The rules of racing state that an animal may not be treated on the day of a race and must be presented at the racetrack as “drug-free”. Drug treatment needs to be stopped prior to ensure compliance with these rules. Many factors influence the drug withdrawal time including, but not limited to, the pharmacokinetics of the substance, the sensitivity of the analytical method, any applicable analytical thresholds, and the accuracy and availability of relevant information. Accurate withdrawal times will greatly benefit the racing industry in terms of animal welfare and the prevention of an unintentional breach of the rules. The purpose of this study was to determine if the application of population modelling approach using these data can aid in the accurate determination of withdrawal times.
Methods: Two drugs, flunixin and furosemide, which are registered therapeutic drugs for the treatment of horses, were studied. Flunixin is a non-steroidal anti-inflammatory drug (NSAID) and is routinely used post-race to assist in recovery and to enable continuous training. Furosemide is a potent high-ceiling diuretic and is often used to prevent exercise-induced pulmonary haemorrhage (EIPH, or “bleeding”). NONMEM 6.1.0 was used for the data analysis.
Results: Flunixin plasma and urine data was best described with a three-compartment PK model, with between-subject variability on clearance and volumes of the central and peripheral compartments, and a combined residual error model. Animal weight was a statistically significant covariate. Confidence intervals on all parameters were obtained from a bootstrap analysis. The clearance of intravenous flunixin was 37 ± 5 L/h and the volume of the central compartment was 58 ± 4 L. Furosemide plasma data was best described by a four-compartment model, however the inclusion of the urine data caused the model to over-estimate the variability in the plasma and under estimate the variability in the urines; the likely causes were flaws in sampling times and loss of urine leading to under-recording the total voids. To counter this a “loss factor” was included in the model, following which the variability in the furosemide plasma and urine data was then described appropriately. No significant covariates were found for any parameter. Confidence intervals on all parameters were obtained using a nonparametric bootstrap analysis. The clearance of furosemide was 270 ± 30 L/h and volume of the central compartment was 57 ± 11 L.
Conclusions: Simulating populations from these models enabled probabilities of breaching the rules to be calculated for a range of times post administration and for a range of analytical thresholds. Application of population modelling has been shown in two examples to accurately determine withdrawal times.