Evaluation of environmental fecal culture for Mycobacterium avium subspecies paratuberculosis detection in dairy herds and association with apparent within-herd prevalence

Abstract

This study evaluated test characteristics of environmental culture (EC) for the detection of Mycobacterium avium subspecies paratuberculosis (MAP) in 32 herds over a 2-year period. Individual fecal samples were collected every 6 mo and environmental samples every 3 mo. Individual fecal culture was performed on samples from positive pools. Samples were cultured in broth, with confirmatory polymerase chain reaction performed on positive fecal samples. Repeated measures were accounted for using GEE logistic models. Relative to a MAP herd-status based on all pooled fecal culture results collected during the study, sensitivity of a set of 6 EC-samples collected from prescribed locations within the herd environment (EC-6) was 71% [95% confidence interval (CI): 49% to 86%] and specificity was 99% (95% CI: 95% to 100%). Sensitivity of EC increased as apparent within-herd fecal culture prevalence (aWHP) increased. The estimated aWHP increased as the proportion of positive EC-samples within an EC-6 set increased. Environmental culture is an acceptable tool for herd diagnosis of MAP in low-prevalence herds.

Introduction

Johne’s disease (JD) is a chronic infectious enteritis of ruminants caused by Mycobacterium avium subspecies paratuberculosis (MAP). Control of JD is motivated by the production-limiting effects of the disease, most notably decreased milk production (1), and premature culling (2) with reduced slaughter weight (3), and by the concern of a possible association between MAP and Crohn’s disease in humans (4).

Determining whether a herd is MAP-positive or MAP test-negative can be challenging. Performing individual cow fecal cultures on an entire herd in order to establish a herd diagnosis for MAP is cost-prohibitive, with each cow fecal culture costing approximately $40 CDN (5). Pooling of individual cow fecal samples (PFC) offers a substantial cost-savings, and has a herd-level sensitivity (Se) of 94% and herd-level specificity (Sp) of 100%, relative to individual fecal culture (6). Another herd-level MAP diagnostic tool is environmental culture (EC), which is relatively cost-effective and non-invasive (7,8).

The majority of studies evaluating EC have assigned herds a MAP test-negative or MAP-positive herd-status against which to evaluate EC (7–9), but there is little information regarding the relationship between EC test characteristics and the within-herd prevalence of these MAP-positive herds. Pillars et al (10) reported that within-herd prevalence levels in 7 study herds ranged from 0% to 42%, but the focus of the research was the distribution of MAP in the environment. Smith et al (11) studied 3 herds, which are described in Pradhan et al (12) as having mean fecal culture within-herd prevalence estimates of 1.5%, 2.5%, and 5.4%, and cautioned against the use of EC in low-prevalence herds, as Se was estimated to be 40%. This is a small number of herds from which to draw this conclusion and further investigation is warranted.

It is important to estimate EC test characteristics in low-prevalence herds because MAP is generally a disease of low within-herd prevalence (13). It was anticipated that EC test characteristics would be lower in dairy herds with low MAP within-herd prevalence, relative to EC test characteristics previously published in which within-herd prevalence of study herds was not considered in the analyses. This study was designed within a longitudinal study frame to allow for a decreased risk of misclassification of low-prevalence, MAP-positive herds, and therefore a strengthened gold-standard herd classification was used to evaluate EC test results. The objectives of this study were to evaluate the test characteristics of EC within MAP test-negative and MAP-positive herds, with the specific focus on evaluating EC test characteristics in herds with low MAP prevalence within purposively selected herds.

Materials and methods

Study design and terminology

A total of 34 herds from the 3 Canadian Maritime provinces participated in this 2-year prospective study. Twenty-seven herds were originally selected to participate based on a non-random selection process. The herds were selected based upon risk assessments completed as part of a previous MAP awareness project at the Atlantic Veterinary College (AVC), with the aim to obtain a mixture of MAP-positive low-prevalence, MAP-positive high-prevalence, and MAP test-negative herds.

