From:  "Li-Pin Kung" lpkung@aol.com
Subject: [NMusers] PD modeling of dataset with opposite values in the measurement effect
Date: Wed, 25 Jan 2006 14:19:35 -0500

Dear friends,

When the population data of a pharmacological effect is distributed across
zero, is it a variability or a dual effect or...? We don't accept the PK
data when it is negative. However, we can not reject a PD data when it
indicates a negative value of the desired effect.
I believe the Hills eq holds when (E-E0 )/Emax is positive. Am I correct?

I am working on a -blocker. I want to build a model for PKPD correlation.
All the PK data are though concentrations. 20% of the corresponding PD data
showed unfavorable to the pharmacological effect, i.e. the individual blood
pressure at the though level was higher than its baseline. Is it reasonable
to fit all the data to one Hills equation? What would be the base model do
you suggest?

Thanks,
Liping
_______________________________________________________

From: Paul Hutson prhutson@pharmacy.wisc.edu
Subject: RE: [NMusers] PD modeling of dataset with opposite values in the measurement effect
Date: Fri, 27 Jan 2006 14:13:14 -0600

Li-Pin:
I think that:
a) your "base" model (Eo) is going to need to be dynamic, either as a function of
drug concentration and/or duration of exposure.  The change in Eo may possibly even
be related to the magnitude of beta blockade achieved, and

b)  it doesn't sound like a very good beta-blocker
Good luck.
Paul 
_______________________________________________________

From: "Pereira, Luis" Luis.Pereira@bos.mcphs.edu
Subject: RE: [NMusers] PD modeling of dataset with opposite values in the measurement effect
Date: Fri, 27 Jan 2006 16:56:23 -0500

Dear LiPing

Some effect data can cross the baseline value over time giving the impression of a
negative effect not translatable into a 'negative' concentration. I suggest you look
at the E vs. C plot since you may have a case of short term tolerance (or rebound
effect) depicted by a clockwise hysteresis loop (or proteresis). In that case you
may minimize the hysteresis defining your PK in terms of the concentration at the
biophase, and then use an Emax, Sigmoid Emax or any other appropriate PD model.

Best luck

Luis

---------------------------------------------------------------
Luis M. Pereira, Ph.D.
Assistant Professor, Biopharmaceutics and Pharmacokinetics
Massachusetts College of Pharmacy and Health Sciences
179 Longwood Ave, Boston, MA  02115
Phone: (617) 732-2905
Fax: (617) 732-2228
Luis.Pereira@bos.mcphs.edu
_______________________________________________________

From: Scott VanWart Scott.VanWart@cognigencorp.com
Subject: RE: [NMusers] PD modeling of dataset with opposite values in the measurement effect
Date: Fri, 27 Jan 2006 18:01:56 -0500

Dear Li-Pin,

If blood pressure is the PD response you are modeling, there is already a wide body of
evidence in the literature that suggests that there is a circadian pattern to this type
of process. Therefore, you may need to model your baseline response as a function of
clock-time prior to trying to study the pharmocology of your drug.  A good example of
one approach that has been done to do this has been published by Hempel et al. (1998),
Clinical Pharmacology and Therapeutics; 64: 622-635.  I would also consider looking at
other factors that may confound your ability to detect a concentration-response
relationship, such as use of other non-study related medications or increases in the
production of other endogenous modulators (if measured).

A second point to consider is that there may be patients, who for any number of reasons, 
simply do not respond regardless of what drug concentration they are exposed to.  There
are several different approaches that you could consider here as well, one of which is
the use of a mixture model within NONMEM to identify "responders" versus "non-responders".

Scott Van Wart 
-- 
-----------------------------------
Scott Van Wart
Assistant Director, Population PK/PD
Cognigen Corporation
395 Youngs Road
Buffalo, NY 14221-5831
(716) 633-3463 ext 241

_______________________________________________________

From:  Gibiansky, Katya 
Subject: RE: [NMusers] PD modeling of dataset with opposite values in the measurement effect
Date:  Wednesday, January 25, 2006 3:28 PM

Liping,

A couple of notes. First, increase in blood pressure compared to baseline is
likely due to variability of baseline and not due to the drug. In the
absence of any drug effect and some systematic change (e.g. circadian
variation), one would expect 50% of PD data points to go up and 50% to go
down compared to baseline. To cope with the negative effect, you need to
account for baseline variability (i.e. estimate it and use estimated
population+individual values in the model rather than the observed value).

Second, if it is not a variability or some systematic difference in the
measurements issue, and you want to correlate it with the drug
concentrations, there is no rational why the correlation should be described
by the same relationship for those who respond negatively and positively to
the drug.   

Katya
_______________________________________________________