Dear NMusers,
I am currently trying to use mixture models on my data because the absorption
rate constant (Ka) appears to follow a bi-modal distribution. But what makes
this complicated is that each subject has two dosing occasions with a washout
period in between, and the absorption tends to vary a lot for the same
individual (the extent of inter occasion variability of Ka is similar between
individual variability). Therefore I would like to allow Ka to vary for the
same subject during different occasions while keeping all the other parameters
the same for a certain subject. I know I could use inter-occasion eta term on
Ka, but I want to allow Ka to have the ability to be assigned to different
subpopulations for the same individual. But in the mixture model structure, the
$MIX is called once per individual and we can only assign each individual to a
subpopulation once.. I wonder if there is a way to assign individual to a
subpopulation every dosing record?
Thank you in advance. I appreciate any comments or suggestions.
Yu
Graduate student in Clinical Pharmaceutical Science
University of Iowa
[mixture models] assign individual to a subpopulation every dosing record?
Yu,
In terms of coding, you can do it using 4 sub-populations:
KA1-KA1
KA1-KA2
KA2-KA1
KA2-KA2
Interpretation would be quite complicated as you would need to assume that each subjects can be in two states, KA1 and KA2, and this state may change from occasion to occasion.
For consistency, you may need to assume that each subject can be in KA1 state with the probability f1, and in KA2 state with the probability f2=1-f1, and then fractions of the four sub-populations defined above would be expressed as
P1=f1*f1
P2=f1*f2
P3=f2*f1
P4=f2*f2
Alternatively, you may want to stay with just one population (same population KA for all subjects) but allow high inter-occasion variability.
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Quoted reply history
On 8/26/2014 1:43 PM, Jiang, Yu wrote:
> Dear NMusers,
>
> I am currently trying to use mixture models on my data because the
> absorption rate constant (Ka) appears to follow a bi-modal distribution.
> But what makes this complicated is that each subject has two dosing
> occasions with a washout period in between, and the absorption tends to
> vary a lot for the same individual (the extent of inter occasion
> variability of Ka is similar between individual variability). Therefore
> I would like to allow Ka to vary for the same subject during different
> occasions while keeping all the other parameters the same for a certain
> subject. I know I could use inter-occasion eta term on Ka, but I want to
> allow Ka to have the ability to be assigned to different subpopulations
> for the same individual. But in the mixture model structure, the $MIX is
> called once per individual and we can only assign each individual to a
> subpopulation once.. I wonder if there is a way to assign individual to
> a subpopulation every dosing record?
>
> Thank you in advance. I appreciate any comments or suggestions.
>
> Yu
>
> Graduate student in Clinical Pharmaceutical Science
>
> University of Iowa
>
> No virus found in this message.
> Checked by AVG - www.avg.com http://www.avg.com
> Version: 2014.0.4716 / Virus Database: 4007/8104 - Release Date: 08/26/14
Dear Lenonid,
Thank you for your suggestion! That is very helpful. But I am not sure how to
implement this in NONMEM. I might be wrong, but from my understanding, the
$MIX block only deals with assigning subpopulation to an individual. But in my
case I need to assign states to each occasion of a given individual first.
Could you maybe be a little more specific on how to do the coding to evaluate
the states from occasion to occasion?
Thanks,
Yu
Graduate student in Clinical Pharmaceutical Science
University of Iowa
Quoted reply history
________________________________________
From: Leonid Gibiansky [[email protected]]
Sent: Tuesday, August 26, 2014 2:14 PM
To: Jiang, Yu; [email protected]
Subject: Re: [NMusers] [mixture models] assign individual to a subpopulation
every dosing record?
Yu,
In terms of coding, you can do it using 4 sub-populations:
KA1-KA1
KA1-KA2
KA2-KA1
KA2-KA2
Interpretation would be quite complicated as you would need to assume
that each subjects can be in two states, KA1 and KA2, and this state may
change from occasion to occasion.
For consistency, you may need to assume that each subject can be in KA1
state with the probability f1, and in KA2 state with the probability
f2=1-f1, and then fractions of the four sub-populations defined above
would be expressed as
P1=f1*f1
P2=f1*f2
P3=f2*f1
P4=f2*f2
Alternatively, you may want to stay with just one population (same
population KA for all subjects) but allow high inter-occasion variability.
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
On 8/26/2014 1:43 PM, Jiang, Yu wrote:
> Dear NMusers,
>
> I am currently trying to use mixture models on my data because the
> absorption rate constant (Ka) appears to follow a bi-modal distribution.
> But what makes this complicated is that each subject has two dosing
> occasions with a washout period in between, and the absorption tends to
> vary a lot for the same individual (the extent of inter occasion
> variability of Ka is similar between individual variability). Therefore
> I would like to allow Ka to vary for the same subject during different
> occasions while keeping all the other parameters the same for a certain
> subject. I know I could use inter-occasion eta term on Ka, but I want to
> allow Ka to have the ability to be assigned to different subpopulations
> for the same individual. But in the mixture model structure, the $MIX is
> called once per individual and we can only assign each individual to a
> subpopulation once.. I wonder if there is a way to assign individual to
> a subpopulation every dosing record?
