Lopinavir/ritonavir population pharmacokinetics in neonates and infants

Abstract

AIMS

Because of immature hepatic metabolism, lopinavir could present specific pharmacokinetics in the first weeks of life. We aimed at determining the optimal dosing regimen in neonates and infants weighing 1 to 10.5 kg.

METHODS

Lopinavir/ritonavir (LPV/r) pharmacokinetics were studied in 96 infants using a population approach.

RESULTS

A one-compartment model described LPV/r pharmacokinetics. Normalized to a 70 kg adult using allometry, clearance (CL/F) and distribution volume (V/F) estimates were 5.87 l h⁻¹ 70 kg⁻¹ and 91.7 l 70 kg⁻¹. The relative bioavailabilty, F, increased with post-menstrual age (PMA) and reached 50% of the adult value at 39.7 weeks.

CONCLUSIONS

Size and PMA explained some CL/F and V/F variability in neonates/infants. Based upon trough concentration limitations, suggested LPV/r dosing regimens were 40 mg 12 h⁻¹, 80 mg 12 h⁻¹ and 120 mg 12 h⁻¹ in the 1–2 kg, 2–6 kg and 6–10 kg group, respectively.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• Lopinavir/ritonavir pharmacokinetics have been fully investigated in adults and children.

WHAT THIS STUDY ADDS

• Lopinavir/ritonavir population pharmacokinetics in 96 neonates and infants from birth to less than 2 years (1.16 to 10.4 kg) showed that CL/F and V/F were dependent on body weight on an allometric basis and post-menstrual age.

Introduction

The protease inhibitor lopinavir/ritonavir (LPV/r) has been shown to be effective in children. Because of the availabilty of the oral formulation, lopinavir is administered to neonates having a high risk of perinatal or postnatal mother-to-child transmission of HIV. In infants and children, studies have shown a fast elimination half-life of LPV/r as compared with adults [1–4]. LPV is a substrate of CYP3A4, immature in neonates compared with infants and children, and could be associated with a specific pharmacokinetic profile in the first weeks after birth [5]. Therefore the objective of this study was to assess LPV/r dosing recommendations in neonates and infants weighing 1 to 10 kg by developing a population pharmacokinetic model, including the investigation of weight, gestational age and postnatal age effects.

Methods

Patients from the hospitals and institutions corresponding to authors' affiliation sites received the liquid formulation of LPV/r for the prevention of mother-to-child HIV transmission. For the youngest infants or when necessary, LPV/r was administered directly in the back of the mouth using a small syringe. The antiretroviral therapy was monitored on a routine basis. For each sampling, the drug dosing regimen, the time from the last dosing, body weight, post-natal age (PNA), gestational age (GA) and combined treatments were recorded. The assays for lopinavir were performed on four different sites: site 1 Saint Vincent de Paul Paris; site 2 UK sites; site 3 Hopital Tenon Paris; site 4, Marseille, according to validated HPLC methods [4]. The quantification limit of the method was 0.05 mg l⁻¹, the inter-assay precision and bias being 7% in the calibration range of 0.1–20 mg l⁻¹.

Data were analyzed using the nonlinear mixed effect modelling software program Monolix version 31s [6] (http://wfn.software.monolix.org). Parameters were estimated by computing the maximum likelihood estimator of the parameters without any approximation of the model (no linearization) using the stochastic approximation expectation maximization (SAEM) algorithm combined with a MCMC (Markov Chain Monte Carlo) procedure. The number of MCMC chains was fixed to 15 for all estimations. A proportional model was used to describe the residual variability (ε), and the between-subject variabilities (BSV or η) were ascribed to an exponential model. Parameter shrinkage was calculated as [1–SD(η)/ω], where SD(η) and ω are the SD of individual η parameters and the population model estimate of the BSV, respectively. The Likelihood Ratio Test (LRT) including the log-likelihood, the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were used to test different hypotheses regarding the final model, covariate effect(s) on pharmacokinetic parameter(s), residual variability model (proportional vs. proportional plus additive error model), and structure of the variance-covariance matrix for the BSV parameters.

