GLP-1 Side Effects and Safety
The published safety evidence for GLP-1 agonists. Includes GI and serious adverse events, head-to-head tolerability, psychiatric signals, patient-reported symptoms, and the FDA's January 2026 suicidality warning reversal.
Last updated: April 21, 2026
Context
Overview
This page compiles adverse event data for GLP-1 receptor agonists across FDA prescribing information, pivotal trials (STEP, SURMOUNT, SELECT, LEADER, SUSTAIN-6, SURPASS, SURMOUNT-5), real-world claims databases (Prime Therapeutics, Truveta, Cleveland Clinic), pharmacovigilance (FAERS), and a large social media analysis.
Other pages in this series cover pricing, persistence and adherence, weight regain after stopping, ROI evidence, and real-world weight loss.
Most common
Gastrointestinal side effects
GI events are the most common adverse effects across all GLP-1 agents. Rates peak during dose escalation (weeks 4–20) and decline substantially thereafter. In the STEP 1-3 pooled analysis, 98.1% of GI events were rated mild or moderate and 99.5% were non-serious.
GI adverse events by agent
| Event | Drug (dose) | On drug | On placebo | Source |
|---|---|---|---|---|
| Nausea | Semaglutide 2.4 mg (Wegovy) | 43.9% | 16.1% | STEP 1-3 pooled |
| Nausea | Oral sema 25/50 mg | ~35% | ~8% | OASIS-4 |
| Nausea | Tirzepatide 15 mg (Zepbound) | 25–31% | 9.5% | SURMOUNT-1 |
| Nausea | Liraglutide 3.0 mg (Saxenda) | 39–40% | 14–15% | SCALE |
| Nausea | Semaglutide 1.0 mg (Ozempic) | 20.3% | 6.1% | Ozempic PI |
| Nausea | Dulaglutide 1.5 mg (Trulicity) | 21.1% | 5.3% | Trulicity PI |
| Vomiting | Semaglutide 2.4 mg | 24.5% | 6.3% | STEP 1-3 pooled |
| Vomiting | Tirzepatide 15 mg | 8–13% | 1.7% | SURMOUNT-1 |
| Vomiting | Liraglutide 3.0 mg | 15.7% | 3.9% | Saxenda PI |
| Diarrhea | Semaglutide 2.4 mg | 29.7% | 15.9% | STEP 1-3 pooled |
| Diarrhea | Tirzepatide 15 mg | 23% | 7.3% | SURMOUNT-1 |
| Diarrhea | Liraglutide 3.0 mg | 20.9% | 9.9% | Saxenda PI |
| Constipation | Semaglutide 2.4 mg | 24.2% | ~11% | STEP 1-3 pooled |
| Constipation | Tirzepatide 10–15 mg | 11–17% | 5.8% | Zepbound PI |
| Abdominal pain | Semaglutide 2.4 mg | ~20% | ~11% | Wegovy PI |
| Dyspepsia | Semaglutide 2.4 mg | ~9% | ~4% | Wegovy PI |
| Eructation (burping) | Semaglutide 2.4 mg | ~7% | ~1% | Wegovy PI |
Sources: Wharton et al., Diabetes Obes Metab 2022 (STEP 1-3 pooled); FDA prescribing information for Wegovy, Ozempic, Zepbound, Saxenda, Trulicity; Jastreboff et al., NEJM 2022 (SURMOUNT-1).
GI events peak during dose escalation (weeks 4–20) and decline substantially thereafter. Slower titration reduces GI AE incidence; survodutide's phase 2 program with rapid escalation saw 20% AE-related discontinuation, which phase 3 reduced through slower titration. Mediation analyses show GI adverse events account for less than 6% of weight loss — efficacy is largely independent of side effects.
Observational data show a signal for gastroparesis and delayed gastric emptying. Sodhi et al. (JAMA, 2023) reported HR 3.67 (95% CI 1.15–11.90) vs. bupropion-naltrexone; a BMI-matched replication (Liu et al., DDW 2024) confirmed HR 2.30 (1.19–4.46). FDA added ileus to the Ozempic label in September 2023. Case reports indicate symptoms improve upon discontinuation. Approximately 5.1% of GLP-1 RA users experienced gastroparesis in a 2024 analysis.
