Predicting percentage of intramuscular fat using two types of real-time ultrasound equipment

In the present study, 500 steers were used to develop models for predicting the percentage of intramuscular fat (PIMF) in live beef cattle. Before slaughter, steers were scanned across the 11th and 13th ribs using Aloka 500V (AL-500) and Classic Scanner 200 (CS-200) machines. Four to five images were collected per individual steer using each machine. After slaughter, a cross-sectional slice of the longissimus muscle from the 12th rib facing was used for chemical extraction to determine actual carcass percentage of intramuscular fat (CPIMF). Texture analysis software was used by two interpreters to select a region for determination of image parameters, which included Fourier, gradient, histogram, and co-occurrence parameters. Four prediction models were developed separately for each of AL-500 and CS-200 based on images captured by the respective machines. These included models developed without transformation of CPIMF (Model I), models based on logarithmic transformation of CPIMF (Model II), ridge regression procedure (Model III), and principal component regression procedure (Model IV). Model R2 and root mean square error of AL-500 Models I, II, III, and IV were 0.72, 0.84%; 0.72, 0.85%; 0.69, 0.91%; and 0.71, 0.86%; respectively. The corresponding R2 and root mean square error values of CS-200 Models I, II, III, and IV were 0.68, 0.87%; 0.70, 0.85%; 0.64, 0.94%; and 0.65, 0.91%; respectively. Initially, AL-500 and CS-200 prediction models were validated separately on an independent data set from 71 feedlot steers. The overall mean bias, standard error of prediction, and rank correlation coefficient across the four AL-500 models were 0.42%, 0.84%, and 0.88, respectively. For the four CS-200 models, the corresponding overall mean values were 0.67%, 0.81%, and 0.91, respectively. In a second validation test, only Model II of AL-500 and CS-200 was evaluated separately based on data from 24 feedlot steers. The overall mean bias, absolute difference, and standard error of prediction of AL-500 Model II were 0.71, 0.92, and 0.98%. For CS-200 Model II, the corresponding values were 0.59, 0.97, and 1.03%. Both AL-500 and CS-200 equipment can be used to accurately predict PIMF in live cattle. Further improvement in the accuracy of prediction equations could be achieved through increasing the development data set and the variation in PIMF of cattle used.     

A spatial risk assessment of bighorn sheep extirpation by grazing domestic sheep on public lands

Bighorn sheep currently occupy just 30% of their historic distribution, and persist in populations less than 5% as abundant overall as their early 19th century counterparts. Present-day recovery of bighorn sheep populations is in large part limited by periodic outbreaks of respiratory disease, which can be transmitted to bighorn sheep via contact with domestic sheep grazing in their vicinity. In order to assess the viability of bighorn sheep populations on the Payette National Forest (PNF) under several alternative proposals for domestic sheep grazing, we developed a series of interlinked models. Using telemetry and habitat data, we characterized herd home ranges and foray movements of bighorn sheep from their home ranges. Combining foray model movement estimates with known domestic sheep grazing areas (allotments), a Risk of Contact Model estimated bighorn sheep contact rates with domestic sheep allotments. Finally, we used demographic and epidemiologic data to construct population and disease transmission models (Disease Model), which we used to estimate bighorn sheep persistence under each alternative grazing scenario. Depending on the probability of disease transmission following interspecies contact, extirpation probabilities for the seven bighorn sheep herds examined here ranged from 20% to 100%. The Disease Model allowed us to assess the probabilities that varied domestic sheep management scenarios would support persistent populations of free-ranging bighorn sheep.     

Models for the dispersal in Australia of the arbovirus vector, Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae)

Culicoides brevitarsis is the main biting midge responsible for the transmission of bluetongue and Akabane viruses to livestock in Australia. Models are given for its dispersal after winter from endemic areas at the southern limit of its distribution in New South Wales (NSW); the models might also be applicable elsewhere. Model 1 shows that dispersal can be explained by distance from a key point just outside the endemic area in mid-northern/northern coastal NSW. The model provides probability data for times of first occurrence at sites within regions down the southern coastal plain or up the Hunter Valley towards (but rarely reaching) the western slopes and tablelands. Model 2 shows that the movement depends on temperature and wind speed from northerly and easterly directions. Preliminary data also are given to suggest a relationship between density in the endemic area and the maximum distance that C. brevitarsis can travel in a given year. The models can be linked to other information which in combination can provide probabilities for winter survival outside the endemic area, times of occurrence at sites where it cannot survive winter and times when activity ceases naturally at these sites at the end of the season. This information can be used to predict the potential for virus transmission and indicate zones of seasonal freedom from both vector and virus for the export of livestock.     