Main covariates of interest in the population were postnatal age (PNA), body weight, postmenstrual age (PMA = PNA + GA) and the formulation. Pharmacokinetic parameters, V and CL, were standardized (STD) for a mean standard bodyweight using the ‘1/4 power’ allometric models, V_i = V_STD × (BW_i/BW_STD)¹ and CL_i = CL_STD × (BW_i/BW_STD)^3/4 for the i^th patient, V_i. CL_i and BW_i are the parameters and bodyweight of the i^th individual [7].

Age-related change functions for CL or V as a function of PNA or post-menstrual age (PMA = GA + PNA) have been described in detail [7]. Other covariates were gender and combined drugs including other antiretroviral agents.

The model was validated by simulation techniques, the normalized prediction distribution errors (npde) metrics [8]. Diagnostic plots and statistics were obtained using the R program [9].

Results

Among the 96 neonates/infants (46 boys, 50 girls), seven were treated for neonatal HIV infection. The median postnatal age of these infants was 2 weeks (range 1 day–102 weeks; interquartile range 3 days–14 weeks), the median gestational age was 38 weeks (range 27.3–41 weeks), the median body weight and body surface area were 3.3 kg (range 1.16–10.4 kg, interquartile range 2.73–5 kg) and 0.22 m² (range 0.13–0.48 m², interquartile range 0.2–0.28 m²). A total of 163 LPV concentrations were available. All samples were at steady-state except one (corresponding to a 1 day newborn). Sampling was performed on more than one occasion for 34 patients. Patients (n) received LPV/r twice (n = 73), thrice (n = 19) or once a day (n = 4). The mean doses were 39 mg kg⁻¹ 24 h⁻¹ (range 11–110) and 590 mg m⁻² 24 h⁻¹ (range 106–1454). As there were only three concentrations below the limit of quantification, they were fixed at LOQ/2.

A one-compartment model described the data. Estimates for K_a and its associated BSV were inaccurate (>100% relative standard error, %rse), so K_a and η_Ka were fixed to 0.564 h⁻¹ and zero without any detrimental effect on the quality of the fit. The covariance term between ηCL/F and ηV/F was significant, the correlation coefficient was 0.84 (rse 20%). Because multiple samples on a same occasion were available for only four patients (four, four, four and two samples per patient), between-occasion variability (BOV) was also investigated. However, the BOV parameters on K_a or V or CL were not accurate and dramatically increased the BIC value.

Weight and age were the main influential covariates. The weight effects on CL/F and V/F were then normalized for a 70 kg body weight adult. This size effect was significant (LRT test P = 0.0002), and the BSVs for CL/F and V/F, η_CL/F and η_V/F, decreased from 0.54 to 0.49 and from 0.66 to 0.55, respectively. Further investigation of age on CL/F and V/F showed a similar decrease of CL/F and V/F as a function of age, PMA or PNA (not shown). Given the similarity of the decrease and the inflated CL/F₇₀ and V/F₇₀ values (relative to expected values in a 70 kg adult) at the intercept, this age effect was thought to influence the relative bioavailability, F, in these very young patients

where F_ADULT is the reference bioavailability in adults, fixed to 1, and PMA₅₀ is the PMA corresponding to F = 0.5. AIC and BIC decreased by 8 and 3 units, respectively, and η_CL/F and η_V/F decreased from 0.49 to 0.46 and from 0.55 to 0.46, respectively. The final model was then

with

where the CL, V and PMA₅₀ typical values (% rse) are 5.87 l h⁻¹ 70 kg⁻¹ (23%), 92 l 70 kg⁻¹ (25%) and 39.7 weeks (48%). The corresponding η_CL/F and η_V/F estimates (%rse) [shrinkage] were 0.46 (11%) [0.20] and 0.45 (41%) [0.34]. The proportional residual variability estimate was 0.37 (rse 18%).

Figure 1 shows the diagnostic plots for the final model. The mean, variance and distribution of the npde metrics (500 Monte Carlo simulations) were not significantly different from zero (Wilcoxon signed rank test, P = 0.92), 1 (Fisher variance test, P = 0.74) and a normal distribution (Shapiro-Wilk test, P = 0.52). The parameters were well estimated with acceptable relative standard errors. Only V/F had notable shrinkage.

Figure 1.

