Exploring adverse drug events at the class level.

Area of Science:

• Pharmacovigilance
• Drug Safety Science
• Computational Biology

Background:

• Adverse drug events (ADEs) are typically identified at the individual drug level.
• However, ADEs are frequently discussed and managed at the pharmacologic class level.
• Existing methods lack detailed exploration of individual drug contributions to class-level ADE signals.

Purpose of the Study:

• To develop and evaluate novel visual and computational approaches for dissecting individual drug contributions to pharmacologic class adverse event signals.
• To enhance the understanding of drug-ADE relationships beyond the class level.
• To provide tools for drug safety professionals and ADE repository curators.

Main Methods:

• Utilized a MEDLINE-derived dataset of adverse drug events (ADEs).
• Aggregated drugs into Anatomical Therapeutic Chemical (ATC) classes and ADEs into Medical Subject Headings (MeSH) terms.
• Computed statistical associations at both individual drug and drug class levels.
• Employed heatmap visualizations for multi-resolution signal exploration and clustering techniques for automated analysis.

Main Results:

• The visual approach successfully identified known drug-class ADE associations, such as fluoroquinolones with tendon injuries and statins with rhabdomyolysis.
• The computational approach systematically analyzed 488 distinct associations between drug classes and adverse drug events.
• Demonstrated the ability to uncover specific drug contributions within broader class signals.

Conclusions:

• The developed visual and computational methods offer valuable insights for drug safety professionals and ADE data curators.
• These techniques facilitate a deeper understanding of drug-ADE relationships by bridging individual drug effects and class-level trends.
• The proposed approaches are adaptable to various drug-ADE datasets, drug classification systems, and signal detection algorithms, offering broad applicability.