Avuyonke Balfour 1, Carene Sima 2, Lufina Tsirizani 1, Caitlin Uren 2,3,4,5, Marlo Möller 2,3,4,5, Geraint Davies 6, James Millard 6,7,8, Paolo Denti 1
1 Division of Clinical Pharmacology, Department of Medicine, University of Cape Town (Cape Town, South Africa), 2 South African Medical Research Council Centre for Tuberculosis Research, Division of Molecular Biology and Human Genetics, Department of Biomedical Sciences, Faculty of Medicine and Health Sciences, Stellenbosch University (Cape Town, South Africa), 3 Centre for Bioinformatics and Computational Biology, Stellenbosch University (Stellenbosch, South Africa), 4 Genomics for Health in Africa (GHA), Africa-Europe Cluster of Research Excellence (Cape Town, South Africa), 5 National Institute for Theoretical and Computational Sciences (Stellenbosch, South Africa), 6 Department of Clinical Infection, Microbiology and Immunology, University of Liverpool (Brownlow Hill, United Kingdom), 7 Imperial NHS Foundation Trust (Paddington, United Kingdom), 8 African Health Research Institute (Durban, South Africa)
Objectives
Isoniazid is used at standard doses of 4–6 mg/kg for drug-susceptible tuberculosis and at higher doses of 10–15 mg/kg in multidrug-resistant tuberculosis (MDR-TB), to overcome low-to-intermediate level resistance, particularly that associated with inhA mutations [2]. Isoniazid pharmacokinetics are variable due to N-acetyltransferase 2 (NAT2) gene polymorphisms [1, 2]. NAT2 single nucleotide polymorphisms (SNPs) classify individuals as slow, intermediate, or rapid acetylators with markedly different isoniazid exposures [3]. This variability may lead to subtherapeutic drug concentrations in rapid acetylators or increased toxicity from overexposure in slow acetylators [4, 5]. The 4-SNP panel (rs1801280, rs1801279, rs1799930, rs1799931) is considered the current gold standard with 94% accuracy in inferring NAT2 phenotype [2], while the tag SNP (rs1495741), recently proposed as a more pragmatic alternative [2, 6]. In the absence of individual genotype data, mixture models can be used to categorise patients into subgroups based on the observed data [5, 7, 8].
We characterised isoniazid pharmacokinetics in adults with RR/MDR-TB and compared these three approaches to identify the optimal strategy for describing NAT2-driven inter-individual variability.
Methods
We characterized isoniazid pharmacokinetics in adults treated for RR/MDR-TB between 2017–2019 in KwaZulu-Natal, South Africa. Participants were enrolled prior to the introduction of BPaLM-based regimens and received an oral short MDR-TB regimen containing weight-based standard- or high-dose isoniazid. Intensive pharmacokinetic (PK) sampling was performed after ≥4 weeks of treatment (pre-dose, 1, 2, 4, 6, 8 hours post-dose). Plasma concentrations were quantified using liquid chromatography-mass spectrometry with a lower limit of quantification (LLOQ) of 0.05 mg/L. DNA was genotyped using the H3Africa array.
Population PK (PopPK) models were developed in NONMEM v7.5.1 using first-order conditional estimation with interaction (FOCE-I). Fat-free mass (FFM) and total body weight (TBW) were compared for allometry. Alternative absorption (first-order with/without delay, transit), distribution (1–2 compartment), and elimination (linear or saturable first-pass using a well-stirred liver model) structures were evaluated. For the saturable model, liver volume was 1 L and hepatic blood flow 90 L/h for a 70 kg individual. Intrinsic clearance (CLint) was defined with respect to unbound drug (fraction unbound, fu = 0.95). NAT2 phenotypes were assigned using tag-SNP, 4-SNP panel, and a mixture model with fixed proportions based on 4-SNP frequencies. Between-occasion variability (BOV) was implemented on absorption parameters and between-subject variability (BSV) on disposition parameters.
Results
Plasma concentrations were collected from 103 participants (69% male, median weight 55 kg [interquartile range, IQR 50–63], FFM 45 kg [39–50]). The tag-SNP classified 33% as slow, 51% intermediate, 16% rapid; the 4-SNP panel classified 40%, 43%, 17%; and the mixture model 42%, 42%, 16%, respectively. Isoniazid pharmacokinetics was best described by a two-compartment model with transit absorption and saturable elimination using a well-stirred liver model [9, 10]. Disposition parameters were best scaled using FFM, which improved model fit compared to using no allometry (ΔOFV = −17.6) more than total body weight (ΔOFV = −13.5).
