Inclusion criteria for the review were clearly defined and several relevant data sources were searched without language restrictions. Publication bias was not assessed and could not be ruled out. Two authors performed study selection and data extraction, minimising the risk of error and bias in the analysis.
The quality assessment of RCTs indicated the low quality of the included trials; there was no quality assessment of the non-randomised studies. The RCTs were combined using meta-analysis and the non-randomised studies were narratively synthesised, which appeared appropriate.
The authors’ conclusions are largely based on the results of the non-randomised studies of unknown quality, which are generally more prone to biases than RCTs. Given this limitation, the authors’ conclusions should be viewed with caution.