This article was originally published in a journal published by Elsevier, and the attached copy is provided by Elsevier for the author’s benefit and for the benefit of the author’s institution, for non-commercial research and educational use including without limitation use in instruction at your institution, sending it to specific colleagues that you know, and providing a copy to your institution’s administrator. All other uses, reproduction and distribution, including without limitation commercial reprints, selling or licensing copies or access, or posting on open internet sites, your personal or institution’s website or repository, are prohibited. For exceptions, permission may be sought for such use through Elsevier’s permissions site at: http://www.elsevier.com/locate/permissionusematerial
Veterinary Parasitology 142 (2006) 159–162 www.elsevier.com/locate/vetpar
Short communication
py
Bayesian estimation of Tritrichomonas foetus diagnostic test sensitivity and specificity in range beef bulls
co
Andre´s Perez a,b,*, Eduardo Cobo c, Alfredo Martı´nez d, Carlos Campero e, Ernesto Spa¨th e a
al
Consejo Nacional de Investigaciones Cientı´ficas y Te´cnicas (CONICET), Instituto Nacional Tecnologı´a Agropecuaria (INTA), Balcarce, Argentina b Center for Animal Diseases Modeling and Surveillance, School of Veterinary Medicine, University of California, Davis, USA c Population Health and Reproduction, School of Veterinary Medicine, University of California, Davis, USA d Laboratorio Azul, Azul, Buenos Aires, Argentina e Instituto Nacional Tecnologı´a Agropecuaria (INTA), Balcarce, Argentina
on
Received 31 March 2006; received in revised form 14 June 2006; accepted 23 June 2006
rs
Abstract
pe
Accuracy of culture for diagnosis of Tritrichomonas foetus was investigated in 2832 naturally exposed range beef bulls from 124 herds. Preputial fluid samples were inoculated into the culture medium, incubated at 37 8C, and daily examined. Diagnostic test was evaluated using Bayesian techniques to estimate sensitivity and specificity without a gold standard. Median posterior test sensitivity was 72.04% (95% probability interval: 58.07–86.38%) and specificity was 95.37% (95% probability interval: 94.07–96.65%). Low diagnostic test accuracy may have resulted from host and/or diagnostic test procedure related factors. Under natural range conditions, more accurate methods for T. foetus diagnostic and repeated preputial samplings of bulls may be necessary on trichomonosis control programs. # 2006 Elsevier B.V. All rights reserved.
r's
Keywords: Tritrichomonas foetus; Diagnostic tests; Culture; Bayesian
Au
th o
Bovine trichomonosis is a venereal disease of cattle due to Tritrichomonas foetus infection. T. foetus causes early embryonic death and transient infertility in cows but can asymptomatically inhabit the bull’s epithelial crypts for years (Parsonson et al., 1974, 1976). Bovine trichomonosis is an economically important reproductive disease in rangeland grazing cattle endemically * Corresponding author at: Sanidad Animal, INTA, CC 276, 7620 Balcarce, Buenos Aires, Argentina. Tel.: +54 2266 439100/439120; fax: +54 2266 439101. E-mail addresses: amperez@balcarce.inta.gov.ar, amperez@ucdavis.edu (A. Perez). 0304-4017/$ – see front matter # 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.vetpar.2006.06.021
recognized in several beef cattle regions (BonDurant et al., 1990; Cobo and Campero, 2002; Kvasnicka et al., 1989). The elective diagnosis of bovine trichomonosis involves the culture of genital secretions (preputial smegma) of bulls into selected media such as Diamond (Parker et al., 2003a), Plastridge (Martinez et al., 1985), In Pouch TF1 (Biomed Diagnostics, San Jose, CA, USA) (Borchardt et al., 1992), or liver infusion broth (Campero et al., 1986). Diamond’s, liver infusion broth, and In Pouch TF1 seemed to be suitable for the growth of T. foetus isolates from distinct geographic regions (Lun et al., 2000). Although diagnostic sensitivity of In Pouch TF1 and Diamond’s for detecting T. foetus in
A. Perez et al. / Veterinary Parasitology 142 (2006) 159–162
al
co
py
