Trends in Poison Center Reports of GLP-1 Weight Loss Drug Exposures After FDA Approval

Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) have been used in the treatment of type 2 diabetes since the 2005 approval of exenatide, the first agent in this class. These agents mimic the endogenous hormone GLP-1, enhancing glucose-dependent insulin secretion, suppressing glucagon, delaying gastric emptying, and reducing food intake [1]. Early trials showed that twice-daily subcutaneous exenatide improved glycemic control and produced sustained weight loss in patients with type 2 diabetes [2–4], and similar effects were later observed in obese individuals without diabetes [5]. Liraglutide, a once-daily subcutaneous GLP-1 RA, also improved glycemic parameters and reduced body weight in both diabetic and non-diabetic populations [6, 7]. These findings supported the 2014 approval by the U.S. Food and Drug Administration (FDA) of liraglutide for chronic weight management in patients aged 12 years and older with obesity and at least one weight-related comorbidity [8].

Semaglutide, a once-weekly subcutaneous formulation, demonstrated greater reductions in hemoglobin A1c and body weight compared with placebo and several active comparators in the SUSTAIN trials [9]. The STEP 1 study later demonstrated a 12.4% mean reduction in body weight from baseline to week 68 with 2.4 mg semaglutide [10], leading to FDA approval for use in adults with obesity [11]. The combination of significant weight loss efficacy, weekly dosing convenience, and expanding clinical adoption has coincided with a substantial rise in GLP-1 RA prescribing across U.S. subpopulations, with marked increases observed between 2020 and 2024, particularly among individuals with obesity or concomitant diabetes [12].

GLP-1 agonists promote weight loss by slowing the rate at which food is emptied from the stomach, which can reduce appetite and promote satiety. This can lead to decreased food intake and result in favorable changes in body weight and composition, including reductions in body fat mass and improvements in insulin sensitivity [13, 14]. Emerging clinical observations suggest that some patients with overlapping obesity and addiction dysregulation may experience benefits in both domains, although the mechanisms remain uncertain [15].

Delayed gastric emptying contributes to hallmark side effects such as nausea and vomiting, which can be severe, persist for several days, and lead to dehydration and electrolyte abnormalities [13, 14]. The relationships between GLP-1 RAs and pancreatitis and, separately, pancreatic cancer remain inconclusive, with retrospective analyses of clinical trial data and observational studies producing inconsistent results; some studies suggest increased risk while others show no change [16–19]. Similarly, an elevated risk of hypoglycemia has been reported, particularly in patients already at risk, such as those prescribed insulin or sulfonylureas [20]. However, some evidence suggests that this hypoglycemia may be specific to individual GLP-1 RAs and not a class effect [21].

The widespread interest in GLP-1 receptor agonists for weight loss, particularly amid drug shortages, has renewed clinical and regulatory attention to their safety profile, yet the clinical profile of acute exposures remains poorly understood. Reports of adverse effects following acute or acute-on-chronic exposures remain incompletely characterized, and existing literature has not clarified whether new toxicity patterns emerge when these drugs are used primarily for obesity rather than diabetes.

Beyond clinical effects, characterizing temporal trends is necessary to contextualize the rise in GLP-1 RA exposures. While simple pre/post comparisons can demonstrate that exposures have increased, such approaches do not account for the strong time-dependent nature of poison center data. Call volumes often exhibit non-linear growth and seasonal fluctuations, which can obscure true inflection points if not modeled properly. Statistical modeling allowed us to formally test whether the increase in GLP-1 RA exposures after the FDA’s approval of semaglutide in 2021 represents a significant change in trajectory beyond what would be expected from underlying prescribing growth. This approach also supports sensitivity analyses using alternative specifications (e.g., Poisson, negative binomial), providing a robust characterization of temporal patterns.

