Social posts reveal missed side effects

Weight-loss and diabetes drugs that mimic a gut hormone can slow digestion and curb appetite, and a new study says patients are also describing side effects online that trials may not fully capture. Researchers at the University of Pennsylvania used artificial intelligence to analyze 410,198 Reddit posts and found reports of menstrual changes, fatigue, chills and hot flashes. (nature.com) The paper, published April 10 in *Nature Health*, covered posts from May 2019 through June 2025 and identified 67,008 users who said they were taking semaglutide or tirzepatide. Of those users, 43.5% described at least one side effect. (nature.com) Many of the most common complaints matched existing drug labels: nausea appeared in 36.9% of users who reported side effects, fatigue in 16.7%, vomiting in 16.3%, constipation in 15.3% and diarrhea in 12.6%. The researchers said that overlap suggests the method was picking up real patient signals, not random chatter. (nature.com) (seas.upenn.edu) Drug labels for semaglutide and tirzepatide already list stomach-related problems among common side effects, and the current United States label for Zepbound also lists fatigue in at least 5% of patients. The Penn team said the social posts pointed to other symptom clusters that are not well reflected in current labeling or trials, especially reproductive and temperature-related complaints. (accessdata.fda.gov) (nature.com) The study did not prove the drugs caused those symptoms. Lead author Neil Sehgal said the findings are signals for follow-up, not evidence of cause and effect, and senior author Sharath Chandra Guntuku said clinicians could treat the posts as patient-generated leads. (seas.upenn.edu) One result stood out: nearly 4% of users in the sample reported menstrual irregularities. Co-author Lyle Ungar said that share would likely be higher in a female-only sample, which is why the team called it a signal worth investigating. (seas.upenn.edu) The researchers argued that social media can complement pharmacovigilance, the system regulators and drugmakers use to track safety after approval. Their paper says large-scale online analysis can help spot patterns that may not surface in clinical trials, doctor visits or formal adverse-event reports. (nature.com) (seas.upenn.edu) That does not make Reddit a perfect sample of patients taking these drugs. The authors said social media is not necessarily representative, but they argued that a large volume of unsolicited posts can still reveal concerns patients are discussing in real time. (seas.upenn.edu) For now, the study leaves the same bottom line it started with: the best-known stomach side effects are still there, and the less familiar symptoms now have a larger paper trail. The next step is not a label change on its own, but more targeted research to test whether those patterns hold up outside social media. (nature.com)