Treatment persistence
Discontinuation due to adverse events
Trial discontinuation rates are substantially lower than real-world discontinuation, and the drivers differ. Trial AE-related discontinuation is typically single-digit; real-world discontinuation is dominated by cost and access in mixed populations, and shifts toward side effects in fully-covered populations.
Trial discontinuation ranges from ~5% in obesity trials (68–72 weeks) to 16.6% in SELECT (3.3 years in older patients with cardiovascular disease). Real-world 1-year discontinuation is 37–67% depending on cohort year and coverage. The "cost = 48% of discontinuation" figure applies to mixed commercial populations; in fully-covered populations, side effects account for ~15–30% of stopping.
Trial discontinuation rates
| Drug / Trial | Duration | On drug | On placebo | Notes |
|---|---|---|---|---|
| Semaglutide 2.4 mg (STEP 1-3) | 68 weeks | 6.8% | 3.2% | Obesity, healthier population |
| Semaglutide 2.4 mg (SELECT) | 3.3 years | 16.6% | 8.2% | Older, CVD population; AEs accumulate over longer exposure |
| Oral semaglutide 25/50 mg (OASIS-4) | 64 weeks | ~8–12% | ~3% | Higher dose of oral formulation |
| Tirzepatide 5/10/15 mg (SURMOUNT-1) | 72 weeks | 4.3% / 7.1% / 6.2% | 2.6% | Dose-dependent |
| Tirzepatide MTD (SURMOUNT-5) | 72 weeks | 6.1% | 8.0% (sema arm) | Lower than semaglutide head-to-head |
| Liraglutide 3.0 mg (SCALE) | 56 weeks | 9.8% | 4.3% | Highest obesity-trial discontinuation |
| Liraglutide 1.8 mg (LEADER) | 3.8 years | ~9.5% | ~7.3% | T2D CVOT |
| Dulaglutide 1.5 mg (REWIND) | 5.4 years | ~9.1% | ~6.3% | T2D CVOT |
Real-world discontinuation by setting
| Setting | 1-year | 2-year | Primary drivers |
|---|---|---|---|
| Trial (STEP) | 7–11% | — | Side effects (controlled context) |
| Academic obesity clinic, no-cost access | ~45% | — | Side effects, plateau, life factors |
| Commercial plans with obesity coverage | ~50–55% | ~75% | Side effects, goal achievement, PA renewals |
| Commercial mixed (Truveta, T2D) | 47% | — | Cost, access |
| Commercial mixed (Truveta, no T2D) | 65% | — | Cost, access, coverage changes |
| Prime Therapeutics 2021 initiators | 67% | 85% | Cost + access + shortages (2021–2023) |
| Prime Therapeutics Q1 2024 initiators | 37% | Not yet available | Improved supply and coverage |
| Cleveland Clinic (Gasoyan et al.) | — | — | 48% cost, 15% side effects, 12% shortages |
Sources: FDA prescribing information; pivotal trial publications (SELECT, STEP, SURMOUNT-1, SURMOUNT-5, SCALE, LEADER, REWIND); Prime Therapeutics Year-3 persistence study (June 2025); Truveta/JAMA 2025 analysis of 125,474 patients; Gasoyan et al., Obesity 2025. See Page 3 for the full persistence analysis.
Some real-world "discontinuation" reflects clinically successful behavior: patients reaching weight goals and transitioning to maintenance doses, seasonal breaks, or switching between GLP-1 products (38% of Prime's 3-year cohort switched at least once). Persistence metrics don't distinguish these from true treatment failure.
Less common events
Serious and rare adverse events
Beyond the common GI effects, several less frequent events warrant attention. Some are well-established (gallbladder disease, thyroid C-cell boxed warning), others are emerging signals (NAION), and one major warning was recently reversed (suicidality).
