INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Zidovudine (ZDV) administration to human immunodeficiency virus (HIV)-infected women during pregnancy and to their infants during the first 6 weeks of life has been demonstrated to reduce the rate of mother-to-infant HIV transmission by approximately two-thirds (8). Guidelines for the use of ZDV to reduce perinatal HIV transmission have been developed, and many neonates now receive ZDV (5). The pharmacokinetics of ZDV have been studied in full-term infants, in whom elimination is reduced compared to that in older infants and children, most likely due to reduced hepatic and renal ZDV clearance (CL) (3). While ZDV elimination pathways are likely to be even less developed in less mature infants, no data describing the pharmacokinetics of ZDV in premature infants have been reported. Traditional pharmacokinetic studies are difficult to perform with premature infants because of limitations on the amount of blood that can be drawn. Population analysis, which allows estimation of pharmacokinetic parameters and interindividual variability with only a few samples per individual, is an attractive alternative (4, 15). The aim of the present study was to use population analysis to delineate ZDV pharmacokinetics in premature infants.

(This study was presented in part to the Society for Pediatric Research, Washington, D.C., 7 May 1996.)

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

We studied 15 premature infants from eight hospitals begun on ZDV by their clinical care givers. Fourteen were treated with ZDV as part of a prophylactic regimen combining antepartum, intrapartum, and newborn therapy for the prevention of mother-to-infant HIV transmission (5). The remaining patient (patient 2; Table 1) had a positive HIV PCR assay result along with thrombocytopenia and lymphopenia and was started on ZDV for the treatment of presumed symptomatic HIV infection. Informed consent for sampling of blood for pharmacokinetic studies was obtained from the parents of two of the patients after approval of the appropriate institutional review board. Serum samples were obtained from the other patients as part of the patients' clinical care to provide therapeutic monitoring of ZDV concentrations.

All patients initially received ZDV intravenously at dosages ranging from 1.5 mg/kg of body weight every 12 h to 2.0 mg/kg every 6 h. Eight of the infants were studied once, and seven of the infants were studied on two separate occasions. Three of the repeat evaluations were performed after the infant had been switched to enteral therapy. Between one and five serum samples were obtained from each patient for ZDV assay during each study period, with an average of 2.7 samples obtained per study period. Samples from 13 of 15 subjects were used to guide therapy and were assayed within 2 weeks of collection. Samples which were not assayed within 24 h of collection were frozen at 20°C or colder until analysis. ZDV concentrations were determined in four laboratories by radioimmunoassay. Three of the laboratories used a commercial kit (ZDV-Trac; INCSTAR Corp., Stillwater, Minn.), and the fourth laboratory analyzed the samples with ZDV antiserum (Sigma Chemical Company, St. Louis, Mo.) and tritiated ZDV (Moravek Biochemicals, Brea, Calif.) by the method of Quinn et al. (13). The lower limits of detection were 1 and 10 ng/ml by the two types of assays, respectively, and the inter- and intraday coefficients of variation for all assays were less than 15%. All four laboratories were certified in the performance of these assays by the AIDS Clinical Trials Group Pharmacology Committee Quality Assurance program, which includes standardization through analysis of blinded samples.

