Pharmacokinetics and Pharmacodynamics - Services
Quantitation of anti-cancer agents and chemopreventive compounds in biospecimens: Pharmacokinetic (PK) evaluation of anticancer agents requires highly sensitive analytical instruments for the quantitation of drugs in biospecimens including blood, plasma, urine, and tissue. Determination of pharmacokinetic parameters such as the area under the concentration-time curve (AUC), half-life, clearance, and volume of distribution require measurement of analytes in the biological sample. Quantitation of analytes in biological samples requires evaluation of the physical and chemical characteristics such as thermal stability, pKa, solubility, spectral characteristics of the analytes, and recovery from the sample with minimum loss. Tissue extraction becomes a significant undertaking if it has not previously been reported. PK/PD personnel have utilized sonication and microwave assisted solvent extraction (MASE) to create a method with quantitative recovery of the analyte as they have already done for analyte extraction from brain tissue. The use of stable isotope labels can validate the extraction recoveries when necessary. The PK/PD shared resource includes all analytical instrumentation required for the quantitation of compounds and their metabolites including high performance liquid chromatography (HPLC), liquid chromatograph-mass spectrometry, atomic absorption spectrophotometry, as well as computers and software for data analyses.
Pharmacokinetic analysis: The drug concentration-time profile data are modeled by either applying parametric compartmental or noncompartmental models to determine PK parameters such as AUC, absorption, and elimination kinetics, t1/2, volume of distribution, and clearance, using WinNonlin software (Pharsight Corporation, Mountain View, CA) and ADAPT-5 software (USC). The PK/PD Shared Resource uses the CINJ Biometrics Shared Resource for the statistical analyses of PK/PD data.
Analytical method development: For analytes (drugs or metabolites) not previously described as being quantified in the literature or those that require a new method, PK/PD specialists will create a method based on their experience and/or adaptation from methods from other analytes with similar properties. They can also generate a theoretical mass spectrum using the Metworks software suite and compare the theoretical to the actual. A process like the one described was used for identification of new tea catachin metabolites in collaboration with Dr. Yang. Development of validated analytical methods for the quantitation of a drug in biological fluids and tissues generally requires a literature search, screening of columns, selection of solvents, identification of interfering compounds in the matrix, and baseline separation of the analyte/metabolites/internal standard in a single run. In 2009, the PK/PD shared resource acquired a triple quad LC-MS (Thermo-Finnigan TSQ Quantum). Similar to HPLC methods, LC-MS method development requires a literature search, setting up the instrument parameters, screening and selection of columns, selection of solvents, identification of interfering compounds in the matrix and the baseline separation of the analyte/metabolites/internal standard in a single run. This also includes the proper ionization conditions and solvent transfer to the appropriate buffer or mobile phase. We have successfully used the instrument for the quantitative analysis of cyclophosphamide and busulfan and plan to use LC-MS for the PK analysis of drugs such as sunitinib and sorafenib.
Population PK/PD: ADAPT-5 and NOMEM. Analyses are performed using NONMEM, Version V level 1.1. The first order conditional estimation (FOCE with INTERACTION) method is used for all analyses. NONMEM provides a mixed effect modeling technique to quantify unexplained inter-individual variability (IIV) and inter-occasion variability (IOV) (random effects) as well as the influence of measured concomitant effects or covariates (fixed effects) on basic (structural) model parameters. The population model can thus be viewed as being comprised of three sub-models: the structural, the statistical and the covariate model. The structural PK/PD sub-model describes the deterministic structure using fixed effect parameters (θ for CL, V, for instance, and other PK/PD parameters). The statistical sub-model accounts for variability using various levels of random effects, IIV, IOV and residual variability. All variability information is characterized by a mean of 0 and a variance (IIV: ω2, IOV: π2 and residual variability: σ2). The differences of the individual PK/PD parameters from the population mean are called η. If parameters show random variation within individuals between study occasions (IOV), this variability is also quantified by using established modeling procedures. Linear relationships are used to relate PK/PD parameters with potential continuous covariates. Categorical covariates are treated using flags for presence or absence of the factor. The next level of random effects called residual variability, which may be the result of assay error, errors in input data and model misspecification, are tested as being additive, proportional (constant CV) or a combination of both. The pharmacokinetic/ pharmacodynamic model structure is optimized with respect to the deterministic structure described by THETAs, to the inter-individual variability (ETAs) and to the residual variability (Epsilons) with the ultimate goal to reduce the unexplained variability. A further modeling success criteria was mechanistic plausibility based upon prior knowledge about the compound. All subsequent statistics and graphics were made using S-PLUS Version 6 or SAS software package (Version 8.2).
Population data is analyzed by ADAPT 5 with or without covariates, incorporating the PK/PD model. ADAPT 5 is a Fortran open-source, free program developed and supported by the Biomedical Simulations Resource at the University of Southern California. The population PK/PD model can be summarized as a hierarchical modeling framework. It includes a PK/PD structural model and a parameter model, which reflects the intra-individual and inter-individual variation, respectively. Parametric maximum likelihood via the expectation maximization algorithm (MLEM) will be used for parameter estimation. Given the initial guesses for the population mean, inter-individual variance and individual parameters, the MLEM algorithm proceeds in two steps. In step 1, the parameters of each individual are estimated. In step 2, the population mean, covariance and inter-individual variance are updated. The two steps are iterated until convergence. Model evaluation is performed by checking the goodness-of-fit plots, covariance matrix and predictive simulation. Further statistical hypothesis testing will be employed as criteria for model selection. All subsequent statistics and graphics will be made by R project (Version 2.10.1) or SAS software package (Version 8.2).