```
From: "Annick Rousseau" annick.rousseau@unilim.fr
Subject: [NMusers] QUESTION ABOUT MIXTURE MODEL
Date: Wed, 20 Jul 2005 15:48:51 +0200

Dear NMUsers

We are trying to model pharmacokinetic profiles after oral administration. Some profiles
(nearly 50%) exhibit two concentration peaks while others show only a single peak. So we
have tested a mixture model in order to describe the profiles either with a simple model
(with first order absorption and lagtime) or a double first-order absorption model. It is
noteworthy that when we tested independently these models, both minimize successfully
with the same  obj function value.

The control file we are using is:

\$SUBROUTINES TOL=4 ADVAN6

\$MODEL

COMP=(1); DOSE, 1

COMP=(2); DOSE, 2

COMP=(3);GA

COMP=(4); CENTRAL

COMP=(5);PERIPH

\$PK

EST=MIXEST

TTT=0

PP=2

IF (MIXNUM .EQ.1) THEN

PP=1

K13=THETA(1)*EXP(ETA(1))

K23=THETA(2)*EXP(ETA(2))

CL=THETA(3)*EXP(ETA(3))

K14=0

K34=THETA(9)*EXP(ETA(9)) ;GA

F1=THETA(10)*EXP(ETA(10))

ALAG2=THETA(7)*EXP(ETA(7))

K45=0

K54=0

F2=1-F1

ELSE

F1=1

K13=0

K23=0

K14=THETA(4)*EXP(ETA(4))

CL=THETA(5)*EXP(ETA(5))

ALAG2=0

K34=0

K45=THETA(12)*EXP(ETA(11))

K54=THETA(13)*EXP(ETA(12))

ENDIF

ALAG1=THETA(6)*EXP(ETA(6))

V4=THETA(8)*EXP(ETA(8))

K40=CL/V4

S4=V4

IF (TIME .GT. THETA(7)) TTT=1

\$MIX

NSPOP=2

P(1)=THETA(11)

P(2)=1-P(1)

\$DES

DADT(1)=-K13*A(1)-K14*A(1)

DADT(2)=-K23*A(2)

DADT(3)=K13*A(1)+TTT*K23*A(2)-K34*A(3)

DADT(4)=K34*A(3)-A(4)*K40+K14*A(1)-K45*A(4)+K54*A(5)

DADT(5)=K45*A(4)-K54*A(5)

\$ERROR

...

\$THETA

....

\$OMEGA

......

\$SIGMA

......

We obtain something we did not explain: (i) the following message :

0MINIMIZATION SUCCESSFUL

NO. OF FUNCTION EVALUATIONS USED: 2053

NO. OF SIG. DIGITS IN FINAL EST.:  3.3

(ii) no estimation of neither pharmacokinetic population parameters nor variability. The run
minimizes successfully at a lower obj function value (-300) than with
the two tested models but unable to run estimation.

We have tried several parameter initial values without success.

Any help would be appreciated.

Regards

Aurelie Premaud

Annick Rousseau

Service de Pharmacologie et Toxicologie
CHU Dupuytren
2 Rue Martin Luther King
87042 Limoges
tel +33 (0) 5 55 05 61 40
fax +33)(0) 5 55 05 61 62

_______________________________________________________

From: Sam Liao sliao@pharmaxresearch.com
Subject: Re: [NMusers] QUESTION ABOUT MIXTURE MODEL
Date: Wed, 20 Jul 2005 10:52:29 -0400

Hi Annick:
There is a bug fix related to the output of mixture model. You problem
may be related to this bug. You can find the bug fix no.III when you go
to globomax' nonmem web site. http://www.globomax.com/nonmem.htm. It is
at the bottom of the page.

Good luck,

Sam Liao
Pharmax Research
_______________________________________________________

```