This retrospective pharmacovigilance study was conducted based on the FDA Adverse Event Reporting System (FAERS) database. FAERS is a publicly available post-marketing safety surveillance database that contains adverse event reports, medication error reports, and product quality complaints reported by health professionals, pharmaceutical manufacturers, lawyers, and individual patients (15, 16). Although FAERS is a US database, it receives AEs reports from all over the world. Therefore, the large size and global coverage of this open database make it particularly suitable for the analysis of spontaneous data reporting (15). FAERS database includes the following eight types of files: demographic and administrative information (DEMO), drug information (DRUG), adverse events (REAC), patient outcomes (OUTC), report sources (RPSR), start and end dates for reported drugs (THER), indications for use (INDI), and invalid reports (DELETED). All files recorded “primaryid” and “caseid” variables; therefore, the information about patients and AEs could be obtained by linking these variables in all files. DRUG and THER files recorded “drug_seq” variables; therefore, information about drug use and therapy could be obtained by linking “drug_seq” variables in these two files. All files are available at the FDA website (https://fis.fda.gov/extensions/FPDQDE-FAERS/FPD-QDE-FAERS.html).

All reports between 1 January 2018 and 30 June 2022 were extracted for this analysis. We chose 2018 as the starting year because the most recent approved GLP-1 RA (semaglutide) except for tirzepatide was marketed in 2017, and the ending date was set as the most recent quarter when data were available. The study period was also the data analysis period, given that our research was a cross-sectional study. During the study period, a total of 7,927,000 reports were retrieved from the FAERS database. According to FDA’s recommendations, a two-step deduplication process was used to ensure the unique report: 1) selecting the higher “primaryid” when the “caseid” and “fda_dt” were the same, 2) selecting the latest “fda_dt” when “caseid” were the same, where “primaryid” was an unique number for identifying a AE report, “caseid” was an unique number for identifying a case, and “fda_dt” was the date when FDA received a case (17). Then, the expurgated data from the DELETED file was downloaded from FAERS to delete the invalid reports, finally reducing the number of reports to 6,803,529 ( Figure 1 ).

FAERS is a publicly available and anonymous database. Therefore, approval and written informed consent were waived by the ethics review board of the Ethical Committee of West China Second University Hospital of Sichuan University.

The generic and brand names of GLP-1 RAs approved by the FDA were used to identify adverse events of GLP-1 RAs in the DRUG files, including exenatide (BYETTA, BYDUREON), liraglutide (VICTOZA, SAXENDA), dulaglutide (TRULICITY), lixisenatide (ADLYXIN, SOLIQUA), and semaglutide (OZEMPIC, RYBELSUS, WEGOVY). Of note, albiglutide (TANZEUM) and tirzepatide (MOUNJARO) were not included in our study, given that albiglutide has been discontinued in the United States market since 26 July 2017, and tirzepatide was approved by the FDA in May 2022. The data for tirzepatide were too immature to support the analysis. The summary of approval dates for these GLP-1RAs is listed in Supplementary Table 1 .

The AEs in REAC files are encoded by the preferred terms (PTs) in the Medical Dictionary for Regulatory Activities (MedDRA) terminology, which is comprised of 27 system organ classes (SOCs) (16). The structural hierarchy of MedDRA terminology has five levels: SOC (system organ class), HLGT (high level group term), HLT (high level term), PT (preferred term), and LLT (lowest level term) (17). Accordingly, the latest version of MedDRA 25.0 was used to classify AEs in reports at the relevant SOC level. In our study, all GLP-1 RA-induced PTs were analyzed, and we found that gastrointestinal AEs were most frequently reported in GLP-1 RAs. Therefore, the PTs of GLP-1 RAs below the SOC for gastrointestinal system disorders (SOC: 10017947) were analyzed. Gastrointestinal-related AEs were defined by 905 PTs, including nausea, vomiting, etc. The role codes for AEs had been assigned by reporters, including primary suspected (PS), secondary suspected drug (SS), concomitant (C), and interacting (I). To guarantee GLP-1 RAs were most likely to cause AEs during drug use, analysis reports were limited to those in which the “role_cod” of the drug was “PS” (primary suspected) in the DRUG files (18).

