Improving results and conclusions about harms in systematic reviews of drugs

Systematic reviews (SRs) of healthcare interventions should identify patient safety concerns; however, many SRs are designed to assess benefits, and preliminary evidence suggests that conclusions about harms in SRs might not be trustworthy. Using >19,000 SRs of drugs for which we have full texts (including >4,000 overlapping reviews), and using innovative methods for analysis (e.g., natural language processing), we will examine whether information about safety in SRs is consistent, whether SRs address safety concerns identified using real-world evidence, and whether SRs are consistent with safety information on drug labels. Based on best evidence, including findings from this study, we will develop contemporary guidance for synthesizing and reporting safety information in SRs.