XPose 2.0 – An S-Plus Based Population Model Building Aid For NONMEM

Niclas Jonsson and Mats Karlsson

Department of Pharmacy, Uppsala University, Sweden

In population analysis we usually deal with data having a complex structure and with models that involves a large number of assumptions. Because of this the analyses benefit from extensive visualization of both the data and different aspects of the model and the fit. In addition it is often helpful with complementary screening type of analyses to identify possible covariates that can contribute to the explanation of variability in the parameters of the model.

Xpose is a collection of Splus functions to facilitate data set checkout, model diagnostics, data exploration and visualization for non-linear mixed effects models. Data set checkout includes visualization of the observed variable(s), covariates and plots to

reveal errors in the data file. Model diagnostic plots includes the usual residual plots but also plots to check the validity of assumptions specific to non-linear mixed effects models. Data exploration is also done by various plots but also includes auxiliary screening analyses such as generalized additive modeling (GAM) and tree based modelling. The stability of the GAM results with respect to covariate model

selection as well as to the impact of influential individuals and certain types of covariate interactions can be explored using a bootstrap resampling procedure. To facilitate documentation Xpose can also produce a one sheet run summary consisting of a combination of goodness of fit plots, parameter estimates and the model file.

The input to Xpose is one or more of six standard NONMEM table files. To produce the summary output Xpose also uses the NONMEM output and model file.

The changes between Xpose version 2.0 and the previous version (1.1) are numerous. The main user visible difference is that Xpose 2.0 uses the Trellis Graphics library instead of the core graphics of Splus. Another change is that it is no longer strictly necessary with separate table files for, e.g. covariates and parameters. New is also

the possibility to save plots, after optional customization of the plot title and axis labels, to a file suitable for inclusion in other programs such as word processors. There are also a large number of new plots specifically designed to check certain assumptions made in

non-linear mixed effects modeling as well as diagnostics for GAM models, including a new Bootstrap of the GAM algorithm.

To run Xpose it is necessary to have Splus version 3.3 (Windows) or 3.4 (Unix) and, for the run summary, a Fortran compiler. Xpose is free of charge and can be obtained from the authors by email or from the Xpose home page at http://xpose.sourceforge.net/.

Reference: PAGE 6 () Abstr 599 [www.page-meeting.org/?abstract=599]

Poster: oral presentation