The wide use of TDM in optimizing the treatment of epilepsy has generated a huge body of data on the serum concentrations of antiepileptic drugs in populations of patients with different physiopathological and clinical characteristics. Such data are a potentially useful source of information for characterizing the kinetic profile of these drugs in specific populations by applying. However, the nature of the data, usually obtained at the same relative sampling times, together with the fact that they are obtained from patients included in dosage optimization programs, -whose doses are adapted to achieving concentrations within the therapeutic range may offer an important limitation when attempting to correctly establish the factors affecting the kinetic behaviour because they may lead to the mistaken belief that the dose administered is a factor governing the clearance (CL) of the drug, regardless of whether its kinetic behaviour is linear or not.

This later aspect has been observed on performing population kinetic studies using the NONMEM program on data from epileptic patients treated with carbamazepine (CBZ), phenobarbital (PB) and sodium valproate (VLP). In these studies, the effect of factors such as weight, age and polytherapy on CL has been observed. Despite this, a correct distribution of weighted residuals versus predicted concentrations and an important decrease in the objective function and in interindividual variability on CL can only be achieved when the dose relative to weight (D/kg) is introduced as a continuous covariable. This can be appreciated in the results shown in the Table, which shows the values of differences in objective function (Dobj) and percentages of reduction in CL variability between the final and basic models and the final model without D/kg.

It seems clear that the effect of the D/kg is more important in the case of CBZ and VLP, drugs with recognized reasons for a possible non-linear kinetics, than for PB, raising the problem of determining whether the effect of the dose on CL is really due to a kinetic effect and/or whether it is due to the TDM itself.

N=number of patients; n= number of data concentrations