Examples of proper reporting for evaluation (Stage 2 validation) of diagnostic tests for diseases listed by the World Organisation for Animal Health.

Reporting and design standards are key indicators of the quality of diagnostic accuracy (validation) studies but, with the exception of aquatic animal diseases and paratuberculosis in ruminants, there is limited guidance for designing these studies in animals. There is, therefore, a need for generic guidelines that are based on disease characteristics, such as mode of transmission, latent period and pathogenesis. Comprehensive, clear and transparent reporting of primary test accuracy studies for diseases listed by the World Organisation for Animal Health (OIE) has value for the end users of diagnostic tests and, ultimately, for decision-makers, who require systematic reviews and meta-analysis of multiple tests for specified diseases and testing purposes. The recent publication of reporting standards for Bayesian latent class models, to analyse test-accuracy data from naturally occurring disease events, fills an important gap as these methods are being increasingly used for OIE-listed diseases. Adherence to design and reporting standards, as well as to guidelines, helps to ensure that research funding for test validation studies is used appropriately and that the strengths and limitations of single tests or test combinations are made clear to test users. The authors provide a review of key points that are often overlooked or misinterpreted in test validation studies, as well as two concrete examples of good practice for use as a reference point for future studies.

Les normes de notification et de conception sont des indicateurs essentiels de la qualité des études de validation des tests destinées à déterminer leur exactitude diagnostique ; or, en dehors des maladies des animaux aquatiques et de la paratuberculose chez les ruminants, il n'existe guère de lignes directrices pour concevoir ce type d'études pour les tests utilisés en santé animale. À la connaissance des auteurs, il n'existe pas non plus de normes de conception applicables aux études de validation en santé humaine. Par conséquent, il conviendrait de disposer de lignes directrices génériques fondées sur les caractéristiques des maladies telles que leurs modalités de transmission, leur période de latence et leur pathogénie. Une notification complète, claire et transparente des études d'exactitude des tests primaires pour les maladies listées par l'Organisation mondiale de la santé animale (OIE) serait une aide précieuse pour les utilisateurs finaux des tests de diagnostic, mais aussi pour les responsables de l'élaboration des politiques, dont les décisions reposent sur des examens et des méta-analyses systématiques couvrant un grand nombre de tests pour certaines maladies ou pour certains usages d'un test. La publication récente des normes de notification applicables aux modèles bayésiens à classe latente pour analyser les données de performance d'un test à partir de foyers naturels de maladie comble une lacune importante dans la mesure où ces méthodes sont de plus en plus utilisées pour les maladies listées par l'OIE. L'adhésion à des normes de conception et de notification ainsi qu'à des lignes directrices en la matière permettra de garantir que les fonds alloués aux études de validation des tests sont bien utilisés et que les atouts et les limitations de certains tests individuels ou associations de tests sont clairement perçus par les utilisateurs. Les auteurs passent en revue certains points essentiels qui sont souvent ignorés ou mal interprétés lors des études de validation des tests et proposent deux exemples concrets de bonnes pratiques qui pourront servir de références pour les études à venir.

Las normas de comunicación y diseño son indicadores básicos de la calidad de los estudios encaminados a determinar la exactitud de diagnóstico (validación) pero, con la excepción de las enfermedades de los animales acuáticos y la paratuberculosis en rumiantes, hay escasas directrices que se apliquen al diseño de esos estudios en animales. Además, hasta donde saben los autores, en el ámbito de la salud humana no hay normas de diseño. De ahí la necesidad de directrices genéricas que estén basadas en las características de las enfermedades, como modo de transmisión, período de latencia o patogénesis. La comunicación exhaustiva, clara y transparente de estudios primarios sobre la exactitud de pruebas de diagnóstico de enfermedades incluidas en las listas de la Organización Mundial de Sanidad Animal (OIE) reviste utilidad no solo para los usuarios finales de la prueba, sino también, en última instancia, para los órganos decisorios, que necesitan metaanálisis y estudios sistemáticos de múltiples pruebas que se apliquen a una u otra enfermedad y sirvan para una u otra finalidad. La reciente publicación de normas de comunicación de modelos bayesianos de clases latentes para analizar los datos de exactitud de pruebas a partir de episodios de enfermedad de origen natural viene a colmar una importante laguna, en la medida en que estos métodos se aplican cada vez más al diagnóstico de enfermedades incluidas en las listas de la OIE. El cumplimiento de las normas de diseño y comunicación, y también de las directrices, ayuda a garantizar que los fondos de investigación destinados a estudios de validación de pruebas sean utilizados debidamente y que el usuario final de una prueba reciba información clara sobre los puntos fuertes y las limitaciones de una prueba o combinación de pruebas. Los autores pasan revista a los principales aspectos que se suelen pasar por alto o malinterpretar en los estudios de validación de pruebas y ofrecen dos ejemplos concretos de buenas prácticas que se pueden utilizar como referencia en futuros estudios.

