Abacavir pharmacokinetics in African children living with HIV: A pooled analysis describing the effects of age, malnutrition and common concomitant medications

Aims Abacavir is part of WHO‐recommended regimens to treat HIV in children under 15 years of age. In a pooled analysis across four studies, we describe abacavir population pharmacokinetics to investigate the influence of age, concomitant medications, malnutrition and formulation. Methods A total of 230 HIV‐infected African children were included, with median (range) age of 2.1 (0.1–12.8) years and weight of 9.8 (2.5–30.0) kg. The population pharmacokinetics of abacavir was described using nonlinear mixed‐effects modelling. Results Abacavir pharmacokinetics was best described by a two‐compartment model with first‐order elimination, and absorption described by transit compartments. Clearance was predicted around 54% of its mature value at birth and 90% at 10 months. The estimated typical clearance at steady state was 10.7 L/h in a child weighing 9.8 kg co‐treated with lopinavir/ritonavir, and was 12% higher in children receiving efavirenz. During coadministration of rifampicin‐based antituberculosis treatment and super‐boosted lopinavir in a 1:1 ratio with ritonavir, abacavir exposure decreased by 29.4%. Malnourished children living with HIV had higher abacavir exposure initially, but this effect waned with nutritional rehabilitation. An additional 18.4% reduction in clearance after the first abacavir dose was described, suggesting induction of clearance with time on lopinavir/ritonavir‐based therapy. Finally, absorption of the fixed dose combination tablet was 24% slower than the abacavir liquid formulation. Conclusion In this pooled analysis we found that children on lopinavir/ritonavir or efavirenz had similar abacavir exposures, while concomitant TB treatment and super‐boosted lopinavir gave significantly reduced abacavir concentrations.


Funding information
Abacavir is indicated for children over 3 months of age at 8 mg/kg twice daily or 16 mg/kg once daily. 4,5 Abacavir is extensively metabolized by the liver, with less than 2% excreted unchanged in urine. 6 The two major pathways of abacavir metabolism involve alcohol dehydrogenase (ADH) and uridine diphosphate glucuronyltransferase (UGT) enzymes, producing inactive carboxylate and glucuronide metabolites. 6 Previous studies report that coadministration with food and formulation has no effect on abacavir exposure. 7,8 However, abacavir solution has been associated with an 11% higher peak serum concentration (C max ) than the tablet formulation. 9 Its binding to plasma proteins is about 50%. 7 Many physiological systems are not functioning optimally in children with malnutrition. Total body water is increased, plasma albumin is decreased, Phase I and II metabolic reactions are considerably reduced as the severity of malnutrition increases. [10][11][12] These physiological alterations may either directly or indirectly influence the pharmacokinetics of abacavir. 13 Drug-drug interactions among antiretrovirals and antituberculosis drugs are common. When adults on abacavir are coadministered LPV/r, abacavir exposure decreases by approximately 30%. 14 In patients co-treated for TB with a rifampicin-based regimen, lopinavir would decrease by up to 90% with the standard LPV/r 4:1 dosing, so additional ritonavir is administered to achieve a 4:4 ratio with lopinavir (super-boosted lopinavir) to counteract the interaction.
We previously showed that abacavir exposure is reduced by 36% when children were co-treated with rifampicin and super-boosted lopinavir. 15 Cohort studies in children raised concern that some abacavir-containing regimens may be less effective than regimens with a different nucleoside reverse transcriptase inhibitor backbone. 16,17 Drug-drug interactions with companion NNRTIs or protease inhibitors (PIs) resulting in reduced abacavir exposures could potentially contribute to such findings.
The purpose of this pharmacokinetic meta-analysis was therefore to pool several available abacavir clinical datasets to take advantage of the increased sample size and perform a more robust analysis to investigate the consequence of differences in body size, age, concomitant TB co-medications, malnutrition and drug formulation on abacavir pharmacokinetics in children.

| Clinical studies and data
This pooled analysis used data from ARROW (Uganda and Zimbabwe), 18 CHAPAS-3 (Uganda and Zambia), 19 DNDi (South Africa) 20 and MATCH (South Africa). 21 Briefly, the objects in regards to PK of the three individual studies were: in ARROW, to compare the pharmacokinetics of once daily vs twice daily dosing of abacavir and lamivudine when given together with nevirapine or efavirenz; in CHAPAS-3, to compare abacavir, stavudine or zidovudine as dual-or triple fixed-dose combination paediatric tablets with lamivudine and What is already known about this subject • The pharmacokinetics (PK) of abacavir have been characterized previously in children; however, the effects of growth and development, drug-drug interactions and other covariates on abacavir PK remain limited.

