From: "INTV (Ivan Nestorov)"
Subject: [NMusers] Placebo in indirect PD models
Date: Wed, 26 Apr 2006 16:36:39 -0700

Dear All,

I would be interested if anybody from the list can share ideas and
experiences with ways to include placebo in indirect PD models.

Considering (one type of such) model - with inhibition of
synthesys, I have seen this done by:


where the term for placebo may range from nothing (leaving all the placebo
observations at the mercy of the residual error), all the way to a complex
periodic function, adding at least three more thetas to estimate and thus
almost always overparametrizing your model. In addition to this, it is a
pain, if at all possible, estimating IIV on both SYN, components of ACTIVE
and of PLACEBO because of the interaction between them in the equation above.

Kind Regards.

Ivan Nestorov

1201 Eastlake Avenue East
Seattle, Washington 98102
Tel: 206 442 6613
Fax: 206 442 6608
Cell: 425 442 9303

From: Mark Sale - Next Level Solutions
Subject: Re: [NMusers] Placebo in indirect PD models
Date: 26-Apr-2006 21:33


General thoughts:
Placebo effect is (IMHO) not generally part of the pharmacology.  This, it occurs to me
is the case in pain.  It is something else, expectations (on the part of the patient or
the observer), other behavioral/dietary/lifestyle changes etc. So, (also IMHO) it wouldn't
be my first choice to model it is part of the drug-related biology.  Rather, it is something
separate.  My favorite placebo model is something that I just simply add to the response

plac = Pmax*Time/(TC50+TIME)
Y = (F+plac)*exp(eps(1))

where Pmax is the maximum possible placebo effect, and TC50 is the time at which half the
maximum is acheived.
I really can't think of a case where I would consider the placebo effect to be part of
the pharmacology.  If you would like something to go up and then down (as, I beleive occurs
in pain placebo models), you might consider just a one-compartment, first order absorption model.

Mark Sale MD
Next Level Solutions, LLC

From: Nick Holford
Subject: Re: [NMusers] Placebo in indirect PD models
Date: Thu, 27 Apr 2006 17:16:19 +1200

Ivan, Mark,

I broadly agree with Mark's suggestion for a placebo response based on the
'Bateman' function as an empirical description. 

But there is a related class of models which describe disease progression that can
be intepreted as having a biological basis e.g. the rate of bone resorption changes
with time; growth of tumour cells (see Holford et al 2001). These models are extensions
to the turnover (aka indirect effect) model family which incorporate time varying
changes in either input or loss which are independent of drug.

Whether or not one can estimates the model parameters is a question of the design of
the study but with suitable designs the structural and randome effects parameters
should be estimable.

Best wishes,


Holford NHG, Mould DR, Peck CC. Disease Progress Models. In: Atkinson A, editor. Principles
of Clinical Pharmacology. San Diego: Academic Press; 2001. p. 253-62.

Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand tel:+64(9)373-7599x86730 fax:373-7556

From: "Piotrovskij, Vladimir [PRDBE]"
Subject: RE: [NMusers] Placebo in indirect PD models
Date: Thu, 27 Apr 2006 09:27:39 +0200

Dear Ivan,

There may be no general rule how to implement a placebo effect. It depends
very much on what is your dependent variable. In case it is a score you may
try the method presented at the PAGE meeting in Pamplona last year:

Best regards,

Vladimir Piotrovsky, PhD
Research Fellow
Clinical Pharmacology & Experimental Medicine
J&J Pharmaceutical Research and Development

Subject: RE: [NMusers] Placebo in indirect PD models
Date: Mon, 1 May 2006 08:30:07 -0400

Mark - Thanks for your general thoughts, but in the case of Ivan's original
posting I do not think they are relevant.  In the case of drug effects that
can be described by a semi-physiological model such as an indirect response
model the placebo response is typically inextricably linked to the pharmacology
and the underlying physiology.  One of the advantages in physiologically-based
models is that you basically "borrow" degrees of freedom from your knowledge of
the system.  So for example, if inter-individual variability in drug response
relates to the measurements under baseline or placebo conditions then you can
perhaps share an ETA for these 2 components that would otherwise be considered
separate in a phenomenological model.

Furthermore, pain is not a good example for the separability of drug/placebo effects. 
How could naloxone be a "placebo antagonist" if this were the case?  For example:

Benedetti, F., C. Arduino, and M. Amanzio. 1999. Somatotopic activation of opioid
systems by target-directed expectations of analgesia. Journal of Neuroscience 19(May 1):3639.
is one example from a large set of work addressing this for pain in particular. 
Meta analyses in pain and other diseases (depression is one that springs to mind)
have shown that the magnitude of placebo response correlates with the magnitude of
pharmacological response.  Unfortunately, these have primarily been for parallel
group trials and thus look at population averages confounded by other factors.
Moreover, I do not yet know of examples that explore this on an individual  basis
in crossover trials, but it nevertheless suggests to us that placebo response is
much more closely linked to drug response than the conventional wisdom of many
decades of drug development has allowed for.  

Getting back to the original question....  If a placebo response model can do any
of the following: credibly borrow from the pharmacological part of the model, if
it can be implemented in such a way as to make the model simpler than the use of
an empirical placebo response model, or if it provides for an improved understanding
of the system (either in the presence or absence of drug) then I think it is worth
pursuing the physiologically based approach.

Now, with all that said I'll expose my mercurial nature and suggest that for an
empirical placebo model the 'bateman' function in which k10=k01 is parsimonious
and has proven to be very flexible and useful in my own experience.

Regards, Jeff

Jeff Wald, PhD
Director, Clinical Pharmacokinetics/Modeling and Simulation
RTP, North Carolina

They always say time changes things, but you actually have to change them yourself.
- Andy Warhol 

From: Mark Sale - Next Level Solutions
Subject: RE: [NMusers] Placebo in indirect PD models
Date: Mon, 01 May 2006 06:37:33 -0700


If the placebo model is inextricably linked to the biology/pharmacology, then
it is easy.  I'm not sure the citation is about a physiologic link.   As you
point out, pain may be a very poor example of seperation of placebo and drug
effect - but I'm searching for another example where they might actually be a
physiologic link between placebo effect and drug effect (I'm imagining endorphins,
but I'm way out of my depth here).  But, I'm not sure that the placebo response
being proportional to the drug response demonstrates that the same mechanism/pathway
is involved.   Does the placebo effect result in greater receptor occupany?   The
discussion uses words like "correlate" and "metaanalysis" and "confounding", not
pharmacologic terms.  I assume that "closely linked" means statistically.  Not
sure what data they might have that could demonstrate a physiologic link.  And I
can think of a number of mechanisms resulting in a statistical link - including
but not limited to being mediated by the same biochemical pathways.


Mark Sale MD
Next Level Solutions, LLC