Date:         Fri, 17 Jul 2009 09:33:35 -0400
Reply-To:     Kevin Viel <citam.sasl@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Kevin Viel <citam.sasl@GMAIL.COM>
Subject:      Basics of pharma studies (EC50)

Greetings,

 I have a question about the basics of time-to-response studies for
various doses within an agent or for various agents.  I will reference the
following paper:

Dose-time-response cumulative multinomial generalized linear model.
Chen DG. J Biopharm Stat. 2007;17(1):173-85.

  For what I gather, a basic shape of a response curve is assumed,
conventionally, it may be the logistic (sigmoid) curve.  To frame my
thoughts, consider that I what to test whether the EC50 is statistically
different between two curves (doses or agents).

  Initially, I thought that I might just add parameters to a model that
would result in different estimates of beta (consider the simple case).  I
am not seeing that in the brief literature review I performed.
Unfortunately, the number of journals to which I have full text articles
is more limited than I desired and I have yet to establish access to a
large library.

  How does this thought strike those in the field?  It might be akin to a
random coefficient model, for instance.

  My question concerning the Chen paper is why would one use probabilities
for what is essentially survival (continuous time-to-event) data?  In
fact, Chen mentions a potential pitfall him- or herself: namely, rates
that change over time.  For instance, for days 0-3 the rate might be x,
for days 4-6 the rate might be y, and for days 7+ the rate might be z.

I would appreciate any comments or references.

Kind regards,

Kevin