From: "David Nix" <nix@pharmacy.arizona.edu>
Subject: Error trap

Date: 19 Dec 1997 15:43:52 -0500

In using Advan3/ Trans3 routines, it has been common to obtain the error V>Vss. This model is parameterized by:

CL
V
Q - distributional clearance
Vss

with the following relationships:
k10 = CL/V
k12 = Q/V
k21 = Q (Vss-V)

The population estimates come out reasonable, but V>Vss for some individuals. As a result, the, the post hoc eta's are not obtained for anyone. I have got around this problem previously by parameterizing the model differently.

Does anyone have and error trap or other method to avoid this problem which makes it possible to parameterize using Vss.

****

From: "Harnisch, Lutz, HMR/DE" <Lutz.Harnisch@hmrag.com>
Subject: AW: Error trap
Date: 19 Dec 1997 16:18:03 -0500

we use these alternatives:

CL = THETA(1)
V1 = THETA(2)
V2 = THETA(3)
VSS = V1+V2
DVZ = THETA(4)
VZ = V1 + V2 + DVZ
Q = CL*(V2/DVZ) * (V2+DVZ)/VZ

or

...
V1 = THETA(2)
VSS = THETA(3)
V2 = VSS-V1 ...

or if (V1<V2)

...

V1 = THETA(2)
DV2 = THETA(3)
V2 = V1+DV2
VSS=V1+V2 ...

but only the first and last one give always POSTHOC estimates. But therefore you have to recalculate the error/omega of VSS. The second method has the same difficulties with V1>VSS, its only for completeness. That's what we use, also to avoid the interpretation of Q itself, because Vz is more meaningful for our understanding.

Best regards

Lutz
______________________________
Lutz Harnisch
Hoechst Marion Roussel
Preclinical Development
Pharmacokinetics
Population Analysis an Modelling
Building H840, Room 449
D-65926 Frankfurt/Main
------------------------------------------------------------
phone: +49 (69) 305-16481
fax: +49 (69) 305-81990
mailto:Lutz.Harnisch@hmrag.com
______________________________

****

From: alison@c255.ucsf.EDU (ABoeckmann)
Subject: Re: Error trap
Date: 19 Dec 1997 19:16:08 -0500

If you really do not want to reparameterize, there is a way to tell NONMEM that a given set of thetas and etas are unacceptable: the EXIT statement. (See Guide IV, Chapt 4, G.2 Implementation). In your case, you might have code that looks like this:

V=TV*EXP(ETA(1))
VSS=TVSS*EXP(ETA(2))
IF (VSS.LE.V) EXIT 1 100

During the posthoc step, NONMEM will try to avoid etas that give rise to this condition.

The second integer that follows EXIT, 100, identifies it in error messages such as

PK SUBROUTINE: USER ERROR CODE = 100

-- Alison Boeckmann

` `