AI may be able to detect unknown side effects of anti-obesity drugs, based on analysis of patient reports on social media.
GLP-1 receptor agonists, which are said to help alleviate obesity by suppressing appetite and lowering blood sugar levels, are generally considered to have few serious side effects. However, there is a lack of sufficient clinical trials regarding the side effects actually experienced by patients. A paper published by computer science researchers at the University of Pennsylvania suggests that using AI to broadly analyze online postings could potentially reveal side effects that were not known through clinical research.
Self-reported side effects of semaglutide and tirzepatide in online communities | Nature Health
AI Finds Potential Ozempic Side Effects Hidden in an Unexpected Data Source : ScienceAlert
https://www.sciencealert.com/ai-finds-potential-ozempic-side-effects-hidden-in-an-unexpected-data-source
Glucagon-like peptide-1 (GLP-1), secreted after meals, stimulates insulin secretion, slows digestion to retain food in the body, and suppresses appetite by creating a feeling of fullness in the brain. GLP-1 receptor agonists were developed to enhance and prolong the function of GLP-1. Examples of GLP-1 receptor agonists include semaglutide, known as the obesity treatment drug 'Ugobi' and the type 2 diabetes treatment drug 'Ozempic,' and tirzepatide, used to treat both type 2 diabetes and obesity. The first GLP-1 receptor agonist was developed in 2005 as a diabetes treatment drug and was approved as an obesity treatment drug in 2014. Subsequently, semaglutide and tirzepatide were introduced as drugs with even greater weight-loss effects.
What are the effects of Ozempic, a drug that may help cure obesity? - GIGAZINE
The research team analyzed over 410,000 posts on Reddit, one of the world's largest online forums, from May 2019 to June 2025, searching for posts that mentioned semaglutide or tilzepatide, major GLP-1 receptor agonists, or brand names such as 'Ozempic' that contain them. They then compiled information on side effects reported by users of these medications.
An analysis of over 410,000 posts revealed that 67,008 users self-reported using GLP-1 receptor agonists. Of those, 43.5% reported at least one side effect.
The most frequently reported side effects were gastrointestinal symptoms, with nausea at 36.9%, fatigue at 16.7%, vomiting at 16.3%, constipation at 15.3%, and diarrhea at 12.6%. Side effects such as nausea, diarrhea, and constipation are known to be mild side effects of GLP-1 receptor agonists.
Furthermore, reports of potentially previously unreported side effects have been confirmed, including reproductive symptoms such as menstrual irregularities and body temperature-related symptoms such as chills and hot flashes. On the other hand, posts have also been found regarding effects that have attracted attention in existing research, such as a decrease in alcohol cravings, in addition to side effects.
Sharath Chandra Guntuk of the University of Pennsylvania, a co-author of the paper, said, 'Clinical trials are the best method, but they are time-consuming due to their structure. Analyzing social media posts is not a substitute for clinical trials, but it is a much faster way to investigate when speed is of the essence. Another noteworthy point is that clues about symptoms that patients might not tell their doctors are obtained spontaneously from the patients themselves.'
This analysis utilized OpenAI's GPT model to extract and classify symptom descriptions in posts, enabling large-scale data processing and pattern discovery. While Reddit's user base is heavily skewed towards young American adults, and the data is based solely on self-reporting, it cannot be definitively concluded that GLP-1 receptor agonists actually caused these symptoms. However, the researchers emphasized its value as a method for quickly collecting specific testimonies and highlighting potential issues that might otherwise be overlooked in further research.
Lyle Unger, co-author of the paper, said, 'Clinical trials can generally identify the most dangerous side effects of a drug, but they may not identify the symptoms that patients are most concerned about. Social media is not necessarily a representative source of information, but if a large number of posts are aggregated, it may reflect further concerns.'
Related Posts:
in AI, Web Service, Science, Posted by log1e_dh