The relativesensitivity of environmental culture was calculated basedon the imperfect FC test, which is known to have low sen-sitivity in low-shedding animals (Collins et al., 2006); this
would lead to a still lower true sensitivity for the VBJDCPprotocol. The S6 samples were unable to significantly pre-dict the presence of MAP in the feces of adult animals on thefarm, nor the FC prevalence of MAP in the herd, althoughboth of these could have been sensitive to the low numberof sampling dates with no positive FC and the overall lowprevalence of MAP in the herds. However, the S6 samples,especially their average MAP cfu’s, were able to predict wellthe average fecal shedding of the herd. This would indi-cate that the S6 method is sensitive to the shedding levelof animals within a herd, and may not be able to detectherds in which fecal shedding is low; this was noted byPillars et al. (2009), who could not culture MAP from envi-ronmental samples taken from herds with a prevalenceof <0.02, and by Raizman et al. (2004), who found thatinfected herds with 2 negative environmental cultures had≤0.04 prevalence by pooled random sampling. As this cat-egory includes newly infected herds, in which the majorityof animals would be latently infected or low-shedding, S6results may not be sufficient to detect herds with new MAPinfections.
Author's personal copy
R.L. Smith et al. / Preventive Veterinary Medicine 102 (2011) 1– 9 9
5. Conclusion
This study shows that the presence of MAP in the envi-ronment of a farm or pen is correlated with fecal sheddingin the cows; however, with our diagnostic methods, MAPmay be absent from environmental samples despite highlevels of fecal shedding in the cows.
Conflict of interest
None.
Acknowledgements
The USDA (Cooperative State Research, Educationand Extension Service, Washington, D.C.) Award Num-ber 2008-35204-04627 provided funding for this study, asdid the USDA-Agricultural Research Service (Agreements.58-1265-3-155, 58-1265-3-156, 58-1265-3-158, and 58-1265-4-020) for the Regional Dairy Quality ManagementAlliance (RDQMA) and the Johne’s Disease Integrated Pro-gram (JDIP, USDA contract 45105). We are grateful to thecomments of the reviewers and the associate editor, whichhave greatly improved our analysis.
References
Aly, S.S., Anderson, R.J., Whitlock, R.H., Fyock, T.L., McAdams, S., Adaska,J.M., Jiang, J., Gardner, I.A., 2009. Reliability of environmental sam-pling to quantify Mycobacterium avium subspecies paratuberculosison California free-stall dairies. J. Dairy Sci. 92, 3634–3642.
Bates, D., Maechler, M., 2010. lme4: linear mixed-effects models using S4classes http://lme4.r-forge.r-project.org/book/ Accessed August 2010.
Benedictus, A., Mitchell, R.M., Linde-Widmann, M., Sweeney, R.W., Fyock,T.L., Schukken, Y.H., Whitlock, R.H., 2008. Calculation of transmis-sion parameters of Mycobacterium avium subspecies paratuberculosisinfections in a dairy herd going through a control program. Prev. Vet.Med. 83, 215–227.
Berghaus, R.D., Farver, T.B., Anderson, R.J., Jaravata, C.C., Gardner, I.A.,2006. Environmental sampling for detection of Mycobacterium aviumssp. paratuberculosis on large california dairies. J. Dairy Sci. 89,963–970.
Collins, M.T., Gardner, I.A., Garry, F.B., Roussel Jr., A.J., Wells, S.J., 2006.Consensus recommendations on diagnostic testing for the detection ofparatuberculosis in cattle in the United States. J. Am. Vet. Med. Assoc.229, 1912–1919.
Lombard, J.E., Wagner, B.A., Smith, R.L., McCluskey, B.J., Harris, B.N., Payeur,J.B., Garry, F.B., Salman, M.D., 2006. Evaluation of environmentalsampling and culture to determine Mycobacterium avium subspeciesparatuberculosis distribution and herd infection status on US dairyoperations. J. Dairy Sci. 89, 4163–4171.
Lovell, R., Levi, M., Francis, J., 1944. Studies on the survival of Johne’s bacilli.J. Comp. Pathol. 54, 120–129.
Lu, Z., Schukken, Y.H., Smith, R.L., Grohn, Y.T., 2010. Stochastic simula-tions of a multi-group compartmental model for Johne’s disease onUS dairy herds with test-based culling intervention. J. Theor. Biol. 264,1190–1201.
