Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PVLens: Enhancing Pharmacovigilance Through Automated Label Extraction

Jeffery L Painter, Gregory E Powell, Andrew Bate

TL;DR

PVLens tackles the problem of outdated pharmacovigilance resources by automating extraction of labeled safety information from FDA SPLs and mapping to MedDRA, RxNorm, and SNOMED CT. The system couples dictionary-based NLP with UMLS-guided mappings and SrLC to produce a scalable, open-source repository of labeled adverse events and indications, validated against expert annotations for 97 drug labels (F1 = 0.882, recall = 0.983, precision = 0.799). It demonstrates strong recall and efficient processing (over 5,000 SPLs processed in under an hour), with a transparent review/adjudication workflow that maintains quality. PVLens represents a practical, real-time complement to existing resources, enabling routine safety surveillance and signal evaluation while laying the groundwork for broader regulatory integration and future AI-assisted adjudication.

Abstract

Reliable drug safety reference databases are essential for pharmacovigilance, yet existing resources like SIDER are outdated and static. We introduce PVLens, an automated system that extracts labeled safety information from FDA Structured Product Labels (SPLs) and maps terms to MedDRA. PVLens integrates automation with expert oversight through a web-based review tool. In validation against 97 drug labels, PVLens achieved an F1 score of 0.882, with high recall (0.983) and moderate precision (0.799). By offering a scalable, more accurate and continuously updated alternative to SIDER, PVLens enhances real-time pharamcovigilance with improved accuracy and contemporaneous insights.

PVLens: Enhancing Pharmacovigilance Through Automated Label Extraction

TL;DR

PVLens tackles the problem of outdated pharmacovigilance resources by automating extraction of labeled safety information from FDA SPLs and mapping to MedDRA, RxNorm, and SNOMED CT. The system couples dictionary-based NLP with UMLS-guided mappings and SrLC to produce a scalable, open-source repository of labeled adverse events and indications, validated against expert annotations for 97 drug labels (F1 = 0.882, recall = 0.983, precision = 0.799). It demonstrates strong recall and efficient processing (over 5,000 SPLs processed in under an hour), with a transparent review/adjudication workflow that maintains quality. PVLens represents a practical, real-time complement to existing resources, enabling routine safety surveillance and signal evaluation while laying the groundwork for broader regulatory integration and future AI-assisted adjudication.

Abstract

Reliable drug safety reference databases are essential for pharmacovigilance, yet existing resources like SIDER are outdated and static. We introduce PVLens, an automated system that extracts labeled safety information from FDA Structured Product Labels (SPLs) and maps terms to MedDRA. PVLens integrates automation with expert oversight through a web-based review tool. In validation against 97 drug labels, PVLens achieved an F1 score of 0.882, with high recall (0.983) and moderate precision (0.799). By offering a scalable, more accurate and continuously updated alternative to SIDER, PVLens enhances real-time pharamcovigilance with improved accuracy and contemporaneous insights.

Paper Structure

This paper contains 11 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. System Overview and Data Extraction
  4. Study Design and Product Selection
  5. Review and Adjudication Process
  6. Results
  7. False Negative Analysis
  8. Discussion
  9. Conclusion
  10. Declarations
  11. Acknowledgments

Figures (1)

  • Figure 1: PVLens Processing Pipeline Overview. [GUID = Global Unique Identifier, SrLC = Safety-Related Label Change]