From: Joern Loetsch
Subject: Placebo data and weighting
Date: Thursday, January 17, 2002 11:15

Dear NONMEM users,
I have fitted placebo data from a standard cross-over study and want to incorporat
the placebo information intothe PK/PD fit of the verum data. However, the blunt
incorporation of the placebo model into the PK/PD fit results in obvious nonsens.
The drug effect would be negative in some subjects. The assumption that placebo
behaviour would be invarate among study occasions is obviously wrong. When I
introduce a weighting factor for the placebo information, the PK/PD fit works fine. 
However, I'm not sure whether a simple weight factor is adequate. Does anyone 
have a better solution for the placebo problem? Fitting placebo and active drug 
data together did not work, the model is probaly too complex, a reliable fit 
could not be obtained.Thanx in advance.

J. Lötsch
Jorn Lotsch, MD
pharmazentrum frankfurt, Department of Clinical Pharmacology
Johann Wolfgang Geothe-University Hospital
Theodor-Stern-Kai 7
D-60590 Frankfurt
Phone: +49-69-6301-4589
Fax: +49-69-6301-7636


From: Gobburu, Jogarao V 
Subject: Placebo data and weighting
Date: Friday, January 18, 2002 8:36 AM

Now, I do not know anything about your drug and the population from the email. 
But I have some general comments to make. 

1.  Why do you consider that getting negative drug effect is nonsense in some
individuals?  It could be that they are non-responders. 
2.  You talk about placebo effect, what about disease progression? May be this is why
in some patients, who either did not receive 'enough' drug or poor responders, the drug
effect is over-shadowed by the disease progression.
3. It would benefit if you explained what 'weight' function was used and its rationale.
4. Ideally simulataneous analysis of all the data is preferred. 

Joga Gobburu,