Bovine paratuberculosis is a chronic and infectious mycobacterial infection of the gastrointestinal tract (13
). At least 22% of U.S. dairy herds are infected with M. paratuberculosis
). Consolidation of the U.S. dairy industry and the ongoing practice of buying cattle without regard for the M. paratuberculosis
infection status of either the purchased cattle or their herd of origin facilitates the spread of the infection among herds. The 20% fecal culture-positive rate seen among the seven infected dairy herds used in the present study is not unusual. Moreover, given the latency of this infection and the diagnostic sensitivity of fecal culture, the true infection rate in these herds is even higher.
For paratuberculosis diagnostics it can be challenging to evaluate an assay's specificity. In dairy cattle the time from infection, usually as a calf, to clinical disease varies from as short as 1 year to as much as 12 years, the upper limit of life span for dairy cattle on most farms (17
). Prolonged disease latency makes it impossible to rely on a negative fecal culture as a “gold standard” to define absence of M. paratuberculosis
infection for animals residing in known infected herds. Instead, the best standard for absence of M. paratuberculosis
infection for an individual animal must be based on absence of infection from the entire herd in which the animal was raised. Moreover, absence of infection from a herd must be based on multiple annual negative tests of all adult cattle, as well as strict adherence to recommended management practices that prevent introduction of M. paratuberculosis
-infected cattle into the herd. These standards are outlined in the Voluntary Bovine Johne's Disease Control Program (35
) and served as the gold standard to qualify the adult cattle in seven dairy herds for ELISA specificity estimation in the present study.
The diagnostic specificity for three of the five ELISAs evaluated was ≥99.8% and not statistically different. This corroborates the findings of similar studies on the same ELISA kits (27
). ELISAs A and D had significantly lower specificities, 94.9 and 84.7%, respectively (Table ). The specificity rates of these two assays also varied among herds, whereas the other three ELISAs' specificity rates did not. Reasons for assay specificity differences among herds are not known but likely relate to qualitative or quantitative differences in the microbial flora the animals in those herds are exposed to, notably the presence of other mycobacteria, as well as the composition of antigens used as solid-phase antigens on ELISA plates and/or antigens contained in serum or milk absorption reagents. Herd-to-herd assay specificity variation is an important consideration in judging assay performance; it may even be necessary to include herds from different geographic areas (and therefore likely varied microbial flora) to obtain a full sense of the range of specificities displayed by the assay. Herd screening for possible M. paratuberculosis
infection by ELISA demands use of high specificity assays to limit false-positive herd classifications and the time and effort spent trying to confirm the diagnosis (7
Among the fecal culture-negative cattle in the infected herds, <2.9% tested positive for serum antibody and <3.7% tested positive for antibody in milk for the three highest-specificity ELISAs. Arguably, these apparently ELISA false-positive cattle were truly infected with M. paratuberculosis, but fecal culture failed to detect the organism because the animal was not shedding it in feces on the day of sampling or was shedding at a level below the culture detection limits. M. paratuberculosis exposure, triggering antibody production without progressive infection, e.g., exposure as an adult when more infection resistant, is also a plausible explanation for ELISA positive findings in fecal culture-negative cattle. Regardless, this was a low-frequency occurrence for the high-specificity assays and of limited concern should these assays be used in a paratuberculosis control program.
The sensitivity of M. paratuberculosis
fecal culture-positive dairy cattle detection by ELISA was the same for four of five assays and was related to the rigor of the case definition. If the sensitivity of these same assays were judged in comparison to the fecal culture results of a single laboratory (Minnesota Veterinary Diagnostic Laboratory), then the estimated sensitivities would be much higher, i.e., 45.7 to 50.0%, and in agreement with previous reports (6
). It is noteworthy that the diagnostic sensitivity of a new direct high-throughput PCR assay applied to fecal samples from cattle used in the present study, using the same single case definition, was also in this range (data not shown). Clearly, the fecal culture methodology and number of samples evaluated to establish a case definition affect the estimates of the sensitivity of other assays being evaluated and are a major study design issue to be considered when evaluating diagnostic tests for paratuberculosis. Given that antibody-based diagnostics detected fewer than one-third of all M. paratuberculosis
-infected and excreting dairy cattle in a herd, the percentage of cattle testing positive could be multiplied by at least a factor of three to get a rough estimate of true within herd prevalence of infection. This is only applicable, of course, for herds that have not previously been testing for paratuberculosis and culling cattle based on test results.
The number of CFU of M. paratuberculosis
in bovine fecal samples is considered a measure of the stage of infection in the animal. In the present study, the ability of an ELISA to detect an infected animal was directly related to the level of M. paratuberculosis
shedding (Table ), a finding consistent with previous reports (6
). Cattle with large numbers of M. paratuberculosis
per gram of feces, so-called “heavy shedders,” were detected by the evaluated ELISAs >72% of the time. This is significant in that these animals logically represent the greatest risk for environmental contamination and thus infection transmission on a dairy farm. Evidence from field studies to support the contention that detection and management of the heavy shedders is critical to control of bovine paratuberculosis are lacking. However, both modeling and field data on human tuberculosis verify that focusing treatment and control efforts on the most infectious subset of Mycobacterium tuberculosis
-infected persons can cause dramatic reductions in the prevalence of the infection in the general population (10
). Whether the cattle shedding large numbers of M. paratuberculosis
but missed by ELISAs quickly progress to clinical paratuberculosis and are therefore culled from the herd or remain in the herd and continue to spread infection is not known. Only well-designed and controlled longitudinal field studies will be able to answer these questions.