Bayesian analysis of animal movements related to factors at herd and between herd levels: Implications for disease spread modeling

A method to assess the influence of between herd distances, production types and herd sizes on patterns of between herd contacts is presented. It was applied on pig movement data from a central database of the Swedish Board of Agriculture. To determine the influence of these factors on the contact between holdings we used a Bayesian model and Markov chain Monte Carlo (MCMC) methods to estimate the posterior distribution of model parameters. The analysis showed that the contact pattern via animal movements is highly heterogeneous and influenced by all three factors, production type, herd size, and distance between holdings. Most production types showed a positive relationship between maximum capacity and the probability of both incoming and outgoing movements. In agreement with previous studies, holdings also differed in both the number of contacts as well as with what holding types contact occurred with. Also, the scale and shape of distance dependence in contact probability was shown to differ depending on the production types of holdings.To demonstrate how the methodology may be used for risk assessment, disease transmissions via animal movements were simulated with the model used for analysis of contacts, and parameterized by the analyzed posterior distribution. A Generalized Linear Model showed that herds with production types Sow pool center, Multiplying herd and Nucleus herd have higher risk of generating a large number of new infections. Multiplying herds are also expected to generate many long distance transmissions, while transmissions generated by Sow pool centers are confined to more local areas. We argue that the methodology presented may be a useful tool for improvement of risk assessment based on data found in central databases.     

Evaluation of the impact of errors in the measurement of backfat depth on the prediction of fat-free lean mass

The development of regression equations to predict carcass composition typically assumes that the independent variables, such as backfat depth, are measured without error. However, technological and operator-specific types of measurement errors do exist. To evaluate the impact of measurement error for backfat depth, Monte Carlo simulation was used to model carcass fat-free lean mass (FFLM) in pigs. In the simulation, FFLM was a linear function of carcass weight and actual backfat depth (ABFD). Carcass weight was assumed to be measured without error, but measurement errors were generated such that the correlation (r(BF)) of the measured backfat depth (BFD) and ABFD ranged from 0.70 to 0.95. Two types of measurement errors were simulated: 1) constant variation that was additive to the variance of ABFD, and 2) variation proportional to the ABFD that was additive to the variance in ABFD. A total of 1,000 replications of 1,000 pigs were simulated. Within each type of measurement error, the absolute values of the regression coefficients and R2 values of the equations decreased as r(BF) decreased. The probability of the backfat depth squared (BFD2) being significant (P < 0.05) in the regression equation was increased when the measurement errors were proportional to ABFD. The occurrence of a significant BFD2 variable was 792 times out of 1,000 replications when r(BF) = 0.95 and increased to 996 times out of 1,000 when r(BF) = 0.85 for BFD with type 2 measurement errors. The inclusion of a CW x BFD variable in the regression equations (P < 0.05) increased (270 to 423 times out of 1,000) as r(BF) decreased from 0.85 to 0.70 for BFD with type 2 errors. Equations developed from BFD with measurement errors resulted in biased predictions of FFLM and changes in FFLM per unit change in BFD. The level and type of measurement errors that exist in the independent variables should be evaluated.     

Metapopulation dynamics and determinants of H5N1 highly pathogenic avian influenza outbreaks in Indonesian poultry

