A project analysing post-marketing reports on the side effects of drugs has uncovered noise and biases within the data that limit its usefulness for patients and healthcare professionals alike.
The joint project, primarily by researchers at Novartis, joined by those at the University of California, San Francisco (UCSF), involved looking at over 8.7 million entries in the Food and Drug Administration Adverse Event Reporting System (FAERS), a constantly growing database used in the US for reporting the side effects of licensed drugs. The team’s method for mining the vast amounts of information, which involved matching compounds with similar chemical properties, could be used to help researchers develop more effective drugs in future.
The results of the project, published in the journal eLife, show that while FAERS provides a solid frame for reporting, it also contains redundancies and biases that affect the ability to analyse and interpret the data. For instance, the average drug has 16 different names in the database, making it difficult to group all of its reported side effects so that users can see the trends and patterns correctly.
The FAERS database lacks a method for linking drugs to the chemical structures of their active ingredients. To address this, the team converted drug names into unique chemical structures, which could be readily searched. To allow for more accurate statistical analysis, the team corrected false associations such as side effect attributions to drugs that are treating the effects of the disease. For example, the team found that diabetes was sometimes listed as a side effect for drugs used to treat diabetes. Another approach included identifying and removing multiple reports of side effects on the same case. Biases were also identified and it was observed that in some cases, side effects were reported more often when drugs were featured in the news, regulatory announcements or during public events. Interestingly, this bias affected both the drug in question and other drugs that acted in the same way or on the same molecular target.
The authors suggest that while FAERS is already a valuable asset for clinicians and pharmaceutical scientists, it could be improved in several ways, such as introducing automatic mapping of drugs and synonyms to ingredients as presented in this study. This would allow the data to be combined and corrected, making it easier to evaluate the safety of different medicines and thereby reduce the occurrence of adverse drug reactions, currently one of the top 10 causes of death in the United States.
There are several limitations to the presented approach, which are more or less general for mining FAERS. These include difficulties in differentiating side effects associated with certain drugs when they were combined with other treatments. Nevertheless, the method could be used to help identify previously unidentified side effects and the biological mechanisms behind them, aiding researchers in developing new drugs with improved side effect profiles.
Media contacts
Emily Packer
eLife
e.packer@elifesciences.org
+441223855373
About
eLife is a unique collaboration between the funders and practitioners of research to improve the way important research is selected, presented and shared. eLife publishes outstanding works across the life sciences and biomedicine – from basic biological research to applied, translational and clinical studies. All papers are selected by active scientists in the research community. Decisions and responses are agreed by the reviewers and consolidated by the Reviewing Editor into a single, clear set of instructions for authors, removing the need for laborious cycles of revision and allowing authors to publish their findings quickly. eLife is supported by the Howard Hughes Medical Institute, the Max Planck Society and the Wellcome Trust. Learn more at elifesciences.org.
