From: Ping Ji ping.ji@bms.com
Subject: [NMusers] NONMEM questions
Date: Fri, 18 Mar 2005 11:53:14 -0500
Dear users,
I am new to this community. I am trying to fit multiple dose data from
healthy subjects (TYPE=0) and patients (TYPE=1). I use
two-compartment with zero- and first-order absorption and absorption
lag time in patients. Shown below is the control file. My problem is
that output is always "minimization terminated". Can anyone help me
out to see what is the problem?
> $PROB MAD&POC
> $INPUT C ID AMT RATE TIME DV EVID TYPE
> $DATA full-data-set.csv IGNORE=C
> $SUBROUTINE ADVAN4 TRANS4
> $PK
>
> IF (TYPE.EQ.0) TVCL=THETA(1)
> IF (TYPE.EQ.1) TVCL=THETA(2)
> CL=TVCL*EXP(ETA(1))
>
> D1=THETA(3)
>
> IF(TYPE.EQ.0) TVV2=THETA(4)
> IF (TYPE.EQ.1) TVV2=THETA(5)
> V2=TVV2*EXP(ETA(2))
>
> Q=THETA(6)
>
> V3=THETA(7)
>
> KA=THETA(8)*EXP(ETA(3))
>
> IF(TYPE.EQ.1) ALAG1=THETA(9)
> S2=V2/1000
> S3=V3/1000
>
> $THETA (0,10) (0, 10) (0,1) (0,20) (0,20) (0,4) (0,160) (0,0.3) (0,0.1)
> $OMEGA BLOCK(2)
> 0.1 0.05 0.2
> $OMEGA 0.5
> $ERROR
> IPRE=F
> Y1=F*(1+EPS(1))+EPS(2)
> Y2=F*(1+EPS(1))+EPS(3)
> Q1=0
> Q2=0
> IF (TYPE.EQ.0) Q1=1
> IF (TYPE.EQ.1) Q2=1
> Y=Q1*Y1+Q2*Y2
>
>
> $SIGMA 0.4 10 10
> $ESTIMATION NOABORT SIG=6 MAX=9999 PRINT=10
> $TABLE ID TIME IPRE DV PRED EVID TYPE
NONMEM questions
From: Bachman, William (MYD) bachmanw@iconus.com
Subject: RE: [NMusers] NONMEM questions
Date: Fri, 18 Mar 2005 12:11:55 -0500
See NONMEM Users Guide V Chapter 13
From: Liping Zhang ZHANG_LIPING@lilly.com
Subject: Re: [NMusers] NONMEM questions
Date: Fri, 18 Mar 2005 13:13:00 -0500
since the two population have many different PK parameters (Cl, V, Tlag,
additive error, etc), why don't you fit them separately, establish good
structural models and get the estimation run successfully, then check if
they are similar enough to fit them together and use Type as a covariate?
Doing so will also give you an idea on which parameter to put Type as a
covariate. BTW, I do not see zero absorption in the control file.
best regards, Liping
Liping Zhang, PhD
From: Johan.Rosenborg@astrazeneca.com
Subject: RE: [NMusers] NONMEM questions
Date: Mon, 21 Mar 2005 10:33:56 +0100
Dear Ji Ping,
I agree with Liping that subdividing the problem could be a
good way forward. To start with, however, I think you should
check your data file - have you accounted for the repeat dose
situation? If not add a steady-state (SS) or Additional Dose
(ADDL) and Interdose Interval (II) variables in the data set
(cf. NM User Guide part V, page 57 and c:\NMV\HELP\SSDOSE.DAT
or c:\NMV\HELP\ADDL.DAT).
/ Johan
Johan Rosenborg, Ph.D.
Senior Pharmacokineticist, Clinical Pharmacology
Medical Science Sweden
AstraZeneca R&D Lund
SE-221 87 Lund, Sweden
From: Ping Ji ping.ji@bms.com
Subject: Re: [NMusers] NONMEM questions
Date: Mon, 21 Mar 2005 09:49:04 -0500
Dear all,
Thanks a lot for the quick reply. I truly appreciate it. To start with
MAD or POC data only, and use the following control file for both
dataset, I can get both minimization successfully.
> $PROB MAD
> $INPUT C ID AMT RATE TIME DV EVID
> $DATA full-data-set-MAD.csv IGNORE=C
> $SUBROUTINE ADVAN4 TRANS4
> $PK
>
> CL=THETA(1)*EXP(ETA(1))
> D1=THETA(2)
>
> V2=THETA(3)*EXP(ETA(2))
>
> Q=THETA(4)
>
> V3=THETA(5)
>
> KA=THETA(6)*EXP(ETA(3))
>
>
>
> S2=V2/1000
> S3=V3/1000
>
> $THETA (0,20) (0,2) (0,100) (0,4) (0,40) (0,1.5)
> $OMEGA 0.1 0.1 0.1
>
>
> $ERROR
>
> IPRE=F
> Y=F*(1+EPS(1))+EPS(2)
>
> $SIGMA 0.4 0.5
> $ESTIMATION NOABORT SIG=2 MAX=9999 PRINT=10 METHOD=0
>
> $TABLE ID TIME IPRED DV CL KA D1 V2 V3 Q
For the two dataset combined, the following control file would work.
> $PROB MAD&POC
> $INPUT C ID AMT RATE TIME DV EVID TYPE
> $DATA full-data-set-1.csv IGNORE=C
> $SUBROUTINE ADVAN4 TRANS4
> $PK
>
> IF (TYPE.EQ.0) THEN
> CL=THETA(1)*EXP(ETA(1))
> ELSE
> CL=THETA(2)*EXP(ETA(2))
> ENDIF
>
> D1=THETA(3)
>
> V2=THETA(4)*EXP(ETA(3))
>
> Q=THETA(5)
>
> V3=THETA(6)
>
> IF (TYPE.EQ.0) THEN
> KA=THETA(7)*EXP(ETA(4))
> ELSE
> KA=THETA(8)*EXP(ETA(4))
> ENDIF
>
>
>
> S2=V2/1000
> S3=V3/1000
>
> $THETA (0,20) ;CL1 1
> $THETA (0, 25) ;CL2 2
> $THETA (0,5) ;D1 3
> $THETA (0,200) ;V2 4
> $THETA (0,20) ;Q 5
> $THETA (0,400) ;V3 6
> $THETA (0,1) ;KA1 7
> $THETA (0,0.75) ;KA2 8
>
> $OMEGA 0.3 0.3 0.7 0.7
>
>
> $ERROR
>
> IPRE=F
>
> Y1=F*EXP(EPS(1))+EPS(2)
> Y2=F*EXP(EPS(3))+EPS(4)
> Q1=0
> Q2=0
> IF (TYPE.EQ.0) Q1=1
> IF (TYPE.EQ.1) Q2=1
> Y=Q1*Y1+Q2*Y2
>
>
> $SIGMA 0.4 0.5 1 1
> $ESTIMATION NOABORT NSIG=2 MAX=9999 PRINT=10 METHOD=0
> $TABLE ID TIME IPRE DV
This is initial step for the fitting. Both would require optimization.
I run into difficulties when I move METHOD=0 to 1, can anyone give me
some suggestions on this? Thanks a lot.
Ping Ji, PHD
Clinical Discovery
Pharmaceutical Research Institute
Bristol-Myers Squibb
Princeton, NJ, 08543
_______________________________________________________