The MAP herd-status used for the purpose of evaluating environmental culture was determined by repeated PFC. For the purpose of EC test evaluation, it was assumed that the MAP herd-status was constant during the study period. A MAP test-negative herd was defined as a herd in which no positive-PFC from the herd was detected during the study. A herd was classified as MAP-positive if a positive-PFC was collected during the study, indicating that MAP infectious cows were present in the herd. For individual cow classification, which was used to establish an apparent within-herd prevalence estimate (aWHP), pools of 5 individual cows were cultured (PFC), with follow-up culture of all individual samples from positive-PFC (PFC-IC). The aWHP was the number of positive PFC-IC divided by the number of cows from which fecal samples were collected in the herd. Using the mean aWHP, a herd was classified as low-prevalence if < 5% of cows within a herd were PFC-IC positive. If ≥ 5% of cows within a herd were PFC-IC positive, the herd was classified as high-prevalence. Because at the first round of sampling there were too few MAP-positive herds, 7 herds were added for the remaining 18 mo of the project. These herds were chosen based on historic MAP-positivity, and all 7 herds were PFC-positive during the study. Because PFC is an imperfect gold-standard test, three or more rounds of cow fecal cultures had to be collected from a herd in order to establish MAP herd-status and be included in the analyses. An EC-6 set was considered positive if there were ≥ 1 EC-samples positive within the EC-6 set. An EC-6 set was considered a test-negative EC-6 (set) if there were no EC-samples positive within the EC-6 set.

Individual cow and environmental manure sample collection

In order to establish a reference MAP herd-status, pooled fecal samples were created by cow age from individual fecal samples collected at 6 mo intervals from all lactating cows, using a clean rectal sleeve lubricated with water. Approximately 30 g of feces were collected per rectum and placed in a clean, labeled 95-mL plastic specimen jar. A set of environmental samples, consisting of 6 manure samples collected from specific sites within the herd environment, was collected every 3 mo. This set of 6 samples is referred to as an EC-6 (set), and each of the 6 samples comprising the set is referred to as an EC-sample. The EC-6 samples were collected following a protocol based on the Voluntary Bovine Johne’s Disease Control Program (14). Two EC-samples were collected from each of the manure storage areas (pits, lagoons, manure piles, or manure spreaders), the mature cow manure concentration areas (alleyways, gutters, adjacent to waterers or feeders) and the mature cow maternity or sick pens if there were 2 or more animals in the pen, and if manure clean-out did not occur between animals. Most farms had only 1 cow in the sick/maternity pens at one time, with manure clean- out between cows. In these cases, 2 additional EC-samples (4 EC-samples total) were collected from the manure concentration areas. In tie-stall barns, manure concentration EC-samples were collected from corners and crevices of the gutter and along the paddles of the stable-cleaner. To create an EC-sample for both tie-stall and free-stall sites, 4 “grabs” of manure within the specified collection site were collected into clean 95-mL specimen cups to form 1 composite EC-sample. Each EC-sample was collected using a clean latex glove. Individual cow and environmental manure samples were kept cool during transport to the laboratory. If samples could not be processed immediately they were frozen. Samples were frozen at −20°C if processing was to occur within 2 wk and were frozen immediately at −80°C if processing was planned between 2 and 12 wk after collection.

Laboratory testing

Fecal culture

All manure samples were processed by the Maritime Quality Milk (MQM) Laboratory (AVC, Charlottetown, Prince Edward Island), which was approved by the United States Department of Agriculture (USDA) proficiency-testing for this technique. Individual cow fecal samples were pooled by age into PFC samples, with samples from 5 cows in each PFC. Individual cow samples that made up positive-PFC samples were thawed and cultured individually (PFC-IC). Fecal samples were processed and inoculated into ESP para-JEM broth (Nova Century Scientific, Burlington, Ontario), according to the manufacturer’s protocol, with the exception that samples were incubated for 49 d, rather than 42 d (15,16). The broth culture bottle was placed in the ESP Culture System II (TREK Diagnostic Systems, Cleveland, Ohio, USA) for incubation up to a maximum of 49 d.

