If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Testing of bulk milk (BM) samples is a convenient, cost-effective strategy that can easily be implemented as part of disease surveillance programs on dairy farms. Here, we performed a scoping review to summarize the literature reporting on the testing of BM samples to detect infectious diseases of dairy cattle caused by bacteria. We also provide a non-exhaustive, albeit significant, list of diagnostic tests that are marketed for BM samples, as well as a list of disease surveillance activities that included testing of BM samples. A literature search was carried out in 5 databases, yielding 8,829 records from which 474 were retained. Overall, 575 eligible bacterial pathogens were screened for using BM samples, ranging from 1 to 6 individual pathogens per study. Staphylococcus aureus, including methicillin-resistant Staph. aureus, were the most studied bacteria (n = 179 studies), followed by Streptococcus agalactiae (86), Mycobacterium avium ssp. paratuberculosis (79), Coxiella burnetii (79), and Mycoplasma spp. (67). Overall, culture-based protocols, ELISA, real-time PCR, and PCR were the most commonly adopted methodologies to screen BM samples. Sensitivity of BM testing for bovine paratuberculosis was generally low and varied greatly according to the ELISA cut-offs adopted and herd-level definition of disease. In general, protocols had low to moderate sensitivities (<50%), which increased for herds with high within-herd seroprevalence. Specificity of BM testing for paratuberculosis was generally high. With respect to mastitis pathogens, BM testing demonstrated high sensitivity and specificity for Strep. agalactiae, in general. However, we observed inconsistency among studies with respect to the sensitivity of BM culture to detect infected herds, which was notably higher if enrolled herds were heavily infected or had history of clinical disease. Among Salmonella spp. pathogens, Salmonella Dublin was the most frequently studied bacterium for which BM testing has been validated. Specificity of BM ELISA was high, ranging from 89.0 to 99.4. In contrast, sensitivity varied greatly among studies, ranging from 50.6% to 100%. Our findings support that one of most important factors affecting sensitivity of BM ELISA for Salmonella Dublin is whether nonlactating cattle are considered in the definition of herd infection status. In general, protocols analyzed in this review suffered from very low sensitivities, which hardly justifies their use as part of disease surveillance as single testing. Nevertheless, test sensitivity can be increased by the adoption of more inclusive definitions of disease-free herds. Further, low-sensitivity and high-specificity methods can be valuable tools for surveillance when used repeatedly over time.
Despite years of concerted efforts by the dairy community, infectious diseases continue to plague the dairy industry worldwide. Diseases such as mastitis, paratuberculosis, and leukosis are prevalent in dairy herds and are responsible for major economic losses globally (
). Furthermore, zoonotic diseases caused by bacteria and transmissible through milk, such as Q fever, brucellosis, salmonellosis, and tuberculosis, are still a burden for the dairy industry in many parts of the world (
), posing considerable health risks to farmers, workers, and people who consume raw milk.
With the implementation of control programs, prevention and eradication of many diseases can be achieved. It is of utmost importance, as part of any control program, to have an effective system to classify herds in terms of disease status. A promising trend toward disease surveillance in veterinary medicine is the testing of aggregated samples. In dairy cattle, testing of bulk milk (BM) samples is a convenient, cost-effective strategy that can easily be implemented as part of disease surveillance programs. Indeed, BM testing has become a cornerstone of several control programs for infectious diseases of dairy cattle in several countries (
Although convenient, testing of BM samples to detect herds with diseased animals has limitations. For one, in general, BM testing suffers from low sensitivity compared with animal-level testing, due to the “dilution effect,” where the milk of infected animals gets diluted with milk from uninfected cows. Additionally, substantial variability in terms of disease definition (e.g., what constitutes a disease-positive herd or positivity threshold in terms of infected animals), interpretation of testing results (e.g., lack specific cut-off values for BM samples), test target (e.g., detection of antigen or antibodies), manufacturers, and herd confounding factors (e.g., herd vaccination status) can hinder interpretation of BM testing results.
Scoping reviews have become increasingly popular for synthesizing research evidence and identifying gaps in knowledge. Similar to systematic reviews, scoping reviews follow a structured process that distinguish them from narrative reviews, requiring rigorous methodology and transparency to ensure that results are trustworthy (
). Hence, scoping reviews are ideal to summarize the complex and highly heterogeneous body of literature reporting on the testing of BM to detect infectious diseases of dairy cattle.
Although the use of aggregate samples such as BM as part of disease surveillance has recently been reviewed (
), no in-depth synthesis of this literature has been performed. Here we performed a scoping review to summarize the literature reporting on the testing of BM samples to detect infectious diseases of dairy cattle caused by bacteria. Results from this review can be used to inform the development of surveillance initiatives that will integrate multifaceted programs aimed to mitigate impacts of endemic and emerging and re-emerging infectious diseases of dairy cattle.
MATERIALS AND METHODS
Because no human or animal subjects were used, this analysis did not require approval by an Institutional Animal Care and Use Committee or Institutional Review Board.
Protocol Registration and Deviations From the Original Protocol
A scoping review protocol was developed following the PRISMA extension for scoping reviews (
). Originally, our goal was to report results from the scoping review as a single publication. However, after the protocol was established, we decided to split the work into 2 parts. The rationale for dividing the work was the unexpectedly large number of eligible studies reporting on the screening of BM samples to detect pathogens that can cause infectious diseases in dairy cattle (n = 769), as well as the substantial number of pathogens or diseases reported in at least one study. The eligibility criteria, search strategy, assessment for eligibility, and data extraction steps were carried out regardless of the infectious agent that was causing disease. For results presentation, studies reporting on diseases caused by bacteria were retained and synthesized.
Eligibility Criteria
Original studies of any design reporting on the testing of farm-level BM samples for detection of infectious diseases of dairy cattle were eligible for inclusion. Articles must have been published in the last 35 years in English, Portuguese, or Spanish to remain eligible.
With respect to sample eligibility, BM samples were necessarily collected before any processing or comingling among farms, regardless of sample provider (e.g., dairy cooperative, farm, milk haulers, milk processors). Farm-level bulk tank milk samples collected for regulatory and payment purposes were eligible. Samples collected at wholesale or retail outlets, or similar, were considered processed or commingled, and therefore were not eligible. In the absence of specific information detailing the origin and commingling status of samples, we assumed that BM samples were collected from bulk tanks on farms before any commingling. Simulated BM samples such as those obtained from the pooling of many cow-level samples at the laboratory were not considered. Likewise, experimentally inoculated and spiked milk samples were not eligible. Additionally, milk filters were not deemed as equivalent to BM, as evidence suggests that bacteriological findings are not necessarily interchangeable between sample types (
Prevalence of Salmonella enterica, Listeria monocytogenes, and Escherichia coli virulence factors in bulk tank milk and in-line filters from U.S. dairies.
Studies failing to describe the methodology used for BM testing were excluded. Microbiome and related techniques (e.g., 16S rRNA metagenomic sequencing) were ineligible. In terms of test target, nonspecific markers of infections (e.g., SCC, lactoferrin) and quality indicators (e.g., total bacterial count, coliform count, yeasts, antibiotic residues, heavy metals) were not considered. Furthermore, studies were eligible for inclusion if at least one of the following was true: (1) BM samples were tested for presence of pathogens that are typically associated with infectious diseases of dairy cattle or specific markers associated with such infections (e.g., antibodies); or (2) The study investigated or provided sufficient data to investigate the use of BM testing to define herd disease status for any given pathogen or disease. Studies that met the second criteria were classified as validation studies. Remaining eligible studies were classified as detection studies. A list of diseases and pathogens considered as typical in the dairy cattle population is available in the study protocol (
). Our rationale for introducing the testing target requirement is to focus on the use of BM testing as part of disease surveillance while providing an inclusive list of tests that were implemented to investigate BM samples across studies.
Studies reporting on the diagnostic characteristics of BM testing to detect herds of varying disease status (e.g., diagnostic sensitivity or specificity, or both) or describing associations between BM and animal-level testing results (e.g., correlation, regression modeling, or R2 values) were eligible for inclusion under the validation category. Minimum data required to investigate the use of BM testing to define herd disease status for any given pathogen or disease included at least one of the following: (1) BM and animal-level testing results from same herds reported on a herd basis for at least 15 herds; or (2) BM testing results reported for at least 15 herds of known disease status. The minimum number of herds was set at 15 because this would allow for the estimation of a test characteristic (e.g., sensitivity, specificity) of 99%, with 95% confidence level and 5% of margin of error. Therefore, case reports, case series, or studies where the diagnostic characteristics of BM testing could be estimated based on a limited number of herds, were not eligible for inclusion under the validation category. Nevertheless, such studies were still considered under the detection category.
