From: "David Nix, Pharm D." <>

Subject: Changing number of observations per indivdual record.

Date: Mon, 14 May 2001 15:55:48 -0700


The installation guide (Chapter III, page 20) provides information about increasing the number of observations per individual record. The line "PARAMETER (NO=50)" needs to be changed in CELS, ELS, INITL, INPT, NCONTR, OBJ, OBJ2, OS PRRES, ROOT, and SIML subroutines. However, I was unable to locate the above line in any of the *.for files listed about in the nm subdirectory. Am I looking in the wrong place? I would appreciate any guidance here.


I have seen this discussion going on in the recent past. The data involved administration of the same drug 4 times separated by 2 weeks washout. Results of the population analysis are considerably different, primarily due to the Tlag estimate. Tlag in the population analysis appears quite short and this leads to a low ka and slower CL compared to individual modeling results. The individual fits (from nonlinear regression) look very good - this leads me to think that the population estimates need to be improved.


One issue is that I deleted all "0.0" concentations even if they were before the first detectable concentration. These concentrations could provide additional help for determining Tlag. The problem is the limit on the number of data points per individual. I have up to 18 samples per individual x 4 treatments = 72.


Best estimates for nonlinear regression modeling based on 1 compt, 1st order input and 1st order elim.


Tx ka V ke Tlag CL

1 0.988 (73.1) 1271 (65.9) 0.159 (67.5) 2.342 (64) 201

2 1.071 (124.6) 845 (88.2) 0.133 (79.7) 1.672 (83.6) 112

3 1.915 (116.3) 1273 (51.9) 0.17 (118.8) 2.135 (75) 216

4 0.807 (68.8) 1358 (46.7) 0.174 (179.7) 1.457 (67.2) 236


F values: Tx 1 = 1.00 Tx 2 = 1.80 Tx 3 = 0.93 Tx 4 = 0.85


Estimates from population analysis (same model, param as ka, V, CL, Tlag)


0.363 (47.5) 1470(45.9) ~0.0477 0.494(48.9) 70.1(17.9)


F values: Tx 1 = 1.00 Tx 2 = 2.32 Tx 3 = 1.04 Tx 4 = 0.982




From: "Bachman, William" <>

Subject: RE: Changing number of observations per indivdual record.

Date: Tue, 15 May 2001 08:15:36 -0400




Yes, you are looking the wrong place. To change the no. of records per individual you need to make the changes to the files in the main directory (usually c:\nmv), delete the original subdirectories, and re-compile using setup.bat to make the changes take effect. Complete instructions below:






NONMEM, PREDPP and NMTRAN need to be recompiled for changes to NSIZES, TSIZES and LSIZES files to take affect.


If NONMEM V was installed using the SETUP.BAT installation utility, here is an easy way to make this change.


In the following, it is assumed that the NONMEM V directory is c:\nmv.


If not, use the appropriate drive and directory.


The steps are:


1. Change to c:\nmv and make the changes to NSIZES, TSIZES and LSIZES (the parameter to changes is NO=50 to a larger number, e.g. 150)






2. Delete the previous versions of these files:


del nm\nsizes


del tr\tsizes


del tl\lsizes


3. Delete the previous object files from all subdirectories:


del *.obj /s


4. Delete the previous nonmem library file and NM-TRAN executable:


del nm\nonmem.lib


del tr\nmtran.exe


5. Determine the setup commands used previously to install NONMEM:


type setup.txt


6. Use SETUP.BAT to recompile all routines. (Use here the same arguments that were used originally, if any.) SETUP.BAT will automatically skip all other steps.


setup a c nmv df y


In the above, it is assumed that SETUP.BAT remains in the NONMEM V directory where it was copied during installation. If not, it can be run from diskette as it was originally:


a:\SETUP a c nmv df y



William J. Bachman, Ph.D.

GloboMax LLC

7250 Parkway Dr., Suite 430

Hanover, MD 21076

Voice (410) 782-2212

FAX (410) 712-0737