Hi Hauke, Joachim,
For a two-compartment linear problem, I would really worry about
non-convergence, warnings, and refused covariance step. The problem was
discussed to death on this list, and the transcript can be found in the
archive, so I will not repeat all the arguments, but the main idea is
that non-convergence (especially for this simple problem) may indicate
over-parameterization, when the data cannot support the model. Often,
too many ETAs are to blame for this.
Objective function difference of 5 is almost negligible. I would go with
the more mechanistic model, model with lower variance estimates, model
with lower standard errors, model with better diagnostic plots, and
disregard OF difference of 5 units.
I am not sure that bootstrap may verify the model for simulations. It
may reveal correlation of parameter estimates, provide confidence
intervals on the parameter estimates, but I am not sure how it can
characterize the predictive power of the model. VPC could be an
important test if one plans to use the model for simulations. Extra ETAs
(e.g., full OMEGA matrix with ETAs on all parameters) often lead to
over-estimation of the actual inter-subject variability, and this can be
seen on VPC plots.
Thanks
Leonid
--------------------------------------
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Grevel, Joachim wrote:
Dear Hauke,
thanks for including portions of your control stream and output. Leonid explained to you the rationale of preferring V1 and V2 over V1 and VSS.
I personally (not a company policy) routinely would use all four etas the way you did, but with a full set of covariates. I would also take out some of the correlation by allometricallly (by a size covariate) scaling all volumes and clearances according to the Holford paradigm (see NMusers on 3/4 rule). A further trick to make it all work is to logarithmically transform all concentrations as recommended by Mats Karlsson.
I go through all this trouble in order to capture "all" between patient variability and to prevent it from spilling over into unexplained (residual) variability.
I am not bothered too much by warnings and refused covariance steps. I verify
my pivotal models with bootstraps. When it comes to simulations with your final
(validated) model you will reap the fruits of your labor.
Hope that is of some use to you,
Joachim
_________________________________
AstraZeneca R&D Charnwood
Clin. Pharmacology and DMPK
Bakewell Road
Loughborough, LE11 5RH
Tel: +44 1509 644035
[email protected]
--------------------------------------------------------------------------
AstraZeneca UK Limited is a company incorporated in England and Wales with
registered number: 03674842 and a registered office at 15 Stanhope Gate, London
W1K 1LN.
Confidentiality Notice: This message is private and may contain confidential,
proprietary and legally privileged information. If you have received this
message in error, please notify us and remove it from your system and note that
you must not copy, distribute or take any action in reliance on it. Any
unauthorised use or disclosure of the contents of this message is not permitted
and may be unlawful.
Disclaimer: Email messages may be subject to delays, interception, non-delivery
and unauthorised alterations. Therefore, information expressed in this message
is not given or endorsed by AstraZeneca UK Limited unless otherwise notified by
an authorised representative independent of this message. No contractual
relationship is created by this message by any person unless specifically
indicated by agreement in writing other than email.
Monitoring: AstraZeneca UK Limited may monitor email traffic data and content
for the purposes of the prevention and detection of crime, ensuring the
security of our computer systems and checking Compliance with our Code of
Conduct and Policies.
-----Original Message-----
From: [email protected]
[mailto:[email protected]]on Behalf Of Hauke Rühs
Sent: 03 August 2009 17:06
To: [email protected]
Subject: Re: [NMusers] OMEGA-Block
Dear NMusers,
thank you for your responses. I parameterized the model by VSS, because
the estimation of the parameter became much better, in terms of
varability, compared to parameterization by V1 and V2.
This is the excerpt from my control stream and output file:
$PK
CL = THETA(1) * EXP(ETA(1))
Q = THETA(2) * EXP(ETA(2))
V1 = THETA(3)* EXP(ETA(3))
VSS = THETA(4) * EXP(ETA(4))
V2=VSS-V1
$OMEGA
0.1 ; OMCL
0.5 ;OMQ
$OMEGA BLOCK(2)
0.05 ;OMV1
0.01 0.05 ;OMVSS
OMEGA - COV MATRIX FOR RANDOM EFFECTS - ETAS ********
ETA1 ETA2 ETA3 ETA4
ETA1
+ 4.08E-02
ETA2
+ 0.00E+00 4.56E-01
ETA3
+ 0.00E+00 0.00E+00 2.62E-01
ETA4
+ 0.00E+00 0.00E+00 -4.34E-02 7.20E-03
Best regards
Hauke
Am 03.08.2009, 16:37 Uhr, schrieb Grevel, Joachim
<[email protected]>:
Hello Hauke,
as always for these specific questions you need to provide us with the
relevant portions of your control stream and of your output.
Thanks for participating,
Joachim
_________________________________
AstraZeneca R&D Charnwood
Clin. Pharmacology and DMPK
Bakewell Road
Loughborough, LE11 5RH
Tel: +44 1509 644035
[email protected]
--------------------------------------------------------------------------
AstraZeneca UK Limited is a company incorporated in England and Wales
with registered number: 03674842 and a registered office at 15 Stanhope
Gate, London W1K 1LN.
Confidentiality Notice: This message is private and may contain
confidential, proprietary and legally privileged information. If you
have received this message in error, please notify us and remove it from
your system and note that you must not copy, distribute or take any
action in reliance on it. Any unauthorised use or disclosure of the
contents of this message is not permitted and may be unlawful.
Disclaimer: Email messages may be subject to delays, interception,
non-delivery and unauthorised alterations. Therefore, information
expressed in this message is not given or endorsed by AstraZeneca UK
Limited unless otherwise notified by an authorised representative
independent of this message. No contractual relationship is created by
this message by any person unless specifically indicated by agreement in
writing other than email.
Monitoring: AstraZeneca UK Limited may monitor email traffic data and
content for the purposes of the prevention and detection of crime,
ensuring the security of our computer systems and checking Compliance
with our Code of Conduct and Policies.
-----Original Message-----
From: [email protected]
[mailto:[email protected]]on Behalf Of Hauke Rühs
Sent: 03 August 2009 15:07
To: [email protected]
Subject: [NMusers] OMEGA-Block
Dear NMusers,
modelling a 2-compartmet model parameterized by CL, V1, VSS and Q, I got
to a problem with which I don’t know how to deal with: After choosing my
structural and statistic model (combined residual error model) I
estimated
the covariance matrix by including an OMEGA-BLOCK(4), which reduced the
OFV by 15. The correlations between the parameters were all estimated to
be minor (< 0.8). But when I model with a BLOCK(2) on VSS and V1, which I
would expect to be positively correlated, the correlation is estimated to
be -0.99. Additionally, the inclusion of BLOCK(2) does not significantly
improve the OFV.
So does it, after all, still make sense to include the BLOCK(2)?
Generally, at which step of model-building would you recommend to test
for
parameter correlation?
Thanking you in advance,
Hauke
-----------------------------
Hauke Rühs
Apotheker
Pharmazeutisches Institut
- Klinische Pharmazie -
An der Immenburg 4
53121 Bonn
Tel: + 49-(0)228 73-5781
Fax: + 49-(0)228 73-9757
www.klinische-pharmazie.info