). Queries were adapted to database-specific terms, as necessary.
Additionally, the International Veterinary Information Service (IVIS; https://www.ivis.org/) and Searchable Proceedings of Animal Conferences (SPAC; https://spac.adsa.org) databases were screened for potentially eligible articles. The query “bulk milk” was used in the 2 databases on October 27, 2020, and titles from all hits were inspected for potential inclusion. Further, proceedings of relevant conferences not indexed by IVIS or SPAC (e.g., ParaTB Forum, International Colloquium on Paratuberculosis) were searched for potential studies. In addition, we screened websites of relevant animal health agencies (e.g., World Organisation for Animal Health, United States Department of Agriculture National Animal Health Monitoring System, Canadian Animal Health Surveillance System), disease surveillance programs (e.g., cattle-related programs listed in the European Commission National Veterinary Programmes and national-level disease surveillance programs identified during the abstract and full-text review stages) for relevant peer-reviewed literature. We accessed disease-specific webpages and searched for peer-reviewed literature potentially reporting on the testing of BM samples. We also screened the Diagnostics for Animal database (https://diagnosticsforanimals.com/list-of-animal-health-diagnostics/), which contains information on nearly 90% of the global animal health diagnostic market. Specifically, information sheets, webpages, or handbooks of each cattle-related diagnostic kit were inspected for potentially relevant peer-reviewed literature. Finally, we inspected websites of veterinary diagnostic providers for products that could be used in either BM or pooled milk samples. In each product sheet, we looked for peer-reviewed literature that supported its use with BM samples. A list of veterinary diagnostic providers that were screened was built using information available from the Diagnostics for Animal database and also using results from this scoping review (manufacturers identified during the abstract and full-text review stages). Product manufacturers were not contacted to request further validation data or access to restricted content.
Data Management and Selection Process
Records identified in the search process were uploaded into EndNote X9 (Clarivate Analytics) and merged as a single database. Duplicate records were flagged and excluded using a 2-step approach. First, we used the Systematic Review Assistant-Deduplication Module developed at Bond University (
). The remaining references were screened for duplicates using EndNote. Screening was conducted at 2 levels. At the first level, 2 reviewers checked titles, abstracts, and keywords of all identified literature, to select studies reporting on the testing of BM samples obtained from dairy herds (
). The initial screening was fairly broad, to encompass all potentially relevant studies. A preliminary assessment was carried out using 20 studies to ensure consistency between reviewers. Thereafter, each reviewer checked all entries independently using a checklist. Agreement between reviewers was almost perfect (κ = 0.94; 95% CI: 0.93–0.95). All conflicts were resolved by consensus.
At the full-text screening stage, full texts were obtained for articles that met the inclusion criteria and assessed for eligibility using a predefined screening checklist (
). This step was carried out by one reviewer (DN) under supervision of a second (DK). Our goal was to retain studies reporting on the testing of farm-level BM samples, collected before any processing or commingling, for presence of infectious diseases of dairy cattle, using a valid methodology, as described in the eligibility criteria.
Data Extraction
From each eligible study, variables extracted were author, year, country or countries of origin of BM samples, eligible disease(s) or pathogen(s) screened, test(s) used, and test(s) manufacturer (when available). Eligible pathogens or diseases were among those considered typical in the dairy cattle population (as described) or for which BM testing could be validated, and for which the testing methodology was specified.
Studies were classified into 2 mutually exclusive categories: detection studies and validation studies, according to previously stated eligibility criteria. For validation studies, we also extracted the following variables, when available:
1.
General characteristics: herd eligibility criteria, number of herds and BM samples tested, BM sampling protocol.
2.
Test characteristics: test target, test cut-off value [e.g., sample-to-positive (S/P) ratio, cycle threshold (CT) value], testing scheme, test interpretation (e.g., parallel, series), sensitivity, specificity, association metric (e.g., R2, correlation coefficient, rate of change).
3.
Disease characteristics: infectious disease or pathogen screened, animal-level testing protocol, interval between BM testing and herd disease assessment, animal-level test used, cut-off value used in the animal-level testing, herd disease definition.
Data were extracted for each test and disease combination, with separate data entries generated to accommodate >1 test or disease reported per study. Additionally, when test characteristics (e.g., sensitivity, specificity) were estimated using different BM test cut-off values, we extracted characteristics from thresholds discussed by authors as most relevant. Further, >1 row per study was used if different herd disease definitions or BM testing protocols existed. For studies reporting on BM testing using samples from >1 country, testing results were pooled by herd disease status regardless of country of origin of samples.
Critical Appraisal of Individual Sources of Evidence
Full-text articles that reported on the sensitivity and specificity of BM testing or that provided sufficient data to estimate these values were critically appraised with respect to potential risk of bias using a simplified version of the Veterinary Quality Assessment of Diagnostic Accuracy Studies (VETQUADAS) checklist (
). In brief, the VETQUADAS tool describes the risk of bias in 4 domains (clarity in reporting, internal validity, external validity, other) using a total of 16 quality items. For the purpose of this study, items no. 1, 3, 4, 8, and 9 were adapted and used to critically appraise individual studies as follows:
1.
Is the spectrum of herds in the study representative of herds that will receive the test in practice? (VETQUADAS item 1.)
2.
Is the herd disease status likely correctly classified? (VETQUADAS item 3.)
3.
Is the time period between herd disease status determination and BM testing short enough to be reasonably sure that the herd disease status did not change before or at BM testing? (VETQUADAS item 4.)
4.
Was the execution of the BM test described in sufficient detail to allow future replication? (VETQUADAS item 8.)
5.
Was the classification of herd disease status described in sufficient detail to allow future replication? (VETQUADAS item 9.)
Studies receiving “no” or “unclear” were classified as “at risk” for the respective item.
Synthesis of Results
Studies reporting on the testing of BM samples for presence of at least one infectious disease of dairy cattle caused by bacteria were retained. Summary tables were used to describe tests and manufacturers detected for each disease surveyed. Because substantial variability existed in terms of disease definition, sampling, and testing protocols, we did not attempt statistical pooling of test characteristics. Rather, we built disease-level tables to summarize characteristics of individual tests as reported or estimated from studies. Finally, results from the gray literature search and abstract or full-text screening steps were used to build a non-exhaustive, albeit inclusive, list of diagnostic tests that are marketed for BM samples, as well as a list of disease surveillance activities that included testing of BM samples.
). In terms of origins of samples, 65, 44, 37, and 30 studies used BM samples originating from the United States, Denmark, Italy, and Iran, respectively (Supplemental Figure S1, https://data.mendeley.com/datasets/cwb8gsyfzr,
). Overall, 575 eligible bacterial pathogens were screened for using BM samples, ranging from 1 to 6 individual pathogens per study. Staphylococcus aureus, including methicillin-resistant Staph. aureus (MRSA), were the most commonly studied bacteria (179 studies; Table 1; Supplemental Table S1), followed by Streptococcus agalactiae (86 studies), Mycobacterium avium ssp. paratuberculosis (MAP; 79 studies), Coxiella burnetii (79 studies), and Mycoplasma spp. (including Mycoplasma bovis; 67 studies). Overall, culture-based protocols, ELISA, quantitative PCR (qPCR), and PCR were the most commonly adopted methodologies (Table 1). Kits manufactured by Svanova (including Boehringer Ingelheim Svanova), IDEXX, LSI, and Thermo Fisher were frequently used to analyze BM samples (Supplemental Table S1). Most commercially available kits that were identified by our search strategy, and that can be used with bulk or pooled milk samples according to manufacturers, were part of the INDICAL Bioscience, Thermo Fisher, or BioSellal portfolios (Supplemental Table S2, https://data.mendeley.com/datasets/j782g9drvh,
Figure 1PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram of the study selection *Studies could be included in >1 category. BM = bulk milk.
Bacteria or disease among those considered typical in the dairy cattle population, or for which the ability of bulk milk testing to classify herd disease status could be evaluated according to previously described eligibility criteria. MAP = Mycobacterium avium subspecies paratuberculosis; Mycoplasma spp. encompass Mycoplasma bovis as well as other Mycoplasma species.