Gallbladder events (cholelithiasis, cholecystitis) are the most consistently elevated serious risk, with absolute rates of 1–3% over trial duration. Pancreatitis risk is not significantly increased in CVOT meta-analyses. Thyroid carcinoma is a rodent-based boxed warning with no confirmed human cases in SELECT. NAION is an emerging signal with very low absolute risk. In January 2026, the FDA requested removal of the suicidal ideation warning based on 91 RCTs and 107,910 patients showing no increased risk.
| Event | Rate (drug vs placebo) | Evidence | Source |
|---|---|---|---|
| Cholelithiasis (gallstones) | 1.6% vs 0.7% (sema 2.4 mg) | Well-established, dose- and weight-loss-dependent | Wegovy PI; He et al., JAMA Intern Med 2022 |
| Cholecystitis | 0.6% vs 0.2% (sema); 0.7% vs 0.2% (tirz) | Well-established | Wegovy PI; Zepbound PI |
| Cholecystectomy (meta-analysis) | RR 1.70 (1.25–2.32) | 27 excess events/10,000 patient-yrs | He et al., JAMA Intern Med 2022 |
| Acute pancreatitis (SELECT) | 0.2% vs 0.3% | Not significant (P=0.28) | SELECT trial |
| Pancreatitis (CVOT meta-analysis) | OR 1.05 (0.78–1.40) | Not significant | Cao et al. 2020; JACC 2025 |
| Pancreatitis (real-world, obesity) | +1 event per 250 treated | Emerging signal; observational | Prime Therapeutics Year-2 |
| Thyroid malignancy (SELECT) | 6 vs 8 cases (<0.1%) | No signal; all 3 MTCs in placebo group | SELECT trial (17,604 pts, 3.3 yr) |
| Thyroid cancer (meta-analysis) | Little or no effect | 48 RCTs, 94,245 patients | Annals of Internal Medicine, Dec 2025 |
| Diabetic retinopathy (SUSTAIN-6) | 3.0% vs 1.8% | HR 1.76 (1.11–2.78); baseline DR is key predictor | Marso et al., NEJM 2016 |
| NAION (semaglutide, T2D) | 8.9% vs 1.8% at 36 mo (single center) | HR 4.28; emerging signal; absolute risk ~1.4–2.5 per 10,000 PY | Hathaway et al., JAMA Ophthalmol 2024 |
| Suicide/self-injury (SELECT) | 0.11% vs 0.11% | Balanced; FDA removed warning Jan 2026 | SELECT; FDA DSC, Jan 13, 2026 |
| Acute kidney injury (SELECT) | 1.9% vs 2.3% | Not significant | SELECT trial |
| Heart rate increase | +1–4 bpm class effect | Despite this, CVOTs show CV benefit (SELECT HR 0.80) | Product labels; SELECT |
| Hair loss (tirzepatide, females) | 7.1% vs 1.3% | Well-established; largely weight-loss-mediated | Zepbound PI |
| Malignant neoplasms (SELECT) | 4.8% vs 4.7% | Not significant (P=0.92) | SELECT trial |
On January 13, 2026, the FDA formally requested removal of the suicidal ideation warning from Saxenda, Wegovy, and Zepbound labels. The decision followed an FDA meta-analysis of 91 RCTs covering 107,910 patients (60,338 on GLP-1, 47,572 on placebo) that found no increased risk of suicidal behavior, ideation, or other psychiatric adverse events. A separate FDA Sentinel System analysis of 2.2M users reached the same conclusion. Multiple independent analyses suggest GLP-1 agonists may have a protective effect on suicidal ideation (Wang et al., Nature Medicine 2024: HR 0.27 for incident cases).
The FDA boxed warning on all GLP-1 agonists (except lixisenatide) is based on rodent carcinogenicity studies showing dose-dependent C-cell hyperplasia. GLP-1 receptors are highly expressed in rodent thyroid C-cells but are low or absent in normal human thyroid. No confirmed human MTC cases from GLP-1 exposure. EMA did not adopt an equivalent warning. The label remains precautionary; contraindicated in personal or family history of medullary thyroid carcinoma or MEN 2 syndrome.
Non-arteritic anterior ischemic optic neuropathy is an emerging signal with 2–8× relative risk increase in some retrospective studies, but absolute risk remains very low — approximately 1 additional case per 4,000–7,000 treated patients per year. No RCT confirmation. No FDA label change to date. Larger general-population analyses (Klonoff, Atropos 66M patients) showed no significant increase. Patients with baseline optic disc crowding, sleep apnea, or prior NAION warrant discussion before initiation.
In October 2024, a multi-society guideline (ASA/AGA/ASMBS/ISPCPO/SAGES) reversed the 2023 recommendation to hold GLP-1 agonists before elective surgery. Most patients may continue the drug with a risk-stratified approach. High-risk patients (escalation phase, significant GI symptoms, higher doses) should follow a liquid-only diet for 24 hours pre-procedure or undergo point-of-care gastric ultrasound.
