Dear all,
in many publications as well as in FDA statistical reviews in the area
of lipid disorders the results are frequently presented as least square
means and confidence intervals for the difference from placebo in the
mean percent change from baseline. As it appears, this is done by first
calculating a % change from baseline within each patient and using these
as dependent variables in an ANOVA. This can probably lead to
considerable violations of the assumptions underlying an ANOVA and I am
therefore wondering why it is so widely used. Or are there other ways
how a difference from placebo in the mean percent change from baseline
can be derived, i.e., from using ANOVA on absolute rather relative
change from baseline?
Any comments would be greatly welcome. Any feedback I receive I'll be
happy to summarize and share with those interested.
Kind regards
Markus
|