In 2008, the Indonesian Government implemented a revised village-level Participatory Disease Surveillance and Response (PDSR) program to gain a better understanding of both the magnitude and spatial distribution of H5N1 highly pathogenic avian influenza (HPAI) outbreaks in backyard poultry. To date, there has been considerable collection of data, but limited publically available analysis. This study utilizes data collected by the PDSR program between April 2008 and September 2010 for Java, Bali and the Lampung Province of Sumatra. The analysis employs hierarchical Bayesian occurrence models to quantify spatial and temporal dynamics in backyard HPAI infection reports at the District level in 90 day time periods, and relates the probability of HPAI occurrence to PDSR-reported village HPAI infection status and human and poultry density. The probability of infection in a District was assumed to be dependent on the status of the District in the previous 90 day time period, and described by either a colonization probability (the probability of HPAI infection in a District given there had not been infection in the previous 90 day time period) or a persistence probability (the probability of HPAI infection being maintained in the District from the previous to current 90 day period). Results suggest that the number of surveillance activities in a district had little relationship to outbreak occurrence probabilities, but human and poultry densities were found to have non-linear relationships to outbreak occurrence probabilities. We found significant spatial dependency among neighboring districts, indicating that there are latent spatial processes that are not captured by the covariates available for this study, but which nonetheless impact outbreak dynamics. The results of this work may help improve understanding of the seasonal nature of H5N1 in poultry and the potential role of poultry density in enabling endemicity to occur, as well as to assist the Government of Indonesia target scarce resources to regions and time periods when outbreaks of HPAI in poultry are most likely to occur.     

Estimation of animal-level prevalence from pooled samples in animal production

Often, the prevalence of an infection in the animal-production sector is determined at the group level. The prevalence at animal level (p) gives more-precise information on the infection status of the sector. This paper shows that pooled-sample data together with mathematical models allow for estimation of p. For this, model assumptions have to be made on the variation of p between groups separated in space and/or time. Formulas were derived for four models that were based on different assumptions. Model 1 assumed that p has the same value for all groups. Models 2–4 assumed that some of the groups were not infected. In addition, model 2 assumed that p has the same value for all infected groups; model 3 assumed that for an infected group, p is equal to either p₁ or p₂; and model 4 assumed that p was Beta distributed among infected groups. The models were applied to data sets on Salmonella infection in broiler flocks, including serotype data dominated by S. Hadar and S. Paratyphi B, var. Java. Based on likelihood-ratio tests, models 3 and 4 consistently fitted significantly better to the data. The applicability of model 4 is numerically bounded, related to the shape of the Beta distribution of p. Model calculations show that flock-level prevalence of Salmonella is much higher after than before slaughter. This difference (which possibly is related to different types of samples) is much smaller at the animal level. An important result of the estimation of p is that it in turn allows for an estimation of the proportion of false-negative groups — which is important in estimating the effect of veterinary or public-health measures.     

Utilizing stochastic genetic epidemiological models to quantify the impact of selection for resistance to infectious diseases in domestic livestock

This paper demonstrates the use of stochastic genetic epidemiological models for quantifying the consequences of selecting animals for resistance to a microparasitic infectious disease. The model is relevant for many classes of infectious diseases where sporadic epidemics occur, and it is a powerful tool for investigating the costs, benefits, and risks associated with breeding for resistance to specific diseases. The model is parameterized for transmissible gastroenteritis, a viral disease affecting pigs, and selection for resistance to this disease on a structured pig farm is simulated. Two genetic models are used, both of which involve selection of sires. The first involves selection with the assumption of continuous genetic variation (the continuous selection model). The second involves selection with the assumption of introgression of a major recessive gene that confers resistance (the gene introgression model). In the base population, the basic reproductive ratio, R0 (i.e., the expected number of secondary cases after the introduction of a single infected animal) was 2.24, in agreement with previous studies. The probabilities of no epidemic, a minor epidemic (one that dies out without intervention), and a major epidemic were 0.55, 0.20, and 0.25, respectively. Selection for resistance, under both genetic models, resulted in a nonlinear decline in the probability of a major epidemic and a decrease in the severity of the epidemic, should it occur, until R0 was less than 1.0, at which point the probability of a major epidemic was zero. For minor epidemics, the probability and severity of the epidemic increased until R0 reached 1.0, at which point the probabilities also fell to zero. The epidemic probabilities were critically dependent on the location on the farm where infected animals were situated, and the relative risks of different groups of animals changed with selection. The main difference between the two genetic models was in the time scale; the introgression results simply depended on how quickly the resistance allele could be introgressed into the population. For the introgression model, the probability of a major epidemic declined to zero when 0.6 of the animals were homozygous for the resistance allele.     

