Order Conditional Estimation Research Articles

Simultaneous versus sequential optimal design for pharmacokinetic-pharmacodynamic models with FO and FOCE considerations

We consider nested multiple response models which are used extensively in the area of pharmacometrics. Given the conditional nature of such models, differences in predicted responses are a consequence of different assumptions about how the models interact. As such, sequential versus simultaneous and First Order (FO) versus First Order Conditional Estimation (FOCE) techniques have been explored in the literature where it was found that the sequential and FO approaches can produce biased results. It is therefore of interest to determine any design consequences between the various methods and approximations. As optimal design for nonlinear mixed effects models is dependent upon initial parameter estimates and an approximation to the expected Fisher information matrix, it is necessary to incorporate any influence of nonlinearity (or parameter-effects curvature) into our exploration. Hence, sequential versus simultaneous design with FO and FOCE considerations are compared under low, typical and high degrees of nonlinearity. Additionally, predicted standard errors of parameters are also compared to empirical estimates formed via a simulation/estimation study in NONMEM. Initially, design theory for nested multiple response models is developed and approaches mentioned above are investigated by considering a pharmacokinetic-pharmacodynamic model found in the literature. We consider design for situations where all responses are continuous and extend this methodology to the case where a response may be a discrete random variable. In particular, for a binary response pharmacodynamic model, it is conjectured that such responses will offer little information about all parameters and hence a sequential optimization, in the form of product design optimality, may yield near optimal designs.

Evaluation of the nonparametric estimation method in NONMEM VI

Purpose In NONMEM VI, a novel method exists for estimation of a nonparametric parameter distribution. The parameter distributions are approximated by discrete probability density functions at a number of parameter values (support points). The support points are obtained from the empirical Bayes estimates from a preceding parametric estimation step, run with the First Order (FO) or First Order Conditional Estimation (FOCE) methods. The purpose of this work is to evaluate this new method with respect to parameter distribution estimation. Methods The performance of the method, with special emphasis on the analysis of data with non-normal distribution of random effects, was studied using Monte Carlo (MC) simulations. Results The mean value (and ranges) of absolute relative biases (ARBs, %) in parameter distribution estimates with nonparametric methods preceded with FO and FOCE were 0.80 (0.1–3.7) and 0.70 (0–3), respectively, while for parametric methods, these values were 23.74 (3.3–97.5) and 4.38 (0.1–17.9), for FO and FOCE, respectively. The nonparametric estimation method in NONMEM could identify non-normal parameter distributions and correct bias in parameter estimates seen when applying the FO estimation method. Conclusions The method shows promising properties when analyzing different types of pharmacokinetic (PK) data with both the FO and FOCE methods as preceding steps.

Population pharmacokinetics of a HER2 tyrosine kinase inhibitor CP-724,714 in patients with advanced malignant HER2 positive solid tumors

To develop a population pharmacokinetic (popPK) model for CP-724,714, a novel HER2 tyrosine kinase inhibitor is under development for the treatment of advanced HER2 positive cancers. Concentration-time data (n = 484) from 30 cancer patients receiving daily oral CP-724,714 at doses of 250 mg QD, 250 mg BID, 250 mg TID, and 400 mg BID in 21-day cycles in an open label First-in-Human dose-escalation study were evaluated. Serum concentrations of CP-724,714 were obtained after single dose and at steady state. Nonlinear mixed effect analysis in NONMEM using first order conditional estimation with interaction (FOCEI) was performed. The effect of covariates was assessed. Diagnostic plots, decrease of objective function value (>7.8), bootstrapping, and predictive check were used as model selection criteria. A 2-compartment first-order absorption pharmacokinetic (PK) model with inter-subject variability (ISV) on the apparent oral elimination clearance (CL/F), apparent central volume of distribution (V1/F), apparent peripheral volume of distribution (V2/F), and first-order oral absorption rate constant (ka), interoccasion variability (IOV) on CL/F, and body weight (WT) as covariate on CL/F was developed. There was no evidence of dose-dependent and/or time-dependent PK. CL/F increased with increasing WT. The ISV of CL/F was reduced by approximately 20% with WT as a covariate. Age, race, and liver metastasis did not influence CP-724,714 disposition. A popPK model was developed that adequately described the pharmacokinetics of CP-724,714. WT was identified as a covariate on CL/F that reduced ISV and improved model performance. Future studies will further assess the importance of WT as a pharmacokinetic covariate. The proposed popPK model can be used to simulate CP-724,714 Phase 2/3 trials.

A pharmacokinetic-pharmacodynamic model for the mobilization of CD34 hematopoietic progenitor cells by AMD3100

AMD3100 is a small-molecule CXCR4 antagonist that has been shown to induce the mobilization of CD34 + hematopoietic progenitor cells from bone marrow to peripheral blood. AMD3100 has also been shown to augment the mobilization of CD34 + cells in cancer patients when administered in combination with granulocyte colony-stimulating factor (G-CSF) (filgrastim). The purpose of this study was to characterize the exposure-response relationship of AMD3100 in mobilizing CD34 + cells when administered as a single agent in healthy volunteers. AMD3100 concentrations and CD34 + cell counts obtained from 29 healthy subjects in a single-dose, intensively sampled pharmacokinetic/pharmacodynamic (PK-PD) study were analyzed by use of nonlinear mixed effects regression with the software NONMEM. FOCE (first order conditional estimation) with interaction was the estimation method, and simultaneous PK-PD fitting was adopted. The pharmacokinetics of AMD3100 was described by a 2-compartment model with first-order absorption. The population estimates (+/-SE) for clearance and central volume of distribution were 5.17 +/- 0.49 L/h and 16.9 +/- 3.79 L, respectively. CD34 + cell mobilization was best described by an indirect effect model that stimulates the entry process of CD34 + from bone marrow to peripheral blood in the form of a sigmoid maximum effect model. The population estimates (+/-SE) of maximum effect, concentration causing 50% of maximum response, and equilibration time were 12.6 +/- 4.89, 53.6 +/- 11.9 mug/L, and 5.37 +/- 1.31 hours, respectively. This study characterizes the exposure-response relationship of AMD3100 in mobilizing CD34 + cells after subcutaneous administration. This PK-PD model will be useful in assessing relevant covariates and for optimizing the use of AMD3100 in various patient populations.

Statistical Evaluation of Clinical Trial Design for a Population Pharmacokinetic Study —A Case Study—

A population pharmacokinetic substudy design of a new chemical entity was evaluated based on the bias in parameter estimates and the power of detecting a specific subpopulation showing different clearance using a clinical trial simulation approach. The effect of analysis algorithms on type I error was also assessed. The design factors included the number of patients (n=100-300) and the number of sampling points per patient (n=2-6). Simulation data were generated from a model developed based on a Phase I study. The power was evaluated for a percentile of test statistics obtained by the simulation study. The clearance (CL) related parameters were estimated with sufficient accuracy in all study designs and all analysis algorithms: the first order (FO), first order conditional estimation (FOCE) and first order conditional estimation with interaction (FOCE-INTER) methods. With the FO and FOCE methods, the type I error rate increased as the frequency of sampling from each patient became higher, but such increase was hardly observed with the FOCE-INTER method. The power tended to depend on the size of the subpopulation. A large difference was found in the power of detecting a specific subpopulation showing a clearance decrease of 30% or 50%. Therefore, the most dominant factors controlling power would be the size of the subpopulation and the decreasing ratio of CL in the subpopulation. These findings obtained by the clinical trial simulation approach are useful for optimization of study design and determination of the limits of evaluation.