Objective: To perform a comparative quantitative evaluation of the predicti
on accuracy for human hepatic metabolic clearance of 5 different mathematic
al models: allometric scaling (multiple species and rat only), physiologica
lly based direct scaling, empirical in vitro-in vivo correlation, and super
vised artificial neural networks.
Methods: The mathematical prediction models were implemented with a publicl
y available dataset of 22 extensively metabolised compounds and compared fo
r their prediction accuracy using 3 quality indicators: prediction error su
m of squares (PRESS), r(2) and the fold-error.
Results: Approaches such as physiologically based direct scaling, empirical
in vitro-in vivo correlation and artificial neural networks. which are bas
ed on in vitro data only, yielded an average fold-error ranging from 1.64 t
o 2.03 and r(2) values greater than 0.77, as opposed to r(2) values smaller
than 0.44 when using allometric scaling combining in vivo and in vitro pre
clinical data. The percentage of successful predictions (less than 2-fold e
rror) ranged from 55% (rat allometric scaling) to between 64 and 68% with t
he other approaches.
Conclusions: On the basis of a diverse set of 22 metabolised drug molecules
, these studies showed that the most cost-effective and accurate approaches
, such as physiologically based direct scaling and empirical in vitro-in vi
vo correlation, are based on in vitro data alone. Inclusion of in vivo prec
linical data did not significantly improve prediction accuracy; the predict
ion accuracy of the allometric approaches was at the lower end of all metho
ds compared.