Mental health signals
Psychiatric and neurological effects
The psychiatric safety profile is more reassuring than early signals suggested. Large prospective studies show neutral-to-beneficial effects on depression and suicidality, and emerging data show reductions in alcohol use, opioid overdose rates, and possibly Alzheimer's disease risk.
Multiple signals suggest potential mental health benefits with GLP-1 agonists. Swedish cohort data show 42% lower risk of worsening mental illness with semaglutide. The first RCT of semaglutide for alcohol use disorder (Hendershot et al., JAMA Psychiatry 2025) showed significant reduction in cravings and heavy drinking days. Observational data suggest 40–70% reduction in Alzheimer's diagnosis risk. "Food noise" reduction reflects genuine GLP-1 receptor–mediated attenuation of reward signaling.
| Effect | Finding | Direction | Source |
|---|---|---|---|
| Suicidality (FDA meta-analysis) | No increased risk in 91 RCTs, 107,910 patients | Null/Protective | FDA Jan 2026 |
| Suicidal ideation (Wang et al.) | HR 0.27 incident; HR 0.44 recurrent | Protective | Nature Medicine 2024 |
| Worsening mental illness (Swedish cohort) | 42% lower with sema; 18% lower with lira | Protective | Lancet Psychiatry 2026 |
| Depression (STEP post-hoc) | No psychiatric harm signal | Null | JAMA Intern Med 2024 |
| Alcohol use disorder (RCT, n=48) | Reduced cravings, heavy drinking days, consumption | Protective | Hendershot et al., JAMA Psychiatry 2025 |
| Opioid overdose (observational) | 40% lower overdose rate | Protective | Addiction 2024 |
| Alzheimer's disease risk | 40–70% reduction vs other T2D drugs | Protective (observational) | Wang et al., Alzheimer's & Dementia 2024 |
| Dementia (CVOT post-hoc) | 53% reduction | Protective | Nørgaard et al. 2022 |
| FAERS psychiatric signals | Insomnia 11.7%, anxiety 11.0%, depression 7.5% | Signal only | FAERS 2004–2023 (8,240 reports) |
"Food noise" reduction — the subjective quieting of intrusive thoughts about food — reflects GLP-1 receptor activation in the mesolimbic reward pathway, which dampens dopamine-driven food seeking. This mechanism appears to extend beyond food to other appetitive behaviors, explaining the alcohol and substance use reduction signals. The alcohol reduction is the most clinically validated neuropsychiatric benefit beyond metabolic effects.
Cross-agent
Head-to-head tolerability
Only one completed head-to-head obesity trial has directly compared GLP-1 agents: SURMOUNT-5 (tirzepatide vs. semaglutide). Despite tirzepatide producing 47% greater weight loss, it had lower AE-related discontinuation. The GIP receptor component may buffer GI tolerability.
In SURMOUNT-5 (Aronne et al., NEJM 2025), tirzepatide at maximum tolerated dose had 2.7% GI discontinuation vs. 5.6% for semaglutide, and 6.1% all-cause AE discontinuation vs. 8.0%. Among the currently-relevant obesity and T2D agents, nausea rates at therapeutic dose roughly rank from lowest to highest: oral semaglutide 14 mg < semaglutide 1.0 mg < tirzepatide 15 mg < oral semaglutide 25/50 mg < liraglutide 3.0 mg < semaglutide 2.4 mg.
| Agent | Nausea (therapeutic dose) | GI discontinuation | Notes |
|---|---|---|---|
| Oral semaglutide 14 mg (Rybelsus) | 16% | ~6% | Lower effective exposure than injectable |
| Semaglutide 1.0 mg (Ozempic) | 20% | ~3% | T2D dose |
| Tirzepatide 15 mg (Zepbound) | 25–31% | 4.3% | Better tolerability vs. sema head-to-head |
| Oral semaglutide 25/50 mg (OASIS-4) | ~35% | ~8–12% | Higher-dose oral formulation |
| Liraglutide 3.0 mg (Saxenda) | 39–40% | 6.2% | Daily injection |
| Semaglutide 2.4 mg (Wegovy) | 43.9% | 6.8% | Highest approved obesity dose |
Sources: FDA prescribing information; Aronne et al., NEJM 2025 (SURMOUNT-5).
