This review addressed a clear question supported by appropriate inclusion criteria. A number of relevant electronic databases were searched without language restrictions. Efforts were made to retrieve unpublished data. Suitable methods were used throughout the review process to minimise the risks of reviewer error and bias. In terms of the generalisability of this review to different clinical situations, it would have been useful if some patient characteristics had been reported.
Results were pooled using meta-analysis. The authors reported that they assessed heterogeneity using the Χ2 and I2 tests. Reporting the results for these analyses may have helped avoid discrepancies: the results report no significant heterogeneity for the primary outcome (data not reported), however, the discussion reports significant heterogeneity for the primary outcome and authors report that it may not have been appropriate to combine studies. This has quite important implications for a review of this size, as there were only three studies included and all had small study numbers.
In terms of methodology, this review was carried out robustly. But, without reporting patient characteristics and results it was difficult to interpret. The authors’ conclusions appeared to be over-optimistic given the limited evidence available and should be interpreted with some caution.