From: "Tzeng, Tsang-Bin" Tsang-Bin.Tzeng@astrazeneca.com
Subject: [NMusers] NM mixture model with three subgroups
Date: Sun, 18 Dec 2005 09:22:47 -0500

Anyone has experience to extend the NONMEM mixture model in clearance from a
two-subgroup (bimodality, as shown below) one to a three-subgroup one. We
are currently working on a drug with polymorphism in metabolism (a primary
elimination pathway) and three subgroups (ultrafast, extensive, and slow
metabolism) have been identified either from previous work or from the
histogram plot of the current data. Thanks!

; MIXTURE MODEL
 EST = MIXEST
 POP1 = 0
 POP2 = 0
 IF (MIXNUM.EQ.1) THEN
 POP1 = 1
 ELSE
 POP2 = 1
 ENDIF
 CL = POP1*CL1 +POP2*CL2

$MIX
 P(1) =THETA(8)
 P(2) =1-THETA(8)
 NSPOP = 2


Sincerely yours,

Tsang-Bin Tzeng, Ph.D.
AstraZeneca Pharmaceuticals
(302) 885-9954
_______________________________________________________

From: "Riley, Steve" Steve.Riley@pfizer.com
Subject: RE: [NMusers] NM mixture model with three subgroups
Date: Mon, 19 Dec 2005 09:23:48 -0500
Tsang-Bin,

I would code the problem as follows:

; MIXTURE MODEL
 EST = MIXEST

 IF (MIXNUM.EQ.1) THEN
CL = CL1

 ELSE IF (MIXNUM.EQ.2) THEN
CL = CL2

 ELSE
CL = CL3
 ENDIF

$MIX
 P(1) = THETA(8)/(1 + THETA(8) + THETA(9))
 P(2) = THETA(9)/(1 + THETA(8) + THETA(9))
 P(3) = 1/(1 + THETA(8) + THETA(9)) 
 NSPOP = 3

Steve
_______________________________________________________

From: "Tzeng, Tsang-Bin" Tsang-Bin.Tzeng@astrazeneca.com
Subject: RE: [NMusers] NM mixture model with three subgroups
Date: Mon, 19 Dec 2005 09:54:36 -0500

 Dear Joern, Sam, Mouksassi, and Steve:

Thanks very much for these wonderful helps and tips! They are more than what
we originally anticipated. I shall share with you if we are lucky enough to
come out with something from this modeling exercise using your suggestions. 

Sincerely yours,

Tsang-Bin Tzeng, Ph.D.
AstraZeneca Pharmaceuticals
(302) 885-9954
_______________________________________________________

From: brendan.m.johnson@gsk.com
Subject: Re: [NMusers] NM mixture model with three subgroups
Date: Mon, 19 Dec 2005 11:27:15 -0500

Hi Tsang-Bin,
I recently had need for a 3 subpopulation model, and used the following

$MIX
;TRIMODAL:
 NSPOP=3          
EXTH10=EXP(THETA(10))
LGT1=EXTH10/(1+EXTH10)
 P(1)=LGT1
EXTH11=EXP(THETA(11))
LGT2=EXTH11/(1+EXTH11)
 P(2)=LGT2*(1-P(1))
 P(3)=1-P(1)-P(2)

you can then apply your conditional statements in a similar fashion to the 2 population model
Brendan
_______________________________________
Brendan Johnson, Ph.D.
Clinical Pharmacokinetics/Modeling & Simulation
GlaxoSmithKline

Ph: +1 919 483 0672 
_______________________________________________________