Risk of bias
We separately assessed the potential for systematic error
within individual studies using the Cochrane risk of bias
tool [30] and the following dimensions of methodological
quality: (1) generation of allocation sequence; (2) allocation
concealment; (3) blinding (participant and researchers); (4)
blinding of outcome assessors; (5) completeness of outcome
data; (6) selective outcome reporting. Studies were
grade as being at, “low”, “high” or “unclear” risk of bias.
Any discrepancies were discussed between the data extractors
until both reached a unanimous decision. Where an
unclear grading was given we contacted trial authors to
obtain further information and searched for the study
protocol to identify sources of reporting bias. We used
standard methods (based on the Cochrane Handbook) to
assess funnel plot symmetry as we had greater than 10
trials in the meta-analysis [30].We employed methods suggested
by Dwan and colleagues [31] to assess the risk of
outcome reporting bias for each of the outcomes using the
Outcome Reporting Bias (ORB) classification whereby
trials are scored as “high risk”, “low risk” or “partial risk”.
The full classification table is available in Additional file 4.