When ELISA results were judged dichotomously (positive or negative), the assays showed comparable accuracy and good agreement. ELISAs A, B, C, and E did not differ in classification of culture-positive cattle as test positive or negative (χ2, P > 0.05), and the kappa statistic indicated a high level of agreement (kappa ≥ 0.66; Table ). Of these assays, B, C, and E were not significantly different in either sensitivity or specificity. ELISA D would have had comparable sensitivity and specificity estimates, i.e., 31.8 and 97.5%, respectively, if an assay cutoff of 250% (sample OD as a percentage of kit positive control) were used instead of 125% as recommended by the kit manufacturer (Table ).
Scatter plots illustrate that serum or milk from individual cattle can respond very strongly when tested by one ELISA but not another. Representative plots are shown in Fig. , and linear regression correlation coefficients for all pairwise assay comparisons are listed in Table . Clearly, cattle can respond to M. paratuberculosis infection with serum antibodies detected by one ELISA kit but not in another. Since the formulation of the M. paratuberculosis solid-phase antigen and M. phlei serum absorption antigens used in these kits is proprietary, the nature of putative antigenic differences in kit components is not known. Identification of the diversity of M. paratuberculosis antigens or epitopes that induce antibody responses in cattle could lead to improved diagnostic kits. Purified single antigen assays will likely suffer from low diagnostic sensitivity.
LR analysis confirmed the earlier report that multilevel paratuberculosis ELISA interpretation is useful (4
). All five ELISAs evaluated showed a direct relationship between the magnitude of ELISA result and the likelihood the tested cattle were fecal culture-positive for M. paratuberculosis
. LRs in combination with estimated within herd prevalence, i.e., the pretest probability of M. paratuberculosis
infection, can be used to calculate the post-test probability of infection (shown for ELISA B in Table ). For ELISAs A, B, C, and E, cattle with results in the highest of the five ELISA result ranges, so called “strong-positive” results, had a post-test probability of concurrent M. paratuberculosis
fecal shedding of >90% if herds had a within herd prevalence >10%. For ELISAs B and C, a high (>90%) post-test probability of M. paratuberculosis
infection was also possible for the next-lower level of assay result when herds had a >15% within-herd infection prevalence. For most M. paratuberculosis
-infected commercial dairy herds, confirmatory testing of such animals by fecal culture would be neither necessary nor cost-effective. Post-test infection probabilities can be used in decision analysis models to optimize economic outcomes from decisions based on ELISA results. However, this requires estimation of the infection transmission risks for animals at each level of M. paratuberculosis
shedding, a factor affected by on-farm management systems. Measurement of infection transmission risks for cows by stage of infection under different farm management conditions by well-controlled epidemiological field studies is an important next step in optimization of paratuberculosis control programs in dairy herds.
An important first step in a national paratuberculosis control program is segregation of cattle herds into infected and not infected categories (1
). To halt the spread of M. paratuberculosis
infection, the next step is to apply voluntary or mandatory biosecurity regulations to prevent movement of cattle from infected (or untested) herds into noninfected or at least test-negative herds. Rapid, low-cost, high-throughput assays are necessary if even a modest proportion of the 95.8 million cattle in the United States are to be tested (34
). The use of high-specificity ELISAs avoids unnecessary follow-up testing of herds due to false-positive tests (7
). Higher assay sensitivity than that available with current ELISAs for detection of infected individual animals would be desirable, but herd-level sensitivity is the more critical factor, and this can be improved by increasing the number of animals tested per herd and by focusing testing on the older animals since they have had sufficient time for a M. paratuberculosis
infection to progress and induce antibody production (15
ELISA E performed as well or better on milk samples than the ELISAs that used sera and are in agreement with a recent report on evaluation of this same commercial test (14
). Assays that use milk samples are ideally suited for paratuberculosis control in dairy herds but impractical for use on beef cattle. Milk samples are normally collected by testing laboratories from all lactating cows in a herd on a monthly basis for measurement of somatic cell counts (a measure of udder health), plus milk fat and protein content. Milk testing laboratories are equipped with automated sample handing equipment that allows thousands of samples to be analyzed daily. In addition, electronic transmission of results downloaded to on-farm computers is common. These data are used on a daily basis for herd management. Milk testing laboratories could thus adapt paratuberculosis ELISA technology to expand the services they provide to dairy producers and more cost-effectively test dairy cattle. Unfortunately, at present there are no USDA-licensed kits validated for testing milk samples for antibodies to M. paratuberculosis
Organism detection tests for paratuberculosis will remain a mainstay for the definitive diagnosis of paratuberculosis. Isolation of M. paratuberculosis
in liquid or on solid medium not only provides the most sensitive and specific diagnostic method available but also offers the ability to complete molecular studies that increase our understanding of the epidemiology and microbial ecology of this pathogen (3
). Genetics-based diagnostic tests have not yet matched the accuracy of culture-based diagnostics for paratuberculosis and fail to provide viable organisms for further study (9
). If M. paratuberculosis
becomes classified as a zoonotic pathogen and for adequate food safety it is decided that dairy and beef products only originate from paratuberculosis test-negative herds, the need for all types of paratuberculosis diagnostics on multiple types of animals and possibly humans will be enormous.