The review addressed a clear question. Inclusion criteria were defined. The literature search was limited to two databases. Only full-text studies published in English were included, so there was a risk of language and publication biases. The authors noted that several relevant studies (one of which was UK based) were published since their search; it was unclear why these had not been included in the review subsequently and the searches updated. Appropriate steps were taken to minimise bias and errors when extracting data. The process of selecting studies was not clearly defined. The authors stated that a quality assessment was carried out but reported neither details of the tool nor the results, so the risk of bias in the included studies was unclear. Only very limited details on the included studies were reported and this made it difficult to determine the generalisability of the review findings. The methods used to pool data appeared appropriate, but confidence intervals around summary estimates (except for overall estimates which could be extracted from the forest plot) and the results of the heterogeneity assessment were not reported. In the discussion, heterogeneity was described as being sufficiently large to preclude meta-analysis, but summary estimates were reported in the results.
Limitations in the review, which included the possibility of missing studies and the unclear quality of the very small and heterogeneous primary studies, mean that the conclusions should be interpreted with caution.