Estimating the herd and cow level prevalence of bovine digital dermatitis on New Zealand dairy farms: A Bayesian superpopulation approach.

A cross-sectional study of 127 dairy herds distributed across four regions of New Zealand (NZ) was conducted to estimate the regional herd-level prevalence of bovine digital dermatitis (BDD) and the prevalence of cows with BDD lesions within affected herds. Each herd was visited once during the 2016-2017 lactating season and the rear feet of all cows in the milking herd were examined to detect the presence of BDD lesions. Of the 127 herds examined, 63 had at least one cow with a detected BDD lesion. Of the 59 849 cows observed, 646 cows were observed with BDD lesions. All of the herds in which BBD was detected were located in three of the four regions (Waikato, Manawatu and South Canterbury). No convincing lesions were observed on the West Coast. The probability of BDD freedom on the West Coast was predicted to be 99.97% using a Bayesian latent class model. For the three regions where BDD lesions were observed, the true herd level and cow level prevalences were estimated using a Bayesian superpopulation approach which accounted for the imperfect diagnostic method. Based on priors obtained from previous research in another region of NZ (Taranaki), the true herd level prevalences in Waikato, Manawatu and South Canterbury were estimated to be 59.2% (95% probability interval [PI]: 44.3%-73.9%), 43.3% (95%PI: 29%-59%) and 65.9% (95%PI: 49.5%-79.9%), respectively, while the true median within-herd prevalences were estimated as 3.2% (95%PI: 2%-5%), 1.7% (95%PI: 0.9%-3.1%) and 3.7% (95%PI: 2.4%-5.5%), respectively. All of these estimates except for the true herd level prevalence in Manawatu were fairly robust to changes in the priors. For Manawatu region, changing from the prior obtained in Taranaki (the best estimate of the herd level prevalence = 60%, 95% sure > 40%) to one where the mode was 50% (95% sure < 80%) reduced the posterior from 43.3% to 35.2% (95%PI: 20.1%-53.5%). The marked variation in BDD prevalence between regions and between farms highlights the need for further exploration into risk factors for disease.

Development and validation of a 3ABC antibody ELISA in Australia for foot and mouth disease.

OBJECTIVE: To measure the diagnostic performance of an Australian-developed ELISA for the detection of antibodies against the non-structural proteins (NSP) 3ABC of the foot and mouth disease (FMD) virus. DESIGN: Test development and validation study. METHODS: The diagnostic specificity was determined using 2535 sera from naïve animals and 1112 sera from vaccinated animals. Diagnostic sensitivity was calculated from the data for 995 sera from experimentally and field-infected animals from FMD-endemic countries in South East Asia. A commercial ELISA detecting antibodies against FMD virus NSP was used as the reference test to establish relative sensitivity and specificity. Bayesian latent class analysis was performed to corroborate results. The diagnostic window and rate of detection were determined at different times using sera from cattle, sheep and pigs before and after infection, and after vaccination and subsequent infection. Repeatability and reproducibility data were established. RESULTS: At 35% test cut-off, the 3ABC ELISA had an overall diagnostic sensitivity of 91.5% and diagnostic specificity of 96.4%. The diagnostic sensitivity in vaccinated and subsequently infected cattle was 68.4% and diagnostic specificity in vaccinated cattle was 98.0%. CONCLUSIONS: The 3ABC ELISA identified field and experimentally infected animals, as well as vaccinated and subsequently infected animals. Diagnostic sensitivity and specificity estimates for other FMD NSP tests are comparable with the results obtained in this study. This NSP ELISA was found to be 'fit for purpose' as a screening assay at the herd level to detect viral infection and also to substantiate absence of infection.

Herd-level prevalence of Mycobacterium avium subsp. paratuberculosis infection in United States dairy herds in 2007.

