The Etiology, Diagnosis, Therapy etc. filters identify Clinical Studies. Systematic Reviews or their variant, Meta-Analyses, are not clinical studies and are not specifically identified by using the clinical studies filters. The terms ‘behind’ the filters for clinical diagnostic studies are:
Sensitiv:.mp. OR diagnos:.mp. OR di.fs. (sensitive)
Sensitiv:.mp. OR predictive value:.mp. OR accurac:.tw. (middling)
The terms used in these diagnostic filters indicate why some Cochrane Systematic Reviews are also identified. A systematic review may well contain e.g. the textword ‘sensitive’ without being a clinical study evaluating a diagnostic procedure. This explains why you may retrieve non-diagnosis entries. More details for the terms used for the various filters and their actual sensitivity, specificity, precision, and accuracy can be found here: http://hiru.mcmaster.ca/hedges/.
For Systematic Reviews the relevant separate SR filters should be used, such as
MEDLINE.tw.OR systematic review.tw.OR meta analysis.pt. (specific)
A comprehensive search on robust EBM-type material should therefore combine a Clinical Study type filter search with a Systematic Review filter, perhaps in the form
[Subject] AND ([clinical study type filter] OR [Systematic Review filter])
On a more general point, the Cochrane DSR is strong when intervention studies have to be identified. For robust diagnostic studies COCHRANE DSR is not the first port-of-call. For instance, only one of at least 130 robust (diagnostic) clinical studies or systematic reviews on the value of ‘peritoneal lavage compared to ultrasound in blunt abdominal trauma’ comes from the Cochrane DSR.