NONMEM Users Network Archive

Hosted by Cognigen

Re: advan8 vs. advan13

From: nonmem
Date: Tue, 03 Nov 2009 02:37:44 +0000 (GMT)

Thank you everyone for help. It seems like everything almost clear now:
1. Low boundary of 0 for bioavalability caused unpredictable results with method=IMP (meaningless objective function, unstable algorithm). When I set it to 0.1, error messages disappeared in many jobs and/or job converged. (I did not implement variability in bioavailability.)
2. Integration difficulties from ADVAN8 caused funny results with method=IMP (for example, exponential increase in objective function). When ADVAN13 was used, most jobs converged. It is nice to implement error messages from IMP when there are intergation difficulties.
3. It looked like integration difficulties from ADVAN8 can cause simulation errors. When ADVAN13 was used, they disappeared in most jobs.
4. Overall, ADVAN13 looks better than ADVAN8, but in few cases ADVAN8 did not have integration difficulties when ADVAN13 did.
5. It looks like there is no bug related to absense or presence of TRANS1. I misenterpreted the outcome. There were intergation problems.

Thanks!
Pavel


----- Original Message -----
From: Nick Holford
Date: Monday, November 2, 2009 3:00 pm
Subject: Re: [NMusers] advan8 vs. advan13
To: nmusers

