Analytical approximations of sensitivities of steady state predictions to errors in parameter estimation

The sensitivity theory is applied to derive a linear approximation to the functional dependence of some steady state quantities of therapeutic significance on pharmacokinetic parameters obtained from the biexponential response to a single drug dose. The error of a steady state prediction depends in general on two terms. The first one may be viewed as an approximate sensitivity of the prediction to the parameter errors, and this depends solely on the algebraic relation between the prediction and the parameters. The second term is the relative error in parameters, and this may be affected by experimental design and the method of data analysis. Comparisons are made with Monte Carlo simulations and "a posteriori" estimates of variance of a prediction.