Navin,
I think you have too many parameters to be estimated from a very small
dataset.
You are trying to estimate 6 Etas and 5 Thetas and you only have 11
patients.
Please see in the NonMem archive the discussion about the "rule of
thumb" being thrown around about the minimum size of the dataset needed
for parameter estimation. It's not completely scientific but it gives
you a reasonable idea where to start. I prefer simulation to design a
pop PK study but started to use this rule of thumb to get a "rough" idea
about how complex I can make the model to fit the data I have. It
definitely is a good starting point.
Below is a copy/paste of an email from Mark Sale on Jan 25, 2007. It was
followed (see archive) by an "interesting" discussion.
Rule of thumb that I found in a guideline when I arrived at Glaxo
(Wellcome) at the time. I think it came from Elion Fuseau, at least she
was the author of the document. minimum 20 samples per parameter
(thetas, omegas and sigmas) - so, for 8 parameters (5 basic parameters,
k,v,ka,k32,k23, 2 omegas, 1 sigma = your 160 samples) minimum 10
patients per eta
if data at good, more if data are noisy
seems reasonably to me.
But, the usual rules to define pk models - 3 terminal half lifes.
I think Steve Dufful has some software as well.
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of navin goyal
Sent: Friday, August 17, 2007 1:28 PM
To: nmusers
Subject: [NMusers] No. of parameters estimated
-
Dear NM Users,
I have been trying to do a POPPK run using nonmem VI.
The part of my control stream is pasted below....
$SUBROUTINE ADVAN6 TRANS1 TOL=3
$MODEL
COMP=(DEPOT,DEFDOSE)
COMP=(CENTRAL);PLASMA
COMP=(PERIPH);PERIPHERAL
$PK
TVF1=THETA(1)
F1=TVF1*EXP(ETA(1))
TVCL=THETA(2)
CL=TVCL*EXP(ETA(2))
TVV2=THETA(3);vol of dist of drug
V2=TVV2*EXP(ETA(3))
K20=TVCL/TVV2
K23=(THETA(6)+THETA(4))+EXP(ETA(4));
K32=THETA(5)+EXP(ETA(5))
TVK12=THETA(6);Abso constant
K12=TVK12+EXP(ETA(6))
S2=V2; OUTPUT IN ng/ml and dose in micrograms
$ERROR
IPRED=F
Y=F+F*ERR(1);+ERR(2)
$DES
DADT(1)=-K12*A(1)
DADT(2)=K12*A(1)-K20*A(2)-K23*A(2)+K32*A(3)
DADT(3)=K23*A(2)-K32*A(3)
$THETA
(0.1,0.3,0.7);F
(46 FIXED);CL
(0,25,200);V2
(0.01,0.1,10);K23
(0.01,0.1,10);K32
(0.01,0.1,2);K12
$OMEGA
$SIGMA
$ESTIMATION METHOD=1 SIGDIGITS=5 INTERACTION MAXEVAL=9999
PRINT=10 POSTHOC
$COV MATRIX=S
The model is very sensitive to the initial estimates and when I
bootstrapped the model about half of the runs failed to minimize. Also
the variability asssociated with parameter estimates is very high. The
data is from 11 subjects and has high variability. I tried various
models, including using FO Method. Am I trying to estimate more
parameters (like KA, F, Vd, K23, K32) with very limited information ? or
is it model misspecification...
Or should I fix other parameters like Vd from IV data. (here I
have fixed Cl from IV data...when I dont not fix CL , the nonmem
estimates very small values of CL ... in decimals)
Any input is appreciated.........
Thanks
-- --
--Navin
*********************************************************************
This message and any attachments are solely for the
intended recipient. If you are not the intended recipient,
disclosure, copying, use or distribution of the information
included in this message is prohibited -- Please
immediately and permanently delete.
