hello all
i'd be grateful for insightful analytical critiques on a problem that
i'm struggling with:
How well, IN GENERAL TERMS, can the following study design answer the
research question?
QUESTION: for specified patients with condition X which of treatments
A, B or C are the most effective in achieving specified outcomes Y, Z
CONSTRAINTS; the ethical constraints are such that a no-treatment-at-all
group is not feasible, so it is a comparison study; treatments A, B or C
could in theory be applied in any combination to the patients. The
therapists are reluctant to assign patients at random to the treatments;
one of the aims of the study is to match realistic practice conditions
as closely as possible.
DESIGN: The centres where patients with condition X are to be treated
are to be allocated at random to one of three treatment groups;
a) not-A; A is NOT allowed but B and C are alone or in any combination
b) not-B, B is not allowed but A and C are alone or in any combination
c) not-C; C is not allowed but A and B are alone or in any combination
Is it possible to obtain useful information about the relative
effectiveness of the treatments from this design?
Does anybody know of any examples of research that have used such a
design?
Intuitively I feel it should, but I cannot find any references to this
type of design in the usual references (Pocock, Robson, Cook &
Campbell); this might be because it is fatally flawed, but if so why?
If not what would be the principles underpinning the analysis?
I'm attracted to the design generally because it seems to provide the
therapists with an increased opportunity to feel that the trial is
ethically acceptable, achieves flexible and therefore relatively
realistic interventions, and controls for differences between the
centres to some extent.
ANY remarks, comments, critical or otherwise would be gratefully
received. I'll summarise back to the list.
regards
Owen P Dempsey
General Practitioner
West Yorks
UK
phone : 01484 - 654504
e-mail: [log in to unmask]
|