From: Rebecca Wrishko <>
Subject: ADVAN3 TRANS3 Problem
Date: Wed, 03 Nov 1999 13:46:58 -0800

Fellow NONMEM Users,

I am a member of a research group that is just commencing utilization of NONMEM for Population and Bayesian analyses. We are novices with both NONMEM and Digital Visual Fortran, but were able to install NONMEM V (1.1) under the MS-DOS prompt launched under F-90 of DF. Further, the installation was tested using the routine outlined in Appendix 5 (cd run nmfe5 etc.). Also, I successfully duplicated the control3 and data3 in microsoft and excel, saved under the .txt and .prn extensions respectively, and generated the same report as that provided in report3 of the sample files.

We are attempting to model vancomycin and gentamicin pharmacokinetics in general pediatric and neonatal patients. Utilizing data from one subject that had been previously modeled with WinNonlin, I created the control stream and data files for both one-compartment (ADVAN1 TRANS2) two compartment (ADVAN3 TRANS3) analyses. Providing the reparameterization line of K=CL/V was included in the $PK block, NONMEM generated estimates of the CL and V I requested. However, I have experienced much difficulty in generating estimates of CL, V, VSS using the ADVAN3 TRANS3 routine. The most common error message I receive is that V is larger than VSS. Should I first implement the TRANS1 routine to generate estimates of K, K21, and K12, in order that these may then be incorporated into the $pk block to provide the variables for the reparameterization lines (K=CL/V; K12=Q/V; K21=Q/(VSS-V))? The following represents a sample of a $PK and $THETA code that generated the above mentioned error message, I did not include OMEGA and SIGMA codes because the data reflects concentration measurements from one individual. Any suggestions you may have to solve this frustrating dilemma would be greatly appreciated.








Date: Wed, 03 Nov 1999 15:53:49 -0800
From: LSheiner <>
Subject: Re: ADVAN3 TRANS3 Problem

You must write your PK code so that the model is a feasible model. That means, among other things, VSS>V. A simple way to accomplish this is to parameterize as follows:

> $PK
> V=THETA(2)
> VSS= V + THETA(3)
> Q=THETA(4)
> K=CL/V
> K12=Q/V
> K21=Q/(VSS-V)
> (0,4)
> (0,7)
> (0,3)
> (0,4)

1. THETA(3) is constrained >0 so VSS is always >V
2. THETA(4) is not the typical value of VSS; the typical value of VSS is THETA(3)+THETA(4)
3. When you add etas and use FO (METHOD=0) above will still work, but if you want to use FOCE (or other METHOD=1) you will have to assure that the individual values of VSS (i.e., when ETAs are nonzero) are > V, which can be tricky.
4. I removed all upper bounds on parameters (in $THETA). These are almost NEVER indicated. Bounds should be used ONLY to constrain a model to a feasible region, as for example the use of lower bounds of zero on physical quantities that cannot physically be negative.

Lewis B Sheiner, MD Professor: Lab. Med., Biopharm. Sci., Med.
Box 0626 voice: 415 476 1965
UCSF, SF, CA fax: 415 476 2796
94143-0626 email:





From: "Piotrovskij, Vladimir [JanBe]" <>
Subject: RE: ADVAN3 TRANS3 Problem
Date: Thu, 4 Nov 1999 13:16:00 +0100

Don't you think TRANS4 solves all the problems, do you?

Best regard,






Date: Thu, 04 Nov 1999 08:12:10 -0800
From: LSheiner <>
Subject: Re: ADVAN3 TRANS3 Problem

Yes, it does, but the point is a general one: parameterizations & constraints must be chosen to limit the parameter space to a feasible region.

Lewis B Sheiner, MD Professor: Lab. Med., Biopharm. Sci., Med.
Box 0626 voice: 415 476 1965
UCSF, SF, CA fax: 415 476 2796
94143-0626 email:






From: Rebecca Wrishko <>
Subject: ADVAN3 TRANS3 Problem Solved
Date: Thu, 04 Nov 1999 14:04:02 -0800

Thank you to everyone who expeditously responsed to the above mentioned problem. Setting VSS=V+THETA(3) permitted the NONMEM execution to provide estimates of CL, V AND VSS whereby VSS>V. Just for clarification, when VSS=THETA(3) does the routine define the variable as volume of peripheral compartment (VSS-V) rather than generating estimates of VSS? Also, Dr. Sheiner wrote that TVVSS (typical value) is Theta(3) + Theta (4), does this suggest that the TVVSS is the sum of a volume and intercompartmental clearance terms (VSS+Q), or the sum of a volume and a slope (ie. VSS*WT)?

Again, thank you for your assistance.

Rebecca Wrishko
Division of Clinical Pharmacy
Faculty of Pharmaceutical Sciences
University of British Columbia
Vancouver, British Columbia, Canada






Subject: ADVAN3 TRANS3 Problem
Date: Sat, 6 Nov 1999 10:21:46 +0100

Dear colleague,

further to modelling vancomycin and gentamycin pharmacokinetics I would like to draw your attention to the USC-Pack, developed at the University of Southern California by R. Jelliffe, A. Schumitzky and many others. It has got implemented population pharmacokinetic models of these and other drugs from this class, compared to NONMEM, learning to handle the programme takes much less time and it has got an added advantage that cannot be overestimated: it uses the nonparametric NPEM algorithm in the fitting process. Consequently, (often unrealistic) assumptions usually imposed on the distribution of e.g. individual pharmacokinetic parameters are not necessary.

For more detailed information you can contact

Roger W. Jelliffe, M.D. Professor of Medicine, USC
USC Laboratory of Applied Pharmacokinetics
2250 Alcazar St, Los Angeles CA 90033, USA
Phone (323)442-1300, fax (323)442-1302, email=
Web site=

So far, we have made some comparison studies and we are very satisfied with the nonparametric approach, which may be considered at least a valuable additional tool in population pharmacokinetic modelling and clinical trial simulation.

Harry Mager
Bayer AG
Clinical Pharmacology
D-42096 Wuppertal
Phone: 0049-202-368891
E Mail: