The objectives and inclusion criteria of the review were clear. Only two databases were searched (so some studies may have been missed) and the restriction to published studies in specific languages meant that the review was prone to publication and language biases. It appeared that publication bias was not formally assessed. Steps were taken to reduce the risk of reviewer bias and error by having more than one reviewer to select studies and assess study validity; it was unclear whether similar precautions applied to data extraction.
The tool used for validity assessment was designed for randomised controlled trials (RCTs) and is unsuitable for validity assessment of observational studies. The authors acknowledged that the included studies differed widely in design, quality, population and treatment regimen, and that it was unclear whether pooling of the data was appropriate. However, the authors did make some attempt to adjust for potential confounding using meta-regression. The authors also acknowledged that the use of rescue medication in the included studies was not quantified, studies may have been of too short duration to show an effect, and that outcome measures were relatively crude. However, much of the analysis was based on untested assumptions and/or used data that had been manipulated by the reviewers. As no studies compared the interventions of interest, the review findings were based on indirect comparisons, and limitations with this type of analysis should be borne in mind.
The limitations of the included studies and issues associated with indirect analysis, as highlighted by the authors, should be borne in mind when interpreting the findings, as should the potential for bias in the review process. However, the authors’ recommendation for further research seems appropriate.