The review included a clear research question and inclusion criteria were appropriate. Three electronic databases were searched and attempts were made to find other relevant studies by searching reference lists of retrieved studies. Search terms were not reported, so it was not possible to assess how comprehensive the search was. The authors did not report whether there was language restriction, so language bias could not be excluded. No explicit attempts were made to search for unpublished studies, but there was no evidence of publication bias from formal assessments. The authors did not fully report methods for study selection or validity assessment, so reviewer error and bias could not be excluded.
The authors noted that three of the included studies were led by clinicians at the same hospital and that this suggested these trials were not completely independent, but they appeared to be separate trials. The median Jadad score for the quality of included trials was 2 out of a maximum of 5, but minimal information was given about the shortcomings of the individual trials, which made it difficult to evaluate the strength of the trial evidence.
The synthesis of included trials by meta-analyses was appropriate, but numbers of cycles and doses varied contributing to heterogeneity in the analyses and details of timings were not reported, which made it difficult to compare interventions. Some estimates were not robust and heterogeneity was significant; exclusion of trials contributing to the heterogeneity and recalculation of estimates resulted in different findings. These exclusions were made post hoc, and the individual characteristics contributing to the heterogeneity not described.
The authors' conclusions reflected the evidence base, but in light of the shortcomings with the way the review was reported and the quality of the included trials, the authors' conclusions should be interpreted with caution.