Self-Reported Side Effects of Semaglutide and Tirzepatide in Online Communities

This study analyzed over 410,000 Reddit posts to reveal that while gastrointestinal symptoms are the most common self-reported side effects of semaglutide and tirzepatide, online communities also highlight emerging concerns like reproductive and temperature-related issues that are often underrepresented in clinical trials and official labeling.

The Gist

Imagine you're trying to understand what it's really like to take a new, popular weight-loss or diabetes medication. The official "user manual" (the clinical trials and FDA labels) tells you the basics: "You might feel a little nauseous or have an upset stomach." It's like reading a weather report that only mentions "sunny" or "cloudy."

But what if you want to know if it also feels like "a sudden, unexpected hailstorm" or "a weird, static electricity shock"? That's where this study comes in.

The Big Picture: Listening to the Crowd

The researchers decided to skip the formal surveys and instead went to the digital "town square" of Reddit. Think of Reddit as a massive, global support group where millions of people anonymously chat in specific rooms (called subreddits) about their health.

They sifted through 410,000 posts from 2019 to 2025, looking for people who said, "I'm taking Semaglutide (like Ozempic/Wegovy) or Tirzepatide (like Mounjaro/Zepbound)." They found 67,000 real people sharing their raw, unfiltered stories.

What They Found: The Known and the Unknown

The study confirmed what doctors already knew, but it also found some "ghosts in the machine"—symptoms that the official manuals haven't really highlighted yet.

1. The "Gut Check" (The Knowns)
Just like the official warnings say, the most common complaints were all about the digestive system. It's like the medication sends a "tornado" through your stomach.

• Nausea was the big boss (37% of people felt it).
• Fatigue (feeling exhausted) was the second most common (17%).
• Vomiting, constipation, and diarrhea followed close behind.
• Analogy: If the medication were a car, the manual says, "It might make the engine sputter." The Reddit users are saying, "Yeah, the engine is sputtering, and it's also making the whole car shake!"

2. The "Hidden Signals" (The New Discoveries)
This is the exciting part. The researchers found symptoms that aren't usually on the "Top 10" list of side effects, but people are talking about them constantly.

• The "Period Problem": About 4% of users mentioned reproductive issues, like irregular periods or heavy bleeding.
  • Analogy: Imagine the medication is a conductor of an orchestra. We knew it changed the tempo of the drums (stomach), but users are noticing it's also messing with the violin section (hormones/reproductive system). The study suggests the drug might be hitting a "control center" in the brain (the hypothalamus) that manages both hunger and your monthly cycle.
• The "Temperature Glitch": People reported feeling sudden chills, hot flashes, or feverishness.
  • Analogy: It's like your body's thermostat got confused. You're freezing one minute and sweating the next, even though the room temperature hasn't changed. This might be linked to how the drug affects how your body burns energy.

The "Reddit Bias" (A Word of Caution)

The authors are very honest about the limitations. They compare their data to a "loud crowd."

• Who is talking? People who are having a bad time are much more likely to post online than people who are having a great time. It's like a restaurant review site: you rarely see a 5-star review from someone who just ate a normal meal, but you see hundreds of 1-star reviews from people who got food poisoning.
• Demographics: Reddit users tend to be younger, more male, and mostly American. This might not represent everyone taking the drug (like older adults or people in other countries).
• No Proof of Causality: Just because someone posted "I took this pill and then my period changed" doesn't 100% prove the pill caused it. Maybe it was stress, or diet, or something else.

The Takeaway: Why This Matters

Think of this study as a radar system.

• Traditional trials are like a lighthouse: they shine a bright, focused beam on the most obvious dangers.
• Social media analysis is like a radar: it picks up the faint, distant blips that the lighthouse might miss.

The Bottom Line:
While the main side effects are still stomach-related, this study suggests doctors and drug makers should start paying closer attention to fatigue, menstrual changes, and temperature swings. It's a call to action to update the "user manuals" and ask patients specifically about these "hidden" symptoms, rather than just waiting for them to show up in a formal clinical trial.

In short: The internet is a giant, noisy, but incredibly useful focus group that helps us see the full picture of what these powerful drugs are really doing to our bodies.

Technical Summary

1. Problem Statement

While clinical trials and regulatory documents (e.g., FDA labeling) have established the safety profiles of glucagon-like peptide-1 receptor agonists (GLP-1 RAs) like semaglutide and tirzepatide, these controlled environments often fail to capture the full spectrum of patient experiences in the real world.

• Gap: Clinical trials may miss rare, long-term, or subjective side effects, particularly those not formally recognized in drug labeling.
• Need: There is a need for large-scale, real-world pharmacovigilance to detect emerging safety signals, understand patient concerns beyond gastrointestinal (GI) symptoms, and identify potential adverse effects that warrant further investigation.

2. Methodology

The study employed a cross-sectional analysis of social media data using Natural Language Processing (NLP) and Large Language Models (LLMs).