Describing temporal variation in reticuloruminal pH using continuous monitoring data.

Reticuloruminal pH has been linked to subclinical disease in dairy cattle, leading to considerable interest in identifying pH observations below a given threshold. The relatively recent availability of continuously monitored data from pH boluses gives new opportunities for characterizing the normal patterns of pH over time and distinguishing these from abnormal patterns using more sensitive and specific methods than simple thresholds. We fitted a series of statistical models to continuously monitored data from 93 animals on 13 farms to characterize normal variation within and between animals. We used a subset of the data to relate deviations from the normal pattern to the productivity of 24 dairy cows from a single herd. Our findings show substantial variation in pH characteristics between animals, although animals within the same farm tended to show more consistent patterns. There was strong evidence for a predictable diurnal variation in all animals, and up to 70% of the observed variation in pH could be explained using a simple statistical model. For the 24 animals with available production information, there was also a strong association between productivity (as measured by both milk yield and dry matter intake) and deviations from the expected diurnal pattern of pH 2 d before the productivity observation. In contrast, there was no association between productivity and the occurrence of observations below a threshold pH. We conclude that statistical models can be used to account for a substantial proportion of the observed variability in pH and that future work with continuously monitored pH data should focus on deviations from a predictable pattern rather than the frequency of observations below an arbitrary pH threshold.

Coxiella burnetii (Q-Fever) Seroprevalence in Prey and Predators in the United Kingdom: Evaluation of Infection in Wild Rodents, Foxes and Domestic Cats Using a Modified ELISA.

Coxiella burnetii, the agent of Q-fever, is recognized as a worldwide zoonosis with a wide host range and potentially complex reservoir systems. Infected ruminants are the main source of infection for humans, but cats and other mammals, including wild rodents, also represent potential sources of infection. There has been a recent upsurge of reported cases in humans, domestic ruminants and wildlife in many parts of the world, and studies have indicated that wild brown rats may act as true reservoirs for C. burnetii and be implicated in outbreaks in livestock and humans. However, investigation of reservoir systems is limited by lack of validated serological tests for wildlife or other non-target species. In this study, serum samples from 796 wild rodents (180 bank voles, 309 field voles, 307 wood mice) 102 wild foxes and 26 domestic cats from three study areas in the UK were tested for the presence of antibodies to C. burnetii using a commercial indirect ELISA kit modified for use in multiple wildlife species. Test thresholds were determined for each species in the absence of species-specific reference sera using a bi-modal latent class mixture model to discriminate between positive from negative results. Based on the thresholds determined, seroprevalence in the wild rodents ranged from 15.6% to 19.1% depending on species (overall 17.3%) and was significantly higher in both foxes (41.2%) and cats (61.5%) than in rodents. This is the first report to quantify seroprevalence to C. burnetii in bank voles, field voles, wood mice, foxes and cats in the UK and provides evidence that predator species could act as indicators for the presence of C. burnetii in rodents. The study demonstrates that wildlife species could be significant reservoirs of infection for both livestock and humans, and the high seroprevalence in domestic cats highlights the potential zoonotic risk from this species.

Quantifying the sources of variability in equine faecal egg counts: implications for improving the utility of the method.

The faecal egg count (FEC) is the most widely used means of quantifying the nematode burden of horses, and is frequently used in clinical practice to inform treatment and prevention. The statistical process underlying the FEC is complex, comprising a Poisson counting error process for each sample, compounded with an underlying continuous distribution of means between samples. Being able to quantify the sources of variability contributing to this distribution of means is a necessary step towards providing estimates of statistical power for future FEC and FECRT studies, and may help to improve the usefulness of the FEC technique by identifying and minimising unwanted sources of variability. Obtaining such estimates require a hierarchical statistical model coupled with repeated FEC observations from a single animal over a short period of time. Here, we use this approach to provide the first comparative estimate of multiple sources of within-horse FEC variability. The results demonstrate that a substantial proportion of the observed variation in FEC between horses occurs as a result of variation in FEC within an animal, with the major sources being aggregation of eggs within faeces and variation in egg concentration between faecal piles. The McMaster procedure itself is associated with a comparatively small coefficient of variation, and is therefore highly repeatable when a sufficiently large number of eggs are observed to reduce the error associated with the counting process. We conclude that the variation between samples taken from the same animal is substantial, but can be reduced through the use of larger homogenised faecal samples. Estimates are provided for the coefficient of variation (cv) associated with each within animal source of variability in observed FEC, allowing the usefulness of individual FEC to be quantified, and providing a basis for future FEC and FECRT studies.

From phenotype to genotype: a Bayesian solution.

