PVLens: Enhancing Pharmacovigilance Through Automated Label Extraction

Area of Science:

• Pharmacovigilance and Drug Safety
• Health Informatics
• Regulatory Science

Background:

• Existing drug safety databases like SIDER are outdated and static, hindering real-time pharmacovigilance.
• Accurate and continuously updated safety information is crucial for monitoring drug risks.
• There is a need for automated systems to efficiently extract and standardize safety data from regulatory documents.

Purpose of the Study:

• To introduce PVLens, an automated system for extracting labeled drug safety information from FDA Structured Product Labels (SPLs).
• To map extracted safety terms to the Medical Dictionary for Regulatory Activities (MedDRA) terminology.
• To provide a scalable, accurate, and continuously updated resource for pharmacovigilance.

Main Methods:

• Developed an automated system (PVLens) to extract safety information from FDA SPLs.
• Integrated term mapping to MedDRA for standardized data representation.
• Implemented a web-based review tool for expert oversight and validation.
• Validated the system against 97 drug labels.

Main Results:

• PVLens achieved a high F1 score of 0.882 in validation.
• The system demonstrated high recall (0.983) and moderate precision (0.799).
• PVLens successfully extracted and mapped safety information, outperforming static resources.

Conclusions:

• PVLens offers a scalable and accurate solution for drug safety data extraction.
• The system enhances real-time pharmacovigilance by providing contemporaneous insights.
• PVLens represents a significant advancement over outdated, static drug safety databases.