Askglp1 maps patient experiences with glp-1 medications online
AskGLP1: A Knowledge Graph LLM Agent for Exploring GLP-1 Experience from Online Patient Narratives
Computation and LanguageArtificial IntelligenceSocial and Information Networks
Summary
GLP-1 drugs help some people lose weight, but the information about what people experience when using them is scattered and hard to understand. The authors built AskGLP1, a tool that looks through millions of online posts and organizes patient stories about the drugs into a structured graph. This helps find common side effects and reasons for taking the drugs, and shows differences between online reports and official medical records. The tool can answer questions based on this information without giving medical advice.
What this means in practice
- •For pharmacovigilance teams: Use patient-reported online experiences structured into a knowledge graph to complement formal drug safety monitoring.
- •For digital health platform developers: Integrate knowledge graph-based question answering about medication experiences from patient narratives into health apps for better user insight.$Commercial implications: Enables selling enhanced patient engagement and drug experience features in consumer health applications.
Authors
Zhaoqian Xue, Anay Somani, Siddhartha Singh, Junting Mu, Zhongyin Wang, Haozheng Liu, Ayesha Khatun, Yeaeun Kwon, Renkai Ma, Siyuan Ma, Zhe He, Danxia Yu, Lingyao Li
Abstract
GLP-1 medications are increasingly used for weight management, and many people share their experiences online. These narratives show why people take the drugs and what they experience, but they remain unstructured and disconnected from formal pharmacovigilance. We therefore present AskGLP1, a system that transforms online patient narratives into a knowledge graph (KG) and answers questions about medication experiences. Through an LLM-assisted pipeline grounded in the MedDRA terminology, AskGLP1 screens nearly 8.5 million Reddit posts for medication experiences and distills 163,135 of them, with 2,506 WebMD reviews, into a KG of 385,790 drug-side-effect and drug-purpose relations. We find that GLP-1 experiences are source- and drug-specific. Online narratives share the most common gastrointestinal symptoms with FAERS, but they foreground everyday burdens such as fatigue and food aversion, whereas FAERS reports more often include injection-site reactions and selected serious diagnoses such as pancreatitis. We further find that posts report less nausea and vomiting for tirzepatide than for semaglutide, more fatigue for Wegovy than Ozempic and for Zepbound than Mounjaro. Evaluations show that the agent answers with verifiable evidence and declines requests for medical advice. Our findings reveal gaps between what FAERS captures and what people report online, and they call for patient-centered approaches to pharmacovigilance. Code, prompts, and evaluation questions are available at https://github.com/ZhaoqianXue/AskGLP1.