Penn AI Study Flags Menstrual and Thermoregulatory Signals in 400k GLP-1 Reddit Posts

The team mined five years of unprompted posts discussing Ozempic, Wegovy, Mounjaro and related drugs, clustering symptoms without relying on predefined lists. Known effects such as nausea appeared as expected, validating the pipeline, while reproductive and thermoregulatory clusters emerged as statistically prominent yet absent from most trial summaries. Senior author Sharath Chandra Guntuku and co-author Lyle Ungar emphasize the method detects patient priorities faster than trials but cannot prove causation.

Reddit data carry selection bias toward younger, English-speaking, female users willing to post publicly, and lack dosage, duration or comorbidity controls. Still, the volume exceeds spontaneous reporting systems like FAERS and can surface quality-of-life issues that trials, powered for serious adverse events, routinely miss. Parallel signals have appeared in smaller pharmacovigilance studies of GLP-1 agents and in post-marketing data for related metabolic drugs.

Next steps require targeted cohort studies measuring menstrual cycle length, ovulation markers and core-temperature logs in female users versus matched controls, ideally within 12-18 months. Regulators could integrate social-media signals into routine surveillance to accelerate label updates if incidence exceeds 3-5% in verified populations.