Confirmation of a positive fecal culture

Confirmatory tests were initiated when the culture system indicated a sample to be positive by headspace pressure change. If confirmatory polymerase chain reaction (PCR) was negative, the sample was returned to the culture system. After the 49 d of incubation, all broth samples were examined microscopically for Mycobacterium using an acid-fast stain. As in McKenna et al (17), final confirmatory PCR was performed on all samples positive by the culture system and/or microscopy. The confirmatory PCR (VetAlert Johne’s Real-Time PCR kit, Tetracore, Rockville, Maryland, USA), which detects the hspX gene, was performed following manufacturer’s instructions. Fecal samples (PFC, PFC-IC, and EC-sample) were classified as positive (PFC positive, PFC-IC positive, and EC-sample positive, respectively) if both broth culture and confirmatory PCR tests were positive.

Statistical analyses

All descriptive and statistical data analyses were conducted using STATA Version 11 (StataCorp LP, College Station, Texas, USA). A P-value ≤ 0.05 was considered significant. In order to evaluate if an EC-sample positive result was more likely from a specific location within the herd environment, a logistic regression model was developed, with EC-sample result as the outcome and EC-sample collection location as a predictor. Sensitivity and Sp of an EC-6 set were calculated using generalized estimating equations (GEE) with an exchangeable correlation structure to account for the repeated measures data. Sensitivity of an EC-6 set, relative to a MAP-positive herd-status based on all PFC results collected during the study, was estimated using a logistic null model. The model outcome was EC-6 results from MAP-positive herds and the equation:

$$\text{Se}=e^{\mu +}/(1+e^{\mu +}),$$

where: μ+ = β₀ + ∑β_j X_j was the linear predictor from the model (18) that was used to calculate Se of EC. Similarly, Sp of an EC-6 set, relative to a MAP test-negative herd-status based on all PFC results collected during the study, was derived from the linear predictor of a logistic null model based on MAP test-negative herds, using the equation:

$$\text{Sp}=1-[e^{\mu -}/(1+e^{\mu -})].$$

Season, housing system, and aWHP were entered into the null models. If the P-value of these predictors on univariable analysis was ≤ 0.15 they were entered into a multivariable logistic GEE model. A P-value of ≤ 0.05 was considered significant for inclusion of a variable in the final model. Lowess smoothers were generated to evaluate linearity and fractional polynomial models were created to explore power transformations of significant predictors in an effort to optimize linearity.

In order to evaluate the impact of repeated EC sampling on EC-6 Se and Sp, a pair-wise combination of consecutive EC-6 results collected were interpreted as positive if either of the 2 consecutive EC-6 sets was positive. This pair-wise interpretation was used as the outcome in null logistic GEE models, and Se and Sp were calculated as described. This was repeated with triplicate and upward combinations of EC-6 results, where a combination of EC-6 results was considered positive if at least 1 EC-6 set in the combination had been EC-6 positive.

A cow-level GEE logistic model was constructed to evaluate if the proportion of positive EC-samples within 2 consecutive EC-6 sets was predictive of the aWHP. Data structure is determined by the outcome variable (semi-annual PFC) and therefore EC-6 results were combined from quarterly into semi-annual measures. The outcome of this cow-level model was the probability for a cow within a herd to be PFC-IC positive. This individual cow-level probability applied to each cow within the herd, and was therefore analogous to the aWHP. Linearity of the predictor was evaluated. Herd level predictors season and herd size were added to the null model. If the P-value of these predictors in univariable analysis was ≤ 0.15 they were entered into a multivariable logistic GEE model.

Results

Herd demographics

In total, 34 herds participated in the project between April 2009 and March 2011. Two herds did not meet the criteria for inclusion in the analyses because of incomplete sampling and were excluded. The sampling schedule for 7 herds consisted of 3 individual cow fecal and 6 environmental collections, and 25 herds had 4 cow fecal and 8 environmental collections. Median herd size was 66 milking cows (mean: 82; range: 28 to 220 cows). Median herd cow-age at testing was 4.1 y (mean: 4.1; range: 2.9 to 5.5 y). Eleven facilities (34%) were tie-stall, and 21 facilities (66%) were free-stall. One herd expanded from 190 to 220 milking cows during the project. This herd had a PFC-positive test prior to the introduction of new animals. Remaining herds, while not necessarily closed, did not have substantial introductions during the project. No farms had sheep or goats on the premises.