What this study adds
• There was a decrease in abacavir exposure when coadministered with LPV/r, LPV/r plus rifampicin, efavirenz.
• Malnourished children had high and variable exposures that normalized as nutritional status resolved.
• Abacavir exposures in children on the recommended 8 mg/kg twice daily or 16  Forty-one of these children also underwent pharmacokinetic evaluation while receiving once daily doses, while three children received daily doses only. A total of 154 children were on concomitant lopinavir/ritonavir and 76 were on efavirenz. Rifampicin-containing anti-TB treatment was administered to 104 children; of these, 101 were on super-boosted lopinavir/ritonavir (4:4) and three on efavirenz. There were 115 malnourished children, characterized in this analysis as having weight-for-age and height-for-age Z-score less than À2.0. The majority of children in our analysis received abacavir with LPV/r (4:1) and therefore were the reference group in the model.

| Analytical methods
The analytical methods have previously been described in depth in the original published articles for each analysis. Plasma abacavir con-

| Population pharmacokinetic analysis
Data from each study were explored separately and added one by one starting from those with more intensive data, as suggested in Svensson et al. 22 After the inclusion of each dataset, the model fit was reassessed and modified if necessary.
The population pharmacokinetics of abacavir was described using The additive error for all samples was set to be at least 20% of the LLOQ of the assay, and the study specific LLOQ can be found in Table 2. Below the limit of quantification (BLQ) concentrations were handled with the M6 method as described by Beal. 25 Briefly, the first BLQ value after the peak (or the last in a series of BLQ values before the peak) was imputed to half the lower limit of quantification (LLOQ/2) and included in the fit with their additive error inflated by LLOQ/2, while any subsequent BLQ values (or preceding if before the peak) were excluded from the fit and only considered for visual predictive check (VPC) diagnostics.
Model building was guided by the drop in the objective function value (ΔOFV; proportional to À2 log-likelihood), inspection of goodness-of-fit plots, VPC, biological plausibility and clinical relevance. A decrease in OFV of more than 3.84 between two nested models after the addition of one parameter was considered significant at P < .05.

| Investigating factors that influence abacavir pharmacokinetics
Allometric scaling by total body weight was introduced on all clearance and volume parameters to account for the known effect of body size on pharmacokinetics with exponents fixed to 3/4 for elimination and intercompartmental clearance and 1 for volumes of distribution. 17,26,27 Total body weight (TBW) and fat-free mass (FFM) 28 were evaluated as alternative size descriptors on both disposition parameters. To account for maturation, a sigmoidal function of postmenstrual age was used (Equation 1): where PMAGE denotes postmenstrual age, PMAGE 50 is the value of PMAGE at which 50% of the maturation is complete, and γ is a parameter determining the shape of the relationship. Since no information on the actual gestational age of the children was available, it was assumed to be 9 months.
After inclusion of weight and age in the model, additional covariates were screened based on inspection of parameter vs covariate plots and physiological plausibility and retained based on statistical significance at P < .01. To describe the time-changing effect of malnutrition that resolves with days on nutritional supplementation, an exponential function was used (Equation 2): where MAL 0 is the initial value of the malnutrition effect at day 0 (before start of supplementation), λ MAL is the half-life of the process (in days) and time is the duration of the nutritional supplementation treatment (in days). The precision of the final parameter estimates was evaluated by sampling importance resampling (SIR). 29  Values reflect the numbers of children on the drugs at PK evaluation. c Weight-for-age and height-for-age z-score <2.0. patient and study characteristics and their distributions in each study are presented in Table 2.