A simulated surveillance program for bovine paratuberculosis in dairy herds in Norway

Monte Carlo simulation models were used to evaluate the feasibility and potential results of a proposed national survey of the prevalence of bovine paratuberculosis (PTB) in dairy herds in Norway. The expected herd prevalence was assumed to be 0.2% in the simulations. The low sensitivity of the ELISA test, the assumed low herd prevalence, the typical low within-herd prevalence of PTB and the small herd sizes all present problems in detection of the disease. Simulations with 500, 1000, 2500 and 6000 herds tested were done. Our results suggest that a national survey would not be feasible at present, due to the low probability of detecting infected herds and because of the high number of false-positive reactions that would be expected to occur.     

Accuracy of genome-wide evaluation for disease resistance in aquaculture breeding programs

Current aquaculture breeding programs aimed at improving resistance to diseases are based on challenge tests, where performance is recorded on sibs of candidates to selection, and on selection between families. Genome-wide evaluation (GWE) of breeding values offers new opportunities for using variation within families when dealing with such traits. However, up-to-date studies on GWE in aquaculture programs have only considered continuous traits. The objectives of this study were to extend GWE methodology, in particular the Bayes B method, to analyze dichotomous traits such as resistance to disease, and to quantify, through computer simulation, the accuracy of GWE for disease resistance in aquaculture sib-based programs, using the methodology developed. Two heritabilities (0.1 and 0.3) and 2 disease prevalences (0.1 and 0.5) were assumed in the simulations. We followed the threshold liability model, which assumes that there is an underlying variable (liability) with a continuous distribution and assumed a BayesB model for the liabilities. It was shown that the threshold liability model used fits very well with the BayesB model of GWE. The advantage of using the threshold model was clear when dealing with disease resistance dichotomous phenotypes, particularly under the conditions where linear models are less appropriate (low heritability and disease prevalence). In the testing set (where individuals are genotyped but not measured), the increase in accuracy for the simulated schemes when using the threshold model ranged from 4 (for heritability equal to 0.3 and prevalence equal to 0.5) to 16% (for heritability and prevalence equal to 0.1) when compared with the linear model.     

Pond-level risk factors associated with columnaris disease on Mississippi commercial catfish farms

A large commercial catfish enterprise encompassing over 500 food fish ponds from five farms covering multiple counties in the Mississippi Delta was included in this analysis of columnaris risk factors. A gram-negative bacterium, Flavobacterium columnare, is the cause of columnaris disease and is considered the second-most prevalent bacterial disease in farm-raised catfish. The objective of this study was to determine if pond-level risk factors reported by farm personnel were associated with columnaris disease mortalities. To identify risk factors affecting susceptibility of farm-raised channel catfish Ictalurus punctatus to columnaris disease, a Catfish Management database was developed. Logistic regression was used to model the relationships between probability of columnaris in ponds and risk factors examined. Generalized linear mixed models incorporating hierarchically structured random effects of ponds and one or more fixed-effects risk factors were fitted. In the screening process, each risk factor was evaluated in the basic model as a single fixed-effects factor, and if associated with the outcome (P ≤ 0.20), was retained for development of multivariable models. Two multivariable logistic regression models were constructed from data collected at the pond level by producers. The first was constructed from data in which water quality was not considered. Pond depth and reduced feed consumption for a 14-d period prior to disease outbreaks measured on a per hectare basis were significantly (P ≤ 0.05) associated with columnaris disease. The second, in which water quality variables were also considered, pond depth, reduced feed consumption, shorter intervals from stocking to disease outbreaks, and total ammonia nitrogen were significantly (P ≤ 0.05) associated with columnaris occurrence. This study showed some commonly recorded production variables were associated with columnaris disease outbreaks and, if monitored, could help identify "at risk" ponds before disease outbreaks occur.     

Variance components and breeding values for growth traits from different statistical models.

