Introduction Harmful effects of medicines should be reviewed with similar rigour as therapeutic benefits. Most evidence on harms is obtained from post marketing surveillance, so the use of meta-analysis to pool safety information presents challenges of inherent biases and differences in study designs. Yet, it's crucial to provide an accurate safety profile of pharmacological interventions. We aimed to describe the data sources of published meta-analysis of adverse drug effects.

Method We searched meta-analysis published in the last 5 years in six medical journals with the highest impact factor. All the meta-analysis focussing primarily on adverse effects of pharmacological interventions, with pooled results, were included and the characteristics assessed.

Results A total of 61 meta-analysis were included, 16 were published in JAMA, 13 in Lancet, 11 in BMJ, 10 in Ann Intern Med, nine in Arch Intern Med and two in NEJM. Of these 90.2% (n=55) included only experimental studies, two included only observational studies and four meta-analysis comprised both type of studies. Less than half (47.5%; n=29) of the meta-analysis assessed the quality of the included studies according to specified recommendation statements, and only 18 (29.5%) considered unpublished studies.

Conclusion The majority of meta-analysis of adverse drug effects included only experimental studies, less than half assessed their quality and few considered unpublished studies. These results reinforce the need for methodological research to clarify the role of meta-analysis in Pharmacovigilance and evaluate how to pool safety information from different surveillance methods, to provide an accurate safety profile of pharmacological interventions.