Understanding the determinants of drug exposure and pharmacokinetic variability in tuberculosis patients

M Muliaditan1, O Della Pasqua1,2. 1UCL School of Life and Medical Science, London, UK, 2GlaxoSmithKline, London, UK

Objectives: Despite promising advances in TB drug development, new integrated methods are still required to efficiently transition the growing number of novel compounds that are needed to shorten current first-line therapy. Clinical trial simulation (CTS) can be a valuable tool for informing different study designs with respect to the population, range of doses and dose intervals. Consequently, a more robust rationale can be developed for selecting doses and study designs before the actual trial is conducted. This project aims to present the first component of a simulation-based framework for the evaluation of combination therapies. A population model for simulation of virtual patients with demographic characteristics similar to a real trial population was developed to explore different dosing designs using rifampin (RIF) as a paradigm compound.

Methods: Virtual patients with varying demographic characteristics were simulated from a database containing 180 patients. Resampling was performed based on the multivariate distribution method1. To account for the effect of metabolic autoinduction, a published RIF PK-enzyme turnover model2 was selected. The enzyme-turnover mechanism assumed increasing enzyme production rate which in turn increased RIF clearance in a nonlinear manner over time. PK profiles were subsequently simulated for two months (intensive phase) of the standard six-month therapy period according to different dosing regimens. The percentage patients reaching a predefined target AUC/MIC ratio (≥271) and Cmax/MIC ratio (≥175) was used as measure of efficacy. Population pharmacokinetic modelling was performed using a non-linear mixed effects approach, as implemented in NONMEM 7.2.

Results: The use of a multivariate distribution method yielded a virtual patient population, with comparable covariate correlations to those observed in the original database. Relevant covariates were identified, including HIV on apparent volume of distribution (V/F) and normal fat mass on apparent clearance (CL/F) and V/F. Simulations suggest that the current dosing regimen of 10 mg/kg (range 8.3-12.1 mg/kg) based on either WHO recommended weight band or 50 kg cut-off point result in unsatisfactory target AUC/MIC and Cmax/MIC ratio, as compared to 16 mg/kg dosing (range 12.7-20.1 mg/kg).

Conclusion: In conclusion, a population model has been developed that can be used for the evaluation of different dosing regimens of RIF. The workflow and multivariate distribution methods proposed here can be easily applied to other drugs used in combination therapy for the treatment of tuberculosis.

References:

(1) Tannenbaum et al. (2006). J Pharmacokinet Pharmacodyn 33: 773-794

(2) Smythe et al. (2012). Antimicrob. Agents Chemother 56: 2091-2098