| Population pharmacokinetics
The population pharmacokinetics of abacavir was best described by a two-compartment disposition model (difference in objective function value, ΔOFV = À728 when compared to a one-compartment model, P < 10 À6 ) with first-order elimination and transit compartments describing absorption (ΔOFV = À148, P < 10 À6 , when compared with simple first-order absorption). To adjust for differences in body size, allometric scaling of TBW was included for all disposition parameters and improved the model fit (ΔOFV = À268). Using FFM instead of TBW did not provide any significant improvements. After adjusting for body size, the effect of age on clearance was captured using a maturation function (ΔOFV = À15, P < 10 À3 ). Clearance was predicted to be at 54% of its mature value at birth and 90% at 10 months. The maturation function of abacavir clearance with confidence intervals is shown in Figure 1. The apparent clearance (CL/F) for a typical 9.8 kg child co-treated with standard LPV/r 4:1 at steady state was estimated at 10.7 L/h.
Clearance of the first abacavir dose was 18.4% lower (ΔOFV = À11.5, P < 10 À4 ) than clearance for a typical child on standard LPV/r 4:1 for over 7 days. Coadministration with anti-TB treatment plus super-boosted lopinavir decreased abacavir bioavailability by 29.4% (ΔOFV = À48, P < 10 À6 ). Decreased abacavir exposures was also seen in studies where abacavir was coadministered with protease inhibitors, as shown in Table 3. An increase in clearance of 12% was seen in children on efavirenz (ΔOFV = À10.9, P < 10 À4 ). Malnourished children had higher and more variable exposures compared to a non-malnourished typical child, as shown in Figure 2.  The final parameter estimates with uncertainty are presented in Table 4 and a VPC stratified by study and visit is shown in  UGT, a primary enzyme involved in abacavir metabolism. [35][36][37] It is uncertain to which extent rifampicin, ritonavir and/or lopinavir contributed to the effect. We previously reported that the lopinavir concentrations were similar during anti-TB treatment and superboosting, 20  Between-subject (BSV), -visit (BVV), and -occasion (BOV) variabilities were assumed as lognormally distributed and are reported as %CV (sqrt [omega] *100).

| Simulations
a All clearances and volumes of distribution were allometrically scaled and the typical values reported here refer to a child weighing 9.8 kg on LPV/r (4:1) at steady state, the median value in the dataset. b PMAGE 50 is the postmenstrual age at which 50% maturation is reached, while ƴ_maturation function is the shape factor in the sigmoidal maturation function. mentioning that although malnutrition does affect plasma protein composition, its impact is very minimal compared to its combined effect with inflammation. [38][39][40] Subsequently, inflammation is also associated with decreased hepatic expression of drug-metabolizing enzymes such as CYP and UGT enzymes. 41,42 Introduction of ART is linked with reduction of inflammation and improvement of malnutrition. The effect of malnutrition on abacavir pharmacokinetics appears to recover faster than children's weight gain. This is evident in the MATCH study, where PK between visits 1 and 2 was different while there was a small improvement in weight-for-age Z-score; see Table 1. This may explain the lack of association between malnutrition and PK in the DNDi study of children co-treated with rifampicin where, although some patients were malnourished, the first study visit was at least 1 month after treatment initiation. Despite including all the above-mentioned covariates, high variability in bioavailability and clearance was still observed in the MATCH study compared to the other studies, possibly reflecting the variability in the severity of malnutrition within the cohort.
In our analysis, abacavir, when formulated in fixed-dose combination tablets with lamivudine (and coadministered with efavirenz), had slower absorption than the liquid formulation, which was coadministered with LPV/r (4:1). This is consistent with prior reports that associated the liquid formulation with an 11% higher C max than the tablet formulation, although the difference was deemed as clinically unimportant. 9 In all the studies in this analysis, food was given at least 2 hours after dose administration, making food an unlikely cause of the observed difference.
The observed abacavir pre-dose concentrations (mostly 12 hours after the previous evening's self-reported time of dose) were often higher than the concentrations observed at 8-12 hours after observed dose intake. This could possibly be due to the night dose being given later than documented, slower absorption due to coadministration with food, or diurnal variation.
In Table 3