For a meta-analysis of interventions composed of only 1 RCT, 5 single-arm (SA) and 5 double-arm observational studies (DA), what is the best way to account for the lack of control arms amongst the SAs? More specifically, is it possible to 'donate' control arms from the RCT / DAs to allow for comparison of data within the recipient SAs?
Some literature proposes matching a control arm from an RCT with an SA (Zhang et al - doi: 10.1016/S1499-3872(12)60209-4), however, I'm unsure if this method adjusts for inherent differences in selection bias between the two designs. Please help?
I would be incredibly grateful for any advice, thank you!
One control arm should only be included in a meta-analysis once. Furthermore, a control arm from another study is unlikely to be comparable with the intervention group in a different study. Therefore, I would strongly advice against this approach.
If there is only one RCT, then I would recommend doing a qualitative synthesis of the studies.
“garbage in garbage out”.
using data at high risk for bias will obtain answers at high risk for bias.
in the setting you described a qualitative synthesis is best Can consider comparison of three groups: sa, da, and rct and explain any differences (ex different populations, different quality of assessment and fu, risk of bias, etc).
I would include the RCT and the 5 double-arm observational studies and exclude all single-arm studies.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Writing