Variability in faecal egg counts – a statistical model to achieve reliable determination of anthelmintic resistance in livestock

Anthelmintic resistance is an increasing challenge for the control of parasites in livestock. The faecal egg count reduction test (FECRT) is the practical gold standard method for evaluating resistance, but the interpretation is complicated due to high levels of variability. A hierarchical statistical model was therefore developed for analysis of FECRT data from multiple farms. Horse age, gender, zip code and pre-treatment egg count were incorporated into the model. Horses and farms were kept as random effects. Resistance classifications were based on model-based 95% lower confidence limit (LCL) values of predicted mean efficacies, and cutoff values were justified statistically. The model was used to evaluate the efficacy of pyrantel embonate paste on 64 Danish horse farms. Of 1644 horses, 614 had egg counts > 200 eggs per gram (EPG) and were treated. The cutoff LCL values used for classifying pyrantel resistance were: >92%: no resistance, 88-92%: suspect resistance, and <88%: resistance. Using model-adjusted LCLs, we classified seven (10.9 %) farms as pyrantel resistant, five (7.8 %) as suspect resistant, and the remainder of farms (81.3 %) as not resistant. In comparison, traditional arithmetic calculations classified nine farms (14.1 %) as resistant and 11 farms (17.2 %) as suspect resistant. Using 10000 Monte Carlo simulated data sets, our methodology provides a reliable classification of farms into different resistance categories with a false discovery rate of 1.02 %. The methodology was shown to be unaffected by single outlier
horses on the farms, while traditional calculations were strongly biased. The statistical model combines information between farms to distinguish between variability and genuine reduction in efficacy and can be adapted to handle FECRT data obtained from other livestock species, drug types, and parasite species.