I would be grateful if anyone could point me towards recent discussions in
the literature and/or offer any observations of their own on the issue of
how best to handle/analyse data which is absent for treatment-related
reasons in clinical trials.
The sort of situation of interest is that in which, for example, there is an
appreciable dropout rate in a placebo group, or a group receiving relatively
ineffective medication, because of poor treatment efficacy (or sometimes
because of unwanted effects of treatment), with the consequence that all of
their data after the time of dropout is absent.
One approach, of course, is to regard rate of dropout as the primary outcome
variable, and to compare treatment groups with respect to that variable.
However, such an approach often demands large sample sizes and often seems
very wasteful of quantitative outcome data - but to simply regard the absent
quantitative data as 'missing' (or to exclude 'dropouts' from analysis)
clearly runs the risk of serious bias. I am therefore interested in current
views on such approaches to this problem as 'carrying forward the last
available observation'.
Thanks in advance for any assistance.
Regards,
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: [log in to unmask]
Buckingham MK18 4EL, UK [log in to unmask]
----------------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|