Methodology used to test bulk milk (BM) samples for presence of pathogen(s) or disease(s). Protocols were described according to method(s) that were primarily used to investigate bulk milk samples. For instance, for protocols employing subsequential testing (e.g., initial culturing of bulk milk samples and species confirmation using quantitative PCR), the method that used milk (e.g., culture) was annotated exclusively.
1 Studies could be included in >1 category of any characteristic.
2 Bacteria or disease among those considered typical in the dairy cattle population, or for which the ability of bulk milk testing to classify herd disease status could be evaluated according to previously described eligibility criteria. MAP = Mycobacterium avium subspecies paratuberculosis; Mycoplasma spp. encompass Mycoplasma bovis as well as other Mycoplasma species.
3 Methodology used to test bulk milk (BM) samples for presence of pathogen(s) or disease(s). Protocols were described according to method(s) that were primarily used to investigate bulk milk samples. For instance, for protocols employing subsequential testing (e.g., initial culturing of bulk milk samples and species confirmation using quantitative PCR), the method that used milk (e.g., culture) was annotated exclusively.
Among included studies, 78 reported on or provided sufficient data to estimate characteristics of BM testing to detect infectious diseases of dairy cattle caused by bacteria and were therefore classified as validation studies (Figure 1). Overall, MAP was the most frequently studied bacteria (n = 27 studies) followed by Salmonella Dublin (9 studies), Strep. agalactiae (8 studies), Brucella abortus (brucellosis; 8 studies), C. burnetii (8 studies), and Staph. aureus (7 studies).
Sensitivity of BM testing for MAP was generally low and varied greatly according to the ELISA cut-offs adopted and herd-level definition of disease (Table 2). Additionally, test sensitivities were generally higher if increased within-herd prevalence of infected animals were used to define true infection status (
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
Association between herd infection level and the detection of Mycobacterium avium subsp. paratuberculosis (MAP) in bulk tank milk tank using real-time PCR in small holder dairy farms in southern Chile.
in: 11th International Colloquium on Paratuberculosis, Sydney, Australia. International Association for Paratuberculosis,
2012: 65
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
Evaluation of IS900-PCR assay for detection of Mycobacterium avium subspecies paratuberculosis infection in cattle using quarter milk and bulk tank milk samples.
). Direct culturing of BM had very low sensitivity (<25.0%), regardless of disease definition. Protocols that included peptide-mediated magnetic separation assays demonstrated increased sensitivity compared with culture-based protocols, and the sensitivity depended on the methodology used, as follow-up (
Application of a peptide-mediated magnetic separation-phage assay for detection of viable Mycobacterium avium subsp. paratuberculosis to bovine bulk tank milk and feces samples.
Sensitive and specific detection of viable Mycobacterium avium subsp. paratuberculosis in raw milk by the peptide-mediated magnetic separation-phage assay.
J. Appl. Microbiol.2017; 122 (28235155): 1357-1367
). Specificity of BM testing was high in general, with the exception of ELISA using low cut-offs that demonstrated low to moderate specificity (50–70%) to distinguish disease-free herds from herds with a low prevalence of infected cows (
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
Bulk tank milk ELISA for detection of antibodies to Mycobacterium avium subsp. paratuberculosis: Correlation between repeated tests and within-herd antibody-prevalence.
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
). Bulk milk ELISA demonstrated moderate to high correlation with within-herd prevalence of infected animals, ranging from 0.39 to 0.91.
Table 2Summary of findings from individual validation studies reporting on the detection of Mycobacterium avium ssp. paratuberculosis (MAP) in bulk milk samples (NA = not available)
Herds previously classified as MAP-positive based on blood testing (ELISA) of cows >1 yr old. Herds of low, moderate, or high prevalence were combined as a single disease-positive category.
Herds previously classified as MAP-positive based on blood testing (ELISA) of cows >1 yr old. Herds of low, moderate, or high prevalence were combined as a single disease-positive category.
Sensitivities of a bulk-tank milk ELISA and composite fecal qPCR to detect various seroprevalence levels of paratuberculosis in cattle herds in Normandy, France.
Sample-to-positive (S/P) ratio = 10%; sensitivity ranged from 12 to 75% depending on the within-herd apparent seroprevalence used to classify disease at the herd-level.
Sample-to-positive (S/P) ratio = 10%; sensitivity ranged from 12 to 75% depending on the within-herd apparent seroprevalence used to classify disease at the herd-level.
Three criteria were used to classify herds: (1) Results of animal-level testing (≥1 positive out of 6 symptomatic animals per herd tested with milk and serum ELISA; 17 herds); (2) previous history of Johne's (8 herds); (3) negative results from annual whole-herd testing for the past 8 yr (1 herd).
Application of a peptide-mediated magnetic separation-phage assay for detection of viable Mycobacterium avium subsp. paratuberculosis to bovine bulk tank milk and feces samples.
Sensitive and specific detection of viable Mycobacterium avium subsp. paratuberculosis in raw milk by the peptide-mediated magnetic separation-phage assay.
J. Appl. Microbiol.2017; 122 (28235155): 1357-1367
The cut-off for a positive test was defined as a test results 2 SD above the mean. Using a cut-off of 3 SD above mean, the sensitivity and specificity were 8.6 and 100%, respectively.
The cut-off for a positive test was defined as a test results 2 SD above the mean. Using a cut-off of 3 SD above mean, the sensitivity and specificity were 8.6 and 100%, respectively.
The cut-off for a positive test was defined as a test results 2 SD above the mean. Using a cut-off of 3 SD above mean, the sensitivity and specificity were 8.6 and 100%, respectively.
The cut-off for a positive test was defined as a test results 2 SD above the mean. Using a cut-off of 3 SD above mean, the sensitivity and specificity were 8.6 and 100%, respectively.
A robust method for bacterial lysis and DNA purification to be used with real-time PCR for detection of Mycobacterium avium subsp. paratuberculosis in milk.
J. Microbiol. Methods.2008; 75 (18694788): 335-340
The evaluation of the utility of bulk tank tests for the surveillance of Johne’s disease and the effect of storage time and temperature on Johne’s milk ELISA results.
Department of Population Medicine, University of Guelph,
Guelph, Canada2011
Optical density (OD) = 0.10. The assay involved modifications to label protocols. When no modifications were implemented, sensitivity was estimated as 0%. When the criteria to define a positive herd increased to ≥2 cows, the relative sensitivity and specificity were 63.3 and 84.2%, respectively.
Optical density (OD) = 0.10. The assay involved modifications to label protocols. When no modifications were implemented, sensitivity was estimated as 0%. When the criteria to define a positive herd increased to ≥2 cows, the relative sensitivity and specificity were 63.3 and 84.2%, respectively.
Evaluation of IS900-PCR assay for detection of Mycobacterium avium subspecies paratuberculosis infection in cattle using quarter milk and bulk tank milk samples.
S/P = 20%. The assay involved modifications to label protocols. When no modifications were implemented, sensitivity and specificity were estimated as 15% and 99%, respectively. For high-prevalence farms (>5% of adult cattle seropositive), the sensitivity was 70%.
S/P = 20%. The assay involved modifications to label protocols. When no modifications were implemented, sensitivity and specificity were estimated as 15% and 99%, respectively. For high-prevalence farms (>5% of adult cattle seropositive), the sensitivity was 70%.
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
Inter-laboratory comparison of radiometric culture for Mycobacterium avium subsp. paratuberculosis using raw milk from known infected herds and individual dairy cattle in Victoria.
Association between herd infection level and the detection of Mycobacterium avium subsp. paratuberculosis (MAP) in bulk tank milk tank using real-time PCR in small holder dairy farms in southern Chile.
in: 11th International Colloquium on Paratuberculosis, Sydney, Australia. International Association for Paratuberculosis,
2012: 65
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
S/P ratio = 12.5%. Sensitivity ranged from 35% to 85% depending on the within-herd apparent seroprevalence used to classify disease at the herd-level. Similarly, specificity ranged from 92 to 99%.
S/P ratio = 12.5%. Sensitivity ranged from 35% to 85% depending on the within-herd apparent seroprevalence used to classify disease at the herd-level. Similarly, specificity ranged from 92 to 99%.
Two milk samples per bulk tank (BT) per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was MAP positive at any given test.
