Covariate Model Building and Evaluation in NONMEM
(A NONMEM Workshop for intermediate and advanced users)
The Uppsala Pharmacometrics Group and is happy to announce a NONMEM Workshop on Covariate Model Building and Evaluation tailored for Intermediate and Advanced NONMEM users. The course will present modeling strategies, techniques and implementations for the handling of covariate information in population models. Traditional methods will be complemented with recent advances in covariate model building methodology and diagnostics for more efficient, robust and informed covariate models. Such new functionality particularly related to covariate modeling has been introduced in recent updates to PsN and Xpose4. Advantages and disadvantages of the different covariate model building methods will be presented and the relationship between modeling aim, data and suitable method discussed. Methods for handling missing covariates, Special considerations for handling time-varying covariates and covariates measured with error will be presented.
The course will consist of both lectures and hands-on computer exercises using NONMEM 7, PsN and Xpose 4. Self-instructive material is provided for further studies and exploration of additional examples within the field. The participants will be provided with a NONMEM library of examples.
The Uppsala group has been involved in developing pharmacometric models and methodology for over twenty years and has extensive experience in development of covariate models and covariate model building methodology. In parallel, model building procedures and informative diagnostics for covariate model building and covariate model evaluation have been developed, evaluated and integrated into the user-friendly software Perl-speaks-NONMEM (PsN) and Xpose 4 .
Course Outline (2-3 days)
The course will last for 2.5 days and consists of both lectures and hands-on computer exercises. Participants will use NONMEM VI, PsN and Xpose 4. PsN and Xpose 4 are freeware and may be installed on participants’ computers for use after the course.
Day 1 -
Time: 8:30 AM to 5:30 PM
- Overview of covariate types
- Motivations for covariate modeling
- Exploratory analysis of covariate distributions and relation to individual parameters
- Introduction to covariate model building and diagnostics using PsN and Xpose
- Parameterization of covariate relationships
- Hypothesis testing and diagnostics for individual covariate-parameter relationships
- Overview of covariate model building methods
- Standard stepwise covariate model building
- Hands-on NONMEM 7, PsN and Xpose4: Exploratory analyses (GAM); Parameterization and individual parameter-covariate relation diagnostics.
Day 2 -
Time: 8:30 AM to 5:30 PM
- Recent advances in stepwise covariate modeling (linearization, cross-validation, bootstrap scm)
- Full covariate modeling approaches
- Other modeling approaches (Lasso, WAM, Genetic algorithms)
- Time-varying covariates and error in covariate measurements
- Missing covariate information
- Evaluation and use of final covariate models
- Hands-on computer exercises with NONMEM 7, PsN and Xpose4. Stepwise covariate modeling, Full model approach/WAM/Lasso, Models for time-varying covariates. Handling of missing covariate information. Evaluation of covariate models.
A prerequisite for the course is basic experience with performing NONMEM analyses.
The course will include hands-on training, with the participants working on their own computers. All programs will run from a USB memory-stick and participants will not be required to install any programs on their computer. NOTE: All participants must bring their own Windows laptops. PsN and Xpose 4 are freeware and may be installed on participants’ computers for use after the course.
Prof. Mats Karlsson