We conducted a retrospective analysis of human exposures to GLP-1 RAs reported to the National Poison Data System (NPDS) from January 1, 2012 through December 31, 2023. Maintained by America’s Poison Centers, the NPDS aggregates de-identified case records from all regional poison centers across the United States and its territories. Case data are coded in near real-time by Specialists in Poison Information, registered nurses and pharmacists with specialized training in clinical toxicology, during telephone consultations with the public and healthcare professionals.

Exposures were included if the primary substance was a GLP-1 RA (exenatide, liraglutide, dulaglutide, albiglutide, semaglutide, tirzepatide, or an unspecified GLP-1 RA). Animal exposures and information-only calls were excluded. The FDA approved once-weekly semaglutide for weight loss on June 4, 2021 [25]. Exposures occurring on July 1 were included in the post-approval group. This cut-point aligns with the start of the standard calendar quarter and accounts for the expected short latency between regulatory approval and the medication’s diffusion into routine clinical use.

Case-level data included demographics (age, sex), exposure circumstances, chronicity (acute, acute-on-chronic, or chronic), exposure site, caller site, management site, level of health care facility (HCF) care, medical outcome, clinical effect duration, clinical effects, and therapies. Patient weight was excluded from the analysis due to a high proportion of missing data (77.7%) and data quality concerns, such as implausible values, and inconsistent units. Similarly, quantitative dosing data were excluded from the analysis due to significant heterogeneity in reporting. Reports frequently conflated volume (mL) with strength (mg), used non-standard units such as “clicks” of a pen device or “units” on an insulin syringe, and often involved compounded formulations with variable or unknown concentrations, making standardization of dose and meaningful quantification impossible.

Clinical effects in NPDS are captured as individual binary indicators coded as “related,” “not related,” or “unknown if related,” and these variables were used to calculate frequencies and percentages for each period. For the primary analysis, symptoms coded as “unknown if related” were grouped with those coded as “related.” This approach prioritized sensitivity by ensuring that rare but clinically meaningful effects were not excluded solely because the initial attribution was equivocal. To assess the impact of this decision, a sensitivity analysis compared the proportion of symptoms coded as “unknown” between the pre- and post-approval periods. Therapy variables, coded as “recommended,” “performed,” “recommended and performed,” or “not recommended/performed,” were collapsed into aggregate summary frequencies for analysis.

This study was determined to be exempt from institutional review board oversight because it involved secondary analysis of de-identified poison center surveillance data. Although the analysis was primarily descriptive, reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) framework in all domains where it was applicable.

A total of 10,033 GLP-1 RA exposures were reported, comprising n = 3,113 before July 1, 2021, and n = 6,920 on and after this date. The post-approval period therefore showed a marked increase in reporting volume, with 2.2 times more exposures than the pre-approval period. Our segmented time-series analysis confirmed that the observed increase in exposures represented a significant change in the trend in reporting volume. The incidence rate of semaglutide calls accelerated by an additional 9.9% per quarter after the July 2021 approval (IRR = 1.10 [95% CI: 1.08,1.11], p < 0.001).

Before July 2021, exposures most frequently involved liraglutide (33.9%), dulaglutide (29.5%), semaglutide (24.6%), exenatide (6.6%), and GLP-1 RAs not otherwise specified (4.9%), with albiglutide rarely reported (0.5%). After July 2021, semaglutide accounted for 64.2% of cases, followed by dulaglutide (12.7%), liraglutide (8.8%), tirzepatide (8.2%), GLP-1 RAs not otherwise specified (5.1%), and exenatide (1.0%). Non-semaglutide products showed only a slight increase in quarterly call frequency (IRR = 1.02 [95% CI: 1.00, 1.03], p = 0.041). These shifts in product distribution over time are shown in Fig. 1.

