NUMERICAL ESTIMATION OF THE NONCOMPARTMENTAL PHARMACOKINETIC PARAMETERS VARIANCE AND COEFFICIENT OF VARIATION OF RESIDENCE TIMES

Noncompartmental investigation of the distribution of residence times from concentration-time data requires estimation of the second noncent ral moment (AUM2C) as well as the area under the curve (AUC) and the a rea under the moment curve (AUMC). The accuracy and precision of 12 nu merical integration methods for AUM2C were tested on simulated noisy d ata sets representing bolus, oral, and infusion concentration-time pro files. The root-mean-squared errors given by the best methods were onl y slightly larger than the corresponding errors in the estimation of A UC and AUMC. AUM2C extrapolated ''tail'' areas as estimated from a log -linear fit are biased, but the bias is minimized by application of a simple correction factor. The precision of estimates of variance of re sidence times (VRT) can be severely impaired by the variance of the ex trapolated tails. VRT is therefore not a useful parameter unless the t ail areas are small or can be shown to be estimated with little error. Estimates of the coefficient of variation of residence times (CVRT) a nd its square (CV2) are robust in the sense of being little affected b y errors in the concentration values. The accuracy of estimates of CVR T obtained by optimum numerical methods is equal to or better than tha t of AUC and mean residence time estimates, even in data sets with lar ge tail areas.