From: Ulrika Simonsson <*ulrika.simonsson*>

Date: Fri, 5 Feb 2010 21:46:12 +0100

Dear all,

My initial comment was not regarding the parameterization as such but =

just that there is still a relationship between CL and half-lives even =

for more complex models than a 1-comp model.

My suggesting of how to approach the data described, although we do not =

know all the details, would be to explore graphically if the WT =

distribution is the same for the two groups. If yes, the same =

relationship for the two groups most likely work. If not, separate WT =

relations for the two groups might be needed. If indicated, I would look =

into nonlinear relationships for CL and WT.

I would then fit all data simultaneously and investigate whether there =

is a statistical difference between the two groups.

As an example, assuming the same linear WT-CL relationship for the two =

groups, I would parameterize TVCL for a Weight and group effect such as:

GP=1 ; for group A

IF(GRP.EQ.1) GP=(1+TH3) ; for group B

TVCL= TH1*(1+TH2*(WT-medianWT))*GP

Typical half-life for the two groups can then be derived based on the =

full expression above of TVCL. I think it is better to test within the =

model whether there is any group effect in any primary PK parameters =

rather than test outside NONMEM based on secondary parameters (half-life =

derived from CL and V) and individual posthoc Bayes estimates.

Best regards,

Ulrika

Ulrika Simonsson, PhD

Assoc Prof of Pharmacometrics

Uppsala Pharmacometrics

Department of Pharmaceutical Biosciences

Uppsala University

BMC, Box 591, 751 24 Uppsala

Sweden

From: owner-nmusers

On Behalf Of Ken Kowalski

Sent: den 5 februari 2010 20:02

To: 'Serge Guzy'; 'Ulrika Simonsson'; varsham

nmusers

Subject: RE: [NMusers] Dilemma with PPK parameters

Hi All,

I think we are getting off the point. What Varsha was asking is how can =

the CL reported as L/kg/hr be so different between the two treatment =

groups while the volume of distribution and half-life could be so =

similar when fitting a one compartment model? I think the answer to =

this question is in the parameterization of CL and Vd with WT. Varsha =

parameterized CL and Vd as

TVCL = THETA(1) + THETA (3) * WT

TVVD = THETA(2) + THETA (4) * WT

and the estimate of CL reported in units of L/kg/hr was obtained from =

the estimate of THETA(3). However, if the WT relationship is not =

proportional to WT for the control group, and is relatively invariant =

over the range of WT, then fitting this linear model in WT could give =

rise to a large (non-zero) value of THETA(1) and a small value of =

THETA(3). Comparing THETA(3) between treatment groups as the difference =

in CL only makes sense if THETA(1)=0. Thus, with the above model, we =

should not interpet THETA(3) as the total CL in L/kg/hr unless =

THETA(1)=0.

The typical individual halflife is

K=TVCL/TVVD

THALF = LN(2)/K

which involves THETA(1) through THETA(4) whereas what Varsha was =

reporting as a difference in CL were the estimates of THETA(3) from =

fitting each treatment separately. So one could obtain different =

estimates of THETA(1) and THETA(3) between the two treatment groups =

while obtaining similar estimates of TVCL for a given value of WT. In =

other words, it is possible that TVCL and TVVD are similar between the =

two treatment groups (which would lead to similar halflives) and yet =

different estimates of THETA(3) for the two treatment groups.

Best regards,

Ken

Kenneth G. Kowalski

President & CEO

A2PG - Ann Arbor Pharmacometrics Group, Inc.

110 E. Miller Ave., Garden Suite

Ann Arbor, MI 48104

Work: 734-274-8255

Cell: 248-207-5082

Fax: 734-913-0230

ken.kowalski

Received on Fri Feb 05 2010 - 15:46:12 EST

Date: Fri, 5 Feb 2010 21:46:12 +0100

Dear all,

My initial comment was not regarding the parameterization as such but =

just that there is still a relationship between CL and half-lives even =

for more complex models than a 1-comp model.

My suggesting of how to approach the data described, although we do not =

know all the details, would be to explore graphically if the WT =

distribution is the same for the two groups. If yes, the same =

relationship for the two groups most likely work. If not, separate WT =

relations for the two groups might be needed. If indicated, I would look =

into nonlinear relationships for CL and WT.

I would then fit all data simultaneously and investigate whether there =

is a statistical difference between the two groups.

As an example, assuming the same linear WT-CL relationship for the two =

groups, I would parameterize TVCL for a Weight and group effect such as:

GP=1 ; for group A

IF(GRP.EQ.1) GP=(1+TH3) ; for group B

TVCL= TH1*(1+TH2*(WT-medianWT))*GP

Typical half-life for the two groups can then be derived based on the =

full expression above of TVCL. I think it is better to test within the =

model whether there is any group effect in any primary PK parameters =

rather than test outside NONMEM based on secondary parameters (half-life =

derived from CL and V) and individual posthoc Bayes estimates.

Best regards,

Ulrika

Ulrika Simonsson, PhD

Assoc Prof of Pharmacometrics

Uppsala Pharmacometrics

Department of Pharmaceutical Biosciences

Uppsala University

BMC, Box 591, 751 24 Uppsala

Sweden

From: owner-nmusers

On Behalf Of Ken Kowalski

Sent: den 5 februari 2010 20:02

To: 'Serge Guzy'; 'Ulrika Simonsson'; varsham

nmusers

Subject: RE: [NMusers] Dilemma with PPK parameters

Hi All,

I think we are getting off the point. What Varsha was asking is how can =

the CL reported as L/kg/hr be so different between the two treatment =

groups while the volume of distribution and half-life could be so =

similar when fitting a one compartment model? I think the answer to =

this question is in the parameterization of CL and Vd with WT. Varsha =

parameterized CL and Vd as

TVCL = THETA(1) + THETA (3) * WT

TVVD = THETA(2) + THETA (4) * WT

and the estimate of CL reported in units of L/kg/hr was obtained from =

the estimate of THETA(3). However, if the WT relationship is not =

proportional to WT for the control group, and is relatively invariant =

over the range of WT, then fitting this linear model in WT could give =

rise to a large (non-zero) value of THETA(1) and a small value of =

THETA(3). Comparing THETA(3) between treatment groups as the difference =

in CL only makes sense if THETA(1)=0. Thus, with the above model, we =

should not interpet THETA(3) as the total CL in L/kg/hr unless =

THETA(1)=0.

The typical individual halflife is

K=TVCL/TVVD

THALF = LN(2)/K

which involves THETA(1) through THETA(4) whereas what Varsha was =

reporting as a difference in CL were the estimates of THETA(3) from =

fitting each treatment separately. So one could obtain different =

estimates of THETA(1) and THETA(3) between the two treatment groups =

while obtaining similar estimates of TVCL for a given value of WT. In =

other words, it is possible that TVCL and TVVD are similar between the =

two treatment groups (which would lead to similar halflives) and yet =

different estimates of THETA(3) for the two treatment groups.

Best regards,

Ken

Kenneth G. Kowalski

President & CEO

A2PG - Ann Arbor Pharmacometrics Group, Inc.

110 E. Miller Ave., Garden Suite

Ann Arbor, MI 48104

Work: 734-274-8255

Cell: 248-207-5082

Fax: 734-913-0230

ken.kowalski

Received on Fri Feb 05 2010 - 15:46:12 EST