> Pavel,
>
> Your model for bioavailability (BIO) (as interpreted by Bob)
> uses a
> random effects model that is based on the normal distribution:
>
> BIO=THETA(1) + ETA(1)
>
> Because there is no constraint on ETA(1) this can lead to
> negative
> values or positive values greater than 1 for individual BIO
> which is not
> realistic.
>
> You could avoid negative individual values of BIO by using this
> parameter random effect model which is based on a lognormal
> distribution:
> BIO=THETA(1) * EXP(ETA(1))
>
> However, the lognormal distribution can generate BIO values
> greater than
> 1 which once again is not realistic.
>
> So Bob has illustrated how to use a different random effects
> model based
> on the logistic distribution. This will constrain BIO to lie
> between 0
> and 1.
>
> MU_1=THETA(1)
> EXPP=MU_1+ETA(1)
> BIO=1/(1+EXP(-EXP(EXPP)))
>
> The value of THETA(1) will be a bit harder to interpret because
> it will
> be the logit of bioavailability. You can get a more
> interpretable value
> of THETA(1) with this code:
>
> LBIO=LOG(THETA(1))
> MU_1=LBIO/(1-LBIO)
> EXPP=MU_1+ETA(1)
> BIO=1/(1+EXP(-EXP(EXPP)))
>
> Bob,
>
> I think you can avoid the test for values of EXPW which could
> cause
> overflow by calculating BIO as shown above. This relies on the
> compiler
> rounding large negative values of exp() to a value of zero which
> is the
> usual behaviour.
>
> Best wishes,
>
> Nick
>
> Bauer, Robert wrote:
> > Pavel:
> > Regarding your statement:
> > Another thing I noted is that when IMP is used, the lower
> boundary for
> > bioavailability cannot be 0
> >
> > Typically, Monte Carlo methods try a large number of etas,
> from
> > negative infinity to positive infinity. You may want to
> remodel your
> > bioavailability to something like:
> >
> > MU_1=THETA(1)
> > EXPP=MU_1+ETA(1)
> > IE (EXPP>100.0) EXPP=100.0 ;protect against floating overflow
> > EXPW=EXP(EXPP)
> > BIO=EXPW/(1.0+EXPW)
> >
> > This way, BIO will always be between 0 and 1, regardless of
> eta value
> > You can now remove the boundaries to the theta.
> >
> >
> >
> > *Robert J. Bauer, Ph.D.
> > Vice President, Pharmacometrics
> > ICON Development Solutions*
> >
> > *Tel:* (215) 616-6428
> > *Mob: *(925) 286-0769
> > *Email: Robert.Bauer
> > *Web:* www.icondevsolutions.com
> >
> >
> >
> >
> >
> >
> >
> > ---------------------------------------------------------------
> ---------
> > *From:* owner-nmusers
> > [mailto:owner-nmusers
> > *Sent:* Sunday, November 01, 2009 1:57 PM
> > *To:* nmusers
> > *Subject:* Re: [NMusers] advan8 vs. advan13
> >
> > Pavel,
> >
> > The TRANSn option is only meaningful for those ADVANs (i.e. 1-
> 4,11,12)
> > which implement a closed form solution of the PK model. TRANS1
> does
> > nothing except keep the linker happy. Because it is the
> default it
> > need never be explicitly specified for ADVAN5-9,13.
> >
> > "TRANS1 Dummy, or null, translator; may be used with all
> ADVAN
> > routines. It is the default." NONMEM7 Help
> >
> > Adding it to those ADVANs which solve ODEs or use matrix
> exponential
> > solutions should make no difference to anything. If you find
> that
> > adding TRANS1 changes the behaviour of ADVAN8 then this is a bug.
> >
> > Nick
> >
> >
> > nonmem
> > > Hello Nonmem Users,
> > >
> > > When I tried simple simulations and changed advan13 to
> advan8, some
> > concentrations changed. Equations were stiff. Then I added
> TRANS1 to
> > the $SUBROUTINE statements. ADVAN13 output did not change.
> ADVAN8
> > output changed and looked exactly like ADVAN13 output. Do you
> know
> > what happened?
> > >
> > > (Another thing I noted is that when IMP is used, the lower
> boundary
> > for bioavailability cannot be 0.)
> > >
> > > Thanks,
> > > Pavel
> > >
> > > ----- Original Message -----
> > > From: nonmem
> > > Date: Sunday, November 1, 2009 8:34 am
> > > Subject: Re: [NMusers] advan8 vs. advan13
> > > To: nmusers
> > >
> > > > It seems like advan8 has integration difficulties when
> both LAG
> > > > time and variability in Ka are implemented. Method=IMP has
> > > > dificulties when advan8 has integration difficulties. Instead
> > > > if reporting issues, it keeps running. The objective function
> > > > is very low even when bioavailability is almost zero. Removing
> > > > LAG and eta of Ka may fix it.
> > > >
> > > > ----- Original Message -----
> > > > From: nonmem
> > > > Date: Sunday, November 1, 2009 12:04 am
> > > > Subject: [NMusers] advan8 vs. advan13
> > > > To: nmusers
> > > >
> > > > > Hello NONMEM users,
> > > > >
> > > > > Because advan13 was recomended for both stiff and nonstiff
> > > > > differential equations, I used it for stiff differential
> > > > > equations. It appeared that some results looked too sensitive
> > > > > to a parameters representing a "slow" processes. I did not
> > > > > observe it with nonmem6. When I used advan8, the objective
> > > > > function changed from 10228.853 (the final value; diagnostic
> > > > > plots looked good) to 5512.594 (the first value; I im still
> > > > > waiting for the final value of the objective function).
> > > > >
> > > > > Does it mean that advan13 should be used with caution
> when the
> > > > > equations are stiff or advan13 cannot replace advan8?
> > > > >
> > > > > Thanks,
> > > > > Pavel
> > > > >
> > > >
> >
> > --
> > Nick Holford, Professor Clinical Pharmacology
> > Dept Pharmacology & Clinical Pharmacology
> > University of Auckland, 85 Park Rd, Private Bag 92019,
> Auckland, New
> > Zealand
> > n.holford
> > mobile: +64 21 46 23 53
> > http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford
> >
> > ICON plc made the following annotations.
> > ---------------------------------------------------------------
> ---------------
> > This e-mail transmission may contain confidential or legally
> privileged information
> > that is intended only for the individual or entity named in
> the e-mail address. If you
> > are not the intended recipient, you are hereby notified that
> any disclosure, copying,
> > distribution, or reliance upon the contents of this e-mail is
> strictly prohibited. If
> > you have received this e-mail transmission in error, please
> reply to the sender, so that
> > ICON plc can arrange for proper delivery, and then please
> delete the message.
> > Thank You,
> > ICON plc
> > South County Business Park
> > Leopardstown
> > Dublin 18
> > Ireland
> > Registered number: 145835
> >
> >
> >
>
> --
> Nick Holford, Professor Clinical Pharmacology
> Dept Pharmacology & Clinical Pharmacology
> University of Auckland, 85 Park Rd, Private Bag 92019, Auckland,
> New Zealand
> n.holford
> mobile: +64 21 46 23 53
> http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford
>
>

Received on Mon Nov 02 2009 - 21:37:44 EST

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request@iconplc.com.

Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.