Herd prevalence

Overall MAP herd-status, based on PFC-results from the study period, was MAP test-negative for 18 herds (56%) and MAP-positive for 14 herds (44%). Nine MAP-positive herds had ≥ 1 PFC-positive result at each round of sampling and 5 MAP-positive herds fluctuated between test-negative PFC results and ≥ 1 PFC-positive result at each herd visit (Table 1). For the 5 MAP-positive herds that did not have ≥ 1 PFC-positive result at every round of sampling, 2 herds had 3 of 4 herd PFC collections with ≥ 1 positive PFC result. The remaining 3 MAP-positive herds that did not have ≥ 1 PFC-positive result at each round of sampling had positive PFC results in the second and third rounds of herd PFC collections. Mean aWHP for all 32 herds, based on PFC-IC results, ranged from 0% to 15.6%, with little clustering of aWHP values between the minimum and maximum aWHP (Figure 1). In the case of a positive PFC with negative PFC-IC results, for determination of aWHP it was considered that 1 cow in the positive-PFC was MAP-positive. The mean aWHP of the 3 herds in which this situation occurred a total of 4 times was 0.9%.

Table 1.

Summary of pooled fecal culture and environmental culture results for 15 herds with a minimum of 1 Mycobacterium avium subsp. paratuberculosis-positive fecal pool or environmental culture set

	Pooled fecal culture (PFC)		Environmental culture (EC)
Milking herd size	Herd visits with ≥1 positive PFC	Mean aWHPa (%)	Herd visits with a positive EC-6 result	Total positive EC-samples (%)
75	3/3b	15.6	6/6c	92
95	3/3	12.3	6/6	94
90	3/3	10.9	6/6	92
220	3/3	10.0	6/6	94
100	3/3	10.0	6/6	100
70	4/4	9.0	6/6	81
150	3/3	5.5	6/6	81
30	4/4	7.1	7/8	46
50	4/4	2.9	6/7	33
47	3/4	1.6	4/8	10
145	3/4	0.7	4/8	14
60	2/4	0.8	1/8	4
50	1/3	0.7	1/6	3
120	1/4	0.2	0/8	0
45	0/4	0.0	1/8	2

^aApparent within-herd prevalence (aWHP) is based on PFC with culture of individual cow samples from positive pools.

^bNumber of herd visits with ≥ 1 positive PFC/total number of herd visits.

^cNumber of positive EC-6 sets/total number of EC-6 collections, where EC-6 represents a set of 6 EC-samples collected from specified sites within the herd environment).

Figure 1.

Mean apparent within-herd Mycobacterium avium subspecies paratuberculosis fecal culture prevalence^a compared to the proportion of positive environmental culture sets in 15 dairy herds that were fecal culture and/or environmental culture-positive. Number markers beside data points indicate the proportion of environmental culture samples that were positive within all environmental culture sets collected from that herd.

^a Apparent within-herd prevalence is based on pooled fecal culture with culture of individual cow samples from positive pools.

Environmental cultures

Summary information of PFC and EC results in herds with ≥ 1 positive PFC or ≥ 1 positive EC-6 is displayed in Table 1. One MAP test-negative herd (no positive-PFC) had 1 positive EC-6 set and 7 test-negative EC-6 sets. One MAP-positive herd had no positive EC-6 sets throughout the study. The MAP-positive herd-status for this herd was the result of 1 positive PFC, from which no positive PFC-IC cows were identified, and the herd’s mean aWHP was 0.2%. Seven of the 14 MAP-positive herds had ≥ 1 test-negative EC-6 sets. The mean aWHP for these 7 herds was 2.1%. Twelve of 14 herds with ≥ 1 positive EC-6 set were EC-6 positive in the first round of testing, one herd was EC-6 positive in the second round and one in the seventh round of EC-6 sampling.