Diagnostic plots for the final population pharmacokinetic model: Observed LPV/r concentrations (DV) vs. model-predicted concentrations PRED in mg l⁻¹ and the corresponding normalized prediction distribution errors (npde) metrics vs. predictions or vs. time after dose, along with the normal quantile–quantile plot for npde. The dotted lines join the data from same individuals. The lines indicate the lines of unity (left, middle) and the y = 0 line (right). npde statistics, mean and variance were not significantly different from zero (Wilcoxon signed rank test, P = 0.92) and 1 (Fisher variance test, P = 0.74) and the distribution was not significantly different from a normal distribution (Shapiro-Wilk test of normality, P = 0.52)

A dosage regimen was derived from the predicted trough concentrations, C_trough. The distribution of C_trough following 20 to 160 mg 12 h⁻¹ or 40–320 mg 24 h⁻¹ was studied by 500 Monte Carlo simulations of the final population model. Figure 2 shows the probability of observing C_TROUGH values between 1 mg l⁻¹ and 8 mg l⁻¹, therapeutic range, and above 8 mg l⁻¹, toxic range, in different body weight groups.

Figure 2.

Probability of observing trough lopinavir concentrations, C_trough in the therapeutic range, between 1–8 mg l⁻¹ (solid lines), below 1 mg l⁻¹ (dashed lines) or above 8 mg l⁻¹ (toxic range, dotted lines) following administration of 20 to 160 mg 12 h⁻¹ LPV/r in paediatric patients weighing from 1 to 10 kg. Predicted trough lopinavir concentrations were obtained from 500 Monte Carlo simulations of the final model for each dosage

Discussion

In this neonate/infant cohort, weighing 1–10.5 kg, LPV/r pharmacokinetics were satisfactorily described by a one-compartment model, including size effects on pharmacokinetic parameters. The allometric body weight scaling of CL and V plus the effect of PMA on the relative bioavailability decreased some variability associated with CL. However, the physiological circumstances in very young infants and neonates probably alter the relationship between body weight and clearance and volume of distribution, explaining the moderate decrease of clearance BSV. The variabilities, BSV and residual, remained at high levels, as previously observed in other childhood studies [1, 2, 4]. In adults, the residual variability, approximated to 0.21 [from the additive 1.15 mg l⁻¹ and proportional 0.075 components reported, 0.21 = square root ((1.15/6)² + 0.075²] was lower and BSVs were estimated via a full variance-covariance matrix, explaining in part a lower BSV(CL) value, 0.17 [3]. The high correlation between CL/F and V/F BSVs (r = 0.86) also supports the importance of bioavailability to explain these high variabilities. Drug intake, with/without concomitant formula milk in infants, as well as the amount of simultaneous milk intake may widely vary. The model predicts that F will be in the order (80%) of the adult value at 160 weeks PMA or 2.3 years PNA. Although the PMA effect on F decreased marginally the BSVs and significantly the AIC and BIC criteria, the final CL/F₇₀ and V/F₇₀ values extrapolated to a 70 kg adult with F = 1 are similar to the adult estimates, 5.73 l h^–1 and 61.6 l, respectively [3], whicht supports this population pharmacokinetic model. Higher apparent CL and V values in the youngest infants of a group have also been observed in other LPV/r childhood studies [1, 2].

Given the available LPV/r formulations, the highest probabilities of C_trough in the normal 1–8 mg l⁻¹ range were obtained with 40 mg 12 h⁻¹ in the 1–2 kg group, 80 mg 12 h⁻¹ in the 2–6 kg group and 120 mg 12 h⁻¹ in the 6–10 kg group (note that in Kaletra*, LPV/r brand name, 40 mg 12 h⁻¹ of lopinavir correspond to 10 mg 12 h⁻¹ of ritonavir). The once a day regimens produced high probabilities, >40%, of observing C_min <1 mg l⁻¹.

Competing Interests

GP has received travel grants, consultancy fees, honoraria or study grants from various pharmaceutical companies, including Abbott, Boehringer-Ingelheim, Bristol-Myers-Squibb, Gilead Sciences, VuV Healthcare, Janssen, Merck and Roche. The hospital institution of SU, CG, DH and J-MT, Cochin Hospital, received a grant of 5000 euros from Abbott for research funding. There are no other competing interests to declare.