Population PK/PD model of GPI 15715 and GPI-derived propofol in sedation and comparison of PK/PD models for ordered categorical observations
E. Gibiansky(1), L. Gibiansky(2)
(1)Guilford Pharmaceuticals, Baltimore MD, USA; (2)Metrum Research Group, Avon CT, USA.
Objectives: AQUAVAN® Injection (GPI) is a water-soluble prodrug of propofol (PR). It was evaluated in an adaptive dose ranging colonoscopy study to produce a desired sedation level (MOAA/S score). A population PK model of GPI and GPI-derived PR and PK/PD model relating propofol concentrations to the MOAA/S scores were developed. Covariate predictors of PK and the effect were identified, and simplified dosing strategies were explored.
Methods: NONMEM analysis was performed using sparse plasma samples from 158 patients pre-medicated by i.v. fentanyl and receiving initial and up to 4 (mean 1) supplemental bolus doses of GPI (total 495-1675 mg). A linear model described PK of GPI (compartments 1, 2), PR (compartments 4, 5) and a GPI/PR concentration delay (compartment 3). Rich MOAA/S data and individual PR concentration predictions were used to develop PK/PD models. The effect compartment described a PK/PD delay. The probabilistic model described probabilities of being at each MOAA/S level while the continuous model described the expected MOAA/S scores. Predictive check simulations compared the models. Back-Step Method [1] was tried to improve the PK/PD estimates.
Results: Lean body weight (LBW) was the best predictor of PK. GPI and PR central volumes, and GPI clearance increased by 1.8%, 2.5%, and 1.4% per kg of LBW, respectively. Predicted PR Cmax (at 4-5 minutes post-dose) was proportional to 1/LBW^0.45. There was no effect of fentanyl, age or gender on PK.
Individual predictions of the PK/PD models were similar and similar to the observed data. Predictive check simulations showed higher fraction of deeply sedated patients than observed, especially for the continuous model. Older patients (> 65 years) were estimated to have approximately 25% stronger effect at the same PR concentrations. No fentanyl or gender effect was detected.
Conclusions: 1. A linear PK model adequately described the data. 2. LBW was the best predictor of PR concentrations. Strictly weight-proportional dosing may overdose overweight individuals. Mg/kg dosing with an upper dose boundary, or fixed-dose (mg) in the ranges of weights may be preferable. 3. Age did not influence PK, but increased the PD effect. A dose reduction of about 25% is needed for patients over 65 years. 4. Fentanyl did not affect PK or PD. 5. Continuous and probabilistic PD models adequately described the data and the covariate effects; simulations demonstrated better predictive abilities of the probabilistic model.
References:
[1] Kjellsson MC, Jönsson S,