A new signal-detection analysis indexed by Europe PMC reports that all five glucagon-like peptide-1 receptor agonists it examined met conventional signal criteria for impaired gastric emptying in the FDA Adverse Event Reporting System. The authors state the analysis is descriptive and hypothesis-generating and makes no claim about clinical risk.

The substances involved include semaglutide (sold as Ozempic and Wegovy) and tirzepatide (Mounjaro and Zepbound), alongside dulaglutide, liraglutide and exenatide. Worth holding onto before the numbers arrive: the authors note that GLP-1 receptor agonists delay gastric emptying as an on-target effect, meaning the slowing is an expected consequence of how the drugs work rather than an unexplained one.

How the signal was measured

The team pulled reports for the five substances from the public openFDA interface to FAERS using the generic-name field, and measured a single outcome: the MedDRA Preferred Term "Impaired gastric emptying" (code 10021518).

For each substance they calculated the proportional reporting ratio, the reporting odds ratio with 95 percent confidence intervals, and the Yates-corrected chi-square. Counts reflect the full database snapshot at the extraction date, and the authors say all figures are reproducible from the public openFDA source using their analysis script. That last point is the part of this paper doing the real work.

Key results

Semaglutide showed the strongest disproportionality, with a proportional reporting ratio of 88.7 (95 percent CI 85.1 to 92.4), based on 3,052 impaired-gastric-emptying reports out of 82,911 total reports. Dulaglutide followed at 37.7 (1,593 of 86,770), then liraglutide at 20.9 (563 of 50,219) and tirzepatide at 20.2 (1,409 of 140,435).

Exenatide came in weakest of the five, at 4.7 (136 of 52,308). The authors say these results reproduce, with a transparent method, the substance-level signal previously described by independent groups analyzing the same database.

Limitations

The authors are explicit that disproportionality quantifies reporting, not incidence or risk, and cannot establish causality. FAERS is a spontaneous reporting database, so the counts reflect what was reported rather than what occurred in any defined patient population.

This is a signal-detection exercise, not a clinical study, and the authors describe it as hypothesis-generating. They state the analysis was carried out without asserting causation.

Reproducibility, not risk

The value here is that someone re-ran an earlier independent finding with a published script and got the same shape of answer at the active-substance level. What the signal implies about actual clinical risk is not something the analysis was designed to answer, and the authors do not pretend otherwise.