Paul,
I am not sure what is the problem you are trying to solve. This is the
estimation step, so results are conditioned on the data. Different
subjects with different dosing schedules have different data; thus, you
see different individual predictions. With the FO method and no POSTHOC,
IPREDs are equal to PREDs: this explains the results you are seeing for
FO. If you like to simulate from the model, then you need to use
simulation rather than estimation step.
Results seems correct to me (and indeed, A2 and A3 should be defined in
the error block): am I missing something? Is this a cross-over study
with long wash out period between daily and weekly dosing regimens? Only
in this case you would expect equality of the end-of-first dose
predictions for each subject. Could you, please, provide the data for
one subject and the output that you deem incorrect?
Thanks
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
On 1/13/2012 1:43 PM, Paul Hutson wrote:
Thank you, Sebastien.
Moving A2 and A3 to the $ERROR block still gives me varying values if I
use FOCE, but I presume that you mention this option for integrated
models, rather than a $DES block.
Thank you for the pointer to Alison's 02 Jun 2011 post. It explains why
the first estimate for the concentrations in the $DES block was so odd.
Can you or others explain why the FOCE method gives different drug
amounts for the different compartments with MAXEVAL=0 when the only
difference is not in the model, but in the dosing frequency? I would
expect the plots of IPRED, A2 and A3 to overlap after the first dose of
each dosing regimen, and diverge only when the regimen with the earliest
second dose finds that dose given. Why is this so with FO, but not with
FOCE?
Thank you
Paul
-------- Original Message --------
Subject: Re: [NMusers] Predicting Drug in Tissue A2 using MAXEVAL=0
Date: Thu, 12 Jan 2012 18:05:24 -0500
From: Sebastien Bihorel <[email protected]>
To: [email protected]
Paul,
The first patient record in the output table reports very unexpected
values of any variable declared in the $DES block. This issue is linked
to the integration process used by NONMEM and was addressed multiple
times on the list (look for Alison Boeckmann posts). This can also
affect other records when model are resest or when MTIME variables are
used. Assuming that you don't have any dummy observation in your dataset
with CMT =2 or 3, you can still get the amounts of drug in your 2nd and
3rd compartments at every observation record (those defined with CMT =1)
by moving the two lines
A2 = A(2)
A3 = A(3)
from the $DES block to the $ERROR block.
As a side note, if you use NONMEM VII, you may want to add the SIGL
option in your $EST record (see section I.14 in the Introduction to
NONMEM 7 guide). There are recommended relationships between NSIG, SIGL,
and TOL when using ADVAN6, 8, 9 and 13.
Hope that helps.
Sebastien BIHOREL
Cognigen Corporation
Paul Hutson wrote:
I re-read Guide IV, III, B.14 and realized that my $EST line is incorrect.
To compare the amount of drug in these tissue compartments with
different dosing regimens, after assigning A2=A(2) in the $DES block,
I should NOT have used
$EST MAXEVAL=0 METHOD=COND POSTHOC NOABORT PRINT=1
Instead, this line below using the first order method and no posthoc
worked:
$EST MAXEVAL=0 METHOD=0 NOABORT PRINT=1
Thanks.
Paul
-------- Original Message --------
Subject: FW: [NMusers] Predicting Drug in Tissue A2 using MAXEVAL=0
Date: Thu, 12 Jan 2012 12:02:16 -0800
From: Zheng, Jenny <[email protected]>
To: [email protected] <[email protected]>
Hi, Paul,
You would expect the values are slightly different since your output
(OBSV) included the residual errors. The IPRED should be the same for
the same individual but different regimens.
Hope this helps.
Jenny
*From:*[email protected]
[mailto:[email protected]] *On Behalf Of *Paul Hutson
*Sent:* Thursday, January 12, 2012 7:29 AM
*To:* [email protected]
*Subject:* [NMusers] Predicting Drug in Tissue A2 using MAXEVAL=0
Colleagues:
I am working with a local investigator to estimate the effects of
daily vs weekly dosing of an antibody. As such, I would like to show
the expected amounts of drug in the two tissue compartments (A(2) and
A(3)). The problem is that using MAXEVAL=0 and the same initial
estimates, I get different values for the first dose, which should be
equal for both the daily and weekly dosing. IPRED in plasma is
different for each dosing regimen, but of course PRED gives the same
end of infusion concentration. How can I code my model so that I can
report out the amount of drug in each respective tissue compartment so
that after dose 1 it is the same for the weekly and daily dosing? (
That is, how can I code it so that A(2) after the first weekly dose is
the same as A(2) after the first of the daily doses.)
CTL follows. Thank you!
Paul
$DATA ..\Qwk_60mg_sim.csv IGNORE=#
$SUBROUTINES ADVAN6 TOL=3
$MODEL NCOMP=3 NPAR=8
COMP=(CENTRAL DEFDOSE DEFOBS)
COMP=(TISU1)
COMP=(TISU2)
$PK
F1=1
IF(DOS1.EQ.40) F1=THETA(8)
TVCL=THETA(1)
TVV1=THETA(2)
TVQ1=THETA(3)
TVV2=THETA(4)
TVQ2=THETA(5)
TVV3=THETA(6)
CL=TVCL*EXP(ETA(1))
V1=TVV1*EXP(ETA(2))
Q1=TVQ1
V2=TVV2
Q2=TVQ2
V3=TVV3
S1=V1
K10= CL/V1
K12= Q1/V1
K21= Q1/V2
K13= Q2/V1
K31= Q2/V3
$DES
DADT(1) = -A(1)*(K10 + K13 + K12) + K21*A(2) + K31*A(3)
DADT(2) = A(1)*K12 - A(2)*K21
DADT(3) = A(1)*K13 - A(3)*K31
A2=A(2)
A3=A(3)
$ERROR
W1=THETA(7)
IPRED = LOG(0.025)
W = SQRT(W1**2)
IF(F.GT.0) IPRED = LOG(F)
IRES = IPRED-DV
IWRES = IRES/W
Y=IPRED+ W * EPS(1)
OBSV=EXP(DV)
$THETA 0.687; POPCL
$THETA 1.19 ; POPV1
$THETA 6.87; POPQ1
$THETA 1.8 ;POPV2
$THETA 0.191; POPQ2
$THETA 29.6; POPV3
$THETA 0.393; W1
$THETA 0.631; F1
$OMEGA BLOCK(2)
1.47E-01 ; IIVCL
1.87E-01 7.03E-01 ; IIVV
$SIGMA 1 FIX ; ERR_CHLO
$EST MAXEVAL=0 METHOD=COND POSTHOC NOABORT PRINT=1
$TABLE ID TIME IPRED A2 A3 NOPRINT ONEHEADER FILE=SimOut.fit
--
Paul R. Hutson, Pharm.D.
Associate Professor
UW School of Pharmacy
777 Highland Avenue
Madison WI 53705-2222
Tel 608.263.2496
Fax 608.265.5421
Pager 608.265.7000, p7856
