Computational framework links Liquorice Root, Mulberry Leaf, and other herbs to GLP-1 RA adverse event patterns

Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) are highly effective for obesity management and type 2 diabetes, yet their widespread use highlights persistent concerns regarding adverse events (AEs). Current clinical understanding often lacks comprehensive insights into potential interactions, particularly with herbal supplements. This gap necessitates advanced methods to identify and characterize AE patterns and explore novel associations, such as those involving herbal compounds, which could impact patient safety and therapeutic outcomes. Understanding these interactions is crucial for optimizing GLP-1 RA therapy.

Researchers conducted a cross-sectional analysis of 142,705 GLP-1 RA AE reports from the FDA Adverse Event Reporting System (FAERS) spanning 2015-2025, with 4,090 reports specifically for obesity indications. Clinical characteristics like outcomes, reporting frequency, demographics, and time-to-onset were analyzed. Signal detection utilized disproportionality metrics and Bayesian approaches. Herb-compound-target-AE networks were constructed using HERB 2.0 and the Comparative Toxicogenomics Database, incorporating drug-likeness and pharmacokinetic filtering. Graph convolutional networks (GCN) were then applied to model herb-AE associations, followed by literature-based contextual evaluation.

Gastrointestinal events predominated among the GLP-1 RA AE reports, with 76% involving female patients and onset clustering within 0-30 days. Semaglutide exhibited a distinct onset distribution, including a higher proportion of late-onset cases (≥ 360 days). Strong signals were detected for biliary, pancreatic, renal, and coagulation events. Specifically, semaglutide-associated pairs showed reporting odds ratios > 10 for these serious AEs. Conversely, tirzepatide demonstrated negative log-transformed reporting odds ratios for several gastrointestinal events, suggesting a potentially different AE profile. Network analysis and GCN modeling prioritized several established medicinal herbs as top candidates associated with AE patterns: > Liquorice Root, Mulberry Leaf, Dahurian Angelica Root, Danshen Root, and Ginkgo Leaf were identified after degree debiasing and exclusion of non-herbal database entries. The GCN achieved an area under the receiver operating characteristic curve (AUC/AR) of 0.85, indicating good predictive performance.