Nonparametric estimation procedures and application to pharmacokinetic/pharmacodynamic analysis.

摘要

Mathematical modeling and parameter estimation are prominent features of pharmacokinetic/pharmacodynamic (PK/PD) analysis. Ambiguity in model selection and problems in parameter estimation complicate the traditional parametric modeling approach making it a subjective analysis. Nonparametric regression provides an attractive and more objective alternative to these parametric approaches. However, currently available nonparametric regression software is not suitable to handle the sparse, unequally spaced PK/PD data typically encountered in drug research, emphasizing the need for new developments in this area.; The main objective of this work is the development and application of some novel, nonparametric methods for improved estimations in PK/PD analysis. A novel, end-constrained cubic spline (ECS) method employing regularization and cross-validation principles specially designed to handle kinetic and physiologic constraints is presented in Chapter 2.; An improved deconvolution method (DDR-ECS) enabling drug absorption rates to be assessed under time-variant disposition conditions is developed and validated through extensive simulations in Chapter 3. The DDR-ECS method provides better input estimations under significant clearance changes but overestimates the input when there were only small clearance changes.; Chapter 4 demonstrates the development and application of a novel cellular deconvolution methodology to quantify the rate of reticulocyte and hemoglobin production following phlebotomy-induced anemia. The methodology showed limited sustained high reticulocyte and hemoglobin production rates in acute anemia.; A "bottom-up" (BU) analysis approach employing convolution/deconvolution principles, hysteresis minimization and special nonparametric estimation procedures is presented in Chapter 5 to determine nonparametrically the population PK/PD transduction function, linking drug effect to drug concentration. The analysis demonstrates that an Emax model best describes the PK/PD function relating the progenitor activation rate to EPO concentrations. In Chapter 6, semiparametric model of EPO activation of the progenitors based on the BU analysis is proposed to estimate the parameters of interest and for prediction purposes.; The PK/PD analysis demonstrates the application of end-constrained splines to determine drug input and cellular production rates, and as an exploratory tool to identify parametric PK/PD transduction functions for further analysis. The methodologies presented are valuable tools to evaluate drug input rates and to resolve the complex PK/PD of drugs governing cellular production.