OBJECTIVES: To assess the reporting and scientific quality of meta-analyses (MAs) of randomized controlled trials (RCTs) as treatments for anxiety disorders. METHODS: We searched EMBASE, EBM Reviews, MEDLINE, Healthstar and IPA from inception to August 2007. Search terms included: MA, RCTs, anxiety, anxiolytic, anti-depressant/antidepressant and cognitive therapy with no language restrictions. Titles and abstracts were assessed by two reviewers independently. Criteria for full-text retrieval included: MA, statistical pooling, anxiety disorder, RCT, pharmacotherapy, cognitive therapy. RCTs examining anxiety symptoms were excluded. A manual search of references was performed. Quality was assessed independently by two raters, using the Quality of Reporting of Meta-analyses checklist (QUOROM) and the Overview Quality Assessment Questionnaire (OQAQ). Kendall's tau measured inter-rater reliability with statistical significance at P=0.05. Means and standard deviations described the overall quality. A time series analysis was performed. RESULTS: In total, 136 titles and abstracts were reviewed; 48 were retrieved, including six from the manual search. Twelve were excluded (not pooled analyses, inappropriate condition or treatment or were duplicates), leaving 36 studies. Publication dates ranged from 1990-2007. Agreement was high; tau=0.856 (P<0.05) for QUOROM and 0.865 (P<0.05) for OQAQ. The mean overall QUOROM score was 65% (SD=18%). The 'results' category yielded the lowest quality scores while the introductions and discussions yielded the highest. The mean overall OQAQ score was 59% (SD=25%). Flaws in scientific quality were observed regarding 'avoidance of bias' and 'validity' while most studies linked the results to the primary objective appropriately. There was a small, non-significant increase in quality of MAs over the years observed. CONCLUSIONS: Reporting and scientific quality of MAs in anxiety were only fair. They were, however, higher than those previously reported for depression and critical care.