Dear all,
I want to build a covariate model for the pk of a drug which is metabolized in
the liver. One of the covariates is liver GVHD which has 5 levels (
0,1,2,3,4). The only way I can think of is to group liver GVHD into two
categories, 0 ( not GVHD) and 1 ( having GVHD).
But there is interest to evaluate the effect of degree of GVHD on pk
parameter, e.g. CL.
Does anyone have any suggestions on building a categorical covariate model of
more than two levels. Thanks for any input.
Best regards,
Hong Li, Ph.D.
Postdoctoral Associate
Department of Pharmaceutical Science
University at Buffalo, SUNY
363 Hochstetter Hall
Buffalo, NY 14260
Tel#: (716)645-4827
categorical covariate with 5 levels
Dear Hong,
I usually use below script for covariates with more than two (n) levels.
Please create n-1 indicator variables.
;GVHD1 = 1 for GVHD=1, 0 otherwise
GVHD1=0
IF (GVHD.EQ.1) GVHD1=1
;GVHD2 = 1 for GVHD=2, 0 otherwise
GVHD2=0
IF (GVHD.EQ.2) GVHD2=1
;GVHD3 = 1 for GVHD=3, 0 otherwise
GVHD3=0
IF (GVHD.EQ.3) GVHD3=1
;GVHD4 = 1 for GVHD=4, 0 otherwise
GVHD4=0
IF (GVHD.EQ.4) GVHD4=1
CL = THETA(A) + THETA(B)*GVHD1 + THETA(C)*GVHD2 + THETA(D)*GVHD3+
THETA(E)*GVHD4 (Additive effect of GVHD on CL)
Or
CL = THETA(A) * (1+THETA(B)*GVHD1) * (1+THETA(C)*GVHD2) * (1+THETA(D)*GVHD3) *
(1+THETA(E)*GVHD4) (Proportional shift in CL for GVHD)
Regards,
Yoshi
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
Yoshimasa Ishida
Chugai Pharmaceutical CO., LTD.
Clinical PK/PD Analysis Group
Clinical Research Planning Dept.
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Hong Li
Sent: Friday, July 09, 2010 1:01 PM
To: [email protected]
Subject: [NMusers] categorical covariate with 5 levels
Dear all,
I want to build a covariate model for the pk of a drug which is metabolized in
the liver. One of the covariates is liver GVHD which has 5 levels (
0,1,2,3,4). The only way I can think of is to group liver GVHD into two
categories, 0 ( not GVHD) and 1 ( having GVHD).
But there is interest to evaluate the effect of degree of GVHD on pk
parameter, e.g. CL.
Does anyone have any suggestions on building a categorical covariate model of
more than two levels. Thanks for any input.
Best regards,
Hong Li, Ph.D.
Postdoctoral Associate
Department of Pharmaceutical Science
University at Buffalo, SUNY
363 Hochstetter Hall
Buffalo, NY 14260
Tel#: (716)645-4827
I have a four levels covariate model. Hope this is useful to you.
Sam Liao
Pharmax Research
==========================================================
RA1=0
RA2=0
RA3=0
IF(RACE.EQ.7) RA1=1
IF(RACE.EQ.8) RA2=1
IF(RACE.GT.8) RA3=1
TVCL= TVCL *(1 + THETA(33)*RA1+ THETA(34)*RA2+ THETA(35)*RA3)
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Hong Li
Sent: Thursday, July 08, 2010 9:01 PM
To: [email protected]
Subject: [NMusers] categorical covariate with 5 levels
Dear all,
I want to build a covariate model for the pk of a drug which is metabolized in
the liver. One of the covariates is liver GVHD which has 5 levels (
0,1,2,3,4). The only way I can think of is to group liver GVHD into two
categories, 0 ( not GVHD) and 1 ( having GVHD).
But there is interest to evaluate the effect of degree of GVHD on pk
parameter, e.g. CL.
Does anyone have any suggestions on building a categorical covariate model of
more than two levels. Thanks for any input.
Best regards,
Hong Li, Ph.D.
Postdoctoral Associate
Department of Pharmaceutical Science
University at Buffalo, SUNY
363 Hochstetter Hall
Buffalo, NY 14260
Tel#: (716)645-4827
Hi Hong,
Other approaches I can think of are:
1.to plot the EBE of CL vs. different GVHD categories to determine
whether you need a different parameter for each level or indeed you can
(and how to) combine several groups together;
2.since your GVHD variable is an ordinal categorical variable instead of
nominal groups like male and female, you may be able to introduce some
functional form (i.e. treating the covariate as continuous)
Assigning each category with a different parameter is a very intuitive
approach, but the underlying assumption is that each group is different
and no clear pattern exist for the relationship between CL and GVHD.
This might be over-parameterized.
Hope this helps,
Yaming
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of Sam Liao
Sent: Friday, July 09, 2010 1:01 AM
To: [email protected]; [email protected]
Subject: RE: [NMusers] categorical covariate with 5 levels
I have a four levels covariate model. Hope this is useful to you.
