Large Language Model-Based Analysis of Statin Therapy Discussions and Sentiment on Social Media: Cross-Sectional...

Area of Science:

• Cardiovascular Medicine
• Health Informatics
• Natural Language Processing

Background:

• Statin therapy is underutilized despite proven cardiovascular benefits.
• Patient perspectives on statins are often underrepresented in clinical settings.
• Social media platforms offer insights into patient experiences with medications.

Purpose of the Study:

• To analyze patient themes, sentiment, and decision factors concerning statin therapy.
• To utilize large language models (LLMs) for analyzing Reddit discussions on statins.
• To understand patient-reported outcomes and concerns related to statin use.

Main Methods:

• Cross-sectional observational study of Reddit posts and comments mentioning statins (January 2022 - May 2025).
• Keyword-based searches via Reddit API to retrieve 5328 discussions.
• LLM-based pipeline for extracting themes, sentiment, clinical relevance, adverse effects, and adherence.

Main Results:

• Prominent themes included adverse effects (31.9%), decision-making factors (51.9% lab results, 38.2% physician advice), and alternative approaches (46.6%).
• Overall sentiment was neutral (34%) and negative (30.9%), with statin-directed sentiment also largely neutral (44.1%) or negative (25.2%).
• Common adverse effects mentioned were muscle pain (7.6%) and fatigue (6.5%); adherence issues were noted in 29.8% of discussions.

Conclusions:

• LLM analysis of Reddit discourse reveals significant negative sentiment, adherence concerns, and adverse effect narratives regarding statin therapy.
• Findings indicate a need for patient-centered communication strategies to address symptom attribution and uncertainty.
• Shared decision-making approaches are recommended to improve patient engagement and information access in digital health environments.