Testing of composite fecal (environmental) samples from high traffic areas in dairy herds has been shown to be a cost-effective and sensitive method for classification of herd status for Mycobacterium avium subsp. paratuberculosis (MAP). In the National Animal Health Monitoring System's (NAHMS) Dairy 2007 study, the apparent herd-level prevalence of MAP was 70.4% (369/524 had ≥ 1 culture-positive composite fecal samples out of 6 tested). Based on these data, the true herd-level prevalence (HP) of MAP infection was estimated using Bayesian methods adjusting for the herd sensitivity (HSe) and herd specificity (HSp) of the test method. The Bayesian prior for HSe of composite fecal cultures was based on data from the NAHMS Dairy 2002 study and the prior for HSp was based on expert opinion. The posterior median HP (base model) was 91.1% (95% probability interval, 81.6 to 99.3%) and estimates were most sensitive to the prior for HSe. The HP was higher than estimated from the NAHMS Dairy 1996 and 2002 studies but estimates are not directly comparable with those of prior NAHMS studies because of the different testing methods and criteria used for herd classification.

A framework for the joint modeling of longitudinal diagnostic outcome data and latent infection status: application to investigating the temporal relationship between infection and disease.

For many diseases the infection status of individuals cannot be observed directly, but can only be inferred from biomarkers that are subject to measurement error. Diagnosis of infection based on observed symptoms can itself be regarded as an imperfect test of infection status. The temporal relationship between infection and marker outcomes may be complex, especially for recurrent diseases where individuals can experience multiple bouts of infection. We propose an approach that first models the unobserved longitudinal infection status of individuals conditional on relevant covariates, and then jointly models the longitudinal sequence of biomarker outcomes conditional on infection status and covariate information through time, thus resulting in a joint model for longitudinal infection and biomarker sequences. This model can be used to investigate the temporal dynamics of infection, and to evaluate the usefulness of biomarkers for monitoring purposes. Our work is motivated and illustrated by a longitudinal study of bovine digital dermatitis (BDD) on commercial dairy farms in North West England and North Wales, in which the infection of interest is Treponeme spp., and the biomarkers of interest are a continuous enzyme-linked immunosorbent assay test outcome and a dichotomous outcome, foot lesion status. BDD is known to be one of the possible causes of foot lesions in cows.

Diagnostic assessment without cut-offs: application of serology for the modelling of bovine digital dermatitis infection.

Bovine digital dermatitis (BDD) is an epidermitis which is a leading cause of infectious lameness. The only recognized diagnostic test is foot inspection, which is a labour-intensive procedure. There is no universally recognized, standardized lesion scoring system. As small lesions are easily missed, foot inspection has limited diagnostic sensitivity. Furthermore, interpretation is subjective, and prone to observer bias. Serology is more convenient to carry out and is potentially a more sensitive indicator of infection. By carrying out 20 serological assays using lesion-associated Treponema spp. isolates, three serogroups were identified. The reliability of the tests was established by assessing the level of agreement and the concordance correlation coefficient. Subsequently, an ELISA suitable for routine use was developed. The benchmark of diagnostic test validation is conventionally the determination of the key test parameters, sensitivity and specificity. This requires the imposition of a cut-off point. For serological assays with outcomes on a continuous scale, the degree by which the test result differs from this cut-off is disregarded. Bayesian statistical methodology has been developed which enables the assay result also to be interpreted on a continuous scale, thereby optimizing the information inherent in the test. Using a cross-sectional study dataset carried out on 8 representative dairy farms in the UK, the probability of infection, P(I), of each individual animal was estimated in the absence of a 'Gold Standard' by modelling I as a latent variable which was determined by lesion status, L as well as serology, S. Covariate data (foot hygiene score and age) were utilized to estimate P(L) when no lesion inspection was performed. Informative prior distributions were elicited where possible. The model was utilized for predictive inference, by computing estimates of P(I) and P(L) independently of the data. A more detailed and informative analysis of the farm-level distribution of infection could thus be performed. Also, biases associated with the subjective interpretation of lesion status were minimized. Model outputs showed that young stock were unlikely to be infected, whereas cows tended to have high or low probabilities of being infected. Estimates of probability of infection were considerably higher for animals with lesions than for those without. Associations were identified between both covariates and probability of infection in cows, but not in the young stock. Under the condition that the model assumptions are valid for the larger population, the results of this work can be generalized by predictive inference.

Use of expert opinion for animal disease decisions: an example of foot-and-mouth disease status designation.

