Tuberculosis (TB) remains one of the leading causes of death from a single infectious agent, with an estimated 10.7 million people developing active disease and 1.23 million dying in 2024. Key challenges in TB management include the limited availability of child-friendly formulations, the need for novel treatment-response biomarkers quantifying bacterial load, and cardiac safety concerns associated with QT-prolonging regimens. The overall aim of this thesis was to apply pharmacometric approaches to address these challenges, spanning cardiac safety assessment, formulation bridging, and the characterisation of treatment-response biomarkers.

Conventional heart rate (HR) corrections of the QT interval assume a fixed QT–HR relationship and will therefore generate biased assessment in patients with TB, in whom an elevated HR normalises during treatment. Using data from 440 patients treated for drug-resistant TB, a population model of HR recovery was developed and used to derive a time-varying correction factor, QTcTBT, which transitions from a TB-specific pretreatment correction exponent towards the Fridericia exponent as HR recovers. QTcTBT maintained independence between corrected QT and HR throughout treatment and was externally validated in two independent cohorts, whereas Bazett’s and Fridericia’s corrections showed systematic over- and under correction, respectively.

Applying QTcTBT to exposure–QT analysis of bedaquiline treatment, the typical QT-prolonging effect of the metabolite M2 was estimated at 7.01 ms (90% confidence interval 5.95–8.16) at clinically relevant concentrations, remaining below the 10 ms regulatory threshold. Estimates obtained with QTcTBT were not less sensitive to assumptions regarding non-drug-related changes in QT over time than those based on Fridericia-corrected QT.

To support paediatric access to the rifapentine–isoniazid preventive treatment regimen, the relative bioavailability of tablets suspended in water was compared with whole tablets in 24 adults. The two modes of administration provided the relative bioavailability within the limits for bioequivalent, with 90% confidence intervals contained within the 80–125% limits, although absorption was more rapid following suspension.

Finally, a joint longitudinal model integrating time to positivity in liquid culture (MGIT-TTP), the TB molecular bacterial load assay (TB-MBLA), and a sputum lipoarabinomannan antigen test (PATHFAST LAM Ag test) was developed using data from 149 participants. The model quantified biomarker- and phase-specific correlations in bacterial clearance and showed the complementarity of the novel biomarkers in addition to MGIT-TTP through exposure–response relationships for sutezolid and delpazolid, although the novel biomarkers were less sensitive than MGIT-TTP for quantifying drug effects.

This thesis demonstrates how pharmacometric approaches can strengthen model-informed development of TB therapeutics by improving cardiac safety assessment, enabling formulation bridging to paediatric populations, and characterizing novel biomarkers can be integrated alongside standard culture-based endpoints to evaluate treatment response.