H. Mager et G. Goller, RESAMPLING METHODS IN SPARSE SAMPLING SITUATIONS IN PRECLINICAL PHARMACOKINETIC STUDIES, Journal of pharmaceutical sciences, 87(3), 1998, pp. 372-378
Toxicokinetic studies often require destructive sampling and the deter
mination of drug concentrations in the various organs. Classically, th
e corresponding information is summarized in one mean concentration-ti
me profile, which is regarded as representative for the animal populat
ion. On the basis of a mean profile, only estimates of the secondary p
harmacokinetic parameters (for example AUC, t(1/2)) but no variability
measures may be obtained. In this paper two resampling techniques are
contrasted to Bailer's approach. The results obtained show that the r
esampling techniques can be considered a reliable alternative to Baile
r's approach for the estimation of the standard error of the AUCl(o)(t
k) in the case of normally distributed concentration data. They can be
extended to the estimation of a variety of other secondary pharmacoki
netic parameters and their respective standard deviations. One disadva
ntage with Bailer's method is its restriction to linear functions of t
he concentrations. On the other hand, using the population approach, p
rior knowledge of the underlying pharmacokinetic model is necessary. T
he resampling techniques discussed here, the ''pseudoprofile-based boo
tstrap'' (PpbB) and the ''pooled data bootstrap'' (PDB), are noncompar
tmental approaches. They are applicable under nonnormal data constella
tions and permit the estimation of the usual secondary pharmacokinetic
parameters along with their standard deviations, standard errors, and
other statistical measures. To assess the accuracy, precision, and ro
bustness of the resampling estimators, theoretical data from three dif
ferent pharmacokinetic models with different add-on errors (up to 100%
variability) were analyzed. Even for the data sets with high variabil
ity, the parameters calculated with resampling techniques differ not m
ore than 10% from the true values. Thus, in the case of data that are
not normally distributed or when additional secondary pharmacokinetic
parameters and their variability are to be estimated, the resampling m
ethods are powerful tools in the safety assessment in preclinical phar
macokinetics and in toxicokinetics where generally sparse data situati
ons are given.