Correct dosing of drugs in neonates, infants and children is hampered by a general lack of knowledge about drug disposition in this population. Suggested methods to improve our knowledge without performing conventional full-scale investigations include population pharmacokinetic studies, allometric scaling of drug disposition according to bodyweight and in silico prediction of pharmacokinetics. The last method entails scaling of pharmacokinetic parameters according to age-dependent changes in drug absorption and elimination capacity, plasma protein binding and physiological characteristics of the subjects. Maturation (or ontogeny) of the drug-metabolising part of the cytochrome P450 (CYP) enzyme system is thus an important factor in the calculations for most drugs. The aim of this commentary is to test and critically examine the proposed methods to estimate hepatic clearance (CL) as a function of age (0-20 years), with CYP3A-mediated metabolism as the case in point. Midazolam and alfentanil were used as model drugs. Allometric scaling failed to predict the CL of midazolam and alfentanil in neonates. Calculations using in vitro findings on CYP maturation gave better estimates for neonates but very divergent ones for older infants and children. This was chiefly due to very different data on CYP3A4/5 ontogeny in three published studies. In the age range where full adult CYP activity per gram of liver could be assumed, allometric scaling and in silico predictions gave similar results. These predictions were also in approximate agreement with clinical data.The findings with the two model drugs can very probably be generalised to most drugs cleared by CYP-dependent hepatic metabolism. Allometric scaling accounts for development of body size and function but not for the fact that the drug-metabolising capacity of the liver is generally low at birth. The crucial question in the prediction of CL is thus when the activity of the applicable CYP isoform(s) attains adult levels. There are still not enough data on this, particularly when different studies even on the same CYP isoform have given very divergent results. It may also be pointed out that CYP ontogeny is an area where we have at least some information. There are several other important developmental changes about which we know practically nothing. Thus, while allometric scaling is generally unreliable for prediction in neonates and infants, the alternative method of in silico prediction can at present be used only to obtain tentative initial estimates of drug CL. Neither of the methods can be used as a substitute for actual clinical studies.