Messam, L.L.M., Branscum, A.J., Collins, M.T., Gardner, I.A., 2008. Frequen-tist and Bayesian approaches to prevalence estimation using examplesfrom Johne’s disease. Anim. Health Res. Rev. 9, 1–23.
Nielsen, S.S., Krogh, M.A., Enevoldsen, C., 2008. Time to the occurence ofa decline in milk production in cows with various paratuberculosisantibody profiles. J. Dairy Sci. 92, 149–155.
Nielsen, S.S., Toft, N., 2009. A review of prevalences of paratuberculosis infarmed animals in Europe. Prev. Vet. Med. 88, 1–14.
Pillars, R.B., Grooms, D.L., Kaneene, J.B., 2009. Longitudinal study of thedistribution of Mycobacterium avium subsp. paratuberculosis in theenvironment of dairy herds in the Michigan Johne’s disease controldemonstration herd project. Can. Vet. J. 50, 1039–1046.
Pradhan, A., Van Kessel, J.S., Karns, J.S., Wolfgang, D.R., Hovingh, E., Nelen,K.A., Smith, J.M., Whitlock, R.H., Fyock, T.L., Ladely, S., Fedorka-Cray,P.J., Schukken, Y.H., 2009. Dynamics of endemic infectious diseases ofanimal and human importance on three dairy herds in the Northeast-ern US. J. Dairy Sci. 92, 1811–1825.
Pradhan, A.K., Mitchell, R.M., Kramer, A.J., Zurakowski, M.J., Fyock, T.L.,Whitlock, R.H., Smith, J.M., Hovingh, E., Van Kessel, J.S., Karns, J.S.,Schukken, Y.H., 2011. Molecular epidemiology of Mycobacteriumavium subsp. paratuberculosis in a longitudinal study of three dairyherds. J. Clin. Microbiol. 49, 893–901.
R Development Core Team, 2010. R: A Language and Environment for Sta-tistical Computing. R Foundation for Statistical Computing, Vienna,Austria.
Raizman, E.A., Wells, S.J., Godden, S.M., Bey, R.F., Oakes, M.J., Bentley, D.C.,Olsen, K.E., 2004. The distribution of Mycobacterium avium ssp. paratu-berculosis in the environment surrounding Minnesota dairy farms. J.Dairy Sci. 87, 2959–2966.
Rogan, W.J., Gladen, B., 1978. Estimating prevalence from the results of ascreening test. Am. J. Epidemiol. 107, 71–76.
Smith, R.L., Grohn, Y.T., Pradhan, A.K., Whitlock, R.H., Van Kessel, J.S.,Smith, J.M., Wolfgang, D.R., Schukken, Y.H., 2009. A longitudinal studyfor the impact of Johne’s Disease status on milk production in individ-ual cows. J. Dairy Sci. 92, 2653–2661.
Smith, R.L., Strawderman, R.L., Schukken, Y.H., Wells, S.J., Pradhan, A.K.,Espejo, L.A., Whitlock, R.H., Van Kessel, J.S., Smith, J.M., Wolfgang, D.R.,Grohn, Y.T., 2010. The effect of Johne’s disease status on reproductionand culling in dairy cattle. J. Dairy Sci. 93, 3513–3524.
Tavornpanich, S., Munoz-Zanzi, C.A., Wells, S.J., Raizman, E.A., Carpen-ter, T.E., Johnson, W.O., Gardner, I.A., 2008. Simulation model forevaluation of testing strategies for detection of paratuberculosis inMidwestern US dairy herds. Prev. Vet. Med. 83, 65–82.
USDA-APHIS-VS-CEAH, 2008. Johne’s Disease on U.S. Dairies, 1991–2007.Center for Epidemiology and Animal Health, Fort Collins, CO.
USDA:APHIS:VS, 2008. NAHMS 2007: Johne’s Disease on U.S. Dairy Oper-ations. USDA:APHIS:VS:NAHMS, Fort Collins, CO, pp. 1–4.
USDA:APHIS:VS, 2010. Uniform Program Standards for the VoluntaryBovine Johne’s Disease Control Program. USDA-Animal and PlantHealth Inspection Service-Veterinary Services, Washington, DC.
Wray, C., Wadsworth, Q.C., Richards, D.W., Morgan, J.H., 1989. A three-year study of Salmonella dublin infection in a closed dairy herd. Vet.Rec. 124, 532–535.
