The nelfinavir-M8 pharmacokinetics was satisfactorily described by the proposed compartmental model. The present study showed a great consistency in the final nelfinavir-M8 population model derived from sequential analyses of nelfinavir and M8, confirming the robustness of the process. The basic one-compartment model was already used in adults for nelfinavir (

21). The pharmacokinetics of the metabolite produced from the parent compound should be described by a two-exponential equation, but the sparse data set (a median of only two M8 concentration-time samples per child) did not allow the identification of two exponential components. So only an integrated modeling of parent-metabolite pharmacokinetics could provide a reliable estimate of M8 elimination, since the information for the fast exponential decay is provided by nelfinavir data. Indeed, in this approach, data on the metabolite may add information to the observations on the parent and vice versa.

The following observations support the use of the proposed pharmacokinetic model. (i) Nelfinavir mean plasma clearance (CL

_{T}/

*F* = 0.92 ± 0.04 liter/h/kg) was consistent with previously reported values: 1.0 to 1.3 liters/h/kg in 18 children (2.1 to 10.8 years) (

25) and 1.57 liters/h/kg in 26 children (0.6 to 16 years) (

8). (ii) Nelfinavir apparent plasma clearance and volume (CL

_{T}/

*F*,

*V*/

*F*) decreased with age, being much higher in children younger than 2 years, in agreement with previous studies. Bergshoeff et al. (

5) showed that the clearance in children aged <2 years was 1.5 times higher than in older children (2 to 18 years). Very high clearance in infants was reported by Litalien et al. (4.2 liters/h/kg for children from 2.3 to 8.5 months) (

16), Capparelli et al. (2.7 liters/h/kg in infants between 15 days and 2 years) (

7), Payen et al. (2.13 liters/h/kg for children younger than 2 years) (

20), and Mirochnick et al. (2.1 liters/h/kg at weeks 1 and 6 of life) (

17). (iii) Finally, the goodness of fit, depicted in Fig. , was also a factor.

The residual error, 1.65 μg/ml, was probably overestimated because it included some part of interoccasion variability that could not be estimated here, since only one sample was available at each occasion. If there was a significant interoccasion variability in the nelfinavir pharmacokinetics, it could result in an underestimation of the ISVs, including CL_{T} intersubject variability.

A major aim of population pharmacokinetics is to determine which measurable pathophysiological factor can cause changes in the dose-concentration relationship and to estimate the degree to which they do so, such that an appropriate dose adjustment can be made. This is particularly relevant for drugs that exhibit an appreciable degree of intersubject variability, such as nelfinavir, in children.

In the present study, age and the NNI drugs influenced the nelfinavir-M8 pharmacokinetics.

As shown, CL_{T} and *V* normalized to BW decreased as an inverse function of age. CL_{T} and *V* were apparent parameters (CL_{T}/*F* and *V*/*F*). In our model, the same age effect was added to increase both *V* and CL_{T} in the younger children, as we supposed that the age effect was due to a decrease in bioavailability (*F*).

A number of factors may explain the decrease of the bioavailability in infants. For instance, a diet which differs in content and calories from that of older children may play a role, as the influence of diet on the bioavailability of nelfinavir is well established (

16). Moreover, newborns have an alkaline gastric pH (pH 6 to 8) and gastric acid production increases over the next 24 to 48 h before declining and remains relatively low in the first months of life. This high gastric pH in the newborn and young infant may reduce the bioavailability of weakly acidic compounds such as nelfinavir. Also, a smaller absorption area and binding of nelfinavir to a baby's inner side bottle may also be suggested. In these young children, an increase in the metabolism did not seem relevant. There is no argument in favor of an overexpression of the P glycoprotein in infants. Moreover, CYP2C19, which metabolizes nelfinavir to M8 (

16), has a low activity during the first year of life (30% of the adult activity) (

23).

The plasma M8 concentrations were 1.9-fold lower in patients treated with efavirenz or nevirapine, consistent with an induction of CYP3A4 by these drugs, with M8 being metabolized via CYP3A4 (

2). Furthermore, very high plasma M8 concentrations were observed in all samples (

*n* = 11) from three children who received ritonavir, a known CYP3A4 inhibitor (

14), but these data were too scanty to reach statistical significance.

It was previously shown that the antiretroviral response was improved in children with a minimum plasma concentration above 0.8 mg/liter (

6). Using a Bayesian approach, we showed that this target concentration was reached more often with an administration every 8 h than an administration every 12 h (Fig. to ). This is in agreement with a previous study which showed that a significantly higher percentage of children in the twice-daily group had subinhibitory minimum plasma concentrations of nelfinavir than the thrice-daily regimen (

9). Nelfinavir has a short half-life (5.5 h), which explains why an administration every 8 h maintained a higher trough concentration than an administration every 12 h. This difference in minimum plasma concentration between administrations every 8 or 12 h is more important in the youngest children. The youngest children, who have a smaller absorption area, may have a better bioavailability with a small dose administered thrice daily than with a higher dose administered twice daily. Therefore, the thrice-daily regimen should be preferred to the twice daily regimen, especially in this group. FDA-recommended doses for nelfinavir were then simulated as doses given with uniform intervals (every 8 h TID and every 12 h BID); however, nelfinavir is taken during a meal to increase bioavailability and children do not eat exactly every 8 or 12 h.

For children from 2 to 13 years, we showed that the new current FDA recommendations, 25 to 35 mg/kg TID or 50 to 60 mg/kg BID, were optimal. We confirmed also that the proposed nelfinavir doses for children younger than 2 years, 40 to 50 mg/kg TID or 60 to 75 mg/kg BID, are optimal for children from 2 months to 2 years. However, more children had a trough concentration above 0.8 mg/liter with the TID than with the BID recommended regimen. For children younger than 2 months, a 40-mg/kg dose of nelfinavir administered twice daily is inadequate. We recommend a nelfinavir dose between 50 and 60 mg/kg administered thrice daily. Mirochnick et al. also proposed further investigations of larger doses, such as 75 mg/kg twice a day for infants younger than 6 weeks (

17). As stated above, the predicted trough concentrations that served to determine this drug dosage recommendation were obtained from a Bayesian approach, and because the residual variability was rather high, these predictions are likely to be close to the mean population trough values. However, the main consequence, i.e., the need to increase dosage in the youngest children, is also supported by direct examination of their observed concentrations at various times after administration that were mainly lower than 0.8 μg/ml. These data confirm the FDA dosage recommendations for children older than 2 months. However, in younger children, it is suggested to increase the dosage and to give it thrice daily. Nevertheless, the results of this population pharmacokinetic analysis should be confirmed by a prospective analysis.