Herbs for GLP-1 side effects? A new study

A new study in Scientific Reports combines FDA pharmacovigilance data with network pharmacology to identify traditional herbal medicines that may share biological targets with the adverse events reported for semaglutide and tirzepatide. The researchers, led by Junkyu Park and colleagues at institutions in South Korea, analyzed 142,705 adverse event reports from the FDA's Adverse Event Reporting System (FAERS) spanning 2015 to 2025, then used a graph convolutional network trained on the HERB database of traditional Chinese medicine to flag five herbal candidates whose known pharmacological targets overlap with the adverse event patterns they observed^[1].

The paper does not claim that any herb prevents or treats GLP-1 side effects. It is a computational screening study, designed to generate hypotheses for future laboratory and clinical testing. That distinction matters, because the gap between a network pharmacology prediction and a verified clinical benefit is large.

What the FAERS analysis found

The team applied disproportionality metrics and Bayesian signal detection to the FAERS database, a publicly available repository of adverse event reports submitted to the FDA since 2004^[2]. The analysis focused on four categories of adverse events linked to GLP-1 receptor agonists: gastrointestinal complications (nausea, vomiting, diarrhoea), biliary and pancreatic events, renal dysfunction, and coagulation disorders. These are well-documented class-effect signals for the GLP-1 receptor agonist category.

One notable finding: semaglutide showed a different adverse-event onset distribution compared with tirzepatide, including a higher proportion of late-onset cases. The authors flag this as potentially relevant for long-term monitoring but do not offer a mechanistic explanation^[1].

From adverse events to herbal targets

The computational step used the HERB database, a large-scale resource that catalogs 7,263 herbs, 49,258 ingredients, and 12,933 gene targets with mapped relationships among them^[3]. The researchers built a graph convolutional network (a type of machine-learning model that operates on network-structured data) to predict which herbs have ingredient-target profiles that overlap with the biological pathways implicated in GLP-1 adverse events.

The model achieved a validation performance of 0.798 for the area under the receiver operating characteristic curve and 0.841 for the precision-recall curve^[1]. Those numbers suggest moderate-to-good predictive accuracy within the dataset, but they measure only how well the model fits its own training data, not whether its predictions translate to real clinical effects.

The five herbal candidates

The network analysis prioritized five medicinal herbs: liquorice root (Glycyrrhiza), mulberry leaf (Morus alba), Dahurian angelica root (Angelica dahurica), Danshen root (Salvia miltiorrhiza), and ginkgo leaf (Ginkgo biloba). Each of these herbs has an established history of use in East Asian traditional medicine and a published pharmacological profile. The authors selected them because their known active compounds target genes and pathways that overlap with those flagged in the FAERS adverse event analysis^[1].

The candidates were filtered for drug-likeness and pharmacokinetic properties using standard computational chemistry tools. The paper reports that several active compounds from these herbs passed absorption, distribution, metabolism, and excretion (ADME) screening, but none has been tested in combination with a GLP-1 agonist in a clinical trial.

What this does not show

Network pharmacology is a hypothesis-generating method, not a proof of efficacy. The study does not demonstrate that any of these herbs reduce GLP-1 side effects in patients. It does not include animal models, cell-based assays, or any clinical data beyond the FAERS reports. The FAERS database itself carries a well-known limitation: reports are voluntarily submitted, may contain duplicates, and cannot establish causation between a drug and an event.

There is also a safety concern that the paper acknowledges only briefly: herbal medicines can interact with prescription drugs. Liquorice root, for instance, is known to affect cortisol metabolism and potassium levels. Ginkgo leaf has documented anticoagulant activity. Patients on GLP-1 therapy who are considering herbal supplements should discuss potential interactions with their prescriber before starting anything new.

Why it matters for the GLP-1 field

The study reflects a broader trend in pharmacovigilance research: using large real-world databases like FAERS as starting points for computational drug discovery and adverse-event mitigation. As GLP-1 prescribing expands, understanding the profile of side effects at population scale becomes more important. The 142,705 reports in this dataset represent one of the larger pharmacovigilance analyses focused specifically on the GLP-1 class.

Whether any of the five herbal candidates will prove useful in practice depends on future bench and clinical work that has not yet been done. For now, the paper adds to the growing body of FAERS-based research mapping GLP-1 adverse events, while introducing a novel angle by connecting those signals to traditional herbal pharmacology. Readers interested in the adverse-event profiles of specific GLP-1 drugs can visit the semaglutide and tirzepatide pages on this site.

This article is for informational purposes only and does not constitute medical advice. Consult a qualified healthcare provider before making any changes to your treatment.