A Bayesian evaluation of three diagnostic assays for the detection of Giardia duodenalis in symptomatic and asymptomatic dogs

Abstract

A Bayesian approach was used to evaluate three commonly used diagnostic assays for the detection of Giardia duodenalis in dogs: microscopical examination (ME), a commercial immunofluorescence assay (IFA: Merifluor^®Giardia test) and a commercial immunochromatographic assay (SNAP: Idexx SNAP^®Giardia test). These assays were evaluated for use in two different settings: in a cross-sectional epidemiological survey in household dogs and in a clinical survey, both conducted in the northern part of Belgium. A total of 272 faecal samples from household dogs and 141 faecal samples from clinically affected dogs were examined using these three diagnostic assays. The Bayesian analysis indicated that all tests were highly specific (specificity above 90%), and that the IFA is more sensitive than SNAP and ME, both in an epidemiological and in a clinical setting. For all three tests, the estimated sensitivity values were higher in the clinical compared to the epidemiological survey, whereas the specificity values were comparable in both studies. The results of the present study indicate that IFA is a highly specific and sensitive technique for the detection of G. duodenalis cysts, both for use in an epidemiological or clinical survey. The SNAP is a specific and fairly sensitive technique for the diagnosis of Giardia in clinically affected dogs. Overall, the ME was found to be a specific diagnostic technique, although lacking sensitivity.

Introduction

Worldwide the intestinal protozoan parasite Giardia duodenalis (Syn. Giardia intestinalis, Giardia lamblia) is commonly identified in the faeces of dogs (Bugg et al., 1999, Barutzki and Schaper, 2003, Papini et al., 2005). Although overt clinical symptoms may be absent in infected animals, giardiosis can result in acute and chronic diarrhoea, along with retarded growth (Zajac, 1992).

Several recent studies focused on the evaluation of different assays for the diagnosis of G. duodenalis in dogs, since data on diagnostic test characteristics are essential, both in epidemiological studies for a reliable estimation of the parasite prevalence and in clinical studies for an adequate diagnosis. Next to the traditional microscopical examination (ME) which is often considered as the reference test for giardiosis, several new diagnostic procedures, such as immunofluorescence assay (IFA) (Rimhanen-Finne et al., 2007), immunochromatography (Gundlach et al., 2005, Dryden et al., 2006) or enzyme-linked immunosorbent assay (ELISA) (Decock et al., 2003, Cirak and Bauer, 2004, Gundlach et al., 2005, Rimhanen-Finne et al., 2007) have been subject to evaluation. In only one of these studies (Rimhanen-Finne et al., 2007) the clinical status of the animal was taken into account, although this might influence to a large extent the estimates of the test characteristics, since clinically affected animals tend to excrete a higher number of cysts in the faeces compared to randomly selected animals with unknown clinical status. In calves, a difference in diagnostic test estimates has previously been reported for the detection of Cryptosporidium in epidemiological and clinical samples (Geurden et al., 2008).

Test evaluation is traditionally performed by contingency table analysis using a gold standard reference test. This might lead to an overestimation of the test characteristics of the gold standard reference test and hence to an underestimation of the test under evaluation (Hopkins et al., 1993, Rimhanen-Finne et al., 2007). Therefore in the present study, a Bayesian approach was used that accounts for non-gold standard data in two different dog populations. The Bayesian methodology has proven its potential in validating diagnostic techniques, when at least three independent diagnostic test results are available (Geurden et al., 2004, Geurden et al., 2008). Next to the traditional ME and IFA, an immunochromatographic assay was included in the present Bayesian evaluation, since it provides a practical, quick and fairly inexpensive alternative to laboratory diagnosis, both in an epidemiological and a clinical setting.