Improving results and conclusions about harms in systematic reviews of drugs

This project is a collaboration with University of North Carolina (main site) and University of Colorado. Halil Kilicoglu serves as a co-PI and the lead at the University of Illinois. 

Systematic reviews are used to develop clinical guidelines, to set research priorities, and to help patients and clinicians make healthcare decisions. Systematic reviews of drugs should identify harms and patient safety concerns; however, they mainly assess benefits, and evidence suggests that conclusions in systematic reviews about adverse effects might not be trustworthy. When systematic reviews reach incorrect conclusions about harms, patients, healthcare providers, and payers might make suboptimal decisions.

In this project, researchers will analyze ~20K systematic review articles for drugs using NLP methods and will examine whether information about adverse effects/safety is consistent between different reviews of the same drug,  between systematic reviews and studies that use real-world evidence (RWE), and between systematic reviews and  drug product labels. Based on best evidence, the team will develop contemporary guidance for synthesizing and reporting safety information in systematic reviews. 

More specifically, the project team will:

• Analyze results and conclusions about patient safety in systematic reviews of the same drug and assess whether the evidence is consistent.
• Analyze whether systematic reviews use RWE and whether their results and conclusions address harms found in RWE. RWEs are defined evidence obtained outside randomized controlled trials and include evidence from electronic health records, claims databases, and surveillance systems.
• Analyze whether information about harms in systematic reviews consistent with information in drug labels.
• Conduct surveys and develop new guidelines for harms reporting in systematic reviews.