All three NAT2 approaches improved model fit compared to the base model. The 4-SNP panel provided the greatest OFV reduction (ΔOFV = −45.6), followed by the tag-SNP (ΔOFV = −33.8) and mixture model (ΔOFV = −15.2). Correspondingly, BSV on clearance decreased from 66% to 51.7% (4-SNP), 57.4% (tag-SNP), and 25.7% (mixture model). BOV on bioavailability decreased from 30.1% to 25.4% (4-SNP), 28.6% (tag-SNP), and 26.0% (mixture model). For the final 4-SNP model, intrinsic clearance for a typical individual (FFM 45 kg) was 13.5, 22.3, and 47.5 L/h for slow, intermediate, and rapid acetylators, respectively, with a volume of distribution of 40.8 L. The Km of 19.1 mg/L from the 4-SNP model was consistent with observed plasma concentrations and previous findings [5], while estimates from alternative approaches were substantially higher, suggesting reduced parameter identifiability.
Conclusions
We developed a PopPK model for isoniazid in adults with RR/MDR-TB. Incorporating the well-stirred liver model captured saturation of isoniazid metabolism at high doses. The 4-SNP panel provided superior model fit compared to tag-SNP or mixture model, with intrinsic clearance varying 3.5-fold across acetylator phenotypes, consistent with previous reports [3, 5]. Improved fit with the 4-SNP panel likely reflects better phenotype resolution, particularly among intermediate acetylators, reducing bias in clearance estimation. Discordance between mixture-derived and genotype-based assignments at individual level highlights uncertainty in probabilistic classification, warranting further investigation.
References:
1. Wattanapokayakit S, Sawaengdee W, Kunhapan P, et al (2025) NAT2 rapid acetylator phenotype and increased risk of tuberculosis retreatment: A TB cohort study in Northern Thailand. International Journal of Infectious Diseases 161:. https://doi.org/10.1016/j.ijid.2025.108123
2. Hein DW, Doll MA (2012) Accuracy of Various Human NAT2 SNP Genotyping Panels to Infer Rapid, Intermediate and Slow Acetylator Phenotypes. Pharmacogenomics 13:31–41. https://doi.org/10.2217/pgs.11.122
3. Thomas L, Batcha JSD, Chaithra, et al (2026) Population pharmacokinetics of isoniazid in adult Indian tuberculosis patients: Evaluation of NAT2 polymorphisms. Infection, Genetics and Evolution 105883. https://doi.org/10.1016/j.meegid.2026.105883
4. Kasamatsu A, Miyahara R, Yoneoka D, et al (2025) One-year mortality of tuberculosis patients on isoniazid-based treatment and its association with rapid acetylator NAT2 genotypes. International Journal of Infectious Diseases 155:. https://doi.org/10.1016/j.ijid.2025.107895
5. Gausi K, Chirehwa M, Ignatius EH, et al (2022) Pharmacokinetics of standard versus high-dose isoniazid for treatment of multidrug-resistant tuberculosis. Journal of Antimicrobial Chemotherapy 77:2489–2499. https://doi.org/10.1093/jac/dkac188
6. García-Closas M, Heind DW, Silvermana D, et al (2011) A single nucleotide polymorphism tags variation in the arylamine N-acetyltransferase 2 phenotype in populations of European background. Pharmacogenet Genomics 21:231–236. https://doi.org/10.1097/FPC.0b013e32833e1b54
7. Keizer RJ, Zandvliet AS, Beijnen JH, et al (2012) Performance of methods for handling missing categorical covariate data in population pharmacokinetic analyses. AAPS Journal 14:601–611. https://doi.org/10.1208/s12248-012-9373-2
8. Calderin JM, Wasserman S, Resendiz-Galvan JE, et al (2025) Population pharmacokinetics of pyrazinamide and isoniazid in plasma and cerebrospinal fluid from South African adults with tuberculous meningitis. Antimicrob Agents Chemother 69:. https://doi.org/10.1128/aac.00099-25
9. Yan Z, Ma L, Carione P, et al (2024) Introducing the Dynamic Well-Stirred Model for Predicting Hepatic Clearance and Extraction Ratio. J Pharm Sci 113:1094–1112. https://doi.org/10.1016/j.xphs.2023.12.020
10. Savic RM, Jonker DM, Kerbusch T, Karlsson MO (2007) Implementation of a transit compartment model for describing drug absorption in pharmacokinetic studies. J Pharmacokinet Pharmacodyn 34:711–726. https://doi.org/10.1007/s10928-007-9066-0
Reference: PAGE 34 (2026) Abstr 12035 [www.page-meeting.org/?abstract=12035]
Poster: Drug/Disease Modelling - Infection