When data representing a preferred measurement of risk cannot be obtained, as is often the case for global animal diseases, decisions that affect millions of people and their animals are typically made based on expert opinion. Expert opinion can be and has been used to address the critical lack of data existing for prevalence and incidence of many global diseases, including foot-and-mouth disease (FMD). However, when a conclusion based on expert opinion applies to a topic as sensitive as FMD, which has tremendous economic, political, and social implications, care should be taken to understand the accuracy of and differences in the opinion data. The differences in experts' opinions and the relative accuracy of an expert opinion elicitation for "diagnosing" country-level FMD presence were examined for the years 1997-2003 using Bayesian methods. A formal survey of eight international FMD experts revealed that individual experts had different opinions as to the probability of finding FMD in a country. However, a weighted average of the experts' responses was relatively accurate (91% sensitivity and 85% specificity) at identifying the FMD status of a country, compared to using a method that employed information available from World Organization for Animal Health (OIE). The most apparent disagreements between individual experts and available information were found for Indonesia, South Korea, and South America, and, in general, the experts seemed to believe that countries in South Asia were more likely to be positive than other countries that reported FMD cases to OIE. This study highlights new methodology that offers a standardized, quantitative, and systematic means by which expert opinion can be used and assessed.

Analytical epidemiology of genomic variation among Pan Asia strains of foot-and-mouth disease virus.

Genetic data from field isolates of foot-and-mouth disease virus (FMDV) have been used to trace the source of recent outbreaks of FMD, to design better vaccines and diagnostic tests for FMDV, and to make conclusions regarding the general variability in the FMDV genome. Though epidemiologic data associated with FMDV isolates are available, these data have been used rarely to explore possible associations of epidemiologic factors with evolution or variation of the FMDV genome. In this study, factors associated with variation in the primary immunogenic peptide gene of FMDV (VP1), for a sample of 147 serotype O, Pan Asia strain sequences were explored using traditional analytical epidemiologic methods: logistic regression and multinomial-response logistic regression. Hypothesized factors included host type (bovine, ovine, buffalo, or porcine) and geographical region (Middle East, South Asia, East Asia, Southeast Asia, and Europe). Results of two regression analyses suggest that host type and region, considered to be possible surrogates for host management, may be associated with selection in the VP1 amino acid sequence in FMDV. For example, isolates from cattle and sheep in South Asia appear to converge with a proposed ancestor sequence, whereas isolates from the same species in the Middle East and Southeast Asia appear to diverge. The methods demonstrated here could be used on a more detailed dataset to explore the selective pressure of host immunity on the evolution of FMDV antigens in an endemic setting. More broadly, epidemiologic methods could be applied extensively to molecular data to explore the causes of genomic variation in disease-causing organisms.

Bayesian spatiotemporal analysis of foot-and-mouth disease data from the Republic of Turkey.

A flexible hierarchical Bayesian spatiotemporal regression model for foot-and-mouth disease (FMD) was applied to data on the annual number of reported FMD cases in Turkey from 1996 to 2003. The longitudinal component of the model was specified as a latent province-specific stochastic process. This stochastic process can accommodate various types of FMD temporal profiles. The model accounted for differences in FMD occurrence across provinces and for spatial correlation. Province-level covariate information was incorporated into the analysis. Results pointed to a decreasing trend in the number of FMD cases in western Turkey and an increasing trend in eastern Turkey from 1996 to 2003. The model also identified provinces with high and with low propensities for FMD occurrence. The model's use of flexible structures for temporal trend and of generally applicable methods for spatial correlation has broad application to predicting future spatiotemporal distributions of disease in other regions of the world.

Estimation of diagnostic-test sensitivity and specificity through Bayesian modeling.

We review recent Bayesian approaches to estimation (based on cross-sectional sampling designs) of the sensitivity and specificity of one or more diagnostic tests. Our primary goal is to provide veterinary researchers with a concise presentation of the computational aspects involved in using the Bayesian framework for test evaluation. We consider estimation of diagnostic-test sensitivity and specificity in the following settings: (i) one test in one population, (ii) two conditionally independent tests in two or more populations, (iii) two correlated tests in two or more populations, and (iv) three tests in two or more populations, where two tests are correlated but jointly independent of the third test. For each scenario, we describe a Bayesian model that incorporates parameters of interest. The WinBUGS code used to fit each model, which is available at http://www.epi.ucdavis.edu/diagnos-tictests/, can be altered readily to conform to different data.

