Evaluation of established and new methods for the detection of Toxocara spp. in cats and dogs without a gold standard - A Bayesian latent class analysis.

Abstract

The commonly used detection method for Toxocara canis and Toxocara cati eggs in faecal samples is the sedimentation-flotation technique (SF). Based on this method, a sequential sieving protocol (SF-SSV) was developed to achieve a higher sensitivity in detecting Toxocara spp. eggs. In addition, two automated DNA extraction methods for faecal samples were applied. Faecal samples of cats (n = 125) and dogs (n = 175) were tested to detect Toxocara spp. with all four methods. Two populations were investigated; one was considered positive based on previous examinations (n = 120), and the other one represented a field population-based sample (n = 180). In general, the parasitological methods (SF, SF-SSV) had in all cases higher estimated sensitivities than the DNA detection methods. For both, cats and dogs, the SF-SSV had significantly higher sensitivities than the DNA detection methods. The estimates for specificity were above 90 % in each analysis, but lowest for the SF-SSV. By DNA detection, T. cati-specific methods tended to have higher diagnostic sensitivities than T. canis-specific methods, but not statistically significant. Traditional methods of evaluating diagnostic tests require a gold standard and a pool of appropriate reference samples. Since there is no gold standard for detecting Toxocara spp. in cats and dogs, Bayesian Latent Class Analysis (BLCA), a probabilistic approach estimating test performance, is a useful tool to estimate sensitivity and specificity based on the available data. This study highlights the challenge of obtaining suitable samples for model estimation, especially in field settings where Toxocara spp. prevalence is typically very low.