University of Pennsylvania researchers used AI to analyze over 400,000 Reddit posts from nearly 70,000 GLP-1 users and found patterns of symptoms — including menstrual irregularities and temperature changes — that aren't prominently featured in official prescribing information. Here's what the study found and what it means.
Clinical trials are tightly controlled and use standardized symptom reporting — which is rigorous, but can miss things patients describe in their own words that don't map cleanly onto a structured trial questionnaire. Online patient communities, the researchers argue, work differently: "like a neighborhood grapevine" where people compare real, unfiltered experiences in real time.
Using large language models to map informal patient language to official medical terminology (MedDRA, the standard used in clinical trial reporting), the team identified two clusters of symptoms discussed far more frequently online than their prominence in official labeling would suggest.
Irregular cycles, heavy bleeding, and intermenstrual bleeding showed up as a clear, repeated pattern. Critically, the study's sample skewed male (typical of Reddit's broader demographics), meaning the true rate among women is likely meaningfully higher than the 4% topline figure. Lead researcher Neil Sehgal called this "one of the most interesting signals" in the entire dataset.
A cluster of users reported feeling unexpectedly cold, or experiencing hot flash-like episodes, independent of any menopausal status. This pattern was distinct enough that researchers flagged it as a category worth dedicated clinical follow-up.
Despite being one of the most frequently mentioned symptoms in the entire 400,000-post dataset, fatigue has historically received relatively little attention in formal trial reporting thresholds.
Nausea and other gastrointestinal symptoms — already well-documented — also appeared prominently, which researchers note actually validates the method: "Some of the side effects we found, like nausea, are well known, and that shows that the method is picking up a real signal," said senior author Dr. Sharath Chandra Guntuku.
The study's authors are explicit that this research shows correlation, not proof of causation. "We can't say that GLP-1s are actually causing these symptoms," said Sehgal. People discussing a symptom while on a medication doesn't prove the medication caused it — but a pattern this consistent, across this many users, is what the researchers call "a signal worth investigating."
This research doesn't change the established safety profile of semaglutide or tirzepatide — both remain extensively studied, FDA-approved medications. What it adds is texture: a more complete picture of what real patients experience day to day, beyond what made it into the original trial reporting thresholds.
If you're experiencing menstrual changes, unusual temperature sensitivity, or persistent fatigue on a GLP-1 medication, this research is a legitimate reason to bring it up directly with your physician — not dismiss it as unrelated, and not assume it's definitely the medication either. It's exactly the kind of symptom pattern worth tracking and discussing.
The team believes this kind of AI-assisted social listening could become a genuinely useful early-warning tool for spotting safety signals faster than traditional reporting systems allow — particularly relevant, they note, for substances gaining rapid popularity in less-regulated markets, like compounded peptides. "The whole point of this kind of approach is that it can move quickly, and that's exactly when it's most valuable," said Guntuku.
Map your full hormone picture — not just what might be GLP-1 related.
Map my hormones free →Sourced, comprehensive breakdown of documented GLP-1 side effects.
Read the full guide →