Predicting the probability of abortion in dairy cows: a hierarchical Bayesian logistic-survival model using sequential pregnancy data.

Although abortion contributes substantially to poor reproductive health of dairy herds, little is known about the predictability of abortion based on age, previous abortion or gravidity (number of previous pregnancies). A poor understanding of effects of maternal factors on abortion risk exists, in part, because of methodological difficulties related to non-independence of multiple pregnancies of the same cow in analysis of fetal survival data. We prospectively examined sequential pregnancies to investigate relationships between fetal survival and putative dam risk factors for 2991 abortions from 24,706 pregnancies of 13,145 cows in nine California dairy herds. Relative risks and predicted probabilities of abortion (PPA) were estimated using a previously described hierarchical Bayesian logistic-survival model generalized to incorporate longitudinal data of multiple pregnancies from a single cow. The PPA increased with increasing dam age at conception, with increasing number of previous abortions, and if the previous pregnancy was aborted >60 days in gestation. The PPA decreased with increasing gravidity and with increasing number of days open. For cows that aborted, the median time to fetal death decreased slightly as gravidity increased. The study considers several methodological issues faced in epidemiologic investigations of fetal health, including multi-modal hazard functions, extensive censoring and non-independence of multiple pregnancies. The model improves our ability to predict bovine abortion and to characterize fetal survival, which have important applications to herd health management.

Bayesian modeling of animal- and herd-level prevalences.

We reviewed Bayesian approaches for animal-level and herd-level prevalence estimation based on cross-sectional sampling designs and demonstrated fitting of these models using the WinBUGS software. We considered estimation of infection prevalence based on use of a single diagnostic test applied to a single herd with binomial and hypergeometric sampling. We then considered multiple herds under binomial sampling with the primary goal of estimating the prevalence distribution and the proportion of infected herds. A new model is presented that can be used to estimate the herd-level prevalence in a region, including the posterior probability that all herds are non-infected. Using this model, inferences for the distribution of prevalences, mean prevalence in the region, and predicted prevalence of herds in the region (including the predicted probability of zero prevalence) are also available. In the models presented, both animal- and herd-level prevalences are modeled as mixture distributions to allow for zero infection prevalences. (If mixture models for the prevalences were not used, prevalence estimates might be artificially inflated, especially in herds and regions with low or zero prevalence.) Finally, we considered estimation of animal-level prevalence based on pooled samples.

Evaluation of brucellosis RB51 vaccine for domestic water buffalo (Bubalus bubalis) in Trinidad.

Thirty-two young domestic water buffalo (Bubalus bubalis) were obtained from a brucellosis-free farm to determine effectiveness of RB51 vaccination for prevention of Brucella infection under natural-exposure conditions in Trinidad. Study animals (20 males and 12 females 5-20 months old) were assigned to vaccination or control groups, using a block randomization design ensuring equal sex distributions between groups. The vaccination group received commercially available RB51 at the recommended calfhood dose of (1.0-3.4)x10(10) colony-forming units (CFU) and controls received 2ml sterile saline. Vaccination did not result in positive serologic results as measured by four traditional agglutination tests: standard tube agglutination test (STAT), standard plate agglutination test (SPAT), buffered plate agglutination test (BPAT), and card agglutination. Study animals were maintained in a brucellosis-positive herd in southern Trinidad with an estimated 56% prevalence to allow for natural exposure to B. abortus, which was evaluated using STAT, SPAT, BPAT, and card tests. Animals were sampled seven times over 2 years and were classified as positive if they had persistent agglutination titers or had Brucella isolated from specimens collected at completion of the study. Five of the original 32 study animals were lost to follow-up during the field trial. Six of the 14 (43%) vaccinated animals completing the study were classified as positive for Brucella infection-as were two of the 13 (15%) control animals (P=0.21). Isolates from four vaccinates and one control were confirmed as B. abortus biovar 1.

Hierarchical Bayesian model for prevalence inferences and determination of a country's status for an animal pathogen.

