Prev: Duplicate Observations
Next: TO MED STAT: Per Protocol Analysis
From: cat.. on 6 Apr 2010 14:57
Hi there,

The primary objective of a double-blind RCT I need to analyze is to
proove the investigational trt is better than placebo in term of
effect on a continuous endpoint (change from baseline of the primary
criterion Y, an interval-type variable). We have 3 active trt groups
with 3 different doses, and a placebo group. The randomization was
stratified on Y', a dichotomized version of Y.

The primary analysis is a set of 3 pairwise ANCOVAs (each comparing a
dose group with placebo), including the following covariates:
treatment group, the baseline value of the primary variable Y, the
center, and the interaction center x trt group.

I still need to find out whether the multiplicity issue was accounted
for in the sample size determination.

I'd like your views about the following:

- What is the point in adding the center in ANCOVAs as the
randomization was not stratified by center ?

- What is the point is stratifying the randomization by a dichotomized
version of the baseline values of the primary endpoint when, in any
case, we need to add the baseline value of Y in the model of the
primary analysis ?

- As the primary analysis involves adjusted ANCOVAs, shouldn't the
sample size calculation be based on an adjusted model rather than just
simple ANOVAs ?

Thanks for your VERY PRECIOUS feedback.

Catherine.
 | 
Pages: 1
Prev: Duplicate Observations
Next: TO MED STAT: Per Protocol Analysis