and daily examined for a week for motile trichomonads by placing a drop of each sample on a glass slide and observing at 100 magnification with a light microscopy (Martinez et al., 1985; Parker et al., 2003c). Sensitivity and specificity of T. foetus culture were estimated using Bayesian procedures to evaluate test accuracy in one single tested population (Branscum et al., 2004, 2005). Let y and p denote the number of culture positives out of n = 2832 sampled bulls and the trichomonosis prevalence, respectively, in the bulls population of this region of Argentina during the study time period; let Se and Sp denote the test sensitivity and specificity. The Bayesian model is yjSe, Sp, p binomial(n, pSe + (1 p)(1 Sp)) with independent beta prior distributions used for Se, Sp, and p. Informative priors, that are required for some of these parameters because of the model lack of identifiability (Branscum et al., 2005), were derived considering the uncertainty on the test sensitivity and specificity. Previous studies published in the peer-reviewed literature estimated that test sensitivity varies from 65% to 94.8%, depending on the culture method (Parker et al., 2003a,b; Schonmann et al., 1994; Skirrow et al., 1985). We assumed that a priori test sensitivity was beta distributed, with a most likely value of 88%, and a 95% certainty that it was higher than 66%. Prior assumption on the culture specificity was based on a study where, from 50 T. foetus-free bulls intrapreputially inoculated with Tetratrichomonas or Campylobacter fetus veneralis, one bull yielded positive results in trichomonadspecific In Pouch TF1 media (Cobo and BonDurant, unpublished data). This prior information on the test specificity was modeled as beta distributed, with a most likely test value of 98%, and a 95% certainty that it was higher than 94%. Trichomonosis prevalence in the Salado River Basin and in Buenos Aires Province were recently estimated in 42% and 28%, respectively (Perez et al., 2005). Because we expect that the prevalence in the sampled herds will be higher than the province average and similar to the prevalence observed in the region under study, we assumed a beta prior distribution for trichomonosis prevalence with most likely value of 42% and a 95% certainty that it was higher than 28%. Prior sensitivity, specificity, and prevalence distributions were consequently estimated using BetaBuster (http://www.epi.ucdavis.edu/diagnostictests/) as beta (13.09, 2.63), (102.4, 3.07), and beta (12.38, 16.72), respectively. The Bayesian model was run using the WinBUGS software (Spiegelhalter et al., 1996). A sensitivity analysis was conducted assuming, alternatively, uniform (0, 1) prior distributions for the sensitivity, specificity, and prevalence. The assumption
Au
th o
r's
pe
rs
bulls diverged among studies (88–98.4%) (Schonmann et al., 1994), 81.6–93.2% (Parker et al., 2003a), differences between the tests sensitivities were not significant (Bryan et al., 1999). In all cases, the overall sensitivity of culture for diagnosis of T. foetus in bulls is limited and false negative results may be obtained if animals are tested only once (Parker et al., 2003c). On the other hand, the specificity of culture for diagnosis of T. foetus in bulls has always been assumed to be close to 100% (Parker et al., 1999, 2003a). However, T. foetus is morphologically similar under low power microscopic observation to other Trichomona genus members, generically known as non-T. foetus trichomonads. Non-T. foetus trichomonads were isolated from preputial secretions of apparently virgin bulls from widely dispersed geographical areas (BonDurant et al., 1999; Campero et al., 2003). For these reasons, sensitivity and specificity of culture for diagnosis of T. foetus in bulls under field conditions is uncertain and this uncertainty most likely affects the accuracy of T. foetus control programs. Bayesian methods, which have recently been applied to estimate accuracy of diagnostic tests in the absence of a gold standard, offer some noteworthy advantages over traditional evaluation methods (Branscum et al., 2005). Briefly, Bayesian approach combines prior information about the sensitivity and specificity of the tests, usually obtained from the literature or expert opinion, with the sampled data, so that uncertainty in the values of sensitivity and