• Data Source: Reddit, selected for its large, anonymous, topic-specific health communities (subreddits) and accessible, researcher-curated corpora.
• Timeframe: May 2019 to June 2025.
• Data Volume: 410,198 posts mentioning semaglutide or tirzepatide.
• Pipeline Architecture:
  1. User Identification: A GPT-4o-mini based classifier identified posts where users disclosed personal use of FDA-approved formulations for type 2 diabetes or weight management.
  2. Side Effect Extraction: A GPT-4.1-mini based classifier extracted self-reported side effects from posts indicating self-use.
  3. Standardization: Extracted symptoms were mapped to MedDRA Preferred Terms (PTs) to ensure standardized medical terminology.
• Cohort: 67,008 unique users were identified as self-reporting medication use. Of these, 29,172 users (43.5%) reported at least one side effect.
• Sub-analyses:
  • Temporal analysis of the top 10 most common symptoms.
  • Comparative analysis between users exclusively mentioning semaglutide vs. tirzepatide.

3. Key Results

A. Prevalence of Side Effects

• Overall: 43.5% of users reporting medication use disclosed at least one side effect.
• Average Burden: Users reported an average of 2.7 distinct MedDRA PTs.
• Top Symptoms (by frequency):
  1. Nausea: 36.9%
  2. Fatigue: 16.7%
  3. Vomiting: 16.3%
  4. Constipation: 15.3%
  5. Diarrhea: 12.6%
• System Organ Classes (SOCs):
  • Gastrointestinal Disorders: 65.8% of users with side effects.
  • General Disorders: 28.8% (driven largely by fatigue).
  • Nervous System: 19.6% (headache, dizziness).
  • Metabolism/Nutrition: 17.8% (decreased appetite).
  • Psychiatric Disorders: 12.9% (anxiety, insomnia, depression).

B. Novel and Under-Reported Signals

The study identified several symptoms not typically highlighted in current prescribing information or clinical trials:

• Reproductive Symptoms: ~3.8% of users reported menstrual irregularities, including intermenstrual bleeding, heavy bleeding, and irregular cycles.
• Temperature Regulation: Reports of chills, feeling cold, hot flashes, and pyrexia (fever) emerged as potential effects, possibly linked to glucagon's role in thermogenesis.
• Fatigue: While common in trials, it was the second most reported symptom on Reddit, often exceeding reporting thresholds seen in placebo-controlled studies.

C. Formulation Comparison (Semaglutide vs. Tirzepatide)

• Semaglutide Users: Reported higher rates of nausea (39.4%), vomiting (18.0%), and fatigue (16.1%).
• Tirzepatide Users: Reported lower rates of nausea (28.6%) and vomiting (11.1%) but showed distinct reports of injection site reactions, myalgia, and temperature-related symptoms (chills/cold) at rates of 1–4%.
• Note: The authors caution that these differences may reflect population demographics rather than pharmacological differences due to a lack of controlled comparison data.

4. Key Contributions

1. Large-Scale Real-World Evidence: Provided the largest analysis to date of self-reported GLP-1 RA side effects, covering over 67,000 users and 410,000 posts.
2. Detection of Emerging Signals: Successfully identified potential safety signals (reproductive issues, temperature dysregulation, severe fatigue) that are under-represented in clinical trial data and regulatory labeling.
3. Methodological Pipeline: Demonstrated a robust, LLM-based pipeline (GPT-4o-mini/4.1-mini) for extracting and mapping unstructured social media text to standardized medical ontologies (MedDRA), offering a scalable approach for pharmacovigilance.
4. Complement to Traditional Surveillance: Showed how social media analysis can act as an early warning system to complement passive reporting systems (like FAERS) by capturing unprompted, context-rich patient narratives.

5. Significance and Limitations

Significance

• Clinical Practice: Alerts clinicians to potential patient inquiries regarding menstrual changes and temperature fluctuations, suggesting these should be part of patient counseling.
• Regulatory Impact: Highlights the need for future prospective trials and post-market surveillance to systematically assess fatigue and reproductive health using validated instruments.
• Patient Safety: Emphasizes that real-world safety profiles may differ from trial data, necessitating updates to patient education materials.

Limitations

• Selection Bias: Reddit users are disproportionately younger, male, and US-based compared to the general GLP-1 RA population.
• Self-Reporting Bias: Users may over-report severe/distressing symptoms and under-report mild ones; causality cannot be established (symptoms may be due to weight loss or comorbidities rather than the drug).
• Lack of Metadata: Dosage, treatment duration, and route of administration were inconsistently reported, preventing dose-response analysis.
• NLP Constraints: While validated, NLP models may misclassify nuanced contexts or map user descriptions to incorrect MedDRA terms.
• Hypothesis-Generating: The findings are signals requiring confirmation through structured clinical studies and representative populations.

Conclusion

This study validates social media as a critical resource for expanding the understanding of GLP-1 RA safety profiles. By leveraging LLMs to analyze massive datasets, the authors identified a broader range of patient-reported side effects—including reproductive and thermoregulatory issues—than currently captured in clinical trials. These findings advocate for a more comprehensive approach to pharmacovigilance that integrates digital phenotyping with traditional surveillance to better serve patient safety.