Estimates of genetic parameters resulting from various analytical models for birth weight (BWT, n = 4,155), 205-d weight (WWT, n = 3,884), and 365-d weight (YWT, n = 3,476) were compared. Data consisted of records for Line 1 Hereford cattle selected for postweaning growth from 1934 to 1989 at ARS-USDA, Miles City, MT. Twelve models were compared. Model 1 included fixed effects of year, sex, age of dam; covariates for birth day and inbreeding coefficients of animal and of dam; and random animal genetic and residual effects. Model 2 was the same as Model 1 but ignored inbreeding coefficients. Model 3 was the same as Model 1 and included random maternal genetic effects with covariance between direct and maternal genetic effects, and maternal permanent environmental effects. Model 4 was the same as Model 3 but ignored inbreeding. Model 5 was the same as Model 1 but with a random sire effect instead of animal genetic effect. Model 6 was the same as Model 5 but ignored inbreeding. Model 7 was a sire model that considered relationships among males. Model 8 was a sire model, assuming sires to be unrelated, but with dam effects as uncorrelated random effects to account for maternal effects. Model 9 was a sire and dam model but with relationships to account for direct and maternal genetic effects; dams also were included as uncorrelated random effects to account for maternal permanent environmental effects. Model 10 was a sire model with maternal grandsire and dam effects all as uncorrelated random effects. Model 11 was a sire and maternal grandsire model, with dams as uncorrelated random effects but with sires and maternal grandsires assumed to be related using male relationships. Model 12 was the same as Model 11 but with all pedigree relationships from the full animal model for sires and maternal grandsires. Rankings on predictions of breeding values were the same regardless of whether inbreeding coefficients for animal and dam were included in the models. Heritability estimates were similar regardless of whether inbreeding effects were in the model. Models 3 and 9 best fit the data for estimation of variances and covariances for direct, maternal genetic, and permanent environmental effects. Other models resulted in changes in ranking for predicted breeding values and for estimates of direct and maternal heritability. Heritability estimates of direct effects were smallest with sire and sire-maternal grandsire models.     

Prognosis in equine colic patients using multivariable analysis

Multiple logistic regression was used to investigate prognosis in 308 horses referred to the University of Minnesota veterinary teaching hospital with colic. Bivariate results identified the following significant individual parameters: absent or hypomotile abdominal sounds, medical or surgical classification, peritoneal fluid total protein, anion gap, serum glucose, capillary refill time, blood pH, heart rate, packed cell volume, base excess, serum chloride, plasma bicarbonate, serum urinary nitrogen and age. Two multivariable prognostic models were developed using logistic regression. Model I (based on 257 cases with a mortality rate of 39%) included age, sex, medical or surgical classification, capillary refill time, packed cell volume and heart rate. Model II (based on 138 cases with a mortality rate of 48%) included age, sex, medical or surgical classification, capillary refill time, serum bicarbonate, serum chloride and respiratory rate. Predictive performance of the models was evaluated by treating the calculated probability of death for each horse as a continuous test result. The influence of varying the probability cutoff point for death on test characteristics (sensitivity, specificity and positive and negative predictive values) was determined. These models have not been validated and thus their performance in a different population is uncertain.     

Evaluation of biosecurity in broiler breeders

The economic impact of management practices designed to limit the introduction of disease into a parent broiler breeder flock (biosecurity) was evaluated using benefit-cost analysis. Equations were developed to quantify the losses resulting from infection with one of four alternative categories of disease representing incremental levels of pathogenicity. Realistic costs and assumed values relating to the probability of infection were used to evaluate the ameliorative effect of three alternative levels of biosecurity. A microcomputer spreadsheet program was used to confirm that expenditure on protective measures can be justified by both the risk of introducing a disease and the magnitude of losses that may occur following infection.     

Estimation of historical effective population size using linkage disequilibria with marker data

Theory hypothesizes that the rate of decline in linkage disequilibrium (LD) as a function of distance between markers, measured by r(2), can be used to estimate effective population size (N(e)) and how it varies over time. The development of high-density genotyping makes feasible the application of this theory and has provided an impetus to improve predictions. This study considers the impact of several developments on the estimation of N(e) using both simulated and equine high-density single-nucleotide polymorphism data, when N(e) is assumed to be constant a priori and when it is not. In all models, estimates of N(e) were highly sensitive to thresholds imposed upon minor allele frequency (MAF) and to a priori assumptions on the expected r(2) for adjacent markers. Where constant N(e) was assumed a priori, then estimates with the lowest mean square error were obtained with MAF thresholds between 0.05 and 0.10, adjustment of r(2) for finite sample size, estimation of a [the limit for r(2) as recombination frequency (c) approaches 0] and relating N(e) to c (1 - c/2). The findings for predicting N(e) from models allowing variable N(e) were much less clear, apart from the desirability of correcting for finite sample size, and the lack of consistency in estimating recent N(e) (<7 generations) where estimates use data with large c. The theoretical conflicts over how estimation should proceed and uncertainty over where predictions might be expected to fit well suggest that the estimation of N(e) when it varies be carried out with extreme caution.     