Herds with ≥1 BTM sample positive for MAP at any test (ELISA or qPCR).
Johne’s disease, Mycoplasma and BVD in Utah-bulk tank milk testing and comparison to previous regional prevalence and individual herd results over time.
Five milk samples per BT per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was MAP positive at any given test.
Herds with ≥1 BTM sample positive for MAP at any test (ELISA or qPCR).
Dairy herd-level prevalence of Johne’s disease and BVD in the intermountain west of the U.S.A. and farm management practices and characteristics for test-positive herds.
Five milk samples per BT per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was MAP positive at any given test.
Herds with ≥1 BTM sample positive for MAP at any test (ELISA or qPCR).
1 Manufacturers as reported from studies. PMS = peptide-mediated magnetic separation. IMS = immunomagnetic separation. qPCR = quantitative PCR.
2 When multiple samples were taken from same herds, results were interpreted in parallel unless stated.
3 Se = sensitivity; Sp = specificity.
4 Characteristic estimated using author-reported data.
5 Sensitivity was 30.1% for single sampling.
6 Characteristic estimated using Bayesian latent class models.
7 Sample-to-positive (S/P) ratio = 10%; sensitivity ranged from 12 to 75% depending on the within-herd apparent seroprevalence used to classify disease at the herd-level.
8 S/P = 10%; data were extracted from figures.
9 The cut-off for a positive test was defined as a test results 2 SD above the mean. Using a cut-off of 3 SD above mean, the sensitivity and specificity were 8.6 and 100%, respectively.
10 Optical density (OD) = 0.10. The assay involved modifications to label protocols. When no modifications were implemented, sensitivity was estimated as 0%. When the criteria to define a positive herd increased to ≥2 cows, the relative sensitivity and specificity were 63.3 and 84.2%, respectively.
11 When the criteria to define a positive herd increased to ≥2 cows, the relative sensitivity and specificity were 6.7 and 99.3%, respectively.
12 S/P = 20%. The assay involved modifications to label protocols. When no modifications were implemented, sensitivity and specificity were estimated as 15% and 99%, respectively. For high-prevalence farms (>5% of adult cattle seropositive), the sensitivity was 70%.
13 OD = 0.02. Depending on the cut-off adopted, sensitivity and specificity ranged from 41 to 100% and 29 to 100%, respectively.
14 Cut-off = 3,000. Specificity was 99.8% if a cut-off of 10,000 was adopted.
15 S/P = 3%. For high-prevalence herds (>10% of seropositive cows) using a cut-off of 10.5%, the relative sensitivity was 85.7%.
16 S/P = −1.44%. For high-prevalence herds (>10% of seropositive cows) using a cut-off of 0.5%, the relative sensitivity was 71.4%.
17 Three culture protocols were used. A single protocol recovered MAP from bulk tank milk (BTM) samples and was therefore extracted.
18 S/P ratio = 12.5%.
19 S/P ratio = 12.5%. Sensitivity ranged from 35% to 85% depending on the within-herd apparent seroprevalence used to classify disease at the herd-level. Similarly, specificity ranged from 92 to 99%.
20 Two milk samples per bulk tank (BT) per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was MAP positive at any given test.
21 OD = 0.1 (ELISA).
22 Five milk samples per BT per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was MAP positive at any given test.
With respect to mastitis pathogens, BM testing demonstrated increased sensitivity and specificity for Strep. agalactiae (Table 3). Protocols based on PCR demonstrated higher sensitivity than culture-based protocols (
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
), although all protocols demonstrated high specificity (>90%) regardless of testing methodology. Testing of BM via qPCR had low to moderate specificity (<70%) for Staph. aureus (
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
Bovine mastitis: The diagnostic properties of a PCR-based assay to monitor the Staphylococcus aureus genotype B status of a herd, using bulk tank milk.
). As for Mycoplasma spp., sensitivity of protocols varied from 15.3% to 76.7%, whereas specificities ranged from 97.3% to 100%. A single study reported on the characteristics of BM for non-agalactiae streptococci. Bulk milk culture had perfect specificity for Streptococcus uberis and Streptococcus dysgalactiae, although protocols suffered from very poor sensitivity (<10%). Conversely, qPCR showed moderate to high sensitivity and specificity (∼80%) for both pathogens (
). Finally, culture of BM demonstrated perfect specificity and moderate sensitivity (77.8%) to classify herds according to presence or absence of intramammary infections caused by Nocardia spp. (
). Correlation between within-herd prevalence of infected cattle and number of colony-forming units (cfu) in BM was higher for Strep. agalactiae and Mycoplasma spp. than for coliforms, coagulase-negative Staphylococcus, Staph. aureus, and non-agalactiae streptococci (
). Correlation between cycle threshold values of BM qPCR and within-herd prevalence of mastitis pathogens were significant for Staph. aureus, Strep. dysgalactiae, and Strep. uberis (
Table 3Summary of findings from individual validation studies reporting on the detection of mastitis pathogens in bulk milk (BM) samples (NA = not available)
One farm contributed 15 milk samples. Five milk samples per bulk tank per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single bulk tank milk (BTM) sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was positive for Mycoplasma spp. at any test.
Herds with ≥1 BTM sample positive for Mycoplasma spp. at any test (culture or qPCR).
Five triplicate milk samples per bulk tank per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was positive for Mycoplasma spp. at any test.
Two samplings were performed. At each sampling, 2 sets of milk samples were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity estimates varied according to sampling. By definition, specificity was set as 100%.
Herds with ≥1 culture-positive BTM sample in any of the 2 culture media.
Characteristics were estimated using latent class models that also included qPCR results. qPCR cut-off value yielding the highest sensitivity [cycle threshold (Ct) <40]. Sensitivity and specificity were also estimated at other cut-off values and ranged from 68.0 to 91.0% and 99.0 to 99.7%, respectively. Characteristics were stratified according to herd-level predictors (herd type, automated milk system, and herd size), and no significant differences were observed between stratum-specific estimates.
Characteristics were estimated using latent class models that also included qPCR results. qPCR cut-off value yielding the highest sensitivity [cycle threshold (Ct) <40]. Sensitivity and specificity were also estimated at other cut-off values and ranged from 68.0 to 91.0% and 99.0 to 99.7%, respectively. Characteristics were stratified according to herd-level predictors (herd type, automated milk system, and herd size), and no significant differences were observed between stratum-specific estimates.
Characteristics were estimated using latent class models that also included culture results. qPCR cut-off value yielding the highest sensitivity (Ct <40). Sensitivity and specificity were also estimated at other cut-off values and ranged from 75.1 to 95.2% and 98.8 to 99.7%, respectively. Characteristics were stratified according to herd-level predictors (herd type, automated milk system, and herd size), and no significant differences were observed between stratum-specific estimates.
Characteristics were estimated using latent class models that also included culture results. qPCR cut-off value yielding the highest sensitivity (Ct <40). Sensitivity and specificity were also estimated at other cut-off values and ranged from 75.1 to 95.2% and 98.8 to 99.7%, respectively. Characteristics were stratified according to herd-level predictors (herd type, automated milk system, and herd size), and no significant differences were observed between stratum-specific estimates.
Characteristics were estimated using latent class models that also included qPCR results. qPCR Ct value <37 and ELISA optical density coefficient (ODC) = 37%. Sensitivity and specificity were also estimated at other ELISA cut-off values and ranged from 33.2 to 94.7% and 65.1 to 99.7%, respectively.
Characteristics were estimated using latent class models that also included qPCR results. qPCR Ct value <37 and ELISA optical density coefficient (ODC) = 37%. Sensitivity and specificity were also estimated at other ELISA cut-off values and ranged from 33.2 to 94.7% and 65.1 to 99.7%, respectively.
Characteristics were estimated using latent class models that also included ELISA results. qPCR Ct value <37 and ELISA ODC = 37%. Sensitivity and specificity were also estimated at other ELISA cut-off values and ranged from 3.8 to 53.3% and 99.2 to 99.7%, respectively.
Characteristics were estimated using latent class models that also included ELISA results. qPCR Ct value <37 and ELISA ODC = 37%. Sensitivity and specificity were also estimated at other ELISA cut-off values and ranged from 3.8 to 53.3% and 99.2 to 99.7%, respectively.
ELISA ODC = 37%. Three control herds were extensively examined for the presence of Mycoplasma bovis (nose, eye, and vaginal swabs, as well as composite milk samples, were cultured from 50 systematically selected cows). Data were extracted from graph.
