12th Grade Student Park Tudor High School, Indiana, United States
Disclosure(s):
Arnay Garhyan: No financial relationships to disclose
Objectives: Assessment of potential liability of prolonged cardiac repolarization for new drugs is crucial 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. Due to practical challenges, only a limited number of pharmacokinetic (PK) samples are collected in these experiments, while extensive QT data is collected every minute using telemetry. This study aims to develop concentration-QTc models using four approaches: Model 1a. Time-matched observed PK and QTc; Model 1b. Time-matched observed PK and QTc averaged over 5 minute intervals; Model 2a. Sequential PK-QTc using population PK model predicted PK to match QTc every minute; and Model 2b. Sequential PK-QTc using population PK model predicted PK to match QTc averaged over 5-minute intervals.
Methods: A cardiovascular telemetry study was conducted in NHPs (n=48, 50% females) with 9 PK samples collected after an 80 mg/kg oral dose of Moxi and QT was collected every minute. A single compartment population PK model with body weight as covariate on clearance, volume, and absorption rate described the data well (1). Direct-effect linear (slope parameter) and non-linear models (Emax) were used to link PK with changes in QTc (corrected for HR). Changes in QTc were estimated by subtracting time-matched pre-Moxi (vehicle period) from post-Moxi (treatment period) QTc values, thus minimizing the circadian pattern in QTc and simplifying the PD model. Modeling was performed using R package nlmixr2 and R Version 4.3.0 was used.
Results: QTc increased after Moxi dosing in all NHPs but was highly variable as data were collected every minute. An Emax model performed better than a linear slope model, indicating saturation of QTc increase at higher Moxi concentrations. Between-subject variability (BSV) was estimated for EC50 . Proportional and additive residual errors were estimated. Fidelity of the model was confirmed using goodness of fit plots, visual predictive check, and parameters had an acceptable standard error. The estimate of Emax was lower for Model 1a (32.3 msec) and Model 1b (34 msec) compared to Models 2a (42.2 msec) and 2b (43 msec), likely due to a lean dataset based on observed PK data only. The point estimates of EC50 were similar, 3610, 3730, 4810, 5020 ng/ml, respectively.
Conclusions: Population PK/PD models to describe the increase in QTc after Moxi dosing in NHPs were developed. The Emax models outperformed linear models, and the parameters were robustly estimated. Lower Emax in time-matched PK-QTc models underestimated QTc; therefore, full sequential PK-PD models are better suited for robustly characterizing QTc. Simulations using these models can be used to identify dose levels to produce desired increase in QTc.
Citations: [1] Garhyan A, Leishman DJ. Development of Population Pharmacokinetics Model and R-Shiny Simulation Platform for Moxifloxacin in Non-Human Primates. ASCPT Annual Meeting, Washington DC, May 2025.