Maternal effects on traits measured during postweaning performance test of swine from four breeds

The objective of this study was to investigate the importance of maternal genetic effects on postweaning performance traits of Yorkshire, Landrace, Duroc, and Hampshire breeds of swine. Data consisted of performance test records collected in a commercial swine operation from 1992 to 1999. Boars from 60% of the litters were culled at weaning based on a combination of maternal and performance indexes that differed by breed. Remaining boars and all females were grown to 100 d of age. At this time all pigs were weighed (WT100) and selected for testing using recalculated breed-specific indexes (n = 15,594, 55,497, 12,267, and 9,782 for Landrace, Yorkshire, Duroc, and Hampshire, respectively). All pigs were weighed at the end of the 77-d test, and backfat (BF) and loin eye area (LEA) were measured over the 12th rib by ultrasound. Average daily feed intake was calculated for boars, and ADG was calculated for all animals. Genetic parameters were estimated for each breed and trait using multiple-trait DFREML procedures. Fixed effects were contemporary groups and either initial or final test age as a covariate. Four models were examined. Model 1 included only the additive genetic effect of the animal. Model 2 added the common litter environmental effect; Model 3 added the maternal genetic value assumed to be uncorrelated with additive genetic effects. Model 4 was the same as Model 3 with additive and maternal genetic effects assumed to be correlated. All models were two-trait models with WT100 as the second trait. Ratios of likelihoods were used to compare models. Maternal effects were important (P < 0.05) for WT100, ADG, ADFI, LEA, and BF in Landrace; for WT100, ADG, LEA, and BF in Yorkshire; for WT100 and ADG in Duroc, and for WT100 in Hampshire. Estimates of heritabilities for direct additive effects using the appropriate model for ADG, ADFI, LEA, and BF were 0.28, 0.34, 0.48, and 0.63 for Landrace; 0.26, 0.31, 0.39, and 0.65 for Yorkshire; 0.14, 0.20, 0.26, and 0.35 for Duroc; and 0.17, 0.23, 0.25, and 0.31 for Hampshire, respectively. Heritability estimates for maternal genetic effects for ADG, ADFI, LEA, and BF were 0.02, 0.05, 0.06, and 0.07 for Landrace and 0.02, 0, 0.04, and 0.06 for Yorkshire, respectively. They were zero for all traits except ADG (0.03) in Duroc and for all traits in Hampshire. Maternal effects may need to be considered in genetic evaluation of performance traits in some breeds of swine.     

Pond-Level Risk Factors Associated with Columnaris Disease on Mississippi Commercial Catfish Farms

Abstract

A large commercial catfish enterprise encompassing over 500 food fish ponds from five farms covering multiple counties in the Mississippi Delta was included in this analysis of columnaris risk factors. A gram-negative bacterium, Flavobacterium columnare, is the cause of columnaris disease and is considered the second-most prevalent bacterial disease in farm-raised catfish. The objective of this study was to determine if pond-level risk factors reported by farm personnel were associated with columnaris disease mortalities. To identify risk factors affecting susceptibility of farm-raised channel catfish Ictalurus punctatus to columnaris disease, a Catfish Management database was developed. Logistic regression was used to model the relationships between probability of columnaris in ponds and risk factors examined. Generalized linear mixed models incorporating hierarchically structured random effects of ponds and one or more fixed-effects risk factors were fitted. In the screening process, each risk factor was evaluated in the basic model as a single fixed-effects factor, and if associated with the outcome (P ≤ 0.20), was retained for development of multivariable models. Two multivariable logistic regression models were constructed from data collected at the pond level by producers. The first was constructed from data in which water quality was not considered. Pond depth and reduced feed consumption for a 14-d period prior to disease outbreaks measured on a per hectare basis were significantly (P ≤ 0.05) associated with columnaris disease. The second, in which water quality variables were also considered, pond depth, reduced feed consumption, shorter intervals from stocking to disease outbreaks, and total ammonia nitrogen were significantly (P ≤ 0.05) associated with columnaris occurrence. This study showed some commonly recorded production variables were associated with columnaris disease outbreaks and, if monitored, could help identify “at risk” ponds before disease outbreaks occur.