Initial estimates of the mean values of the pharmacokinetic parameters for the population and their variances were generated with the program NONMEM (version IV, level 1.0) and its first-order subroutine. A one-compartment model with first-order absorption and elimination (ADVAN2, TRANS2) was used during model development. Our population had a large variation in size (264% by weight), representing significant differences in lean body mass. Due to the strong correlations of size with measures of maturation and pharmacokinetic parameters, we incorporated weight into CL and the volume of distribution (V) before inclusion of other maturational parameters. We then evaluated the impacts of gestational age at birth, postnatal age, postconceptional age, and changes associated with repeat sampling on the weight-adjusted parameters graphically and by making changes in the objective function. This was first performed in a univariate manner followed by a forward selection with inclusion of those covariates that resulted in at least a 3.86 reduction in the objective function (P < 0.05 based on a chi-square distribution and a loss of 1 degree of freedom). After incorporating all potential covariates, the significance of each covariate was verified by assessing the impact of its removal from the full model. Only covariates which maintained at least a reduction of 8 in the objective function (P < 0.005) were retained in the final model. Evaluation of maturation components for CL were included in the model (incorporated as both a fixed effect and a random effect). Maturational changes in CL confounded bioavailability (F) in individual infants who received enteral therapy such that F could not be modeled independently of CL and V for individual patients. Therefore, for patients for whom repeat evaluations were performed during enteral therapy, the apparent CL (CL/F) and V (V/F) are listed. Additive, proportional, and combination residual error models were evaluated, including models with separate residual error terms for individual assay performance sites. The final model was rerun with the first-order conditional estimation subroutine. It produced results (average 4.5% difference from mean parameters) nearly identical to those obtained with the first-order subroutine. Bayesian estimates of the values of the pharmacokinetic parameters for individual patients were calculated by the post hoc subroutine. The bias and precision of the post hoc evaluation were calculated as mean absolute error and mean error and were expressed as a percentage of individual predicted concentrations for individual patients.

Average steady-state ZDV concentrations in individual subjects were calculated as F · dose/CL · Tau, where tau is the dosing interval. The typical ZDV concentration profile in premature infants receiving 1.5 mg/kg every 12 h was constructed with a Monte Carlo simulation of 200 subjects by using NONMEM and the final population model, parameters, and variability.

Data are presented as means ± standard deviations. Linear regression analysis was used to correlate pharmacokinetic parameters with gestational age. Pharmacokinetic parameters for infants studied twice were compared by paired Student's t test or the Wilcoxon signed rank test.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

The study population had a mean gestational age of 29.4 ± 2.1 weeks and a mean weight of 1,230 ± 402 g at birth. Patient descriptions and pharmacokinetic data are listed in Table 1. Two of the patients (patients 5 and 6; Table 1) were twins. Initial sampling for the 15 infants occurred at a mean postnatal age of 5.5 ± 3.5 days. Sampling for pharmacokinetic analysis was repeated for seven of the infants at a mean postnatal age of 17.7 ± 6.9 days.

The impact of specific covariates on model development are listed in Table 2. Weight explained a significant proportion of the interpatient variability for both V and CL (Fig. 1 to 3). While gestational age at birth, postconceptional age, postnatal age, and different clearances at the second sampling all improved the model during the univariate analysis, due to the high degree of correlation with weight, only data from the second sample collection were retained in the final model. A proportional residual error model was selected on the basis of graphical analysis and a reduction in the objective function. Inclusion of separate residual error terms for individual assay performance sites did not improve the model. Values for typical parameters, interpatient and residual errors, and their standard errors from the final population model are as follows. Typical values of the pharmacokinetic parameters were 1.57 liters/kg for V and 0.150 liters/kg/h (2.5 ml/min/kg) for CL, and at the second sampling, the CL was 1.66 times the prior estimate of CL. Interpatient and residual errors were 47% for V and 18% for CL, and at the second sampling the error for CL was 25%. The residual error was 20.8%. For patients sampled on two occasions, the CL during the second sampling period was significantly increased. The weighted residuals of the final population model are presented in Fig. 4.

View larger version (10K): [in this window] [in a new window]   FIG. 1.   Correlation between weight and individual estimates of V from a Bayesian post hoc evaluation of the initial model.

View larger version (9K): [in this window] [in a new window]   FIG. 2.   Correlation between weight (WT) and CL from a Bayesian post hoc evaluation of the initial model. Solid triangles, initial evaluation; open triangles, subsequent evaluation.

View larger version (10K): [in this window] [in a new window]   FIG. 3.   Correlation between postconceptional age and CL from a Bayesian post hoc evaluation of the basic model with weight included. Solid triangles, initial evaluation; open triangles, subsequent evaluation.

View larger version (12K): [in this window] [in a new window]   FIG. 4.   Weighted residuals versus predicted concentration from the final model.