Over the course of the project, 1 EC-6 positive set and 138 test-negative EC-6 sets were collected from the 18 MAP test-negative herds. In the 14 herds that were MAP-positive, over the course of the project there were 67 positive EC-6 sets and 29 test-negative EC-6 sets. In total, 235 EC-6 sets were collected, and 231 EC-6 sets (1386 EC-samples) had collection location identified. On average, 20% of the EC-samples from these rounds were positive (Table 2). Numerically, MAP was recovered less than 40% as frequently from cow concentration areas (i.e., maternity and sick cow pens) compared to manure concentration (i.e., alleyways or gutters) or manure storage areas. However, when herd was accounted for within the model, collection location was not significantly associated with the EC-sample result (P = 0.13).

Table 2.

Summary statistics for environmental culture samples collected and percent Mycobacterium avium subsp. paratuberculosis culture-positive by location on 32 dairy herds from the 3 Maritime provinces of Canada.

Location	Number of samples	Percent positive
Manure storage (e.g., lagoons, manure piles)	410	21.6%
Manure concentration (e.g., alleyways, manure gutters)	878	20.7%
Cow concentration (e.g., calving/sick pens)	98	8.2%
Total	1386	20.4%

Test characteristics of environmental culture

Based on null logistic GEE models (Table 3), the Se of an EC-6 set was 71% (95% CI: 49% to 86%) and Sp was 99% (95% CI: 95% to 100%). Sensitivity increased with increasing aWHP (Table 3 and Figure 2). Being MAP-positive overall did not preclude the herd from having aWHP values of 0% at some times throughout the study. Sensitivity was not affected by the season of EC-6 set collection. When the results of successive EC-6 collections were evaluated for accuracy of determining MAP herd-status, there was no significant effect of using more than 1 EC-6 set on either the Se or Sp of an EC-6 set, regardless of the number of successive EC-6 sets that was used to classify herds. However, a trend toward improved Se was observed (Table 4) with each successive EC-6 set added. The greatest numerical increase in Se occurred between a single EC-6 set and 2 consecutive EC-6 sets (71% versus 81%) (Figure 3).

Table 3.

Three logistic generalized estimating equation models for Mycobacterium avium subsp. paratuberculosis environmental culture sensitivity and specificity. Environmental culture result was the outcome in all models, with null models estimating overall sensitivity (MAP-positive herds) and specificity (MAP-negative herds) of environmental culture, and the multivariable model predicting the impact of fecal culture within-herd prevalence on the sensitivity of environmental culture

		95% Confidence interval
Model	Estimate	Lower	Upper	P-value
Null logistic model of MAP-negative herds (Sp)
(Intercept)	−4.93	−6.85	−3.02	< 0.001
Null logistic model of MAP-positive herds (Se)
(Intercept)	0.91	−0.03	1.86	0.058
Multivariable logistic model of MAP-positive herds (Se)
(Intercept)	−1.53	−2.40	−0.65	0.001
Fecal culture test prevalence (proportion)	78.08	41.42	114.73	0.000

Figure 2.

Sensitivity and 95% confidence intervals (CI) of single sets of 6 environmental culture samples (EC-6) by apparent Mycobacterium avium subspecies paratuberculosis within-herd test prevalence^a, using a logistic generalized estimating equations model.

^a Apparent within-herd prevalence is based on pooled fecal culture with culture of individual cow samples from positive pools.

Table 4.

Results of logistic generalized estimating equation models to evaluate sensitivity and specificity of environmental culture (EC), with increasing numbers of EC sets used to determine herd status

Number of combined EC setsa	Models using MAP-positive herd			Models using MAP test-negative herds
	Intercept (P-value)	#obs (#herds)	Sensitivity (95% CI)	Intercept (P-value)	#obs (#herds)	Specificity (95% CI)
1	0.914 (0.058)	97 (14)	71% (49%–86%)	−4.93 (< 0.001)	140 (18)	99% (95%–100%)
2	1.45 (0.019)	83 (14)	81% (56%–93%)	−4.11 (< 0.001)	122 (18)	98% (90%–100%)
3	1.56 (0.017)	69 (14)	83% (57%–94%)	−3.95 (< 0.001)	104 (18)	98% (88%–100%)
4	1.64 (0.015)	55 (14)	84% (58%–95%)	−3.76 (< 0.001)	86 (18)	98% (86%–100%)
5	1.77 (0.012)	41 (14)	85% (59%–96%)	−3.53 (< 0.001)	68 (18)	97% (82%–100%)
6	2.24 (0.013)	27 (14)	90% (62%–98%)	−3.23 (< 0.001)	50 (18)	96% (72%–100%)
7 and 8	No convergence of models			No convergence of models

^aAn EC set consisted of 6 samples collected from 6 sites within the farm environment.