ELISA ODC = 37%. Three control herds were extensively examined for the presence of Mycoplasma bovis (nose, eye, and vaginal swabs, as well as composite milk samples, were cultured from 50 systematically selected cows). Data were extracted from graph.
≥1 culture-positive sample out of quarter-level milk samples collected from 30 cows per herd [10 primiparous, 10 multiparous, and 10 high-risk cows (high SCC or previous reports of Staph. aureus)].
Bovine mastitis: The diagnostic properties of a PCR-based assay to monitor the Staphylococcus aureus genotype B status of a herd, using bulk tank milk.
Johne’s disease, Mycoplasma and BVD in Utah-bulk tank milk testing and comparison to previous regional prevalence and individual herd results over time.
Five milk samples per bulk tank per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was Mycoplasma-positive at any given test.
Herds with ≥1 BTM sample positive for Mycoplasma spp. at any test (ELISA or qPCR).
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
Characteristics were estimated using latent class models that also included quantitative culture and qPCR results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were equal to 98%.
Characteristics were estimated using latent class models that also included quantitative culture and qPCR results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were equal to 98%.
Characteristics were estimated using latent class models that also included qualitative culture and qPCR results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were equal to 99%.
Characteristics were estimated using latent class models that also included qualitative culture and qPCR results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were equal to 99%.
Characteristics were estimated using latent class models that also included quantitative and qualitative culture results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were 92 and 94%, respectively.
Characteristics were estimated using latent class models that also included quantitative and qualitative culture results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were 92 and 94%, respectively.
Characteristics were estimated using latent class models that also included quantitative culture and qPCR results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 94 and 95%, respectively.
Characteristics were estimated using latent class models that also included quantitative culture and qPCR results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 94 and 95%, respectively.
Characteristics were estimated using latent class models that also included qualitative culture and qPCR results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 92 and 93%, respectively.
Characteristics were estimated using latent class models that also included qualitative culture and qPCR results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 92 and 93%, respectively.
Characteristics were estimated using latent class models that also included quantitative and qualitative culture results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 99 and 67%, respectively.
Characteristics were estimated using latent class models that also included quantitative and qualitative culture results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 99 and 67%, respectively.
1 Manufacturers as reported from studies. qPRC = quantitative PCR.
2 When multiple samples were taken from same herds, results were interpreted in parallel unless stated.
3 Se = sensitivity; Sp = specificity.
4 Characteristic estimated using author-reported data.
5 One farm contributed 15 milk samples. Five milk samples per bulk tank per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single bulk tank milk (BTM) sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was positive for Mycoplasma spp. at any test.
6 Five triplicate milk samples per bulk tank per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test-positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was positive for Mycoplasma spp. at any test.
7 Two samplings were performed. At each sampling, 2 sets of milk samples were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity estimates varied according to sampling. By definition, specificity was set as 100%.
8 Characteristics were estimated using latent class models that also included qPCR results. qPCR cut-off value yielding the highest sensitivity [cycle threshold (Ct) <40]. Sensitivity and specificity were also estimated at other cut-off values and ranged from 68.0 to 91.0% and 99.0 to 99.7%, respectively. Characteristics were stratified according to herd-level predictors (herd type, automated milk system, and herd size), and no significant differences were observed between stratum-specific estimates.
9 Characteristics were estimated using latent class models that also included culture results. qPCR cut-off value yielding the highest sensitivity (Ct <40). Sensitivity and specificity were also estimated at other cut-off values and ranged from 75.1 to 95.2% and 98.8 to 99.7%, respectively. Characteristics were stratified according to herd-level predictors (herd type, automated milk system, and herd size), and no significant differences were observed between stratum-specific estimates.
10 Characteristics were estimated using latent class models that also included qPCR results. qPCR Ct value <37 and ELISA optical density coefficient (ODC) = 37%. Sensitivity and specificity were also estimated at other ELISA cut-off values and ranged from 33.2 to 94.7% and 65.1 to 99.7%, respectively.
11 Characteristics were estimated using latent class models that also included ELISA results. qPCR Ct value <37 and ELISA ODC = 37%. Sensitivity and specificity were also estimated at other ELISA cut-off values and ranged from 3.8 to 53.3% and 99.2 to 99.7%, respectively.
12 Depending on the herd disease definition adopted, sensitivity and specificity ranged from 94.8 to 100% and 43.1 to 100%, respectively.
13 Depending on the herd disease definition adopted, sensitivity and specificity ranged from 89.7 to 100% and 49.0 to 100%, respectively.
14 ELISA ODC = 37%. Three control herds were extensively examined for the presence of Mycoplasma bovis (nose, eye, and vaginal swabs, as well as composite milk samples, were cultured from 50 systematically selected cows). Data were extracted from graph.
15 Nasal swabs, udder cleft swabs, teat liners, human, and dust samples were not used to define herd disease status.
16 Five milk samples per bulk tank per herd (3- to 4-d intervals) were used to establish herd disease status and estimate the sensitivity of testing a single BTM sample of each herd. Sensitivity was defined at the sample level, as the number of test positive BTM samples divided by total BTM samples collected from herds where ≥1 BTM sample was Mycoplasma-positive at any given test.
17 Characteristics were estimated using latent class models that also included quantitative culture and qPCR results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were equal to 98%.
18 Characteristics were estimated using latent class models that also included qualitative culture and qPCR results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were equal to 99%.
19 Characteristics were estimated using latent class models that also included quantitative and qualitative culture results. qPCR Ct = 50, model without covariance terms. Using a qPCR Ct = 40, the sensitivity and specificity were 92 and 94%, respectively.
20 Characteristics were estimated using latent class models that also included quantitative culture and qPCR results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 94 and 95%, respectively.
21 Characteristics were estimated using latent class models that also included qualitative culture and qPCR results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 92 and 93%, respectively.
22 Characteristics were estimated using latent class models that also included quantitative and qualitative culture results. qPCR Ct = 50, model with covariance terms. Using a qPCR Ct = 38, the sensitivity and specificity were 99 and 67%, respectively.
Validation studies reporting on Salmonella spp. used ELISA to detect antibodies against the bacteria of interest. Among Salmonella spp. pathogens, Salmonella Dublin was the most frequently studied bacterium for which BM testing has been validated (Table 4). Specificity was high, ranging from 89.0 to 99.4. In contrast, sensitivity of BM testing varied greatly among studies, ranging from 50.6% to 100%. When a positive herd was defined based on previous clinical history or blood test of adult lactating cattle, sensitivity was higher compared with circumstances where disease was also considered among calves (
). Bulk milk testing had low sensitivity for Salmonella spp. other than Salmonella Dublin, ranging from 26.3% for Salmonella spp. to 37.5% for Salmonella Typhimurium (
), whereas the within-herd prevalence of infected cattle accounted for 51% to 67% of the variance of the Salmonella Dublin ELISA OD values of BM samples (
Sensitivity and specificity were also estimated at other cut-off values and ranged from 90.0 to 100%. Although the study protocol included animal-level sampling and also defined disease at the herd-level based on animal-level testing results, bulk tank milk test characteristics were estimated based on the assumed disease status of enrolled herds.
Herd level diagnosis for Salmonella Dublin infection in bovine dairy herds.
in: 9th Symposium of the International Society for Veterinary Epidemiology and Economics (ISVEE), Breckenridge, CO. International Society for Veterinary Epidemiology and Economics,
2000: 179
Herds from the Netherlands and Sweden were pooled for specificity estimation. Sensitivity and specificity were also estimated at other cut-off values and ranged from 20.3 to 96.2% and 42.5 to 100%, respectively. When the 2 ELISA were tested in combination, sensitivity was estimated as 64.6 and 53.1% for parallel and serial testing, respectively. Likewise, specificity was estimated as 96.8 and 100% for parallel and serial testing, respectively.
Herds from the Netherlands and Sweden were pooled for specificity estimation. Sensitivity and specificity were also estimated at other cut-off values and ranged from 20.3 to 96.2% and 42.5 to 100%, respectively. When the 2 ELISA were tested in combination, sensitivity was estimated as 64.6 and 53.1% for parallel and serial testing, respectively. Likewise, specificity was estimated as 96.8 and 100% for parallel and serial testing, respectively.
