Aim: The aim of this study was to develop a population PK model for propofol in adults that could be used to determine optimal sampling times for a prospective pharmacokinetic study.
Methods: Data from 15 pharmacokinetic studies of propofol in adults were obtained from the OpenTCI Initiative [1]. A total of 7711 concentrations from 405 subjects were used for analysis. The participants in the studies were either surgical patients or healthy volunteers given bolus or infusion dosing schedules. 13 of 15 datasets (326 subjects) included patient covariates with demographics as follows (mean [range]): weight 71.1 [40-160]kg, age 42.4 [18-92]yrs, lean body weight (LBW) 52.8 [29-84]kg, 72% male/28% female. Samples for the determination of propofol concentrations were either taken from arterial (80% of subjects) or venous sites (20% of subjects) over a period from 5min-25hr after the start of dosing. The number of propofol concentrations ranged from 3-42 samples/subject. A population pharmacokinetic model was developed in NONMEM VI using the first order conditional estimation (FOCE-I) method. Both two and three compartment zero-order input models were tested with proportional and combined residual error models. Between subject variability (BSV) was modeled as log-normally distributed. The NONMEM objective function value (OBJ) and diagnostic plots were used to discriminate between models. A visual predictive check was used to assess the final model. Preliminary screening of covariates was undertaken to evaluate the relationships between parameters and available covariates.
Results: A three compartment model with between subject variability on clearance (CL), inter-compartmental clearances (CL2 & CL3), volume of distribution of the central (V1) and peripheral compartments (V2 & V3) with combined additive and proportional residual error best described the concentration-time data. The population mean estimates of CL, CL2, CL3, V1, V2 and V3 were 106L/h, 92.8L/h, 67.1L/h, 7.5L, 21.7L and 305L, respectively, with BSVs of 46% for CL, 49% for CL2, 95% for CL3, 63% for V1, 68% for V2 and 49% for V3. Proportional residual variability was estimated as 25% and 20% for subjects with arterial and venous sampling, respectively, and additive error was estimated as 0.008μg/mL for all subjects. Standard errors were low for all parameters (<6% for fixed effects, <20% for BSV and <22% for RUV). The potential influence of patient covariates on PK parameters, such as a relationship between LBW and CL, will be further investigated.
Conclusions: A base model for propofol was developed with estimated PK parameters comparable to those previously reported in the literature. Separating proportional RUV terms according to sampling method resulted in a significant drop in OBJ, with arterial sampling showing larger variability than venous sampling.
References:
- The OpenTCI Initiative (available from http://opentci.org/) © 2008, Minto, C. & Schnider, T.