Drug-safety monitoring has a blind spot, and a team at the University of Pennsylvania is testing whether social media can help close it. The researchers turned AI loose on a large body of online conversation to produce what they describe as the first systematic look at symptoms that self-described users of retatrutide attribute to the drug. Roughly 40,000 Reddit contributors, whose posts and comments number close to 150,000, formed the raw material.
Retatrutide belongs to the next generation of GLP-1 weight-loss drugs and has not won approval for human use. That has not stopped a gray market from forming, where people buy versions never meant for human consumption and then trade notes about the results in public forums.
Testable hypotheses, not a finished safety profile, are what the team says it set out to produce. Recurring patterns in what people reported were surfaced by the AI, and some of those patterns may slip past the conventional channels that track drug safety after approval.
One contrast stood out. Among participants the analysis could classify, appetite increase came up most often — a result that runs against the drug’s Phase 2 clinical trials, where gastrointestinal problems topped the list of reported side effects.
Menstrual changes and temperature-related complaints were also flagged across the corpus, symptoms the group says clinical trials have never fully characterized.
Sharath Chandra Guntuku, who directs Penn’s Computational Social Science Lab, led the work. Because the Reddit sample carries built-in limits, the authors caution that their results should be read as hypothesis-generating rather than conclusive.