Certification that a country, region or state is "free" from a pathogen or has a prevalence less than a threshold value has implications for trade in animals and animal products. We develop a Bayesian model for assessment of (i) the probability that a country is "free" of or has an animal pathogen, (ii) the proportion of infected herds in an infected country, and (iii) the within-herd prevalence in infected herds. The model uses test results from animals sampled in a two-stage cluster sample of herds within a country. Model parameters are estimated using modern Markov-chain Monte Carlo methods. We demonstrate our approach using published data from surveys of Newcastle disease and porcine reproductive and respiratory syndrome in Switzerland, and for three simulated data sets.

Selection bias in epidemiological studies of infectious disease using Escherichia coli and avian cellulitis as an example.

In epidemiological studies of infectious disease, researchers often rely on specific cues of the host, such as clinical signs, as surrogate indicators of pathogen presence. A selection bias would manifest if the specific visual cues used in sampling for the pathogen were not representative of the full range of signs caused by the strains of that pathogen. In our molecular epidemiological studies of Escherichia coli associated with avian cellulitis in broilers, we collect carcasses at the processing plant based on visual cues of lesion morphology. Therefore, the objectives of this study were to: (1) explore the potential impacts of selection bias in an application of infectious disease epidemiology, and (2) utilize a validation protocol to assess the potential for selection bias in our molecular epidemiological studies of E. coli and avian cellulitis. In two different trials, E. coli DNA fingerprints were compared between birds that our observers collected and the birds that the observers missed. Using Fisher's exact tests and simulation models, we determined that the isolates collected by the observers were not significantly different from the isolates missed by the observers (P > 0.60 in both trials). Our method of selecting birds suspected of having cellulitis did not significantly bias our inferences about the population of E. coli associated with cellulitis in the flock. We encourage more investigators to critically assess the relationship of the sample to the target population in epidemiological studies of infectious disease.

Screening without a "gold standard": the Hui-Walter paradigm revisited.

The authors consider screening populations with two screening tests but where a definitive "gold standard" is not readily available. They discuss a recent article in which a Bayesian approach to this problem is developed based on data that are sampled from a single population. It was subsequently pointed out that such inferences will not necessarily be accurate in the sense that standard errors for parameters may not decrease as n increases. This problem will generally occur when the data are insufficient to estimate all of the parameters as is the case when screening a single population with two tests. If both tests are applied to units sampled from two populations, however, this particular difficulty disappears. In this article the authors further examine this issue and develop an approach based on sampling two populations that yields increasingly accurate inferences as the sample size increases.

Risk factors for outbreaks of disease attributable to white sturgeon iridovirus and white sturgeon herpesvirus-2 at a commercial sturgeon farm.

OBJECTIVE: To determine management, fish, and environmental risk factors for increased mortality and an increased proportion of runts for white sturgeon exposed to white sturgeon iridovirus (WSIV) and white sturgeon herpesvirus-2 (WSHV-2). ANIMALS: White sturgeon in 57 tanks at 1 farm and observations made for fish at another farm. PROCEDURE: A prospective cohort study was conducted. Data on mortality, proportion of runts, and potential risk factors were collected. Five fish from each tank were examined for WSIV and WSHV-2 via inoculation of susceptible cell lines and microscopic examination of stained tissue sections. An ANCOVA was used to evaluate effects of risk factors on mortality and proportion of runts. RESULTS: Major determinants of number of dead fish (natural logarithm [In]-transformed) were spawn, source (90% confidence interval [CI] for regression coefficient, 0.62 to 2.21), and stocking density (90% CI, 0.003 to 0.03). Main predictors of proportion of runts (In-transformed) were spawn, mortality incidence density (90% CI, 0.004 to 0.03), age (90% CI, -0.012 to -0.004), and the difference in weight between the largest and smallest nonrunt fish (90% CI, 0.0002 to 1.24). Additional observations indicated a possible protective effect attributable to previous exposure to the viruses. CONCLUSIONS AND CLINICAL RELEVANCE: Mortality and proportion of runts for white sturgeon after exposure to WSIV and WSHV-2 may be reduced for a farm at which the viruses are endemic by selection of specific broodstock, stocking with fish that survived outbreaks of viral disease, using all-in, all-out production, and decreasing stocking densities.

Persistence of cellulitis-associated Escherichia coli DNA fingerprints in successive broiler chicken flocks.

