Statistical and Population Approaches to Ocular Pharmacokinetics

Optimally, one would like to extend the ocular pharmacokinetics knowledge base to patients. This has been made possible by the recent transfer of adaptive control theory and optimization methods from control systems engineering to the field of pharmaceutical pharmacokinetics/pharmacody-namics (PK/PD). These combined mathematical and statistical tools enable the robust determination of population pharmacokinetic mean parameter values and their dispersions from fragmentary data. There is now software available to support these population-modeling approaches, such as the nonlinear mixed effects modeling (NONMEM) (89), nonparametric expectation maximization (NPEM2) (88), an iterative two-stage (IT2S) population modeling software package (37,38,93) and WinNonMix (Pharsight Corporation, Mountain View, CA). These algorithms have been validated and employed to obtain population pharmacokinetic parameter values from sparse pharmacokinetic data sets (13,29), and comparisons of these packages are available in literature (57,76,77,98).

The utility of these mathematical models can be readily extended to the patient care situation. Most notably, adaptive population models require only a single pair of data where a prior complete kinetic data set exists to begin analysis, no matter how many system parameters are available in the population pharmacokinetic model. Thus, each new data set supplements the population data, which keeps the general population database growing. Most importantly, the more data samples that are obtained for an individual patient, the more patient-specific the model parameter estimation becomes, thereby resulting in customized therapy.

We have used a population modeling approach to characterize ocular pharmacokinetics. We used IT2S population modeling software to characterize and validate this approach following systemic administration of cipro-floxacin in a rabbit model (29). Sequential serum samples were obtained in

Figure 4 Simulation of mean concentrations of sparfloxacin in the serum and vitreous humor of rabbits following a single intravenous bolus (40 mg/kg) using mean pharmacokinetic parameter estimates from observed data. Penetration was expressed as a cumulative percentage by dividing the area under the concentration versus time curve (AUC) in the vitreous by that in the serum.

Figure 4 Simulation of mean concentrations of sparfloxacin in the serum and vitreous humor of rabbits following a single intravenous bolus (40 mg/kg) using mean pharmacokinetic parameter estimates from observed data. Penetration was expressed as a cumulative percentage by dividing the area under the concentration versus time curve (AUC) in the vitreous by that in the serum.

different subjects along with a single, sequential vitreous sample from one eye in each subject. We simultaneously also obtained sequential vitreous samples from the contralateral eye of all subjects. Data were analyzed by IT2S to determine the robustness of pharmacokinetic estimates obtained from single data point experiments. Penetration and derived pharmacokinetic constants were equivalent when single or complete ocular data sets were used. This finding suggests that intensive plasma sampling and optimized single sample experimental design holds promise for the analysis or penetration of drugs into privileged spaces, such as the eye and CSF, from which only single samples can be obtained.

Was this article helpful?

0 0

Post a comment