Nichada Seriniyom 1, Peter Velickovic 1,3, Alessandro Di Deo 1,3, Candie Joly 2, Julien Lemaitre 2, Emma Jougla 2, Oscar Della Pasqua 1,3
1 University College London (London, United Kingdom), 2 Université, Paris-Saclay, Inserm, CEA, Center for Immunology of Viral, Auto-Immune, Hematological and Bacterial Diseases (IMVA-HB/IDMIT) (Fontenay-aux-Roses, France), 3 Consiglio Nazionale Delle Ricerche (CNR) (Rome, Italy)
Introduction: Tuberculosis (TB) remains a leading cause of death from a single infectious agent worldwide[1]. Efforts in shortening anti-TB therapy have led to the developing of a 4-month regimen with isoniazid (H), rifapentine (P), moxifloxacin (M), and pyrazinamide (Z) (HPMZ), which has demonstrated non-inferiority relative to the standard-of-care regimen in drug susceptible TB[2].
Macaques are widely used in antitubercular research as they recapitulate key features of human TB pathology[3]. Assessing the predictive and translational value of TB infection models in non-human primates (NHPs) may improve clinical translation in to humans, reduce attrition and optimise early clinical trial design. To achieve this, data generation should be combined with quantitative pharmacokinetic-pharmacodynamic (PKPD) modelling approaches, taking into account the complex pathophysiology of TB in humans[4].
However, the use of NHPs in research carries significant ethical responsibilities. In accordance to the 3Rs (replacement, reduction, and refinement) principles, their use must be minimised and refined[5]. By characterising the pharmacokinetics (PK) of HPMZ using satellite PK data in cynomolgus macaques, we aimed to inform dose selection, dosing regimens, sampling schedule and sample sizes determination for subsequent efficacy studies in this species, whilst ensuring human-equivalent exposure ranges are achieved.
Methods: Pharmacokinetic data were obtained at steady state after multiple dose oral administration of the different drugs to uninfected macaques (H: 50mg/kg, P: 6mg/kg, M: 60mg/kg, and Z: 120mg/kg). Model development was implemented by integrating these data with historical information. PK models were developed, with one- and two-compartment models tested for each drug, including the effect of body weight based on allometric principles. Model diagnostics included statistical and graphical criteria (including OFV reduction, goodness-of-fit plots, and visual predictive checks).
The final PK models were subsequently used to simulate individual concentration vs. time profiles at steady-state and identify suitable dose range to be used for each drug in a prospective study. Systemic exposure was summarised in terms of AUCss, and compared with observed pharmacokinetic data in tuberculosis patients[6-9]. If significant differences were observed between observed and predicted exposure, doses were adjusted and simulations re-run until AUCss estimates were within the therapeutic range, assuming pharmacokinetic linearity across the simulated dose range in macaques.
Available pharmacokinetic models were applied in conjunction with D-optimality principles, as implemented via the $DESIGN tool on NONMEM[10], to optimise sparse blood sampling in infected macaques. Multiple scenarios, with varying time points, number of samples and animals per treatment arm, were evaluated. Optimisation procedures aimed to minimise the relative standard error (RSE) of clearance, which we selected as the main parameter of interest for optimisation purposes.
All modelling and simulation steps were implemented in NONMEM 7.5 and PsN 5.2.6. Data handling, graphical and statistical summaries were performed in R 4.3.0[10-12].
Results: The pharmacokinetics of P, M and Z was best described by a one-compartment model, whereas a two-compartment model structure was required to describe the pharmacokinetics of H. All four drugs could be parameterised using first order absorption and elimination. Based on the initial model predicted AUCss (95%CI) for P, M, Z and H were 512.77mg∙h/L (model did not include variability), 50.8 mg∙h/L (41.1-63.3mg∙h/L), 487.2mg∙h/L (253.9-897.0mg∙h/L) and 28.2mg∙h/L (16.9-47.7mg∙h/L), the following doses were determined to produce comparable exposure in macaques as seen in humans: H (50mg/kg), P (6mg/kg), M (60mg/kg), and Z (120mg/kg).
In a prospective study with infected animals, to capture Cmax and trough concentrations following 2 weeks of dosing, sampling should take place on two occasions: once after single dose and once after steady-state. Each sampling occasion should consist of 3 sampling windows, with 3 samples taken in 15-minute intervals per window. Furthermore, our analysis showed that the use of informative priors would allow significant reduction the required number of animals, with a total of four macaques being sufficient for the evaluation of PKPD relationships and efficacy of HPMZ.
Conclusions: Whilst early evidence of the efficacy of novel anti-TB regimens in macaques may be valuable, the use of non-human primates in tuberculosis research should not be accepted without a model-based translational framework that ensures informative data generation and full adherence to the 3R principles [5].
This work has received support from the Innovative Medicines Initiatives 2 Joint Undertaking (grant No 853989).
References:
[1] World Health Organization. Global tuberculosis report 2024. Geneva: WHO; 2024.
[2] Galvis ME, Gao GE, Salerno MM, Whitehead M, Juste D, Buchanan C, et al. Challenges to initiate and complete a 4-month rifapentine-moxifloxacin TB treatment regimen. Int J Tuberc Lung Dis. 2025; 29(7):318–24.
[3] Peña JC, Ho WZ. Non-human primate models of tuberculosis. Microbiol Spectr. 2016 ;4(4):TNMI7-0029-2016.
[4] Yang HJ, Wang D, Wen X, Weiner DM, Via LE. One size fits all? Not in in vivo modeling of tuberculosis chemotherapeutics. Front Cell Infect Microbiol. 2021; 11:664.
[5] Poh WT, Stanslas J. The new paradigm in animal testing – 3Rs alternatives. Regul Toxicol Pharmacol. 2024; 153:105705.
[6] Savic RM, Lu Y, Bliven-Sizemore E, et al. Population pharmacokinetics of rifapentine and desacetyl rifapentine in healthy volunteers: nonlinearities in clearance and bioavailability. Antimicrob Agents Chemother. 2014;58(6):3035-3042.
[7] Pranger AD, van Altena R, Aarnoutse RE, et al. Evaluation of moxifloxacin for the treatment of tuberculosis: 3 years of experience. European Respiratory Journal. 2011;38(4):888-894.
[8] Xu AY, Velásquez GE, Zhang N, et al. Pyrazinamide Safety, Efficacy, and Dosing for Treating Drug-Susceptible Pulmonary Tuberculosis: A Phase 3, Randomized, Controlled Clinical Trial. American Journal of Respiratory and Critical Care Medicine. 2024;210(11):1358-1369.
[9] Naidoo A, Chirehwa M, Ramsuran V, et al. Effects of genetic variability on rifampicin and isoniazid pharmacokinetics in South African patients with recurrent tuberculosis. Pharmacogenomics. 2019;20(4):225-240.
[10] Bauer RJ, Hooker AC, Mentré F. Tutorial for $DESIGN in NONMEM: Clinical trial evaluation and optimization. CPT Pharmacometrics Syst Pharmacol. 2021;10(9):1000–12.
[11] Beal SL, Sheiner LB, Boeckmann AJ, Bauer RJ. NONMEM 7.5 User’s Guide. ICON plc; 1989–2020. Available from: https://nonmem.iconplc.com/nonmem750
[12] R Core Team (2023). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
Reference: PAGE 34 (2026) Abstr 12216 [www.page-meeting.org/?abstract=12216]
Poster: Drug/Disease Modelling - Infection