Introduction to NONMEM 7

Date: Thursday, October 8, 2009

Registration fee: Non-student ($500), Student/fellow ($50)

Maximum class size: 75

Sponsor: ICON Development Solutions

Instructors/Teaching Assistants: Bob Bauer and Bill Bachman

Workshop summary:

This one day workshop will cover the description and use of new features in NONMEM 7. The NONMEM 7 software has been significantly upgraded from NONMEM VI to meet the demands of population PK/PD modeling. The classical NONMEM algorithm first order conditional estimation method (FOCE) has been improved by reducing the occurrence of computational problems that result in abnormal termination.

Workshop attendees will be instructed how to specify gradient precision, which improves the efficiency of optimization and increases the incidence of successful completion of the problem. Workshop attendees will also be instructed on how to use the new estimation methods, such as iterative two stage (ITS), importance sampling expectation maximization (EM), Markov chain Monte Carlo (MCMC) stochastic approximation EM (SAEM), and three hierarchical stage MCMC Bayesian method using Gibbs and Metropolis-Hastings algorithms. All set-up parameters for these new methods may be specified in the standard NMTRAN control stream file format. Demonstrations will show that NONMEM 7 has the ability to handle more data file items, longer labels, and initial parameters may be expressed in any numerical format. Output files that are readily transferred to post- processing software are also produced, and the number of significant digits reported may be specified by the user. Attendees will also learn how to obtain diagnostic results such as inter-subject and residual variance shrinkage, conditional weighted residuals, Monte Carlo assessed exact weighted residuals, and normalized probability distribution errors.

The features of PDx-POP 4.0, the companion software to NONMEM 7, will also be demonstrated, which provides real-time graphical monitoring of the objective function, and various toggle switches for console printing of iterations, ending a problem gracefully, and ending a NONMEM run gracefully. In addition, PDx-POP 4.0 provides extended summary output, graphical displays of parameter sampling histories (for Bayes analysis), setting up and running multiple analysis chains simultaneously, graphical and tabular summary of multiple analysis chains, and testing problem stability using multiple sets of initial estimates.

Schedule:

9:00 am - Welcome and Introduction, Bob Bauer

9:15 am - Modifications and Enhancements to NONMEM7, Bill Bachman

  • Conversion of Fortran 77 to Forttran 90/95
  • Centralized Error Processing
  • Improvements in Gradient Methods
  • Decreased Incidence of Estimation Failure Due to Numerical Problems
  • Added Option to Specify Step-Size for Gradient Calculation
  • Increased Number of Data Items and Label Lengths
  • Flexible Numerical Formats for Input and Output
  • Added Information in Standard Results File
    • Identifier Tags for Certain Sections
    • Shrinkage of Variance
  • Additional Output Files Easily Readable by Post-Processing Software
  • Additional Weighted Residuals outputs (conditional, exact versions)

10:00 am - New Methods Available in NONMEM7, Overview, Bob Bauer

  • Monte Carlo Importance Sampling Expectation Maximization (EM) (IMP)
  • Markov Chain Monte Carlo (MCMC) Stochastic Approximation EM (SAEM)
  • Iterative Two Stage (ITS)
  • Model Modifications That Improve Efficiency of EM Methods (Mu Modeling)

10:45 am - Coffee Break

11:00 am - Hands-on Examples for EM Methods

  • Basic two compartment model problem, incorporate Mu Model
  • Two compartment model with age and gender covariates

12:00 pm - Lunch

1:00 pm - Hands-on Examples for EM Methods (continued...)

  • Population mixture model problem
  • Receptor mediated clearance model

2:00 pm - New Methods (continued...), Robert Bauer, PhD

  • MCMC Bayesian Analysis (BAYES)
  • Additional control switches for this analysis (Gibbs vs. Metropolis-Hastings)
  • Revisit two compartment model, adding Bayesian analysis

3:00 pm - Creating Random Initial Parameters for Multiple Chains (CHAIN)

  • Chain command syntax
    • Thetas: univariate or normal randomization
    • Sigmas: univariate randomization
    • Omegas: Wishart randomization
    • Using random samples for immediate, or later problems

3:30 pm - PDx-Pop Interface For NONMEM7, Bill Bachman

  • Real-Time Graphical Monitoring of Objective Function
  • Interaction with NONMEM Run
    • Toggle Switch for Console Printing of Iterations
    • Switch to End a Problem Gracefully
    • Switch to End a NONMEM Gracefully
  • Extended Summary Output
  • Graphical Display of Parameter Sampling History (BAYES)
  • Setting Up and Running Multiple Analysis Chains Simultaneously
  • Graphical and Tabular Summary of Multiple Analysis Chains (BAYES)
  • Initial Parameters Variation Test
  • PDx-Pop on Linux and MAC OS X

5:00 pm - Question and Answer, Demonstration