Herds from the Netherlands and Sweden were pooled for specificity estimation. Sensitivity and specificity were also estimated at other cut-off values and ranged from 12.7 to 68.4% and 91.2 to 100%, respectively. When the 2 ELISA were tested in combination, sensitivity was estimated as 64.6 and 53.1% for parallel and serial testing, respectively. Likewise, specificity was estimated as 96.8 and 100% for parallel and serial testing, respectively.
Herds from the Netherlands and Sweden were pooled for specificity estimation. Sensitivity and specificity were also estimated at other cut-off values and ranged from 12.7 to 68.4% and 91.2 to 100%, respectively. When the 2 ELISA were tested in combination, sensitivity was estimated as 64.6 and 53.1% for parallel and serial testing, respectively. Likewise, specificity was estimated as 96.8 and 100% for parallel and serial testing, respectively.
When the criteria to define disease at the herd level was the presence of clinical signs only among animals <2 yr old, sensitivity was estimated as 31.0%. When the criteria to define disease at the herd level was the presence of clinical signs only among animals ≥2 yr old, sensitivity was estimated as 78.9%.
A positive bulk milk test was defined as a 4-measurement moving average of ≥25 optical density coefficient (ODC)% or having a difference of >20 ODC% between the current measurement and the average of the previous 3 measurements.
≥1 Salmonella culture-positive fecal sample or ≥5% within-herd prevalence based on antibody measurements in serum or milk
Characteristics were estimated using simulation models for a true herd-level infection prevalence of 15%. Specificity varied with prevalence, ranging from 69 to 98%.
Characteristics were estimated using simulation models for a true herd-level infection prevalence of 15%. Specificity varied with prevalence, ranging from 69 to 98%.
Sensitivity and specificity were also estimated at other cut-off values and ranged from 39.0 to 100% and 26 to 100%, respectively.
1 Manufacturers as reported from studies. ELISA: GP = flagellar antigen ELISA.
2 Se = sensitivity; Sp = specificity.
3 Sensitivity and specificity were also estimated at other cut-off values and ranged from 90.0 to 100%. Although the study protocol included animal-level sampling and also defined disease at the herd-level based on animal-level testing results, bulk tank milk test characteristics were estimated based on the assumed disease status of enrolled herds.
4 Characteristics were estimated using simulation models.
5 Characteristics estimated using author-reported data.
6 Blood and milk samples were collected from individual cattle in a subset of participating herds but were not used to define herd disease status.
7 Three to five samplings per herd (3-mo intervals) were carried out. Characteristics were estimated at the sample level.
8 Dates were extracted from graph.
9 Herds from the Netherlands and Sweden were pooled for specificity estimation. Sensitivity and specificity were also estimated at other cut-off values and ranged from 20.3 to 96.2% and 42.5 to 100%, respectively. When the 2 ELISA were tested in combination, sensitivity was estimated as 64.6 and 53.1% for parallel and serial testing, respectively. Likewise, specificity was estimated as 96.8 and 100% for parallel and serial testing, respectively.
10 Herds from the Netherlands and Sweden were pooled for specificity estimation. Sensitivity and specificity were also estimated at other cut-off values and ranged from 12.7 to 68.4% and 91.2 to 100%, respectively. When the 2 ELISA were tested in combination, sensitivity was estimated as 64.6 and 53.1% for parallel and serial testing, respectively. Likewise, specificity was estimated as 96.8 and 100% for parallel and serial testing, respectively.
11 When the criteria to define disease at the herd level was the presence of clinical signs only among animals <2 yr old, sensitivity was estimated as 31.0%. When the criteria to define disease at the herd level was the presence of clinical signs only among animals ≥2 yr old, sensitivity was estimated as 78.9%.
12 Characteristics were estimated using simulation models for a true herd-level infection prevalence of 15%. Specificity varied with prevalence, ranging from 69 to 98%.
13 A positive bulk milk test was defined as a 4-measurement moving average of ≥25 optical density coefficient (ODC)% or having a difference of >20 ODC% between the current measurement and the average of the previous 3 measurements.
14 Sensitivity and specificity were also estimated at other cut-off values and ranged from 39.0 to 100% and 26 to 100%, respectively.
Other bacterial diseases or pathogens for which BM testing has been validated to detect animal-level infections include brucellosis (8 studies), C. burnetii (8 studies), bovine tuberculosis (2 studies), digital dermatitis (2 studies), and Listeria spp. (1 study). Three assays were used in BM samples to classify herds according to presence of cattle infected by B. abortus: milk ring test, ELISA, and fluorescence polarization assay. Specificity was high regardless of test methodology (
Comparative evaluation of the indirect enzyme-linked immunosorbent assay in milk for the detection of cattle infected with Brucella abortus, in herds located in the province of Cundinamarca, Colombia.
). Five studies described associations between within-herd prevalence of C. burnetii-positive cattle and BM testing results. Studies reported significant correlations between numbers of C. burnetii-seropositive cattle and BM ELISA (
Dynamics of relationship between the presence of Coxiella burnetii DNA, antibodies, and intrinsic variables in cow milk and bulk tank milk from Danish dairy cattle.
Dynamics of relationship between the presence of Coxiella burnetii DNA, antibodies, and intrinsic variables in cow milk and bulk tank milk from Danish dairy cattle.
). Additionally, a BM ELISA demonstrated significant but low correlation with an average clinical score, which incorporated presumptive clinical signs associated with C. burnetii infections such as abortion, stillbirth, weak calves, irregular repeat breeding, decreased milk production, infertility, and mastitis (
A touch-down IS6110 PCR returned positive results in nearly 38% of herds assumed to be free of tuberculosis in Argentina, which yielded an estimated specificity of 62% (
). Additionally, analysis of BM samples using an IS1081 PCR demonstrated perfect sensitivity when tested in infected herds from Mexico, and a commercial ELISA had 100% specificity when tested in tuberculosis-free herds from Michigan, in the United States (
demonstrated that herds free from DD, as reported by farmers and hoof trimmers, had low OD values in BM ELISA against 3 antigens (PrrA, VpsA, and VspB) and were therefore considered disease negative (100% specificity). In contrast, 3 out of 15 herds with clinical evidence of DD also had low OD values in BM, which corresponded to a sensitivity of 80%. Likewise,
described the use of the PrrA ELISA to estimate within-herd prevalence of DD. Significant herd-level correlations were detected between within-herd prevalence of cattle with any DD lesion and BM S/P values, as well as between within-herd prevalence of cattle with active DD lesions and BM S/P values. Further, herds with low S/P ratio in BM ELISA had correspondingly low prevalence (<10%) of DD-positive cattle (sensitivity = 97% and specificity = 100%).
A single study provided sufficient data to validate BM testing for Listeria spp. (
). Ninety-eight herds from Spain were enrolled. From each herd, fecal samples were obtained from 3 apparently healthy lactating cows and pooled for culture. In addition, each herd contributed 1 BM sample as well as a variety of environmental samples. Analysis of BM detected ∼25% of herds with fecal culture-positive cows and ∼88% of herds where enrolled animals were culture negative. Tiestall systems and improper milking order were strongly associated with presence of Listeria spp. in BM.
Study Quality
A total of 56 full-text articles that reported on the sensitivity or specificity of BM testing, or provided sufficient data to estimate these characteristics, were critically appraised with respect to potential risk of bias (Table 5). Nearly half of studies were considered at risk for at least one checklist item. Most studies at risk of bias did not provide a detailed description of how herd infection status was determined. Further, it was unclear whether herd-level disease classification based on herd history (e.g., herds where disease has never been diagnosed) was accurate to determine herd disease status at the time the study was carried out. Additionally, most studies at risk of bias failed to specify the time interval between herd disease status assessment and BM testing. Without a clear description of how disease status was determined at the herd level, it was challenging to assess whether BM testing was accurate or not. Finally, it should be mentioned that although the validation of BM testing was not necessarily the primary goal of studies, studies were critically reviewed for that specific goal. Therefore, studies considered at risk in this review were not necessarily biased with respect to their primary goals.