Avian cellulitis in broiler chickens is primarily caused by Escherichia coli. Previous research found that the E. coli isolates of cellulitis origin were unique to each ranch, suggesting that these E. coli were endemic within the ranch environment. To test the hypothesis that the E. coli associated with cellulitis are endemic in the litter of the broiler house, we designed a study to determine whether E. coli DNA fingerprints associated with cellulitis persist over successive flocks that are grown in the same house. In addition, we assessed the impact of different cleaning and disinfection strategies on this persistence. Two broiler houses were followed on each of five farms over 3-4 flocks. A total of 353 E. coli isolates from cellulitis lesions were analyzed in this study, and 314 of these isolates (89%) were DNA fingerprinted by PFGE. In each ranch, there were several DNA fingerprint patterns that were present over successive flocks, regardless of the cleaning and disinfection strategy utilized. Isolates persisted as long as 191 days, implying that these E. coli are capable of persisting in the broiler house environment for long periods of time. In addition, these E. coli isolates were associated with cellulitis lesions in successive flocks. Thus, the isolates of E. coli that are associated with cellulitis in broiler chickens appear to be endemic in the litter environment of the broiler house.

Pooled-sample testing as a herd-screening tool for detection of bovine viral diarrhea virus persistently infected cattle.

The study was conducted to develop methodology for least-cost strategies for using polymerase chain reaction (PCR)/probe testing of pooled blood samples to identify animals in a herd persistently infected with bovine viral diarrhea virus (BVDV). Cost was estimated for 5 protocols using Monte Carlo simulations for herd prevalences of BVDV persistent infection (BVDV-PI) ranging from 0.5% to 3%, assuming a cost for a PCR/probe test of $20. The protocol associated with the least cost per cow involved an initial testing of pools followed by repooling and testing of positive pools. For a herd prevalence of 1%, the least cost per cow was $2.64 (95% prediction interval = $1.72, $3.68), where pool sizes for the initial and repooled testing were 20 and 5 blood samples per pool, respectively. Optimization of the least cost for pooled-sample testing depended on how well a presumed prevalence of BVDV-PI approximated the true prevalence of BVDV infection in the herd. As prevalence increased beyond 3%, the least cost increased, thereby diminishing the competitive benefit of pooled testing. The protocols presented for sample pooling have general application to screening or surveillance using a sensitive diagnostic test to detect very low prevalence diseases or pathogens in flocks or herds.

Estimation of sensitivity and specificity of diagnostic tests and disease prevalence when the true disease state is unknown.

The performance of a new diagnostic test is frequently evaluated by comparison to a perfect reference test (i.e. a gold standard). In many instances, however, a reference test is less than perfect. In this paper, we review methods for estimation of the accuracy of a diagnostic test when an imperfect reference test with known classification errors is available. Furthermore, we focus our presentation on available methods of estimation of test characteristics when the sensitivity and specificity of both tests are unknown. We present some of the available statistical methods for estimation of the accuracy of diagnostic tests when a reference test does not exist (including maximum likelihood estimation and Bayesian inference). We illustrate the application of the described methods using data from an evaluation of a nested polymerase chain reaction and microscopic examination of kidney imprints for detection of Nucleospora salmonis in rainbow trout.

Log-linear and logistic modeling of dependence among diagnostic tests.

We developed log-linear and logistic-modeling approaches to investigate dependence among diagnostic tests. To illustrate the approaches, we used published data for swine toxoplasmosis, bovine paratuberculosis, and swine brucellosis. These diseases were selected because each animal's true disease status was known, at least five tests were used, and the serologic tests had been previously shown to have moderate-to-high pairwise dependence in test sensitivities (and sometimes in test specificities). Log-linear and logistic modeling yielded similar results for swine toxoplasmosis and swine brucellosis. However, logistic modeling could not be used to investigate test dependence for bovine paratuberculosis because of quasi-separation in the data attributable to two fecal-based tests having specificities of 100%. Findings from our modeling indicated that 3 (modified agglutination, enzyme-linked immunosorbent assay (ELISA), latex agglutination) of 5 serologic tests for toxoplasmosis and 2 (rivanol and particle concentration fluorescence immunoassay) of 6 serologic tests for brucellosis were adequate for diagnosis. For bovine paratuberculosis, both fecal-based tests (Herrold's egg-yolk culture and radiometric culture) and 1 (ELISA) of 3 serologic tests were necessary in serial and parallel testing schemes.