Related Experiment Video
Updated: Aug 6, 2026
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
Real-world pharmacovigilance for anti-Aβ therapies using a large language model
Allan Fong1,2, Azade Tabaie1,2,3, Samantha Paylor4
1Center for Biostatistics, Informatics, and Data Science MedStar Health Research Institute, MedStar Health Washington District of Columbia USA.
Alzheimer'S & Dementia (Amsterdam, Netherlands)
|July 23, 2026
Summary
Large language models (LLMs) show promise for improving Alzheimer's disease drug safety monitoring. An LLM-augmented workflow demonstrated high sensitivity in detecting adverse events (AEs) from clinical documents, aiding pharmacovigilance.
Area of Science:
- Neuroscience
- Pharmacology
- Artificial Intelligence
Background:
- Alzheimer's disease (AD) anti-amyloid beta (Aβ) therapies necessitate robust safety surveillance.
- Adverse event (AE) data is often fragmented across electronic health records (EHRs).
- Large language models (LLMs) offer potential for scalable pharmacovigilance.
Purpose of the Study:
- To evaluate the efficacy of an LLM-augmented workflow for detecting therapy-related AEs in early AD patients.
- To compare LLM performance against standard expert review in identifying AEs.
Main Methods:
- Analysis of 20,123 clinical documents from 46 patients receiving anti-Aβ therapy.
- Comparison of LLM-augmented review with standard expert annotation of AEs.
- Adjudication of discordant cases to establish a reference standard.
Main Results:
- 76% of patients experienced at least one AE.
- The LLM-augmented workflow achieved 100% sensitivity with 89.7% PPV.
- Expert review showed 88.6% sensitivity with 100% PPV.
Conclusions:
- LLMs show potential as pharmacovigilance tools for AE detection.
- Further validation and clinical integration studies are required.
- LLM-augmented workflows may enhance the safety monitoring of AD therapies.