>
> Thank you in advance. I appreciate any comments or suggestions.
>
> Yu
>
> Graduate student in Clinical Pharmaceutical Science
>
> University of Iowa
>
> No virus found in this message.
> Checked by AVG - www.avg.com http://www.avg.com
> Version: 2014.0.4716 / Virus Database: 4007/8104 - Release Date: 08/26/14
>
This is straightforward, as you would do for 2 populations:
$MIX
FR1 = THETA(1)
FR2 = 1-THETA(1)
P(1)=FR1**2
P(2)=FR1*FR2
P(3)=FR1*FR2
P(4)=FR2**2
$PK
....
; occasion (dose number) should be defined somehow,
; either in the data file, or based on TIME.
; KA1 and KA2 should be expressed via THETAs
MIXN = MIXEST
IF(MIXNUM.EQ.1) TVKA = KA1
IF(MIXNUM.EQ.2.AND.OCC.EQ.1) TVKA = KA1
IF(MIXNUM.EQ.2.AND.OCC.EQ.2) TVKA = KA2
IF(MIXNUM.EQ.3.AND.OCC.EQ.1) TVKA = KA2
IF(MIXNUM.EQ.3.AND.OCC.EQ.2) TVKA = KA1
IF(MIXNUM.EQ.4) TVKA = KA2
KA=TVKA*EXP(ETA(1))
.....
$THETA
(0,0.5,1) ; FR1
.....
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Quoted reply history
On 8/27/2014 12:21 PM, Jiang, Yu wrote:
> Dear Lenonid,
>
> Thank you for your suggestion! That is very helpful. But I am not sure how to
> implement this in NONMEM. I might be wrong, but from my understanding, the
> $MIX block only deals with assigning subpopulation to an individual. But in my
> case I need to assign states to each occasion of a given individual first.
> Could you maybe be a little more specific on how to do the coding to evaluate
> the states from occasion to occasion?
>
> Thanks,
>
> Yu
>
> Graduate student in Clinical Pharmaceutical Science
>
> University of Iowa
> ________________________________________
> From: Leonid Gibiansky [[email protected]]
> Sent: Tuesday, August 26, 2014 2:14 PM
> To: Jiang, Yu; [email protected]
> Subject: Re: [NMusers] [mixture models] assign individual to a subpopulation
> every dosing record?
>
> Yu,
> In terms of coding, you can do it using 4 sub-populations:
>
> KA1-KA1
> KA1-KA2
> KA2-KA1
> KA2-KA2
>
> Interpretation would be quite complicated as you would need to assume
> that each subjects can be in two states, KA1 and KA2, and this state may
> change from occasion to occasion.
>
> For consistency, you may need to assume that each subject can be in KA1
> state with the probability f1, and in KA2 state with the probability
> f2=1-f1, and then fractions of the four sub-populations defined above
> would be expressed as
> P1=f1*f1
> P2=f1*f2
> P3=f2*f1
> P4=f2*f2
>
> Alternatively, you may want to stay with just one population (same
> population KA for all subjects) but allow high inter-occasion variability.
>
> Leonid
>
> --------------------------------------
> Leonid Gibiansky, Ph.D.
> President, QuantPharm LLC
> web: www.quantpharm.com
> e-mail: LGibiansky at quantpharm.com
> tel: (301) 767 5566
>
> On 8/26/2014 1:43 PM, Jiang, Yu wrote:
>
> > Dear NMusers,
> >
> > I am currently trying to use mixture models on my data because the
> > absorption rate constant (Ka) appears to follow a bi-modal distribution.
> > But what makes this complicated is that each subject has two dosing
> > occasions with a washout period in between, and the absorption tends to
> > vary a lot for the same individual (the extent of inter occasion
> > variability of Ka is similar between individual variability). Therefore
> > I would like to allow Ka to vary for the same subject during different
> > occasions while keeping all the other parameters the same for a certain
> > subject. I know I could use inter-occasion eta term on Ka, but I want to
> > allow Ka to have the ability to be assigned to different subpopulations
> > for the same individual. But in the mixture model structure, the $MIX is
> > called once per individual and we can only assign each individual to a
> > subpopulation once.. I wonder if there is a way to assign individual to
> > a subpopulation every dosing record?
> >
> > Thank you in advance. I appreciate any comments or suggestions.
> >
> > Yu
> >
> > Graduate student in Clinical Pharmaceutical Science
> >
> > University of Iowa
> >
> > No virus found in this message.
> > Checked by AVG - www.avg.com http://www.avg.com
> > Version: 2014.0.4716 / Virus Database: 4007/8104 - Release Date: 08/26/14
>
> -----
> No virus found in this message.
> Checked by AVG - www.avg.com
> Version: 2014.0.4716 / Virus Database: 4007/8110 - Release Date: 08/27/14