The study of biological systems commonly depends on inferring the state of a 'hidden' variable, such as an underlying genotype, from that of an 'observed' variable, such as an expressed phenotype. However, this cannot be achieved using traditional quantitative methods when more than one genetic mechanism exists for a single observable phenotype. Using a novel latent class Bayesian model, it is possible to infer the prevalence of different genetic elements in a population given a sample of phenotypes. As an exemplar, data comprising phenotypic resistance to six antimicrobials obtained from passive surveillance of Salmonella Typhimurium DT104 are analysed to infer the prevalence of individual resistance genes, as well as the prevalence of a genomic island known as SGI1 and its variants. Three competing models are fitted to the data and distinguished between using posterior predictive p-values to assess their ability to predict the observed number of unique phenotypes. The results suggest that several SGI1 variants circulate in a few fixed forms through the population from which our data were derived. The methods presented could be applied to other types of phenotypic data, and represent a useful and generic mechanism of inferring the genetic population structure of organisms.

Associations between the presence of virulence determinants and the epidemiology and ecology of zoonotic Escherichia coli.

The severity of human infection with pathogenic Escherichia coli depends on two major virulence determinants (eae and stx) that, respectively, produce intimin and Shiga toxin. In cattle, both may enhance colonization, but whether this increases fitness by enhancing cattle-to-cattle transmission in the field is unknown. In E. coli O157, the almost uniform presence of the virulence determinants in cattle isolates prevents comparative analysis. The availability to this study of extensive non-O157 E. coli data, with much greater diversity in carriage of virulence determinants, provides the opportunity to gain insight into their potential impact on transmission. Dynamic models were used to simulate expected prevalence distributions for serogroups O26 and O103. Transmission parameters were estimated by fitting model outputs to prevalence data from Scottish cattle using a Bayesian Markov chain Monte Carlo (MCMC) approach. Despite similar prevalence distributions for O26 and O103, their transmission dynamics were distinct. Serogroup O26 strains appear well adapted to the cattle host. The dynamics are characterized by a basic reproduction ratio (R(0)) of >1 (allowing sustained cattle-to-cattle transmission), a relatively low transmission rate from environmental reservoirs, and substantial association with eae on transmission. The presence of stx(2) was associated with reduced transmission. In contrast, serogroup O103 appears better adapted to the noncattle environment, characterized by an R(0) value of <1 for plausible test sensitivities, a significantly higher transmission rate from noncattle sources than serogroup O26, and an absence of fitness benefits associated with the carriage of eae. Thus, the association of eae with enhanced transmission depends on the E. coli serogroup. Our results suggest that the capacity of E. coli strains to derive fitness benefits from virulence determinants influences the prevalence in the cattle population and the ecology and epidemiology of the host organism.

Comparison of three alternative methods for analysis of equine Faecal Egg Count Reduction Test data.

The Faecal Egg Count Reduction Test (FECRT) is the most widely used method of assessing the efficacy of anthelmintics, and is the only in vivo technique currently approved for use with horses. Equine Faecal Egg Count (FEC) data are frequently characterised by a low mean, high variability, small sample size and frequent zero count observations. Accurate analysis of the data therefore depends on the use of an appropriate statistical technique. Analyses of simulated FECRT data by methods based on calculation of the empirical mean and variance, non-parametric bootstrapping, and Markov chain Monte Carlo (MCMC) are compared. The MCMC method consistently outperformed the other methods, independently of the distribution from which the data were generated. Bootstrapping produced notional 95% confidence intervals containing the true parameter as little as 40% of the time with sample sizes of less than 50. Analysis of equine FECRT data yielded inconclusive results in 53 of 63 (84%) datasets, suggesting that the routine use of prior sample size calculations should be adopted to ensure sufficient data are collected. The authors conclude that computationally intensive parametric methods such as MCMC be used for analysis of FECRT data with sample sizes of less than 50, in order to avoid erroneous inference about the true efficacy of anthelmintics in the field.

The distribution of the pathogenic nematode Nematodirus battus in lambs is zero-inflated.

Understanding the frequency distribution of parasites and parasite stages among hosts is essential for efficient experimental design and statistical analysis, and is also required for the development of sustainable methods of controlling infection. Nematodirus battus is one of the most important organisms that infect sheep but the distribution of parasites among hosts is unknown. An initial analysis indicated a high frequency of animals without N. battus and with zero egg counts, suggesting the possibility of a zero-inflated distribution. We developed a Bayesian analysis using Markov chain Monte Carlo methods to estimate the parameters of the zero-inflated negative binomial distribution. The analysis of 3000 simulated data sets indicated that this method out-performed the maximum likelihood procedure. Application of this technique to faecal egg counts from lambs in a commercial upland flock indicated that N. battus counts were indeed zero-inflated. Estimating the extent of zero-inflation is important for effective statistical analysis and for the accurate identification of genetically resistant animals.