Demographics (gender, age), reporting characteristics (reporting region, year, occupation of reporters), and indications of reports of GLP-1 RA-associated gastrointestinal toxicities were analyzed. In addition, we calculated the time-to-onset of gastrointestinal AEs and the proportion of AE outcomes that were caused by different GLP-1 RAs. The calculation method of time-to-onset is the interval between the start time of GLP-1 RAs’ utilization and the time of AE occurrence (19). Reports with date errors (drug use time later than event occurrence time) and missing date data were excluded. Severe outcomes included death (grade 5), life-threatening conditions (grade 4), and outcomes causing hospitalization, disability, or a congenital anomaly (grade 3) (20). The proportion of reports with severe outcomes was calculated by dividing the number of reports with severe outcomes by the total number of reports.

All characteristics and outcome measures of AE reports regarding GLP-1 RAs were evaluated descriptively. Categorical variables were reported as frequencies and percentages, and continuous variables were summarized as means and standard deviations or medians with interquartile ranges depending on the data distribution.

A disproportionality analysis model was performed to detect the potential signals of gastrointestinal AEs caused by GLP-1 RAs at both the class level and the drug generic name level (21). When a target drug is more likely to induce a target AE than all other drugs, it will typically get a higher score due to higher disproportionality (e.g., when semaglutide causes more gastrointestinal AEs than other drugs, it will get a higher disproportionality score) (21). Both frequentist methods (reporting odds ratio [ROR] (22) and proportional reporting ratio [PRR] (23)) and Bayesian methods (information component [IC] (24) and empirical Bayes geometric mean [EBGM] (25)) of disproportionality analysis were applied to investigate the association between a gastrointestinal AE and GLP-1 RAs. The equations and corresponding criteria of the four disproportionality algorithms are listed in Supplementary Table 2 .

To improve the reliability of signals and avoid occurring false positive signals, a signal was regarded as positive only when it met all criteria of four algorithms simultaneously (26). Then, the signal strength of the top six most frequently reported gastrointestinal AEs of overall GLP-1 RAs in the FAERS database was ranked among different GLP-A RAs, and the top six AEs are nausea, vomiting, diarrhea, upper abdominal pain, constipation, and pancreatitis, respectively. A Kruskal–Wallis test was performed to compare the time-to-onset of gastrointestinal outcomes in different GLP-1 RAs. The severe proportion of gastrointestinal events in different GLP-1 RAs was compared using Fisher’s exact test or Pearson’s chi-squared test. Two-sided P-values of <0.05 were considered statistically significant. Data extraction was conducted by MySQL 8.0, and statistical analyses were performed using R 4.10.

From January 2018 to June 2022, a total of 231,730 GLP-1 RA-associated AEs reports were recorded, of which 21,281 reports of gastrointestinal AEs were identified. Of these, 1,956 gastrointestinal reports were for exenatide (9.19%), 2,891 for liraglutide (13.58%), 10,757 for dulaglutide (50.55%), 182 for lixisenatide (0.86%), and 5,567 (26.16%) for semaglutide ( Figure 1 ).

The characteristics of AE reports for different GLP-1 RAs are presented in Table 1 . More male patients reported gastrointestinal AEs from GLP-1 RAs (54.19%). The median age (interquartile range) of patients was 62 (54–70) years. The region with the highest number of reports was North America (91.40%), followed by Europe (4.10%), and Asia (2.55%). The number of gastrointestinal AEs steadily increased from 4,163 in 2018 to 5,048 in 2021, which reflected the increasingly clinical application of GLP-1 RAs. These AEs were mainly reported by consumers (74.28%) and physicians (11.04%). Gastrointestinal AEs were most frequently reported for unknown indications (50.37%) and diabetes mellitus (41.07%). The median time-to-onset of gastrointestinal AEs was 1 (IQR 0–24) days, and 77.44% of the AEs occurred within 30 days after excluding invalid reports. Outcomes causing hospitalization, disability, or a congenital anomaly (2,933, 13.78%) account for the most frequent outcomes in all cases.

