Semaglutide Panic Reports Rose After Publicity, but Studies Show No Consistent Link
Panic reports rose after scrutiny of GLP-1 drugs. Adjusted comparisons and a 12-study review found no consistent link, but cannot exclude one.
Summary
Reports of panic attacks with semaglutide (Ozempic, Wegovy) looked alarming at first. Researchers examined 33,267 semaglutide reports in the FDA's adverse event database and compared them with two other diabetes drugs. Panic attacks were reported in 0.26% of semaglutide reports versus 0.05% for each comparator. After adjusting for age, sex and reason for taking the drug, the gap shrank. Before July 2023, when European regulators began a safety review, the adjusted difference was no longer statistically significant. Reports then rose about sixfold, which fits a pattern where publicity prompts more reporting. A review of 12 studies, including randomized trials, found no consistent rise in panic or anxiety. These databases cannot prove or rule out harm, but the evidence does not support a clear link.
Detailed Summary
Semaglutide is among the most widely used drugs in the world, taken for type 2 diabetes, obesity and, increasingly, off-label weight loss. Because so many people take it, any hint of psychiatric side effects spreads quickly. Earlier pharmacovigilance analyses flagged panic attacks as a possible signal, but spontaneous-report databases are prone to distortion. This study asked how much of the semaglutide–panic attack signal reflects real pharmacology and how much reflects who takes the drug and when reports are filed.
The authors analyzed the FDA Adverse Event Reporting System (FAERS) from 2018Q1 to 2024Q4. They included only reports where the drug was the primary suspect: 33,267 for semaglutide, 54,282 for dulaglutide (a related GLP-1 drug) and 25,380 for empagliflozin (an SGLT2 inhibitor used in a similar diabetes population). Panic attack was defined by a single MedDRA term. Logistic regression adjusted for age, sex, and diabetes or obesity indication. Temporal analyses split the data at July 1, 2023, when the European review began. A sensitivity analysis used all other drugs as the reference.
Panic attacks appeared in 87 semaglutide reports (0.26%), 28 dulaglutide reports (0.05%) and 13 empagliflozin reports (0.05%). Crude reporting odds ratios were 5.08 (95% CI 3.32–7.78) versus dulaglutide and 5.12 (2.86–9.17) versus empagliflozin. Adjustment reduced these to 2.82 (1.74–4.56) and 2.75 (1.41–5.36). Semaglutide users were more often female (61.6%) and more often had an obesity indication (12.3%), compared with 0.4% for dulaglutide and 0.1% for empagliflozin. This shows how differences between patient groups can inflate a crude signal.
The temporal analysis was the most telling. Before July 1, 2023, crude RORs were 2.24 (1.23–4.07) and 2.15 (1.00–4.58). Adjusted estimates were 1.38 (0.69–2.75) and 1.14 (0.48–2.68), neither statistically significant. Panic attack reports per 10,000 semaglutide reports rose from 6.7 in 2018–2022 to 42.0 in 2023–2024 (drug-by-period interaction p = 0.007). Dulaglutide rose from 3.7 to 10.9, while empagliflozin fell from 5.8 to 3.7. The semaglutide reporting odds ratio for 2023–2024 versus 2018–2022 was 6.31 (3.26–12.20).
The sensitivity analysis against 9,518,893 reports for all other drugs showed the same pattern. The adjusted ROR was 1.48 (1.18–1.87) overall but 0.76 (0.48–1.18) before July 2023, rising to 2.15 (1.57–2.93) afterward. A parallel systematic review of 12 studies, registered in PROSPERO and covering randomized trials, cohorts, pharmacovigilance analyses and case reports, found that randomized trials and active-comparator observational cohorts did not show a consistent increase in panic or anxiety-related outcomes. Taken together, the evidence does not support a consistent pharmacological link.
The authors stress that this does not exclude a true effect. Disproportionality analysis cannot estimate incidence or establish causality, and FAERS has no exposure denominators. Age was missing for 44.8–65.6% of reports, and the panic attack counts were small (87, 28 and 13). Trials were not powered for rare acute events. The study mainly shows how an active-comparator design, temporal analysis and evidence triangulation can keep a widely publicized safety signal in perspective.
Key Findings
- Panic attacks appeared in 0.26% of 33,267 semaglutide FAERS reports versus 0.05% of 54,282 dulaglutide and 0.05% of 25,380 empagliflozin reports
- Crude reporting odds ratios of about 5.1 versus both comparators fell to 2.82 (95% CI 1.74–4.56) and 2.75 (1.41–5.36) after adjusting for age, sex and indication
- Before July 1, 2023, adjusted RORs were 1.38 (0.69–2.75) versus dulaglutide and 1.14 (0.48–2.68) versus empagliflozin, with no statistically significant signal
- Semaglutide panic attack reports per 10,000 rose from 6.7 (2018–2022) to 42.0 (2023–2024); drug-by-period interaction p = 0.007, OR 6.31 (3.26–12.20)
- Against all other drugs (9,518,893 reports), the aROR was 0.76 (0.48–1.18) before July 2023 and 2.15 (1.57–2.93) afterward
- Reporting proportions stayed flat for empagliflozin (0.051% in both periods) and the all-drug reference set (0.142% vs 0.140%) but rose for semaglutide (0.108% to 0.262%)
- Across 12 reviewed studies, randomized trials and active-comparator cohorts showed no consistent increase in panic attacks or anxiety outcomes
Methodology
This was an active-comparator restricted disproportionality analysis of FAERS (2018Q1–2024Q4), restricted to primary-suspect reports for semaglutide, dulaglutide and empagliflozin and reported under READUS-PV guidance. Crude and adjusted reporting odds ratios came from logistic regression adjusting for age, sex and indication, with unknown values kept as separate categories. Notoriety bias was tested with a pre-July 2023 restriction, a drug-by-period interaction and annual reporting proportions per 10,000, plus a sensitivity analysis against all other drugs. A PRISMA-compliant systematic review (PROSPERO CRD420251183110; MEDLINE, Embase, CENTRAL through November 2025) used Cochrane RoB 2, ROBINS-I and Naranjo for quality assessment.
Study Limitations
FAERS lacks exposure denominators and clinical detail, so it cannot estimate incidence or causality. Age was unrecorded in 44.8–65.6% of reports, and panic attack counts were small (87, 28 and 13), which limits stable estimation. The authors say a pharmacological effect cannot be excluded, and trials were not designed to detect rare acute neuropsychiatric events. Conflict-of-interest and funding statements were not included in the supplied text.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: