Development of limited sampling strategies for characteristics of a pharmacokinetic profile.

New approaches for empirical and model-based development of constrained, limited sampling strategies for the estimation of one or more characteristics of a pharmacokinetic (PK) profile are evaluated. The methods (i) permit a specification of an overall risk function weighted by each estimation objective, (ii) require only a few sampling times for each of one or more new individuals, (iii) permit tight time constraints, such as those imposed by outpatients, and (iv) recognize variations across individuals, but determine optimal sampling times that are the same for all individuals. The methods are applied to a 4-way crossover pharmacokinetic study of two formulations of cyclosporin G, under fed and fasted conditions, in renal transplant patients. This application used average percentage absolute error for estimating AUC and Cmax with an overall risk function that first put all weight on the error for estimating AUC and second put equal weights on the errors for estimating AUC and Cmax. Empirical and model-based methods identified constrained, 3-point designs with acceptable precision for estimating either AUC alone or AUC and Cmax simultaneously. Model-based sampling times were obtained by software for minimizing a general objective function subject to linear constraints where the objective function was evaluated by computer-intensive sampling under a nonlinear mixed-effects model. The model-based approach permitted a direct comparison of the precision of limited and full sampling strategies.