Figure 3.

Predicted sensitivity (Se) and specificity (Sp), with 95% confidence intervals (CI), of sets of 6 environmental culture samples (EC-6) for Mycobacterium avium subspecies paratuberculosis, by the number of EC-6 sets used to determine herd status.

Proportion of positive EC-samples within positive EC-6 sets

Sixty-eight of the 235 EC-6 sets collected (29%) were positive. In 43% of the EC-6 positive sets, all 6 EC-samples within the set were positive. The mean percentage of positive EC-samples within a positive EC-6 set was 71%, meaning a positive EC-6 set contained, on average, 4 to 5 positive EC-samples. The proportion of positive EC-samples within an EC-6 set was a significant predictor of the probability for a cow within a herd to be PFC-IC positive. This cow-level probability is equal for each cow in the herd and is analogous to the aWHP. As the proportion of positive EC-samples increased, this within-herd cow-level probability (aWHP) increased (Figure 4). For best fit, the model required a log transformation of the predictor. No other herd level predictors were significant.

Figure 4.

The effect of the proportion of positive Mycobacterium avium subspecies paratuberculosis environmental culture samples on the probability^a of a cow within a herd being positive based on pooled fecal culture with individual fecal culture follow-up (PFC-IC).

^a The probability of a cow to be PFC-IC positive is equal for each cow within the herd, and is analogous to the apparent within-herd prevalence.

Discussion

Distinguishing a MAP test-negative herd from a MAP-positive, low-prevalence herd can be difficult (19). The test characteristics of EC in MAP-positive, low-prevalence herds have previously been questioned (11), and since most MAP-positive dairy herds are expected to be low-prevalence (20), it is critical to understand how EC will perform in these herds. The wide range of MAP aWHP in this study, and in particular the high proportion of low-prevalence herds, fill a knowledge gap regarding EC test characteristics.

Repeated sampling will maximize the identification of MAP-positive herds (21) and thereby minimize the influence of mis-classification bias on herd Se and Sp (22). In the current study, 5 of the 14 MAP-positive herds had at least 1 herd test with no positive PFC samples. The mean aWHP in these herds was 0.8%. As a comparison, the mean aWHP in the 9 MAP-positive herds that had ≥ 1 positive PFC at every herd test was 9.1%. These results indicate that low-prevalence herds are more susceptible to misclassification of MAP herd-status based on cross-sectional sampling. One of the assumptions made in this analysis was that MAP herd-status was stable over the study period. A MAP test-negative herd-status should be based on multiple negative tests from adult cattle in the herd (23). All 18 study herds classified as MAP test-negative had 4 whole-herd PFC collections with no positive-PFC results, and while they were not designated as closed herds, none of the MAP test-negative herds increased in size during the study. There are limited field data supporting the short-term efficacy of eradication programs, even in MAP-positive low-prevalence herds (24). As a result, a change from MAP-positive to MAP test-negative herd status was considered unlikely in a 2-year time window. Nine of the 14 MAP-positive herds were PFC-positive at all herd collections and the 5 herds with intermittent MAP PFC-positivity were not more likely to be positive at the beginning or end of the 2-year study window.

In general, studies involving EC focus on MAP-positive herds (10,11,25), and the Se of EC has been examined more frequently than Sp. Raizman et al (8) studied 108 herds, 28 of which had been historically classified as uninfected. An EC sampling program, from which 2 individual environmental samples were collected, resulted in 1 of these 28 herds being classified positive by EC. Lombard et al (9) sampled 98 herds. Ten of the 60 herds tested with cow fecal culture were considered negative based on a single fecal culture sampling. Of these 10 herds, 2 were EC-positive. The herds classified as overall MAP test-negative in this study are a good subset from which to calculate EC Sp due to the repeated PFC sampling. Specificity estimates from the current research may be somewhat underestimated, as individual cow fecal samples were only collected from the milking herd. Dry cows, which typically represent 15% of the adult cow population in a herd, were not included in the individual cow sampling in order to facilitate sample collection. Therefore, the cows represented in the EC samples were not identical to the individual cows sampled. Repeated sampling increased the opportunity for all cows to be included in the PFC collected during the study.