specificity can be accommodated into the modeling. The result of Bayesian analysis is an estimate of the median and the 95% probability intervals of the sensitivity and specificity posterior distributions, which can be used as a proxy for test accuracy. A comprehensive introduction to Bayesian analysis with particular emphasis in inference of tests sensitivity and specificity is available elsewhere (Branscum et al., 2004, 2005; Suess et al., 2002). The aim of this study was to estimate sensitivity and specificity of T. foetus culture for diagnosis of naturally exposed bulls in the absence of a gold standard test. A total number of 2832 mature (>4 years old) bulls from 124 beef herds located in the Salado River basin, Province of Buenos Aires, Argentina, suspected to be T. foetus infected based on herd and laboratory historical results were genitally sampled by 72 veterinary practitioners between April and October 2003. In each bull, preputial fluid samples were taken by aspiration or scraping of the preputial cavity, inoculated directly into a modified Plastridge medium (Tricoazul1) (Martinez et al., 1985), and submitted to the diagnostic laboratory within 24 h. Cultured samples were incubated at 37 8C
on
160
A. Perez et al. / Veterinary Parasitology 142 (2006) 159–162
on
al
co
py
the attachment of parasites to the male genital mucosa, altering the probability of detecting animals that are actually infected (Parker et al., 1999). As diagnostic technical factors, inappropriate storage of culture media and samples handling (i.e. interval and temperature between sample collection and culture, incubation, and microscopic examination) may also decrease test sensitivity (Parker et al., 2003c). Moreover, sampling technique, including speed and expertise of the operator, vigor of scraping and side of the bull from which the sample was obtained – higher probability of getting positive samples from the right side than from the left side of the prepuce, Parker et al. (2003c) – may have also affected the test sensitivity. The estimated low test sensitivity suggests that one single negative result is not sufficient to demonstrate freedom from disease and that repeated preputial samples of bulls decrease the probability of false negative tests. However, no minimum number of negative direct microscopic and cultural examination tests is specified in the Office International des Epizooties Terrestrial Animal Health Code for countries importing breeding bulls (natural service or artificial insemination) (OIE, 2005). Limitations of the diagnostic test sensitivity should be considered when designing control and eradication programs. Our findings are consistent with recommendations of two or four consecutive negative cultures, depending on whether the bull comes from a negative or endemic herd, respectively, required to certify uninfected status (Parker et al., 1999). In the present study, the culture specificity to detect T. foetus in naturally infected bulls (95.4%) was lower than earlier estimates (Parker et al., 1999, 2003a). Failure in culture specificity may be expected since nonT. foetus trichomonads, morphologically similar and difficult to be distinguished from T. foetus under light microscopy, can occasionally grow in cultured bull genital secretions (BonDurant et al., 1999; Campero et al., 2003). Novel diagnostic methods may increase the probability of accurately detect infected and uninfected bulls. In the case of T. foetus and related trichomonad, DNA identification by PCR targeting ribosomal RNA (rRNA) gene, principally the small subunit rRNA (1571 bp) and the 5.8S rRNA (159 bp), has been described as more sensitive than culture in detecting T. foetus in cultured bulls smegma (Felleisen et al., 1998). Thus, PCR in addition to culture was recommended as a confirmatory test for T. foetus positive cultures (Campero et al., 2003). In summary, the sensitivity and specificity of culture for detecting naturally infected T. foetus bulls, estimated here as 72% (59–87%) and 95% (94–96%), respectively,
Au
th o
r's
pe
rs
