>I appreciate Dave Sackett's point about inconclusive (under powered) studies
>and his decision to reject them for a journal like EBM - the single study
here
>is uninformative and therefore not useful.
>
>I wonder wether he (or anyone else who edits a journal) might want to
comment
>further on what he feels should be the fate of such studies. I for one
struggle
>with the idea of power as a quality issue in an era when meta-analysis
appears
>to offer the possibility of making all those small studies useful!
>
I suppose Dave Sackett would publish a review of the meta-analysis rather
than of the individual underpowered primary studies. But that presupposes
that the meta-analysts (systematic reviewers) have been able to unearth
such underpowered, unpublished studies. In the past that has been difficult
but in the future, prospective registration of studies would make it easier.
I think the editors of EBM would be more likely to publish a meta-analysis
in which the possibility of publication bias towards positive studies had
been properly addressed.
BTW, even positive studies which are small may not be informative. A wide
confidence interval might embrace the possibilities that (a) the treatment
carries a clinically important benefit , and (b) that it carries a benefit
which is not big enough to be clinically important.
Richard W Morris, PhD
Senior Lecturer in Medical Statistics
Department of Primary Care & Population Sciences
Royal Free and University College Medical School
Royal Free Campus
Rowland Hill Street
London NW3 2PF
Tel: 0171 830 2239 or 0171 794 0500, X4144
FAX: 0171 794 1224
EMail: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|