The largest pharmacovigilance study of GLP-1 drugs wasn't run by the FDA, Novo Nordisk, or Eli Lilly. It was mined from Reddit by four researchers at the University of Pennsylvania.
Their study, published in Nature Health, analyzed more than 400,000 posts from nearly 70,000 users who self-reported taking semaglutide or tirzepatide. About 44% of them described at least one side effect. Large language models translated informal language into the Medical Dictionary for Regulatory Activities, the standardized vocabulary used to classify adverse events in clinical settings. Then the team looked for patterns the clinical trials missed.
The known side effects showed up right on schedule. Nausea at 36.9%. Vomiting at 16.3%. Constipation at 15.3%. This is the control group, in a sense. The method picks up real signal.
Then came the ones that aren't in the drug labels.
Nearly 4% of users who reported side effects described menstrual irregularities: intermenstrual bleeding, heavy periods, irregular cycles. Temperature dysregulation appeared too. Chills. Hot flashes. Feeling persistently cold. Fatigue was the second most frequently reported symptom overall, at 16.7%, though it rarely reaches clinical trial reporting thresholds.
The biological explanation is disarmingly simple. GLP-1 receptor agonists suppress appetite by acting on the hypothalamus. The hypothalamus also regulates the menstrual cycle. It also regulates body temperature. It also coordinates hormonal rhythms involved in energy and fatigue. Jena Shaw Tronieri, a co-author from Penn's Center for Weight and Eating Disorders, put it directly: "These drugs are thought to work by engaging part of the brain called the hypothalamus, which helps regulate a wide variety of hormones."
You can't pharmacologically reach into the body's thermostat and only turn down hunger. The thermostat has other settings.
Clinical trials are designed to test whether a drug works and to flag dangerous complications. They run for defined periods in carefully selected populations, typically a few thousand participants. What they don't capture well are symptoms that patients consider tolerable enough to live with: the persistent chill, the period that comes a week early, the fatigue that never quite lifts. Those complaints go to Reddit.
The FDA's Adverse Event Reporting System, the official channel for post-market drug safety data, captures roughly 6% of actual adverse drug reactions according to published estimates. The median underreporting rate is 94%. Patients forget, doctors don't file, mild symptoms don't feel worth the paperwork. The system was designed for catching serious and rare events, not for mapping the full terrain of what a drug does to a body.
Reddit fills that gap by accident. "Online patient communities work a lot like a neighborhood grapevine," said Lyle Ungar, a Penn professor and co-author who helped pioneer the use of internet posts for drug safety signals in 2011. "People who are living with these medications are swapping notes with each other in real time, sharing experiences that rarely make it into a doctor's office visit or an official report."
Before large language models, connecting hundreds of thousands of informal posts to standardized medical vocabulary required enormous manual effort. One person writes "I've been freezing all day." Another says "constant chills." A clinician might categorize both under a specific MedDRA code. GPT and Gemini can now perform that translation at scale, consistently, across five years of posts. The Penn team's senior author, Sharath Chandra Guntuku, calls it "computational social listening."
The sample has clear limits. Reddit users skew young, male, and American. The study establishes association, not causation. Tronieri disclosed an investigator-initiated grant from Novo Nordisk. None of this invalidates the signal. It contextualizes it.
The investment question is whether the signal reaches the prescribing office. Roughly 12% of American adults now take a GLP-1 drug. Eli Lilly's Mounjaro and Zepbound generated approximately $36 billion in 2025. Novo Nordisk trades around $43 per ADR as of September 14, down roughly 70% from its mid-2024 peak, holding about 40% of the weight-loss market to Lilly's 60%. Women of reproductive age are a core demographic for these prescriptions. If the FDA adds label warnings for menstrual disruption based on real-world evidence, prescribing behavior could shift. Label changes don't require proof of causation. They require enough signal to justify a warning.
A separate study from NYU Grossman School of Medicine, published in the Journal of Investigative Dermatology the same week, found that men genetically predisposed to male-pattern baldness face a 7% higher risk of hair loss on GLP-1 drugs. Two papers, two universities, one conclusion: the side-effect profile patients experience doesn't match the one printed on the label.
Clinical trials test whether a drug works. The grapevine captures what living on it feels like.