Advanced Population Analysis Features in the S-ADAPT/MCPEM Program
Robert J. Bauer
XOMA (US) LLC
Objective: To demonstrate advanced population analysis and simulation abilities of S-ADAPT.
Methods: S-ADAPT is a Fortran 95 open-source, free program distributed by U. of Southern California, Biomedical Simulations Resource department (USC, BMSR), and has been successfully used to analyze clinical data for Raptiva, consisting of six differential equations and 16 model parameters . The S-ADAPT Program provides an environment for performing population analysis of data, with or without covariates, using complex PK/PD models with extensive simulation tools.
The program provides several interface types: Interactive command line, interactive menu, or execution of a series of commands from a script file, allowing complete batch-processed control . The command line allows one to evaluate algebraic expressions, and store the results in user defined variables. The run-time environment also includes a built-in database for storing and retrieving data, and maintaining analysis results.
A population analysis validation system is available that generates simulated data sets, analyzes each data set, and outputs bias and precision statistics. The open-source code may be compiled by Intel or Compaq Visual Fortran for Microsoft Windows, Intel compiler for Linux, and g95 (MingW), a free FORTRAN compiler. The user fills out a FORTRAN model file from a template, providing the code necessary to describe the PK and PD model functions, residual error functions, parameter transformations, covariate models for the population parameters, and differential equations if needed.
Template model files for basic PK models based on the various Advan/Trans algorithms in NONMEM are available, to be used as is, or modifiable by the user. Data may be imported and used in NONMEM format. Nonlinear mixed effects population analysis may be performed at two or three hierarchical stages (incorporating prior information), and maximization methods include iterative two-stage and Monte-Carlo Parametric Expectation-Maximization (MCPEM) methods .
Deterministic and Monte-Carlo algorithms may be simultaneously used to increase efficiency of the analysis while retaining accuracy. Inter-occasion variability, population mixtures, and below quantification limit (BQL) data may also be modeled in S-ADAPT. The population analysis provides standard error analyses, post-hoc analyses, and graphical viewing of post-hoc results [2,3,4]. In addition, two or three-stage hierarchical Bayesian analysis may be performed in S-ADAPT, to provide quantile ranges for the estimation of the population parameters. Extensive capabilities for importing and exporting data and/or analysis results are provided, including easy export of data, dosing information, initial parameters values and prior information to WinBUGS data files. S-ADAPT also has the ability to distribute the computation effort of a single analysis or multiple analyses across several computers to reduce analysis time.
Results: Comparisons of results from S-ADAPT with NONMEM and WinBUGS have been performed, as well as validation analyses using multiple sets of simulated data [3,4,5]. S-ADAPT's performance was very stable, and provided population means, inter-subject variances, and their standard errors with little bias. S-ADAPT tended to perform slowly with simple one and two compartment PK models, but performed more efficiently with more complex PK/PD models involving differential equations.
Conclusions: The S-ADAPT program using MCPEM methods offers a robust and versatile environment for PK/PD modeling and population analysis.
 Ng CM, Joshi A, Dedrick R, Garovoy M, Bauer R. Pharmacokinetic-pharmacodynamic-efficacy analysis of efalizumab in patients with moderate to severe psoriasis. Pharmaceutical Research. 2005;22(7):1088-1100.
 S-ADAPT/MCPEM User's Guide [computer program]. Version 1.52. Berkeley, CA.; 2006. http://bmsr.usc.edu/Software/Adapt/sadapt.html.
 Bauer RJ, Guzy S. Monte Carlo parametric expectation maximization (MC-PEM) method for analyzing population pharmacokinetic/pharmacodynamic data. In: D'Argenio DZ, ed. Advanced Methods of Pharmacokinetic and Pharmacodynamic Systems Analysis. Vol 3. Boston: Kluwer Academic Publishers; 2004:135-163.
 Bauer RJ, Guzy, S, and Ng, C. A survey of population analysis methods for complex pharmacokinetic and pharmacodynamic models with examples. AAPS Journal 2007; 9(1) Article 7, E60-E83.
 Girard P, Mentre F. A comparison of estimation methods in nonlinear mixed effects models using a blind analysis. PAGE Meeting, Pamplona, Spain. 2005. Abstract 834.