AI Analysis of Reddit Posts Uncovers Patient-Reported GLP-1 Drug Symptoms

A study from the University of Pennsylvania used artificial intelligence to analyze hundreds of thousands of Reddit posts discussing GLP-1 drugs. The research identified patient-reported symptoms, including some not frequently highlighted in clinical trials, but the authors caution the findings show correlation, not causation, and are based on a non-representative sample.

Study Overview and Methodology

Researchers at the University of Pennsylvania School of Engineering and Applied Science published a study in the journal Nature Health. The research analyzed more than 400,000 Reddit posts from nearly 70,000 users over a period exceeding five years. The posts discussed the GLP-1 agonist drugs semaglutide and tirzepatide.

The team employed large language models (LLMs) in a method they describe as "computational social listening." This approach systematically mapped user descriptions of symptoms to standardized medical terminology from the Medical Dictionary for Regulatory Activities (MedDRA).

Reported Findings

The analysis found that approximately 44% of users in the study sample described experiencing at least one side effect. The most commonly reported issue was gastrointestinal distress.

The research highlighted two categories of symptoms that were reported by users but are not fully captured in existing clinical trial data or regulatory documents:

Reproductive Symptoms: These included reports of irregular menstrual cycles, intermenstrual bleeding, and heavy bleeding. Nearly 4% of users who reported any side effect described symptoms in this category.
Temperature-Related Complaints: Users reported symptoms such as chills, feeling cold, hot flashes, and fever-like sensations.

Fatigue was ranked as the second most common complaint among the Reddit users analyzed.

Researcher Statements and Context

Researchers involved in the study provided context for the findings:

Sharath Chandra Guntuku, the study's senior author, stated that the method's ability to identify well-known side effects like nausea validates its function. He noted the underreported symptoms "are leads that came from patients themselves, unprompted."
Lyle Ungar, a co-author, said clinical trials are designed to identify the most dangerous side effects but may not capture all symptoms patients find concerning. He added that a large collection of social media posts may reflect these additional concerns.
Neil Sehgal, the study's first author, explicitly stated the research cannot establish that GLP-1 drugs cause the reported symptoms. He characterized the report of menstrual irregularities as "a signal worth investigating."
Jena Shaw Tronieri, another co-author, noted that GLP-1 drugs engage the hypothalamus, a brain region involved in hormone regulation. This mechanism, she suggested, makes reports of menstrual changes and body temperature fluctuations worthy of more systematic study.

Study Limitations and Future Research

The authors emphasized several important limitations:

The findings demonstrate a correlation but do not prove the GLP-1 drugs cause the reported symptoms.
The Reddit user population is not representative of the general population. The sample was described as younger, more likely to be male, and disproportionately based in the United States.

The research team expressed hope that the findings will encourage further clinical investigation into the symptoms patients discuss online. They plan to expand the work to other social platforms and non-English language communities to test if similar patterns emerge across different populations. The researchers suggest AI-assisted social media analysis could become a tool for spotting early signals around emerging drugs.

Background and Disclosures

Lyle Ungar participated in one of the earliest efforts to mine online content for drug adverse effect information in 2011. The current study received no outside funding. Author Jena Shaw Tronieri reported receiving an investigator-initiated grant from Novo Nordisk (a manufacturer of semaglutide) and consulting fees from Currax Pharmaceuticals, LLC. The other authors reported no conflicts of interest.