210 likes | 235 Views
PKfit - A Pharmacokinetic Data Analysis Tool in R. Speaker: Chun-ying Lee 1 Advisor: Yung-jin Lee 2 1 Pharmacy Department, Changhua Christian Hospital Changhua, Taiwan 2 College of Pharmacy, Kaohsiung Medical University, Kaohsiung, Taiwan. Motivation.
E N D
PKfit - A Pharmacokinetic Data Analysis Tool in R Speaker: Chun-ying Lee1 Advisor: Yung-jin Lee2 1Pharmacy Department, Changhua Christian Hospital Changhua, Taiwan 2College of Pharmacy, Kaohsiung Medical University, Kaohsiung, Taiwan
Motivation • Using available packagesin R to create theprogram • Selecting menu-driven mode as the user interface • Studying the application ofgenetic algorithm to PK data analysis • Comparing with WinNonlin and Boomer • WinNonlin http://www.pharsight.com • Boomer http://www.boomer.org
User Interface • Menu-driven mode • Users need not require to be familiar with programming of R • Analyses data step by step
PK models included • Fourteen PK models are currently available (all single-dosed) • Intravenous drug administrations with i.v. bolus or i.v. infusion • Extravascular drug administrations, linear PK with 1st-order absorption/elimination or with nonlinear (Michaelis-Menten) elimination models
Packages required • odesolve (by R. Woodrow Setzer) • lsoda function: solving differential equations • rgenoud (by Walter R. Mebane and Jasjeet S. Sekhon ) • genoud function: Genetic algorithm • stats • nls function: Gauss-Newton algorithm (by Douglas M. Bates and Saikat DebRoy) • optim function: Nelder-Mead simplex method • logLik function: Log-Likelihood (by Jose Pinheiro and Douglas Bates) • AIC function: Akaike’s Information Criterion (by Jose Pinheiro and Douglas Bates) • stats4 • BIC function: Bayesian information criterion
Coding process 7 2020/1/1
Output information • Summary table • Time, observed and calculated concentrations • Weighted residuals • Area under plasma concentration curves (AUC) • Area under the first moment curves (AUMC) • Final values of PK parameters • Model selection criteria • Log-Likelihood (LL) • Akaike’s Information Criterion (AIC) • Schwarz’s Bayesian Criterion (SBC) • Diagnostic plots • Linear plots • Semi-log plots • Residual plots
Validation • All the conditions are set the same in these three software • Data sets • Selected models • Initial values for parameters • Fitting algorithms and numerical integration tools • WinNonlin: Nelder-Mead Simplex method / RKF5 • Boomer: Nelder-Mead Simplex Gauss-Newton / RKF5 • Criteria • Prediction (absolute) error (PE) • Percentage of prediction (absolute) error (%PE)
Comparison of software Model 1: One-Compartment PK Model I.V. Bolus Single-Dose with Linear Elimination Model 2: One-Compartment PK Model I.V. Bolus Single-Dose with Nonlinear Elimination Model 3: One-Compartment PK Model I.V. Infusion Single-Dose with Linear Elimination Model 4: One-Compartment PK Model I.V. Infusion Single-Dose with Nonlinear Elimination : Running O.K. (AE% 5%)
Comparison of software Model 5: One-Compartment PK Model Extravascular Single-Dose with First-Ordered Absorption and Linear Elimination without Lag Time Model 6: One-Compartment PK Model Extravascular Single-Dose with First-Ordered Absorption and Nonlinear Elimination without Lag Time Model 7: One-Compartment PK Model Extravascular Single-Dose with Zero-Ordered Absorption and Linear Elimination without Lag Time Model 8: One-Compartment PK Model Extravascular Single-Dose with Zero-Ordered Absorption and Nonlinear Elimination without Lag Time : Running O.K. (AE% 5%) X: Not acceptable for final PK parameters (AE% > 5%)
Comparison of software Model 9: Two-Compartment PK Model I.V. Bolus Single-Dose with Linear Elimination Model 10: Two-Compartment PK Model I.V. Infusion Single-Dose with Linear Elimination Model 11: Two-Compartment PK Model Extravascular Single-Dose with First-Ordered Absorption and Linear Elimination without Lag Time : Running O.K. (AE% 5%) X: Not acceptable for final PK parameters (AE% > 5%)
Comparison of software Model 12: One-Exponential Term Model 13: Two-Exponential Term Model 14: Three-Exponential Term : Running O.K. (AE% 5%) X: Not acceptable for final PK parameters (AE% > 5%) X: Crashed
Packages required • stats • runif function: random uniform distribution derivates • rnorm function: random normal distribution derivates
Coding process 16 2020/1/1
Output information • Time and concentration • Linear plot • Semi-log plot
Validation • All the simulation results in these three software are very similar with “Error type = No error” with four significant digits A one-compartment PK model with extravascular, single-dose with first-ordered absorption without lag time and linear elimination Dose= 500 (mg) ka= 0.32 (1/hr) kel= 0.11 (1/hr) Vd= 5.8 (L)
Thanks to… • Dr. Woodrow Setzer for odesolve • Dr. Jasjeet Sekhon for rgenoud • Dr. Anthony Rossini for scripting
Reference • R Installation and Administration, Version 2.0.1, 2001. (http://cran.r-project.org/doc/manuals/R-admin.html) • R Language Definition, Version 1.9.1, 2004. (http://cran.r-project.org/doc/manuals/R-lang.html) • Sekhon, J.S. and Mebane, W.R.Jr., Genetic Optimization Using Derivatives: Theory and Application to Nonlinear Models. Political Analysis, 7:187-210, 1998. • Setzer, R.W., The odeslove Package, 2004. • The R Project for Statistical Computing. (http://www.r-project.org) • Writing R Extensions, Version 1.9.1, 2004. (http://cran.r-project.org/doc/manuals/R-exts.html)