Overall, the post-approval data describes a younger, predominantly female population experiencing acute therapeutic errors involving semaglutide, accompanied by an increased likelihood of seeking emergency care for their symptoms. Findings from the segmented Poisson model support a significant inflection in the reporting trajectory after the 2021 approval, with semaglutide exposures increasing nearly 10% per quarter compared with only a 1.5% quarterly increase among all other GLP-1 RAs. This divergence is consistent with a reporting inflection temporally aligned with the new weight-management indication and its rapid clinical adoption, rather than background growth in the drug class. The proportion of exposures requiring care in or referral to a health care facility increased from 23.0% before approval to 33.5% after approval (RR = 1.46 [95% CI: 1.36, 1.57], p < 0.001), indicating higher health care utilization despite the predominance of mild gastrointestinal effects. Several factors may contribute to this trend. One hypothesis involves patient behavior in the expanding weight-loss population; brisk public demand and rapid clinical adoption could plausibly lead to premature dose escalation or simultaneous use of multiple products (“dose stacking”) [26], although NPDS data cannot confirm this mechanism.

However, this pattern is unlikely to be explained by a single mechanism. Although therapeutic errors remained the predominant exposure circumstance, their proportion decreased slightly, which suggests an evolving exposure profile in which adverse drug reactions and anxiety-related concerns may account for a growing share of the overall volume. The post-approval period was also marked by substantial drug shortages, which forced many patients across both weight-loss and diabetes indications to navigate supply disruptions. These conditions often required switching between formulations, using compounded products with non-standard or unknown concentrations [27, 28], or rationing doses, any of which could increase the likelihood of error independent of patient intent [29]. Patients prescribed GLP-1 RAs for diabetes may be particularly vulnerable to adverse outcomes during such disruptions, given their concurrent use of other glucose-lowering agents and their intrinsic risk of hypoglycemia. Because NPDS does not reliably capture quantitative dosing, product strength, patient weight, or clinical indication, we could not determine which subgroup or mechanism contributed most to the increase in health care utilization.

Additionally, although greater emergency department utilization could suggest increasing case severity, several findings indicate that changes in patient and public reporting behavior likely played a role. Early clinical trials reported that 5–10% of patients experienced mild symptoms of depression or anxiety [10], which may lower the threshold for seeking urgent care for otherwise manageable gastrointestinal or neurologic complaints. Intense media attention and rapid cultural adoption of GLP-1 RAs may have further reduced the threshold for contacting a poison center, contributing modestly to the rise in call volume independent of clinical severity. On the other hand, some patients may also be unaware that poison centers routinely provide guidance for adverse drug events, resulting in direct emergency department presentation rather than initial poison center consultation. Standard management of these adverse effects typically involves symptomatic treatment for acute presentations, such as antiemetics and intravenous fluids, followed by patient-directed non-pharmacologic strategies (smaller meals, slower eating, increased meal frequency) and prescriber-directed pharmacologic adjustments (slower titration, temporary interruption, dose reduction) [29]. Taken together, these behavioral and system-level factors support the conclusion that nondiscretionary utilization of emergency health care resources increased substantially after FDA approval.

The clinical profile of these exposures remained aligned with the expected pharmacologic effects of GLP-1 RAs, with gastrointestinal and neurological symptoms predominating. Nausea and vomiting were the most frequently reported effects, followed by abdominal pain, dizziness, and headache. To evaluate the reliability of these symptom patterns, we examined the certainty of attribution coding. The proportion of symptoms classified as “unknown if related” decreased in the post-approval period. This finding should be interpreted cautiously; increased exposure frequency may have contributed to more consistent symptom attribution, although changes in reporting behavior or underlying case severity cannot be excluded. Nevertheless, the decline in equivocal attribution supports the conclusion that the post-approval rise in reported effects likely reflects true clinical effects rather than misclassification or declining data quality.

Most adverse clinical effects were not serious. However, one death was reported. The patient had a history of obesity and had recently undergone liposuction, with tirzepatide initiated three weeks post-procedure. Ten days later, the patient presented to the emergency department with acute abdominal pain and abdominal distention after several days of constipation. Exploratory laparoscopy revealed a markedly distended colon from rectum to cecum with three areas of ischemia, and surgical resection of the sigmoid, left colon, and splenic flexure was performed. Postoperatively, the patient remained hemodynamically unstable, and nasogastric tube output was concerning for gastrointestinal bleeding. The clinical course progressed to ventricular tachycardia followed by asystole, and resuscitation efforts were unsuccessful. Causality cannot be determined from this single report, but the temporal sequence raises important clinical questions regarding the safety of GLP-1 RA use in the perioperative setting [30].