Received September 16, 2011; accepted February 9, 2012     

Zero-inflated models for identifying disease risk factors when case detection is imperfect: Application to highly pathogenic avian influenza H5N1 in Thailand

Logistic regression models integrating disease presence/absence data are widely used to identify risk factors for a given disease. However, when data arise from imperfect surveillance systems, the interpretation of results is confusing since explanatory variables can be related either to the occurrence of the disease or to the efficiency of the surveillance system. As an alternative, we present spatial and non-spatial zero-inflated Poisson (ZIP) regressions for modelling the number of highly pathogenic avian influenza (HPAI) H5N1 outbreaks that were reported at subdistrict level in Thailand during the second epidemic wave (July 3rd 2004 to May 5th 2005). The spatial ZIP model fitted the data more effectively than its non-spatial version. This model clarified the role of the different variables: for example, results suggested that human population density was not associated with the disease occurrence but was rather associated with the number of reported outbreaks given disease occurrence. In addition, these models allowed estimating that 902 (95% CI 881–922) subdistricts suffered at least one HPAI H5N1 outbreak in Thailand although only 779 were reported to veterinary authorities, leading to a general surveillance sensitivity of 86.4% (95% CI 84.5–88.4). Finally, the outputs of the spatial ZIP model revealed the spatial distribution of the probability that a subdistrict could have been a false negative. The methodology presented here can easily be adapted to other animal health contexts.     

Gaussian covariance graph models accounting for correlated marker effects in genome‐wide prediction

Several statistical models used in genome‐wide prediction assume uncorrelated marker allele substitution effects, but it is known that these effects may be correlated. In statistics, graphical models have been identified as a useful tool for covariance estimation in high‐dimensional problems and it is an area that has recently experienced a great expansion. In Gaussian covariance graph models (GCovGM), the joint distribution of a set of random variables is assumed to be Gaussian and the pattern of zeros of the covariance matrix is encoded in terms of an undirected graph G. In this study, methods adapting the theory of GCovGM to genome‐wide prediction were developed (Bayes GCov, Bayes GCov‐KR and Bayes GCov‐H). In simulated data sets, improvements in correlation between phenotypes and predicted breeding values and accuracies of predicted breeding values were found. Our models account for correlation of marker effects and permit to accommodate general structures as opposed to models proposed in previous studies, which consider spatial correlation only. In addition, they allow incorporation of biological information in the prediction process through its use when constructing graph G, and their extension to the multi‐allelic loci case is straightforward.     

Development of a physiologically based pharmacokinetic model to predict tulathromycin distribution in goats

Physiologically based pharmacokinetic (PBPK) models, which incorporate species- and chemical-specific parameters, could be useful tools for extrapolating withdrawal times for drugs across species and doses. The objective of this research was to develop a PBPK model for goats to simulate the pharmacokinetics of tulathromycin, a macrolide antibiotic effective for treating respiratory infections. Model compartments included plasma, lung, liver, muscle, adipose tissue, kidney, and remaining poorly and richly perfused tissues. Tulathromycin was assumed to be 50% protein bound in plasma with first-order clearance. Literature values were compiled for physiological parameters, partition coefficients were estimated from tissue:plasma ratios of AUC, and the remaining model parameters were estimated by comparison against the experimental data. Three separate model structures were compared with plasma and tissue concentrations of tulathromycin in market age goats administered 2.5 mg/kg tulathromycin subcutaneously. The best simulation was achieved with a diffusion-limited PBPK model and absorption from a two-compartment injection site, which allowed for low persistent concentrations at the injection site and slower depletion in the tissues than the plasma as observed with the experimental data. The model with age-appropriate physiological parameters also predicted plasma concentrations in juvenile goats administered tulathromycin subcutaneously. The developed model and compilation of physiological parameters for goats provide initial tools that can be used as a basis for predicting withdrawal times of drugs in this minor species.