This review answered a clearly defined review question. A number of relevant databases were searched for relevant studies, but only published English language studies were included in the review, so there was a risk of both publication and language bias. The risk of reviewer error and bias appeared to be relatively low, given the involvement of more than one reviewer when selecting studies and assessing their methodological quality; however it was unclear how many reviewers extracted the study data.
Relevant criteria were used to assess the quality of the trials, but the results of the assessment were not reported. Therefore it was difficult to assess the reliability of the individual trials, although the authors stated that they only included trials of sufficient quality. Given the use of indirect methods to compare the included interventions, few trial and population details were reported. This made it difficult to assess the reliability of the effect sizes, as analysis methods rely on the assumption of homogeneity and consistency amongst the included trials. Levels of statistical heterogeneity were reported to be low and further analyses were carried out to investigate the effects of other potential modifying factors. However, the results of indirect comparisons are only suggestive of relative effectiveness and are not as reliable as evidence from good quality head-to-head RCTs.
Overall, the evidence presented supported the review conclusions, but the reliance on indirect evidence and the poor reporting of trial characteristics, make it difficult to assess the reliability of the findings.
Three of the reviewers disclosed that they were full-time employees of AstraZeneca UK Ltd (the review sponsors).