The following issues have arisen in connection with a systematic reviews =
dissertation that I have been supervising and I would welcome some =
discussion of them on the list.
1. The majority of systematic reviews to date are around therapies. =
There are some good reasons for this e.g. high % of clinical questions =
(circa 75%), preeminence of RCT design and relative ease of analysis =
(eg. Cf. With diagnosis).=20
However would it not be the case that it is not worth doing a systematic =
review of a therapy unless there is already a very reliable method of =
diagnosis for the condition? Should systematic reviews of diagnosis be =
equally/more important? I don't recall having seen much discussion of =
this (any refs would be appreciated). Do any of the review planning =
handbooks tell us that we ought to think about how reliable/uniform the =
diagnosis of a condition is before we choose to do a review on it? Or is =
this just an academic issue? - I mean if a condition is likely to be =
poorly differentially diagnosed in a trial then presumably it is more =
likely to be poorly diagnosed in health practice. In which case as it is =
the pragmatic results we are interested in (i.e. effectiveness, not =
efficacy) perhaps this doesn't really matter?
I appreciate this could be a chicken and egg debate, especially as rules =
for screening tell us that conditions should only be screened for if =
there is the potential to change clinical management (rough paraphrase). =
However to take a review that I am currently undertaking namely =
"acupuncture in migraine" I was speculating whether the review is =
potentially strengthened by the existence of the International Headache =
Society's criteria for diagnosing migraine, or maybe this is irrelevant?
2. For any of you who think that the above is too irrelevant, rambling =
or ill-conceived to merit a response- a slightly more focussed question =
- Does anyone have a reference for the "lumping" versus "splitting" =
debate in systematic reviews?> i.e. whether a review should lump =
together trials to address a broad question (at the risk of more aspects =
of heterogeneity) or whether it should split into very focused =
questions but at the risk of retrieving fewer relevant studies and =
therefore being inconclusive. I would particularly appreciate a =
reference to any article that actually uses the terms "lumping" and =
"splitting".
As always, gratefully yours.
Andrew Booth
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|