Back to the Top
Hello all,
Do any of you have ideas of how to simulate adverse effects in a phase
II or III clinical trial simualtion? As I see it, it is (almost)
impossible to simulate an ADR that occurs with a low frequency. It
will not have been found during the earlier phases of drug
development, and thus how does one build a model? For the more
frequent ADR's, I still don't see how to build proper models. Of
course it can be necessary to take data from non-clinical trials, but
this seems like a very unceartain route to follow. I cant' recall that
this issue was discussed in the recent workshop in Arlington.
Regards,
Thomas Senderovitz, M.D.
Dpt. Of Clinical Pharmacology Q7642
Rigshospitalet
University of Copenhagen
Tagensvej 20
DK-2200 Copenhagen N
Denmark
Phone: +45 35 45 69 44
Fax: +45 35 45 27 45
E-mail: senderovitz.-a-.rh.dk
Back to the Top
[A few replies - db]
Reply-To: "Stephen Duffull"
From: "Stephen Duffull"
To:
Subject: Re: PharmPK Simulation of adverse effects
Date: Fri, 26 Mar 1999 09:41:55 -0000
MIME-Version: 1.0
X-Priority: 3
Hi Thomas
This is obviously a very general question. It seems to me
that the fact that the you know the ADR occurs means that
you have some data already. Therefore this data can
presumably be modelled in some manner and inference gained.
It might well be that this data arises from pre-clinical
observations or indeed from another similar drug that is
already being used. In either of these cases you will
probably need to make certain assumptions within the model
(eg scaling parameters, covariate relationships) and the
sensitivity of the model to each of these assumptions will
need to be tested.
Is this the sort of thing you were getting at?
Regards
Steve
=====================
Stephen Duffull
School of Pharmacy
University of Manchester
Manchester, M13 9PL, UK
Ph +44 161 275 2355
Fax +44 161 275 2396
---
X-Server-Uuid: c7fabf50-5a2e-11d2-968c-00805fc1f894
From: "Sale, Mark "
To: "'PharmPK.-a-.pharm.cpb.uokhsc.edu'"
Subject: RE: PharmPK Simulation of adverse effects
Date: Fri, 26 Mar 1999 08:53:45 -0500
MIME-Version: 1.0
X-WSS-ID: 1AE54EE115055-01-01
Thomas,
Interesting question, and the topic of a session at AAPS this fall.
First you have to determine what sort of ADR you mean. If you mean the
"type 1", (exaggeration of known effect of drug), then this can fall
naturally out of you drug effect model and the intersubject/residual
variability. For example, I did a simulation of a class 3 anti arrhythmic
drug, known to prolong the QTc interval. We had data on the pk and QTc
effect of the drug and established a model. We could then estimate the QTc
effects in a trial, and even the risk/incidence of Torsade, which has a
known relationship as well. This actually applies to a lot of ADE's,
basically to any concentration or dose dependent effect.
The type 2 ("idiosyncratic", not concentration related, things like rash,
agranulocytosis) are a lot harder. Basically, one needs an empiric model,
supported by data, which in phase 2 at least one won't have, and for rare
events, there will be insufficient data in phase 3 as well.
Mark
---
From: "Sasso, Patrick [PRI]"
To: "'PharmPK.aaa.pharm.cpb.uokhsc.edu'"
Subject: RE: PharmPK Simulation of adverse effects
Date: Fri, 26 Mar 1999 10:32:57 -0500
MIME-Version: 1.0
A group in Wilmington Delaware simulated allergic response in dogs after
ADR's were reported with a batch of a protease inhibitor in phase I. Should
be literature references.
---
Reply-To:
From: "Jeff Wald"
To:
Subject: RE: PharmPK Simulation of adverse effects
Date: Fri, 26 Mar 1999 13:05:26 -0500
MIME-Version: 1.0
X-Priority: 3 (Normal)
Importance: Normal
Thomas - When there is no data, there is no way of reliably simulating a
response. One trick is to be clever about finding data. For example, use
information based on similar compounds, use literature sources, etc... In
the absence of data however, simulation has been shown to be a very powerful
tool to assess ones ability to measure a response. For example, if your
criteria for a clinical trial is to measure adverse events that occur at a
rate of N per 1000 patients per week. Simulation can be used to compare
trial designs that achieve that goal with a prespecified degree of
confidence. Alternately, if you have a trial design (which has been
optimized for other criteria perhaps) you can compute the N associated with
your desired degree of confidence.
Regards, Jeff
Jeff Wald, Ph.D. | Manager, Scientific Support Services
Pharsight Corporation | www.Pharsight.Com
Direct: Jeff.Wald.-a-.Pharsight.com | Technical inquiries:
MAD.Support.aaa.Pharsight.com
Back to the Top
Dear Thomas,
We have done some clinical trial simulations for the adverse effects of one
of our drugs in development.
Healthy subjects were treated with a fixed total daily dose for 1 week in a
cross-over trial (3 periods: placebo, bid or od). As a result we obtained
multiple AE measurements for each subject. We grouped the AE's in two
groups: acute and delayed.
The acute AE's were correlated to Cmax and the delayed to AUC's.
It was feasible to build a model because the AEs were relatively frequent
for the doses and regimens used.
We fitted a logistic regression model using the GLIMMIX-macro of SAS. This
routine allowed us to assess the population mean EC50 with the
interindividual variability for the different AE's. In addition, this method
takes into account the repeated measurements characteristic of the data.
Steepnesses of the s-shaped curves where assumed to be the same.
Interestingely, we observed that the frequencies of AE's diminshed over
time. Therefore, we included time as a covariate showing an increase of the
EC50's over time, indicating tolerance towards the AE's of this drug.
Subsequently, we simulated the AE profiles of this drug for different doses
and dosing regimens (once daily vs twice daily). We tabulated the
AE-frequencies and presented them to the clinicians. Based on the predicted
AE profiles a dosing regimen and a dose range was chosen for a Phase II
study (n=300).
Kind regards,
Kees Bol (Pharmacokinetics)
Filip de Ridder (Biostatistics)
Janssen Research Foundation
Beerse, Belgium
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)