Table 5Risk of bias as assessed using a simplified version of the Veterinary Quality Assessment of Diagnostic Accuracy Studies (VETQUADAS) checklist from validation manuscripts that reported on the characteristics (sensitivity and specificity) of bulk milk testing to detect infectious diseases of dairy cattle or that provided sufficient data to estimate characteristics
External validity: Is the spectrum of herds in the study representative of herds that will receive the test in practice? (VETQUADAS item 1.) Accuracy of index: Is the herd disease status likely correctly classified? (VETQUADAS item 3.) Time between tests: Is the time period between herd disease status determination and bulk milk (BM) testing short enough to be reasonably sure that the herd disease status did not change before or at BM testing? (VETQUADAS item 4.) Test description: Was the execution of the BM test described in sufficient detail to allow future replication? (VETQUADAS item 8.) Index description: Was the classification of herd disease status described in sufficient detail to allow future replication? (VETQUADAS item 9.)
Sensitivities of a bulk-tank milk ELISA and composite fecal qPCR to detect various seroprevalence levels of paratuberculosis in cattle herds in Normandy, France.
Application of a peptide-mediated magnetic separation-phage assay for detection of viable Mycobacterium avium subsp. paratuberculosis to bovine bulk tank milk and feces samples.
Sensitive and specific detection of viable Mycobacterium avium subsp. paratuberculosis in raw milk by the peptide-mediated magnetic separation-phage assay.
J. Appl. Microbiol.2017; 122 (28235155): 1357-1367
A robust method for bacterial lysis and DNA purification to be used with real-time PCR for detection of Mycobacterium avium subsp. paratuberculosis in milk.
J. Microbiol. Methods.2008; 75 (18694788): 335-340
The evaluation of the utility of bulk tank tests for the surveillance of Johne’s disease and the effect of storage time and temperature on Johne’s milk ELISA results.
Department of Population Medicine, University of Guelph,
Guelph, Canada2011
Evaluation of IS900-PCR assay for detection of Mycobacterium avium subspecies paratuberculosis infection in cattle using quarter milk and bulk tank milk samples.
Bulk tank milk ELISA for detection of antibodies to Mycobacterium avium subsp. paratuberculosis: Correlation between repeated tests and within-herd antibody-prevalence.
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
Inter-laboratory comparison of radiometric culture for Mycobacterium avium subsp. paratuberculosis using raw milk from known infected herds and individual dairy cattle in Victoria.
Comparative evaluation of the indirect enzyme-linked immunosorbent assay in milk for the detection of cattle infected with Brucella abortus, in herds located in the province of Cundinamarca, Colombia.
Bovine mastitis: The diagnostic properties of a PCR-based assay to monitor the Staphylococcus aureus genotype B status of a herd, using bulk tank milk.
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
Johne’s disease, Mycoplasma and BVD in Utah-bulk tank milk testing and comparison to previous regional prevalence and individual herd results over time.
Dairy herd-level prevalence of Johne’s disease and BVD in the intermountain west of the U.S.A. and farm management practices and characteristics for test-positive herds.
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
2 External validity: Is the spectrum of herds in the study representative of herds that will receive the test in practice? (VETQUADAS item 1.) Accuracy of index: Is the herd disease status likely correctly classified? (VETQUADAS item 3.) Time between tests: Is the time period between herd disease status determination and bulk milk (BM) testing short enough to be reasonably sure that the herd disease status did not change before or at BM testing? (VETQUADAS item 4.) Test description: Was the execution of the BM test described in sufficient detail to allow future replication? (VETQUADAS item 8.) Index description: Was the classification of herd disease status described in sufficient detail to allow future replication? (VETQUADAS item 9.)
3 “Very representative” and “partially representative” were considered yes.
Most disease surveillance programs detected by our search strategy that incorporated active testing of BM samples were designed as part of brucellosis control programs (Supplemental Figure S2, https://data.mendeley.com/datasets/y4w24y6xgt,
). In Denmark, the following 3 programs included testing of BM samples to detect diseases caused by bacteria: the Danish paratuberculosis control program, the Danish surveillance program of Strep. agalactiae, and the Danish surveillance program for Salmonella Dublin. Additionally, the M. bovis disease eradication program includes monthly testing of bulk samples from all dairy farms supplying milk in New Zealand (
Tests rarely have perfect diagnostic performance characteristics. Nonetheless, they provide useful information to help rule in or rule out a condition in a pragmatic manner. In the context of BM testing, defining the condition to be diagnosed is essential and yet often not straightforward. Even for the same diseases, studies employ a variety of definitions of what constitutes a disease-positive herd. This scoping review plays an important role in providing an inclusive list of conditions to be detected within herds and respective characteristics of BM testing, as well as a summary of tests that were used across studies. The findings can be used toward the establishment of disease control programs based on BM testing.
Culture-based protocols have long been the gold standard to diagnose Johne's disease at the cow level (
Predicting sensitivity of repeated environmental sampling for Mycobacterium avium subsp. paratuberculosis in dairy herds using a Bayesian latent class model.
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
The evaluation of the utility of bulk tank tests for the surveillance of Johne’s disease and the effect of storage time and temperature on Johne’s milk ELISA results.
Department of Population Medicine, University of Guelph,
Guelph, Canada2011
Association between herd infection level and the detection of Mycobacterium avium subsp. paratuberculosis (MAP) in bulk tank milk tank using real-time PCR in small holder dairy farms in southern Chile.
in: 11th International Colloquium on Paratuberculosis, Sydney, Australia. International Association for Paratuberculosis,
2012: 65
). Here we reviewed different PCR-based protocols with respect to their ability to correctly classify true infection status. In general, protocols had low to moderate sensitivities (<50%), which were notably higher when enrolled herds had an elevated within-herd seroprevalence (
Association between herd infection level and the detection of Mycobacterium avium subsp. paratuberculosis (MAP) in bulk tank milk tank using real-time PCR in small holder dairy farms in southern Chile.
in: 11th International Colloquium on Paratuberculosis, Sydney, Australia. International Association for Paratuberculosis,
2012: 65
). Thus, it can be used for detection of high-prevalence herds but not herds with a low prevalence of infected cattle (e.g., a negative test result most likely implies that the herd of origin does not contain an elevated within-herd proportion of infected cattle). The poor sensitivity of BM PCR assays is probably a consequence of commingling of milk in tanks as well as the intermittent nature of MAP shedding in milk (
Evaluation of IS900-PCR assay for detection of Mycobacterium avium subspecies paratuberculosis infection in cattle using quarter milk and bulk tank milk samples.
). Specificities of PCR-based protocols were less frequently reported and ranged from 50% to 100%. Low specificity values, as reported from some studies (
Evaluation of IS900-PCR assay for detection of Mycobacterium avium subspecies paratuberculosis infection in cattle using quarter milk and bulk tank milk samples.
Mycobacterium porcinum strains isolated from bovine bulk milk: Implications for Mycobacterium avium subsp. paratuberculosis detection by PCR and culture.
). In short, the practical utility of BM PCR-based testing protocols for MAP, particularly those utilizing nonspecific primers, seems limited at the time of the review.
Mycobacterium avium ssp. paratuberculosis can gain access to bulk tanks through milk of infected cows or through the environment that animals live in. Hence, it is challenging to determine the exact cause of BM contamination if MAP is detected in samples by either PCR-based protocols or direct culture of milk. In this context, ELISA-based technologies that detect antibodies against MAP have gained an important place in Johne's disease surveillance. Here we demonstrate that, in contrast to PCR-based protocols, specificity of ELISA was consistently high (>90%) with few exceptions.
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
), and it is likely that the same data analyzed using revised cut-off values, comparable to other studies, would yield specificities higher than reported. Conversely, sensitivity of BM ELISA was low, even using cut-offs lower than defined by manufacturers. As observed for PCR-based protocols, sensitivity increased according to within-herd prevalence of affected cattle used to define true infection status (
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
Short communication: Correlation between within-herd antibody-prevalence and bulk tank milk antibody levels to Mycobacterium avium ssp. paratuberculosis using 2 commercial immunoassays.
), which means that bulk milk ELISAs will be useful to detect high-prevalence herds, where nearly 5% of seropositive cattle will be required to obtain a positive BM test result (
Diagnostic performance of the Pourquier ELISA for detection of antibodies against Mycobacterium avium subspecies paratuberculosis in individual milk and bulk milk samples of dairy herds.
). The characteristics of ELISA-based protocols for MAP support their use as part of quality assurance or surveillance programs, particularly in settings where MAP is endemic, as positive BM test results will rarely be observed in MAP-negative or low-seroprevalence herds.