Despite tirzepatide producing 47% greater weight loss, head-to-head it had lower GI-related discontinuation (2.7% vs 5.6%) and lower all-cause AE discontinuation (6.1% vs 8.0%) compared to semaglutide. In SURPASS-2 (T2D), nausea rates were comparable (17–22% tirz vs 18% sema 1 mg) but discontinuation still favored tirzepatide at comparable glycemic efficacy. This is the most directly relevant head-to-head tolerability data currently available for formulary or prescribing decisions.
Qualitative signal
Patient-reported symptoms
Social media analysis captures experiences that trials under-measure, but the signal is qualitative, not epidemiological. The largest systematic analysis (Sehgal et al., Nature Health, April 2026) used LLMs to extract symptoms from 410,198 Reddit posts, identifying 67,008 self-identified GLP-1 users, of whom roughly 44% described at least one side effect.
This is qualitative social-media data, not a representative sample. Reddit users skew younger, male, and US-based. Reporting bias favors negative experiences; satisfied patients rarely post. The authors themselves emphasize these are signals worth investigating, not conclusions. Do not cite these numbers as rates. The concordance between common GI symptoms here and in trials is a useful validity check, but every symptom without a clinical correlate should be treated as a lead for further study.
Reported GI symptom frequencies track the direction of trial data for nausea, vomiting, diarrhea, and constipation. Symptoms that appear under-captured in trials include fatigue, sulfur burps, taste changes or food aversions, and temperature dysregulation. Reproductive symptoms (menstrual irregularities) were mentioned by a small but meaningful slice and lack a trial correlate — the most notable new signal the study surfaced.
| Symptom | Reported frequency band | Trial concordance | Classification |
|---|---|---|---|
| Nausea | Most common (~1 in 3 reporters) | Matches trial data | Established |
| Fatigue | Common (~1 in 6 reporters) | Partial — in PI but low trial reporting thresholds | Possibly under-recognized |
| Vomiting | Common (~1 in 6 reporters) | Matches trial data | Established |
| Constipation | Common (~1 in 6 reporters) | Matches trial data | Established |
| Psychiatric (anxiety, insomnia) | Moderate | STEP showed no harm signal; FAERS shows signals | Mixed evidence |
| Diarrhea | Moderate | Matches trial data | Established |
| Sulfur burps | Frequently mentioned | "Eructation" in PI obscures sulfur-specific quality | Possibly under-captured |
| Taste changes / food aversions | Frequently mentioned | "Decreased appetite" in PI | Possibly under-captured |
| Abdominal pain | Mentioned | Matches trial data | Established |
| Hair loss | Mentioned | In Wegovy/Zepbound PI; largely weight-loss-mediated | Established |
| Menstrual irregularities | Small but meaningful slice (~1 in 25 reporters; higher among female reporters) | No trial correlate; plausible HPO axis mechanism | Hypothesis-generating |
| Temperature dysregulation (chills, hot flushes) | Emerging signal | No trial correlate | Hypothesis-generating |
| Libido / sexual changes | Occasional | One case report with rechallenge evidence | Hypothesis-generating |
| Sleep changes | Occasional | Equal to placebo in STEP (2.4% vs 2.4%) | Likely secondary |
Source: Sehgal NKR, Tronieri JS, Ungar L, Guntuku SC, "Self-reported side effects of semaglutide and tirzepatide in online communities," Nature Health, April 2026. Frequency bands are paraphrased from paper data and deliberately expressed qualitatively; do not cite as population prevalence.
Subgroups
Special populations
Adolescents, older adults, patients with chronic kidney disease, pregnant patients, and those on oral contraceptives each warrant specific consideration. One interaction is under-recognized: tirzepatide reduces oral contraceptive exposure.
Semaglutide is FDA-approved for adolescents ≥12 (STEP TEENS); pediatric cholelithiasis signal was 4% vs 0%. Elderly subgroups tolerate treatment well with slower titration. FLOW demonstrated a 24% reduction in the composite kidney endpoint in CKD patients. Pregnancy is not recommended with a 2-month pre-conception washout. Tirzepatide reduces oral contraceptive exposure by ~20% — backup contraception is required for 4 weeks after initiation or each dose increase. GLP-1s may be contraindicated in advanced HFrEF based on the FIGHT trial signal.