of constant test accuracy was evaluated by running the model in three 20-herds randomly selected subsets. Changes in the mean posterior distribution <5% and overlapping of the probability intervals were considered evidence of model robustness. The median number of bulls sampled per herd was 14 (range 1–161). Positive cultures were identified in 113 of the 124 sampled herds (91.1%) and the number of positive bulls in infected herds ranged from 2 to 23 (median prevalence 12.5%). Bayesian median posterior estimates of test sensitivity, test specificity, and prevalence were, respectively, 72.43% (95% probability interval: 58.79–87.25%) and 95.24% (95% probability interval: 94–96.39%) (Fig. 1) and 45.94% (95% probability interval: 35.27–57.55%). The model was not sensitive to the selection of the prior distributions (parameters posterior means changed by <5% and 95% probability intervals overlapped when non-informative priors were alternatively assumed for each parameter), indicating that the results are not affected by the selection of the prior distributions. The posterior median sensitivity and specificity changed by <5% when the model was run using only data from 20 randomly selected herds. Sensitivity of culture for T. foetus diagnostic in naturally infected bulls was noticeably lower than the sensitivity estimated in previous studies. The low sensitivity estimated in our study may have been associated with factors related with the host and/or with the diagnostic test procedure. For instance, reduction in the T. foetus population in the preputial cavity during service period or after frequent samplings may reduce sensitivity (Parker et al., 1999). Moreover, age of the males and different immune capabilities may modify
Fig. 1. Prior (dashed lines) and posterior (solid) probability density function of Trichomonas foetus diagnostic test sensitivity and specificity in Argentine range beef bulls.
161
162
A. Perez et al. / Veterinary Parasitology 142 (2006) 159–162
may be lower than expected in experimental conditions and suggest a need for developing more accurate techniques for T. foetus diagnostic in field.
Au
th o
r's
co
al
on
pe
rs
BonDurant, R.H., Anderson, M.L., Blanchard, P., Hird, D., DanayeElmi, C., Palmer, C., Sischo, W.M., Suther, D., Utterback, W., Weigler, B.J., 1990. Prevalence of trichomoniasis among California beef herds. J. Am. Vet. Med. Assoc. 196, 1590–1593. BonDurant, R.H., Gajadhar, A., Campero, C.M., Johnson, E., Lun, Z.R., Nordhausen, R.W., Van Hoosear, K.A., Villanueva, M.R., Walker, R.L., 1999. Preliminary characterization of a Tritrichomonas foetus-like protozoan isolated from preputial smegma of virgin bulls. Bovine Pract. 33, 124–127. Borchardt, K.N.B., Thomas, M., Harmon, W., 1992. Evaluation of a new culture method for diagnosing Tritrichomonas foetus infection. Vet. Med. 104–112. Branscum, A.J., Gardner, I.A., Johnson, W.O., 2004. Bayesian modeling of animal- and herd-level prevalences. Prev. Vet. Med. 66, 101–112. Branscum, A.J., Gardner, I.A., Johnson, W.O., 2005. Estimation of diagnostic-test sensitivity and specificity through Bayesian modeling. Prev. Vet. Med. 68, 145–163. Bryan, L.A., Campbell, J.R., Gajadhar, A.A., 1999. Effects of temperature on the survival of Tritrichomonas foetus in transport, Diamond’s and In Pouch TF media. Vet. Rec. 144, 227–232. Campero, C., Catena, M., Medina, D., 1986. Caldo infusion higado para el cultivo de Tritrichomonas foetus. Vet. Arg. 3, 80–81. Campero, C.M., Rodriguez Dubra, C., Bolondi, A., Cacciato, C., Cobo, E., Perez, S., Odeon, A., Cipolla, A., BonDurant, R.H., 2003. Two-step (culture and PCR) diagnostic approach for differentiation of non-T. foetus trichomonads from genitalia of virgin beef bulls in Argentina. Vet. Parasitol. 112, 167–175. Cobo, E.R., Campero, C.M., 2002. Nuevos aspectos inmunolo´gicos y vacunales de la tricomoniasis bovina. Rev. Med. Vet. 83, 203–208. Felleisen, R.S., Lambelet, N., Bachmann, P., Nicolet, J., Muller, N., Gottstein, B., 1998. Detection of Tritrichomonas foetus by PCR and DNA enzyme immunoassay based on rRNA gene unit sequences. J. Clin. Microbiol. 36, 513–519. Kvasnicka, W.G., Taylor, R.E., Huang, J.C., Hanks, D., Tronstad, R.J., Bosomworht, A., Hall, M.R., 1989. Investigations of the incidence of bovine trichomoniasis in Nevada and of the efficacy of immunizing cattle with vaccines containing Tritrichomona foetus. Theriogenology 31, 963–971.