All adverse events signals of GLP-1 RAs were detected by using four algorithms and their corresponding criteria. A total of 619 positive signal for GLP-1 RAs were observed, and the most AEs of positive signal were distributed in gastrointestinal system ( Figure 2 ).

Within the SOC level, gastrointestinal system disorders in overall GLP-1 RAs were overreported compared with the background frequency (ROR 1.46, PRR 1.33, IC 0.41, EBGM 1.33), which were further attributed to liraglutide (ROR 2.39, PRR 1.84, IC 0.88, EBGM 1.33), dulaglutide (ROR 1.39, PRR 1.28, IC 0.36, EBGM 1.25), and semaglutide (ROR 3.00, PRR 2.10, IC 1.07, EBGM 2.10), while exenatide (ROR 0.57, PRR 0.63, IC −0.68, EBGM 0.63) and lixisenatide (ROR 0.51, PRR 0.57, IC −0.81, EBGM 0.57) did not show the association with gastrointestinal AEs ( Table 2 ).

Furthermore, the positive signals of gastrointestinal AEs were classified as PT level ( Supplementary Table 3 ). The top six most common gastrointestinal reports were nausea (8,988, 42.23%), diarrhea (4,666, 21.93%), vomiting (4,660, 21.90%), upper abdominal pain (2,082, 9.78%), constipation (1,790, 8.41%), and pancreatitis (1,752, 8.23%). All gastrointestinal AEs in the label of GLP-1 RAs were found in our study, which confirmed the reliability of this study. However, there were five disproportional signals that were not covered on the label that were detected for liraglutide (gastroesophageal reflux disease, gastritis, bezoar, breath odor, and intra-abdominal hematoma), one for dulaglutide (duodenogastric reflux), and three for semaglutide (breath odor, pancreatic failure, and mesenteric panniculitis).

Simultaneously, the signals of the top six common gastrointestinal AEs (nausea, diarrhea, vomiting, upper abdominal pain, constipation, and pancreatitis) for different GLP-1 RAs were detected and compared ( Table 3 ). The results showed that semaglutide had the greatest risk of nausea (ROR, 7.41; 95% CI, 7.10–7.74), diarrhea (ROR, 3.55; 95% CI, 3.35–3.77), vomiting (ROR, 6.67; 95% CI, 6.32–7.05), and constipation (ROR, 6.17; 95% CI, 5.72–6.66); liraglutide had the greatest risk of upper abdominal pain (ROR, 4.63; 95% CI, 4.12–5.21). In addition, it was observed that use of any GLP-1 RA was associated with the risk of pancreatitis, and the signal strength of pancreatitis was ranked as follows: liraglutide (ROR, 32.67; 95% CI, 29.44–36.25) > semaglutide (ROR, 19.10, 95% CI, 17.26–21.13) > dulaglutide (ROR, 12.63; 95% CI, 11.76–13.56) > lixisenatide (ROR, 6.78; 95% CI 4.26–10.80) > exenatide (ROR, 4.91; 95% CI, 4.12–5.85).

The time-to-onset of gastrointestinal system disorders for each GLP-1 RA regimen is shown in Table 4 . The result showed the time-to-onset of GLP-1 RAs had a statistical difference (p <0.001, Kruskal–Wallis χ² = 109.90). The longest median time-to-onset was 4 (IQR 0–40) days for semaglutide, while the shortest was 0 (IQR 0–11) day for lixisenatide, 1 (IQR 0–19.75) day for exenatide, 1 (IQR 0–7) day for dulaglutide, and 3 (IQR 0–36.75) days for liraglutide, respectively.

To explore the prognosis of patients with gastrointestinal AEs after using GLP-1 RAs, this study evaluated the outcome of reports by severe proportions of outcomes. The corresponding results were presented in Table 1 and Figure 3 , and they showed a statistical difference in the severe proportions of GLP-1 RA-related gastrointestinal AEs among different GLP-1 RAs (p <0.001, Pearson’s χ² = 11,227). In all GLP-1 RAs, liraglutide (23.31%, 657 severe outcomes out of 2,819 cases) had the highest severe proportion of GLP-1 RA-associated gastrointestinal AEs, and the lowest for dulaglutide (12.29%, 1,322 severe outcomes out of 9,435 cases).