Sam Liao
Pharmax Research
==========================================================
RA1=0
RA2=0
RA3=0
IF(RACE.EQ.7) RA1=1
IF(RACE.EQ.8) RA2=1
IF(RACE.GT.8) RA3=1
TVCL= TVCL *(1 + THETA(33)*RA1+ THETA(34)*RA2+ THETA(35)*RA3)
-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of Hong Li
Sent: Thursday, July 08, 2010 9:01 PM
To: [email protected]
Subject: [NMusers] categorical covariate with 5 levels
Dear all,
I want to build a covariate model for the pk of a drug which is
metabolized in the liver. One of the covariates is liver GVHD which
has 5 levels ( 0,1,2,3,4). The only way I can think of is to group
liver GVHD into two categories, 0 ( not GVHD) and 1 ( having GVHD).
But there is interest to evaluate the effect of degree of GVHD on pk
parameter, e.g. CL.
Does anyone have any suggestions on building a categorical covariate
model of more than two levels. Thanks for any input.
Best regards,
Hong Li, Ph.D.
Postdoctoral Associate
Department of Pharmaceutical Science
University at Buffalo, SUNY
363 Hochstetter Hall
Buffalo, NY 14260
Tel#: (716)645-4827
Notice: This e-mail message, together with any attachments, contains
information of Merck & Co., Inc. (One Merck Drive, Whitehouse Station,
New Jersey, USA 08889), and/or its affiliates Direct contact information
for affiliates is available at
http://www.merck.com/contact/contacts.html) that may be confidential,
proprietary copyrighted and/or legally privileged. It is intended solely
for the use of the individual or entity named on this message. If you are
not the intended recipient, and have received this message in error,
please notify us immediately by reply e-mail and then delete it from
your system.
Another consideration is that GVHD is likely to have a rank-ordered effect on
clearance (though I agree with the earlier suggestion that 5 levels may be
over-parameterized).
You could try making a cumulative sum of the effects on clearance extending on
the samples from Sam and Yoshi. This is essentially a piecewise-linear,
monotonic function. This is not going so far as having a continuous function,
but it is intuitive and would probably simplify identifying which parameters
you really need. Once you plot out this empirical function, a parsimonious,
continuous relationship may present itself to you.
Jeff
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Hang, Yaming
Sent: Friday, July 09, 2010 10:00 AM
To: Sam Liao; [email protected]; [email protected]
Subject: RE: [NMusers] categorical covariate with 5 levels
Hi Hong,
Other approaches I can think of are:
1.to plot the EBE of CL vs. different GVHD categories to determine
whether you need a different parameter for each level or indeed you can
(and how to) combine several groups together;
2.since your GVHD variable is an ordinal categorical variable instead of
nominal groups like male and female, you may be able to introduce some
functional form (i.e. treating the covariate as continuous)
Assigning each category with a different parameter is a very intuitive
approach, but the underlying assumption is that each group is different
and no clear pattern exist for the relationship between CL and GVHD.
This might be over-parameterized.
Hope this helps,
Yaming
-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of Sam Liao
Sent: Friday, July 09, 2010 1:01 AM
To: [email protected]; [email protected]
Subject: RE: [NMusers] categorical covariate with 5 levels
I have a four levels covariate model. Hope this is useful to you.
Sam Liao
Pharmax Research
==========================================================
RA1=0
RA2=0
RA3=0
IF(RACE.EQ.7) RA1=1
IF(RACE.EQ.8) RA2=1
IF(RACE.GT.8) RA3=1
TVCL= TVCL *(1 + THETA(33)*RA1+ THETA(34)*RA2+ THETA(35)*RA3)
-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of Hong Li
Sent: Thursday, July 08, 2010 9:01 PM
To: [email protected]
Subject: [NMusers] categorical covariate with 5 levels
Dear all,
I want to build a covariate model for the pk of a drug which is
metabolized in the liver. One of the covariates is liver GVHD which
has 5 levels ( 0,1,2,3,4). The only way I can think of is to group
liver GVHD into two categories, 0 ( not GVHD) and 1 ( having GVHD).
But there is interest to evaluate the effect of degree of GVHD on pk
parameter, e.g. CL.
Does anyone have any suggestions on building a categorical covariate
model of more than two levels. Thanks for any input.
Best regards,
Hong Li, Ph.D.
Postdoctoral Associate
Department of Pharmaceutical Science
University at Buffalo, SUNY
363 Hochstetter Hall
Buffalo, NY 14260
Tel#: (716)645-4827
Notice: This e-mail message, together with any attachments, contains
information of Merck & Co., Inc. (One Merck Drive, Whitehouse Station,
New Jersey, USA 08889), and/or its affiliates Direct contact information
for affiliates is available at
http://www.merck.com/contact/contacts.html) that may be confidential,
proprietary copyrighted and/or legally privileged. It is intended solely
for the use of the individual or entity named on this message. If you are
not the intended recipient, and have received this message in error,
please notify us immediately by reply e-mail and then delete it from
your system.