py
References
Lun, Z., Parker, S., Gajadhar, A.A., 2000. Comparison of growth rates of Tritrichomonas foetus isolates from various geographic regions using three different culture media. Vet. Parasitol. 89, 199–208. Martinez, A.H., Bardon, J.C., Noseda, R.P., Cordeviola, J., 1985. Diagnostico de trichomoniasis en toros. Propuesta de un esquema de diagnostico. Vet. Arg. 2, 966–977. OIE, 2005. Trichomonosis. In: Terrestrial Animal Health Code, OIE. Parker, S., Campbell, J., Gajadhar, A., 2003a. Comparison of the diagnostic sensitivity of a commercially available culture kit and a diagnostic culture test using Diamond’s media for diagnosing Tritrichomonas foetus in bulls. J. Vet. Diagn. Invest. 15, 460–465. Parker, S., Campbell, J., McIntosh, K., Gajadhar, A., 2003b. Diagnosis of trichomoniasis in ‘virgin’ bulls by culture and polymerase chain reaction. Can. Vet. J. 44, 732–734. Parker, S., Campbell, J., Ribble, C., Gajadhar, A., 1999. Comparison of two sampling tools for diagnosis of Tritrichomonas foetus in bulls and clinical interpretation of culture results. J. Am. Vet. Med. Assoc. 215, 231–235. Parker, S., Campbell, J., Ribble, C., Gajadhar, A., 2003c. Sample collection factors affect the sensitivity of the diagnostic test for Tritrichomonas foetus in bulls. Can. J. Vet. Res. 67, 138–141. Parsonson, I.M., Clark, B.L., Dufty, J., 1974. The pathogenesis of Tritrichomonas foetus infection in the bull. Aust. Vet. J. 50, 421–423. Parsonson, I.M., Clark, B.L., Dufty, J.H., 1976. Early pathogenesis and pathology of Tritrichomonas foetus infection in virgin heifers. J. Comp. Pathol. 86, 59–66. Perez, A.M., Martinez, A.H., Cordeviola, J.M., Fiscalini, B., Bardon, J.C., Combessies, G., Noseda, R.P., Spa¨th, E.J., 2005. Factores de riesgo y distribucio´n espacial de trichomoniasis bovina en la provincia de Buenos Aires, Argentina. In: Montevideo, Uruguay. Schonmann, M.J., BonDurant, R.H., Gardner, I.A., Van Hoosear, K., Baltzer, W., Kachulis, C., 1994. Comparison of sampling and culture methods for the diagnosis of Tritrichomonas foetus infection in bulls. Vet. Rec. 134, 620–622. Skirrow, S., BonDurant, R., Farley, J., Correa, J., 1985. Efficacy of ipronidazole against trichomoniasis in beef bulls. J. Am. Vet. Med. Assoc. 187, 405–407. Spiegelhalter, D., Thomas, A., Best, N., Gilks, W., 1996. BUGS: Bayesian Inference Using Gibbs Sampling, Version 0.50. MRC Biostatistics Unit, Cambridge. Suess, E.A., Gardner, I.A., Johnson, W.O., 2002. Hierarchical Bayesian model for prevalence inferences and determination of a country’s status for an animal pathogen. Prev. Vet. Med. 55, 155–171.