Adolescents
STEP TEENS (Weghuber et al., NEJM 2022): semaglutide 2.4 mg in ages 12–17 (n=201, 68 weeks) produced BMI reduction of −16.1% vs. +0.6% placebo. GI AEs were 62% vs. 42%. Cholelithiasis: 4% vs. 0% (8 cases vs 0 in N=201) — a pediatric signal worth noting, though the small sample means the rate estimate has wide confidence intervals. No new safety concerns versus adults. FDA approved semaglutide for ages ≥12 in December 2022. SCALE TEENS (liraglutide) showed a similar GI profile with no signals for bone age, Tanner stage, height, or hormone levels. The ELLA trial (liraglutide in children 6–<12) is ongoing.
Elderly (>65)
Generally well-tolerated across SELECT, STEP, and FLOW subgroups. Considerations: slower titration for GI sensitivity, and monitoring lean mass against baseline sarcopenia risk. The SEMALEAN study found improved handgrip strength despite weight loss. REWIND showed 11% AE discontinuation in patients ≥65 vs. 7% in younger patients. FLOW enrolled patients with a mean age of 67 with consistent benefit.
Chronic kidney disease
The FLOW trial (Perkovic et al., NEJM 2024) enrolled 3,533 patients with T2D and CKD (mean eGFR 47) on semaglutide 1.0 mg. Over median 3.4-year follow-up, semaglutide produced 24% reduction in the composite kidney endpoint (HR 0.76, 95% CI 0.66–0.88). Total adverse events were lower with semaglutide than placebo (49.6% vs. 53.8%). No dose adjustment needed down to eGFR 15.
Pregnancy and contraception
All GLP-1 agonists are not recommended in pregnancy. The recommended washout is 2 months before planned conception (semaglutide half-life ~1 week). An observational cohort of ~50,000 pregnant women with T2D showed no significant increase in major congenital malformations among >900 patients exposed at conception.
Tirzepatide reduces oral contraceptive exposure by ~20% (Kapitza et al., JCP 2015; Mounjaro PI). Women initiating or dose-escalating tirzepatide should use backup barrier contraception for 4 weeks after initiation or each dose increase. This does not apply to semaglutide, liraglutide, dulaglutide, or exenatide.
Heart failure
STEP-HFpEF (semaglutide) showed +7.8-point KCCQ improvement and −13.3% body weight in HFpEF. SUMMIT (tirzepatide) showed 38% reduction in CV death or worsening HF (HR 0.62). Caution: FIGHT trial showed liraglutide trended toward worse outcomes in HFrEF — GLP-1 agonists may be contraindicated in advanced heart failure with reduced ejection fraction.
Quality concerns
Compounded GLP-1 products
During the 2023–2025 shortage period, compounded GLP-1 products saw widespread use. FDA enforcement discretion ended after the shortage resolved (February 2025), but a meaningful cohort of patients experienced events tied to compounded formulations.
FDA received 520 adverse event reports for compounded semaglutide and 480 for compounded tirzepatide through April 2025 (likely underreported). Disproportionality analysis of 81,078 FAERS reports showed compounded products had higher reporting odds ratios for abdominal pain (2.84), vomiting (3.21), cholecystitis (3.39), and suicidality (6.34). The "10× overdose" phenomenon, caused by unit/mL/mg confusion with self-administration, produced ~3,000 poison control calls for semaglutide in Jan–Nov 2023 — a 1,500%+ increase vs. 2019.
| Issue | Data | Source |
|---|---|---|
| FDA adverse event reports | 520 compounded sema, 480 compounded tirz (through Apr 2025) | FDA.gov |
| Higher reporting odds (compounded vs. non-compounded) | Abdominal pain ROR 2.84, vomiting 3.21, cholecystitis 3.39, suicidality 6.34 | McCall et al., Expert Opin Drug Safety 2025 |
| 10× overdose phenomenon | Patients self-administering 5–20× intended dose due to unit/mL/mg confusion | Lambson et al., JAPhA 2023; FDA Jul 2024 alert |
| Poison control calls | ~3,000 semaglutide calls Jan–Nov 2023 (>1,500% increase vs. 2019) | America's Poison Centers |
| Salt form concerns | Semaglutide sodium and acetate are chemically different from approved base | FDA.gov |
| Sterility / quality | Contamination, counterfeit labels, injection site infections | FDA enforcement records |
| FDA enforcement | 50+ warning letters Sept 2025; enforcement discretion ended Apr–May 2025 | FDA.gov |
Risk factors
Predictors of side effects
A handful of factors predict who is likely to experience significant side effects. These inform dose selection, titration pace, and patient counseling.
