The withdrawal time estimation of veterinary drugs revisited.

The aim of this article is to review current issues concerning the statistical determination of a withdrawal time for a veterinary drug. Special attention has been paid to the Food and Drug Administration (FDA) approach which consists of using a regression method to estimate a 99th percentile of the population with a 95% confidence level. The different assumptions of the simple regression techniques are reviewed. It is shown, both on theoretical grounds and using Monte Carlo simulations, that the main assumptions for the correct use of linear regression do not hold when a tissue depletion curve is under investigation. Finally, it is suggested that methods other than the regression technique must be evaluated, especially a revised version of the so-called simpler approach, which consists of using a non-parametric strategy to estimate a withdrawal time.