The review question was defined but the inclusion criteria with regard to the type of steroid and study design were unclear. The search strategy seemed appropriate, but the limitation to trials published in English, Dutch and German may have resulted in a language bias. As there was no description of methods used to identify unpublished studies (publication bias was not investigated), this may have been an additional bias in the review. Two authors selected studies, assessed validity, and extracted data, although it was not clear how differences were resolved.
The criteria used to assess trial quality seemed appropriate, although the standardisation criteria were not defined. The quality of the included trials may have also resulted in a biased estimate of effect. One of the trials was not fully randomised and outcome assessment was not blind in two trials. Information on the characteristics of included trials was limited. The methods used to synthesise outcome data appeared acceptable. The authors did not report why meta-analysis was not used or how heterogeneity was assessed.
The authors’ conclusions reflected the findings of the review, but the small number and size of included trials, potential for language and publication bias, limited reporting of quality criteria in included trials, and a narrative synthesis of the literature may limit the reliability of these findings.