Biologically, EC would be expected to have an Sp of 100%. In this project, a positive culture was confirmed with PCR analysis to ensure the positive culture was not another Mycobacterium species. In this study there was 1 MAP test-negative herd with 1 positive EC-sample within an EC-6 set. The other 7 EC-6 sets collected from this herd were EC-6 test-negative. A low within-herd prevalence is one of the most likely reasons a herd would be EC-positive yet MAP test-negative.

Previous studies using fecal culture to define MAP herd-status reported Se values for EC ranging from 40% to 74% (7,9–11,25). Variation in reported Se is due, in part, to the herd in which the test is applied, and Se is expected to be higher in high-prevalence herds (11). This was consistent with study results, and although EC Se was relatively low at very low aWHP values (≤ 2% aWHP), it increased quickly and approached 100% Se at moderate aWHP levels of approximately 8%. The aWHP estimates may be an underestimation of the true within-herd prevalence for these herds, as PFC-IC was performed only on cows from positive-PFC. A single cow shedding low levels of MAP may not have been detected in the PFC, and therefore would have been a false-negative result, contributing to an underestimation of the true within-herd prevalence. Individual PFC-IC samples also underwent an additional freeze-thaw cycle, which could have contributed to an underestimation of the aWHP.

In agreement with previous studies (9), no other herd predictors were significantly associated with Se or Sp. Only 1 MAP-positive herd had tie-stall facilities, precluding statistical evaluation of the relationship between housing type and EC Se. In this 50-cow tie-stall herd, 6 of 7 EC-6 sets collected were positive, and mean aWHP was 2.9%. Although not statistically significant, this suggests it is possible to detect MAP-positive, low-prevalence herds in tie-stall facilities using EC. Further study into the Se of EC for tie-stall facilities is required.

When evaluating collection location in all study herds, manure storage and concentration areas had a numerically greater, although not statistically significant, number of positive EC-samples than cow concentration areas. Previous studies have also reported manure storage and shared alleyways as the sites most likely to be positive (8,10,11). Collecting EC-samples from manure storage and concentration sites may provide the optimum chance to detect MAP.

Repeated herd testing to determine MAP herd-status is frequently used in control programs, such as the Voluntary Bovine Johne’s Disease Control Program (14) in the United States. In this dataset, information from additional, consecutive EC-6 sets did not significantly change the estimated Se and Sp of EC. Numerically, the greatest increase in Se occurred with a second set of EC-6 samples, with minimal change in Se with additional sets of EC-6. Even though this was a large study with 32 herds, when analyzing data at the herd level confidence intervals tend to be very large, making it difficult to establish statistically significant differences between point estimates and predictors. The difference in the 2 point estimates is such that the reader may want to consider the difference relevant for their purposes.

As the proportion of positive EC-samples within an EC-6 set increased, the probability increased for a cow within that herd to be PFC-IC positive. This cow-level probability is analogous to the aWHP. When 1 of the 6 EC-samples in an EC-6 set was positive, estimated aWHP was 0.7%, while when all 6 EC-samples were positive, estimated aWHP was 12.7%. The proportion of positive EC-samples within an EC-6 set provides an indication of MAP aWHP. This knowledge is advantageous in MAP-control programs. For example, aggressive testing may be more beneficial and cost-effective in high-prevalence herds (6). When a herd had no positive EC-samples over 2 consecutive EC-6 samplings, the model predicted the probability of a cow being PFC-IC positive to be < 1%. Previous studies have estimated herd prevalence to be ≤ 2% when EC was negative (10).

Environmental culture is an economical and non-invasive method of determining MAP herd-status. Knowledge of MAP herd-status and an estimate of aWHP are valuable tools for herd MAP control programs. Environmental culture is an acceptable herd test for classification of MAP herd-status in MAP test-negative and MAP-positive herds.