Hylke Waalewijn1,2, Lufina Tsirizani1, Vanessa Rouzier3,6, Paolo Denti1, Veronique Dartois4, M Zimmerman5, Myung Hee Lee7, Nadalette Alcenat6, Nao Haba7, Daniel Fitzgerald7, Jean William Pape6, Jyoti Mathad7
1University of Cape Town, 2Certara, 3Weill Cornell Medical College, 4Rutgers University, Department of Medicine, 5Center for Discovery and Innovation, Hackensack Meridian Health , 6GHESKIO, Medicine, 7Weill Cornell Medical College, Center for Global Health
Introduction: Current dosing of tuberculosis drugs in children is largely based on extrapolation from adult data, and does not correctly account for physiological differences between the two populations (1). Additional factors, including malnutrition, HIV and concomitant medications (including antiretroviral treatment), may further alter drug exposure in children (2,3). Generating pharmacokinetic data in children, especially those with malnutrition, is hampered by frequent blood draws for intensive pharmacokinetic sampling, and less invasive blood sampling techniques for children are needed. We conducted a prospective study to investigate the impact of malnutrition and HIV on the pharmacokinetics of first-line tuberculosis drugs in children <5 years in Port-au-Prince, Haiti. We also evaluated a newly developed blood collection device using capillary blood in young children, comparing it to the standard intravenous method. Methods: We enrolled children with and without HIV, receiving treatment for drug-sensitive tuberculosis as per WHO guidelines at GHESKIO’s pediatric clinic in Port-au-Prince, Haiti. All drugs were dosed using WHO weight-bands. Between 2 and 8 weeks after treatment initiation, we collected blood at 0, 1, 2, 6, and 24 hours post-dosing for pharmacokinetic testing. Using nonlinear mixed-effects modelling in NONMEM, we characterized the population pharmacokinetics of rifampicin (RIF), isoniazid (INH), pyrazinamide (PZA), and ethambutol (EMB). For each drug, we tested one- and two-compartment disposition models with first-order elimination and absorption (with or without delay with either transit absorption compartments or lag time). Modelling was performed using data from all three matrices (plasma, venous whole blood and capillary whole blood) and comparing the concentrations attained. We used the L2 data item in NONMEM to estimate correlations between the residual error terms of the observations collected at the same time, employing a Cholesky decomposition to improve the stability of correlation estimates. We tested the effect of body size, age, malnutrition, HIV, and ART. For INH, a mixture model was used to assign children to either slow, or intermediate/fast N-Acetyltransferase 2 (NAT2) genotype. Results: Data for 39 children were analyzed, 17 with malnutrition (weight-for-age z-score < -2), and 22 without malnutrition. Median baseline weight (IQR) was 8.22 (6.94-10.4) kg and age 1.46 (1.02 – 1.86) years. 8 children were living with HIV (5 on LPV/r and 3 on dolutegravir-based regimens). A one-compartment disposition model with first-order elimination best described RIF and PZA data, while a two-compartment model with first-order elimination best described INH and EMB data. Accounting for body size with allometric scaling and characterizing the maturation of clearance using a sigmodal function (with different maturation parameters for each drug and using priors where needed) improved the model fit to the data for all drugs. Furthermore, a sigmoidal function describing an increase of bioavailability with age significantly improved the model fit for all drugs except pyrazinamide. We did not find a significant impact of HIV status. Malnutrition was found decrease pyrazinamide absorption by 32% and, led to 48% lower isoniazid bioavailability. The correlation between the drug concentrations observed in plasma and capillary and venous whole blood was high, with r2 values ranging between 0.94 and 0.98. Conclusions: The pharmacokinetics of 1st-line tuberculosis drugs in this cohort are in line with previous reports (2,4). We observed that malnutrition significantly decreased isoniazid exposure, but this result requires further investigation due to potential limitations in our study. Firstly, the absence of NAT2 genotype data may have introduced bias, since there could be an imbalance between NAT2 genotype and malnutrition. Our findings would therefore need confirmation in children with NAT2 genotype data. Additionally, the dosing information was not directly recorded in all children, with that information needing imputation based on the weight-band of each child on the day of the PK visit. Lastly, and importantly, tuberculosis drug concentrations were comparable between samples collected via capillary sampling versus venous sampling, thus making the capillary sampling technique an attractive alternative for PK studies and monitoring of TB drug concentrations in children.
1. World Health Organization. WHO Consolidated Guidelines on Tuberculosis, Module 4: Treatment – Drug-Susceptible Tuberculosis Treatment. 2022. 2. Galileya LT, Wasmann RE, Chabala C, Rabie H, Lee J, Njahira Mukui I, et al. Evaluating pediatric tuberculosis dosing guidelines: A model-based individual data pooled analysis. PLOS Med. 2023;20(11):e1004303. 3. Gafar F, Wasmann RE, McIlleron HM, Aarnoutse RE, Schaaf HS, Marais BJ, et al. Global estimates and determinants of antituberculosis drug pharmacokinetics in children and adolescents: a systematic review and individual patient data meta-analysis. Eur Respir J. 2023 Mar;61(3):2201596. 4. Tikiso T, McIlleron H, Abdelwahab MT, Bekker A, Hesseling A, Chabala C, et al. Population pharmacokinetics of ethambutol in African children: a pooled analysis. J Antimicrob Chemother. 2022 Jun 29;77(7):1949–59.
Reference: PAGE 33 (2025) Abstr 11709 [www.page-meeting.org/?abstract=11709]
Poster: Drug/Disease Modelling - Paediatrics