PI-117 - DEVELOPMENT OF POPULATION PHARMACOKINETICS MODEL AND R-SHINY SIMULATION PLATFORM FOR MOXIFLOXACIN IN NON-HUMAN PRIMATES

D. Leishman¹, A. Garhyan²; ¹Eli Lilly and Company, Indianapolis, IN, USA, ²Park Tudor High School, Indianapolis, IN, USA.

Background: Assessment of potential liability of prolonged cardiac repolarization for a new drug is critical in drug research and development. Moxifloxacin (Moxi) is used as a positive control in well controlled in-vivo experiments to evaluate cardiac safety in preclinical species. A limited number of pharmacokinetic (PK) samples are collected in these experiments due to practical challenges, thus limiting the full utilization of extensive QT data collected every minute. The objectives of this work were to develop a population PK model of Moxi in non-human primates (NHP) to enable concentration-QTc evaluation of all available telemetered ECG data, and a user-friendly R Shiny application to predict full PK profiles using limited PK data.

Methods: A cardiovascular telemetry study was conducted in NHPs (n=48, 50% females) with pharmacokinetic samples collected on 4 separate periods after an 80 mg/kg oral dose of Moxi. Periods 1 (at 24 hours) and 2 (at 6 and 24 hours) had sparse PK samples and Periods 3 and 4 collected serial PK samples for 48 hours post dose. Population PK modeling was performed using R package nlmixr2, R package rxode2 (Version 2.1.3) was used for simulations, and R Version 4.3.0 was used. The simulation platform was created using Shiny for R.

Results: Moxi PK in NHPs (body weight (BW) range: 2.83 - 6.67 kg) was best described with a one compartment, first-order absorption and elimination model. Between-subject variability was estimated for absorption rate (ka), clearance (CL), and volume (V) with a correlation between CL and V. BW was used to allometrically scale CL and V. BW was a covariate on ka; ka decreased with increase in BW (Table 1). Fidelity of the model was confirmed using goodness of fit plots, visual predictive check, and parameters had an acceptable standard of error. As only one dose level of Moxi was used in modeling, PK data in literature (1, 2) at other dose levels (10 to 90 mg/kg) were overlaid on simulations to confirm dose-linearity and externally validate the PK model.

Conclusion: A comprehensive population PK model to describe Moxi PK in NHPs was developed and validated against published data not used in the model development. A R Shiny app implemented the developed model allowing users to generate full PK profiles for the NHP from serial or sparse PK samples based on estimated post-hoc PK parameters.