Nearly 40% of studies included in this scoping review reported on Staph. aureus in BM samples. Yet, only 7 studies reported on or provide sufficient data to estimate characteristics of BM testing to classify herds with respect to presence or absence of cows with mastitis caused by Staph. aureus. Despite being an important causative agent of mastitis, Staph. aureus bacteria have the potential to cause milk-borne intoxications in humans and are frequently reported as major contaminants of dairy products (
), which explains the increased number of studies focused on Staph. aureus detected by our scoping review. With respect to their potential in infectious disease surveillance, bacteriological culture of BM for presence of Staph. aureus demonstrated inconsistent results, with sensitivity values ranging from 33% to 94% (
) and very high specificity. Thus, given that mastitis from Staph. aureus is prevalent, culture-positive BM samples will rarely be observed in herds without infected cattle.
Conversely, PCR-based technologies demonstrated very high sensitivities and decreased specificities compared with culture-based protocols (
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
), which implies that PCR-negative BM samples will likely come from disease-free herds. The low specificity could be a consequence of contamination of BM by non-udder-associated Staph. aureus or a limited sensitivity of methods used to establish herd disease status, which can be particularly problematic if enrolled herds have a low prevalence of affected cattle. We had no reason to suspect that methods used across studies to determine herd disease status suffered from low sensitivity and directly influenced BM testing specificity. Culture of quarter-level non-repeated milk samples has high sensitivity to detect IMI caused by Staph. aureus (
). In addition, sensitivity increases as more animals are sampled and those results are used to inform herd disease status. Most validation studies reporting on Staph. aureus enrolled a representative number of cows per herd to classify farms as either positive or negative for presence of infected cattle. In contrast, Staph. aureus can be isolated from sources other than mastitic milk, which includes teat skin, housing, feedstuffs, equipment, and humans working on dairy farms (
). Non-udder-associated Staph. aureus may eventually contaminate the BM even in herds where cows are free from IMI caused by Staph. aureus, which would in turn affect BM testing specificity. Indeed, qPCR had 100% specificity for the udder-associated Staph. aureus genotype B (
Bovine mastitis: The diagnostic properties of a PCR-based assay to monitor the Staphylococcus aureus genotype B status of a herd, using bulk tank milk.
). Additionally, milk from cows with clinical mastitis is usually not included in the BM, which will potentially explain false negatives on culture and qPCR. Nevertheless, the 3 methods demonstrated very high specificity (>95%); that is, positive BM samples will most likely come from infected herds, particularly in endemic areas. For Strep. agalactiae, we observed inconsistency among studies with respect to the sensitivity of BM culture (
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
Short communication: Comparing real-time PCR and bacteriological cultures for Streptococcus agalactiae and Staphylococcus aureus in bulk-tank milk samples.
). Nevertheless, as observed for Mycoplasma bovis, testing of BM samples had very high specificity (>95%) for Strep. agalactiae, regardless of methodology.
Salmonella Dublin has been a cause of concern for the European dairy industry for many decades (
Narrative review comparing principles and instruments used in three active surveillance and control programmes for non-EU-regulated diseases in the Danish cattle population.
). In addition, Salmonella Dublin can cause severe infections in humans, and studies have suggested that its incidence in people has been increasing over the last decades (
). Our findings reinforce that one of most important factors affecting sensitivity of BM ELISA to detect Salmonella Dublin herds is whether nonlactating cattle such as heifers and calves will be considered in the definition of true infection status (
). Additionally, as specificity of included protocols was generally high (>95%), positive BM tests will likely denote infection in endemic settings. Nevertheless, in disease-free or low-prevalence areas, most positive tests will still come from true negative herds (e.g., false positives), particularly when tests with low sensitivities are employed (
Estimation of the accuracy of an ELISA test applied to bulk tank milk for predicting herd-level status for Salmonella Dublin in dairy herds using Bayesian Latent Class Models.
In the context of dairy cattle disease surveillance, testing of BM samples is extremely convenient, as they are collected routinely from a significant number of herds in many parts of the world. However, in general, protocols analyzed in this review suffered from very low sensitivities, hardly justifying the use of BM as part of disease surveillance. Low-sensitivity and high-specificity methods can be valuable tools for surveillance when used repeatedly over time. In New Zealand, a BM sample is collected from all milk-recorded farms monthly and screened for Mycoplasma bovis using ELISA (
). Positive results are followed by on-farm investigations to determine the true infection status of herds. Likewise, the Danish Strep. agalactiae eradication program was initially based on repeated culture of BM samples from all Danish herds (
). Furthermore, the Danish Salmonella Dublin surveillance program is based on 4 assessments per year of BM samples collected from all milk-recorded herds in Denmark using an in-house ELISA (
Narrative review comparing principles and instruments used in three active surveillance and control programmes for non-EU-regulated diseases in the Danish cattle population.
). Additionally, repeated testing of BM samples is an important element of many brucellosis surveillance and eradication programs, including the Dutch national B. abortus eradication program (
Most studies included in this review analyzed BM samples using ELISA, PCR-based technologies, or bacteriological culture; the same technologies are still being used to detect infectious diseases of dairy cattle in many parts of the world. Conventional methods for diagnostics of bacterial diseases of dairy cattle are, in general, time consuming and can hardly be automated, with a few exceptions. Recent advances in technology have led to the development of a new family of tools, such as biosensors (
). Biosensors, as diagnostic devices, offer a low-cost solution for rapid detection of infectious diseases. In theory, testing with biosensors can be automated and used directly on farms. Historically, the dairy industry has been increasing its uptake of technology to optimize costs and minimize the amount of labor required (
). In this context, we believe that further research should investigate the use of modern, nonconventional diagnostic tools that can be used on farms, such as biosensors, which were not used in any study included in this scoping review.
Here we provide an inclusive list of diseases that can be tracked using BM samples, tests that were used for that purpose, as well as a detailed description of their characteristics. Nevertheless, our findings should be considered in the face of study limitations and caveats. Although we included 474 studies in this scoping review, we have excluded several studies based on language and absence of full text. Further, we did not screen proceedings from all disease-specific conferences; neither have we requested access to internal validation reports from diagnostic industries, as the majority of these documents are not peer-reviewed. Hence, we potentially missed studies and tests that would have been included in this review. However, we must emphasize that we also screened the Diagnostics for Animal database, which contains information on nearly 90% of the current global animal health diagnostic market. Hence, it is unlikely that non-included studies employed a diagnostic method that is currently available to dairy producers and other stakeholders. Finally, the list of pathogens that we provided should not be regarded as a complete list of pathogens that can be tracked using BM.
CONCLUSIONS
This scoping review focused on the use of BM testing to detect infectious diseases of dairy cattle caused by bacteria. In general, protocols analyzed had low sensitivities and high specificities, which varied according to the pathogen screened as well as the testing methodology employed. For MAP, BM PCR and ELISA protocols demonstrated increased sensitivity and decreased specificity compared with culture-based protocols, and sensitivity increased according to within-herd prevalence of affected cattle used to define infection status of herds. Likewise, PCR had higher sensitivity and decreased specificity compared with culture to detect herds with cows infected by Staph. aureus, although qPCR demonstrated excellent specificity to detect herds free of udder-associated Staph. aureus. We found inconsistencies with respect to the sensitivity of BM culture to detect herds infected with Strep. agalactiae, which varied according to the spectrum of enrolled herds. For Salmonella Dublin, our findings support that one of most important factors affecting sensitivity of BM ELISA is whether nonlactating cattle are considered in the definition of herd infection status. Finally, testing of BM samples can be an important element of infectious disease surveillance programs, particularly if repeated testing is implemented over time.
ACKNOWLEDGMENTS
DN received funding and support from Mitacs (Toronto, Canada), the Dairy Farmers of Ontario (DFO; Mississauga, Canada) and the Natural Sciences and Engineering Research Council of Canada (NSERC; Ottawa). Project funding was provided by the Ontario Ministry of Agriculture Food and Rural Affairs (OMAFRA; Guelph, Canada) through the Ontario Agri-Food Innovation Alliance and Dairy Farmers of Ontario. The authors have not stated any conflicts of interest.
REFERENCES
Agger J.F.
Christoffersen A.
Rattenborg E.
Nielsen J.
Agerholm J.S.
Prevalence of Coxiella burnetii antibodies in Danish dairy herds.
Dynamics of relationship between the presence of Coxiella burnetii DNA, antibodies, and intrinsic variables in cow milk and bulk tank milk from Danish dairy cattle.
00032-2&id=gr1.jpg){kind=link}