Slower titration is the single most actionable risk-reduction lever — survodutide's phase 3 program sharply reduced discontinuation from phase 2 through titration changes alone. Dose level and titration speed are the strongest modifiable predictors; indication (obesity carries higher gallbladder risk than T2D), age ≥65, baseline GI conditions, baseline diabetic retinopathy, and concurrent insulin or sulfonylurea use are the main non-modifiable risk factors.
| Predictor | Evidence | Strength |
|---|---|---|
| Dose level | Tirzepatide GI d/c: 1.9% (5 mg), 3.3% (10 mg), 4.3% (15 mg) | Well-established |
| Titration speed | Survodutide Ph2 rapid: 20% d/c; Ph3 slower titration much lower | Well-established |
| Indication (obesity vs T2D) | Gallbladder RR 2.29 (obesity) vs 1.27 (T2D) | Well-established |
| Sex | Hair loss: tirz females 7.1% vs males 0.5%; women overrepresented in FAERS | Moderate |
| Age ≥65 | REWIND: 11% AE d/c (≥65) vs 7% (<65) | Moderate |
| Baseline GI conditions | Pre-existing gastroparesis, GERD, or IBS likely increase susceptibility | Clinical consensus |
| Magnitude of weight loss | Hair loss: 5.3% at >20% loss vs 2.5% at <20%; gallbladder risk proportional to rate | Well-established |
| Baseline diabetic retinopathy | SUSTAIN-6: DR worsening 8.2% (with baseline DR) vs 0.7% (without) | Well-established |
| Concurrent insulin / sulfonylurea | Hypoglycemia risk much higher; monotherapy near placebo | Well-established |
Transparency
How this page was made
Selection criteria. We included FDA prescribing information, peer-reviewed randomized trials (STEP, SURMOUNT, SELECT, LEADER, SUSTAIN-6, FLOW, STEP-HFpEF, SUMMIT), large pharmacovigilance analyses (FDA meta-analyses, FAERS disproportionality studies), real-world claims analyses (Prime Therapeutics, Truveta, Cleveland Clinic), and the largest systematic patient-reported symptom analysis available. Where findings differ between trial and real-world data, both are shown side by side.
Where the numbers come from. GI event rates are drawn directly from FDA prescribing information or the STEP 1–3 pooled analysis (Wharton et al., 2022). SELECT data is from Lincoff et al. (NEJM 2023). Head-to-head tolerability is from SURMOUNT-5 (Aronne et al., NEJM 2025). Patient-reported symptom frequencies are expressed as qualitative bands (most common, common, moderate, mentioned) rather than percentages, because the source data is from self-reported social media posts and should not be treated as population prevalence.
What's not included. Retatrutide and survodutide safety profiles (pre-approval; phase 3 still accruing). Individual adverse event case reports below a disproportionality threshold. Drug-drug interactions beyond oral contraceptives (for a complete interaction list, consult the relevant prescribing information). Injection site reactions (generally mild and well-characterized in product labels).
Known limitations. First, most trial-derived rates come from populations healthier than real-world users: STEP excluded patients with significant baseline GI disease, and trial discontinuation is consistently lower than real-world discontinuation. Second, the compounded GLP-1 disproportionality data (higher reporting odds ratios vs approved products) reflects a mix of product-quality issues and dosing errors; separating the two requires more granular data than FAERS provides. Third, the Sehgal et al. social media analysis is qualitative; it identifies signals worth investigating, not rates for clinical citation.
Author and update cadence. Compiled by Ray Wu, MD/MBA — physician-founder working on metabolic health technology. Last updated April 21, 2026. Corrections welcome via the contact page.
References
Sources
Looking for clinical, ROI, or benefits-design guidance on GLP-1 coverage?
Talk to Key to Health