No cases of pancreatitis were identified in this dataset, as evidenced by the absence of documented elevations in lipase or amylase. However, poison center records and follow-up protocols do not routinely capture laboratory results, limiting the ability to confirm or exclude this diagnosis with certainty. Even so, the lack of reported pancreatitis among more than 10,000 GLP-1 RA exposures suggests that acute pancreatitis related to short-term therapeutic misadventure is uncommon. This observation aligns with prior studies demonstrating inconsistent associations between GLP-1 RAs and pancreatitis risk. The true incidence of pancreatitis or pancreatic cancer from chronic therapeutic use, however, remains unresolved.

Our findings are broadly consistent with prior single-center studies describing GLP-1 RA exposures. In those reports, most cases stemmed from therapeutic errors, and patients typically experienced no symptoms or only mild effects such as nausea, vomiting, abdominal pain, or weakness, with the majority managed safely at home without hospitalization [23, 24]. Hypoglycemia was uncommon, although sporadic cases were noted. Direct comparison across studies should be interpreted cautiously, as poison center follow-up practices and case ascertainment vary across institutions. The present national dataset corroborates these observations at scale and supports the conclusion that isolated GLP-1 RA exposures rarely result in serious toxicity. Because medication reconciliation is incompletely captured in NPDS data, we cannot fully exclude contributions from concomitant agents such as insulin or sulfonylureas, although the lower proportion of diabetic patients in the post-approval cohort likely contributed to the low overall incidence of hypoglycemia. These findings reinforce existing guidance that GLP-1 RAs have a low intrinsic risk of hypoglycemia but may pose increased risk when combined with agents such as insulin or sulfonylureas.

Beyond the specific findings for GLP-1 RAs, this study illustrates the utility of poison center data for near real-time pharmacosurveillance. Raw call volumes are influenced by multiple external factors, including prescribing trends and media attention, which can make it difficult to distinguish true safety signals from background variability. The modeling framework used here provides a formal method for detecting inflection points in reporting trajectories while accounting for underlying long-term growth. This offers a rapid and scalable approach for monitoring post-marketing signals following events such as new FDA approvals, supply disruptions, or shifts in clinical practice. Importantly, the descriptive trends were complemented rather than replaced by the statistical modeling. The regression analyses served to test whether observed increases exceeded what would be expected from prescribing patterns alone, thereby strengthening interpretability while acknowledging that poison center data cannot establish causality.

Below is the link to the electronic supplementary material.

All authors contributed to the development of this article. All authors have read and approved the final version and take responsibility for its contents. Concept and design: JM, RM. Methodology and statistical analysis: JM, DH. Data acquisition: RM, SV. Drafting of the manuscript: JM, RM. Major editing and revisions: RM. Critical revision of the manuscript for important intellectual content: all authors. Clinical supervision and editing: SV. Statistical supervision and editing: DH.

This study was funded by the University of Texas at San Antonio Provost’s Undergraduate Research Fellowship.

America’s Poison Centers (APC) maintains the National Poison Data System (NPDS), which houses de-identified case records of self-reported information collected from callers during exposure management and poison information calls managed by the country’s poison centers. NPDS data do not reflect the entire universe of exposures to a particular substance as additional exposures may go unreported; accordingly, NPDS data should not be construed to represent the complete incidence of U.S. exposures to any substance(s). Exposures do not necessarily represent a poisoning or overdose, and APC is not able to completely verify the accuracy of every report. Findings based on NPDS data do not necessarily reflect the opinions of APC. The de-identified dataset used for this analysis, including both pre- and post-processing spreadsheets, annotated R scripts (markdown files), and supplemental material will be made publicly available at DOI: [10.5281/zenodo.17259427].