The description of the data produced by the Bayesian post hoc analysis was unbiased (mean error, 2.4%; 95% confidence interval [CI] of the mean, 7.0 to 1.7%) and precise (mean absolute error, 12.6%; 95% CI of the mean, 10.0 to 15.2%). Measured and individual predicted concentrations are plotted in Fig. 5. The mean values of the pharmacokinetic parameters from the initial sampling period were as follows: CL, 2.50 ± 0.49 ml/min/kg; V, 1.54 ± 0.40 liters/kg, and half-life (t_1/2), 7.1 ± 1.5 h. There were no significant correlations between gestational age at birth or postconceptional age and CL, V, or t_1/2 (Table 3).

View larger version (12K): [in this window] [in a new window]   FIG. 5.   Observed ZDV concentrations versus predicted ZDV concentrations from a Bayesian post hoc analysis of the final model.

For the subgroup of patients sampled twice, pharmacokinetic parameters from the first sampling period, at a mean postnatal age of 6.3 ± 2.1 days, were similar to those for the entire study population, with a mean CL of 2.33 ± 0.59 ml/min/kg and a mean t_1/2 of 7.3 ± 1.9 h. However, by the second sampling period, at a mean postnatal age of 17.7 ± 6.9 days, the mean CL had increased to 4.35 ± 1.80 ml/min/kg (P = 0.016 compared to the value at the initial sampling period) and the mean t_1/2 had decreased to 4.4 ± 1.5 h (P = 0.003) (Fig. 6). CL increased for patients continuing to receive ZDV intravenously (range, 36 to 129%) as well as those switched to enteral dosing (range, 46 to 160%).

View larger version (10K): [in this window] [in a new window]   FIG. 6.   Estimated ZDV CL for individual patients from the final model versus postnatal age. Dotted lines connect evaluations from individual patients studied more than once.

The simulated ZDV concentration-versus-time curve for premature infants given 1.5 mg/kg every 12 h is presented in Fig. 7. Predicted average steady-state ZDV concentrations at the first sampling (postnatal age, 5.5 days) are 8.7 ± 2.2 µM when the patients are receiving 2 mg/kg every 6 h, whereas they are 3.3 ± 0.8 µM when the patients are receiving 1.5 mg/kg every 12 h. Average steady-state ZDV concentrations when the patients are receiving these doses are predicted to be 6.0 ± 3.8 and 2.1 ± 0.8 µM, respectively, at the second sampling (postnatal age, 17.7 days).

View larger version (11K): [in this window] [in a new window]   FIG. 7.   Average (± 1 standard deviation) simulated steady-state ZDV concentrations in a typical premature infant receiving 1.5 mg/kg every 12 h.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

While the pharmacokinetics of ZDV have been well described in adults, children, and full-term newborns, no previous data describing the pharmacokinetics of ZDV in premature infants have been reported. In adults, ZDV CL averages 21.7 ml/min/kg and t_1/2 averages 1.1 h because ZDV is rapidly and extensively glucuronidated, and both the parent drug and the glucuronide are excreted in the urine (11). Boucher et al. (3) studied the pharmacokinetics of ZDV in full-term infants during the first months of life and found reduced ZDV elimination in those younger than 14 days, with CL averaging 10.9 ml/min/kg and t_1/2 averaging 3.12 h. Hepatic and renal elimination pathways are known to be even less well developed in premature infants than in full-term infants, and it is not surprising that ZDV elimination was reduced further among the premature infants in the current study (9, 10).

In full-term infants, ZDV elimination increases rapidly during the first weeks of life. ZDV CL averaged 19.0 ml/min/kg and t_1/2 averaged 1.87 h among the full-term infants older than age 14 days studied by Boucher and colleagues (3, 7), and postnatal age was the most important determinant of total body clearance. Similarly, among those premature infants sampled on two occasions in the current study, ZDV elimination increased with advancing postnatal age. While there was no correlation between gestational age at birth and ZDV pharmacokinetic parameters in the current study, the significance of this negative finding is limited by the small number of patients studied and the low variability in postnatal age at the time of study. However, gestational age must play an important role in determining ZDV elimination, because the ZDV CL in even the oldest of these premature infants was well below that seen in full-term infants during the first 2 weeks of life.

Premature infants receiving the standard dosing regimen for full-term infants, i.e., 2 mg/kg every 6 h, accumulated elevated serum ZDV concentrations. The most common toxicities of ZDV, anemia and neutropenia, appear to be dose related (12, 16). In children receiving continuous infusions of ZDV, neutropenia has been observed more commonly when steady-state concentrations exceed 3.0 µM (2). Using the values of the pharmacokinetic parameters derived from our analysis, we predict that the average steady-state concentrations in our newborns at the age of their first pharmacokinetic evaluation would be 8.7 µM with the administration of 2 mg/kg every 6 h, whereas they would be 3.3 µM with the administration of 1.5 mg/kg every 12 h. Even with the increase in CL seen at the second pharmacokinetic evaluation (at between 11 and 32 days of life), administration of 1.5 mg/kg every 12 h should produce average concentrations of 2.1 µM. The latter concentration is comparable to that seen in full-term infants during the first 2 weeks of life (1.7 µM). We recommend that initial ZDV dosing be reduced to 1.5 mg/kg every 12 h in premature infants to avoid the accumulation of potentially toxic serum ZDV concentrations. However, we evaluated only two patients older than 15 days of age, preventing us from determining if ZDV dosing needs to be increased as preterm infants mature beyond age 2 weeks. Further studies of the pharmacokinetics of ZDV in premature infants are needed to evaluate our recommended initial dosing regimen, to delineate changes in the values of the pharmacokinetic parameters for ZDV as the preterm infant develops, and to evaluate dose modifications which may be necessary to maintain therapeutic drug concentrations in these infants.

Our study is also limited by the variability in the ZDV dosing regimens administered to these patients. Three patients were receiving enteral therapy at the time of the second pharmacokinetic evaluation, and we may have overestimated systemic CL and V for these infants. However, ZDV is well absorbed in newborn full-term infants (greater than 90% of the dose), with bioavailability decreasing over the first weeks of life as hepatic glucuronidation and first-pass metabolism increase (3). Our data suggest that hepatic glucuronidation of ZDV is even more underdeveloped in preterm infants than in full-term infants. Therefore, ZDV bioavailability in premature infants is likely to be high and oral CL may approximate systemic CL.

Due to the sparse and varied nature of sample collection for pharmacokinetic analyses, we used a population approach to analyze these data. Although this data set is smaller than those typically used in population pharmacokinetic analyses, the population analysis methodology has been successfully applied to populations with fewer than 20 individuals (6). The large discrepancy in the actual values compared with the expected levels on the basis of pharmacokinetics in full-term infants led us to report this preliminary investigation with data from our current database. The fact that the predicted concentrations from the post hoc analysis fit the observed data well (mean absolute error, <15%) and generated mean parameter values and variabilities similar to those generated by the initial population model suggest that the data in the database were well described by the population analysis. However, due to the limited size and scope of the data, this analysis lacked the power to describe changes in the pharmacokinetics of ZDV with maturation and must be interpreted with caution.

The reduction of in the mother-to-infant HIV transmission through the administration of ZDV to the mother during pregnancy and labor and to the infant during the first 6 weeks of life is a major advance in the fight against HIV disease. It is hoped that further reductions in mother-to-infant transmission can be achieved through the use of combination regimens including ZDV and newer antiretroviral agents. The rate of preterm delivery may be increased in HIV-infected pregnant women, and the risk of mother-to-infant HIV transmission appears to be greater in preterm infants than in full-term infants (1, 14). In one recent study, delivery before 37 weeks of gestation occurred for 33% of HIV-infected pregnant women whose infants became infected with HIV compared with 19% of those women whose infants did not become infected (1). A detailed understanding of the developmental changes in the pharmacokinetics of antiretroviral agents will be necessary for their safe and effective use for both the prophylaxis and treatment of HIV infection in premature infants.