Genes and gut microbes shape who tolerates GLP-1 drugs, but never together

Nausea, vomiting, diarrhea and constipation are the price many people pay for the remarkable weight loss and blood sugar control delivered by GLP-1 receptor agonists such as semaglutide, liraglutide and tirzepatide.

The scale of the tolerability problem is well documented. Pooling data from 101 randomized controlled trials involving 57,684 participants, researchers found that GLP-1 receptor agonist therapy produced nausea in 22.8 percent of patients, more than twice the placebo rate, vomiting in 9.1 percent, diarrhea in 13 percent and constipation in 7.4 percent. In the pooled STEP 1–3 trials of semaglutide, 4.3 percent of participants permanently stopped treatment because of a gastrointestinal adverse event — a modest proportion that compounds across the tens of millions of people now taking these drugs.

Crucially, the review’s authors, a team from Digbi Health in Mountain View, California, emphasize that these symptoms are not the mechanism of benefit. Less than one percentage point of the additional 7.6 to 14.4 percent weight loss achieved on semaglutide was mediated by gastrointestinal adverse events. The physiological driver of the symptoms — delayed gastric emptying and reduced small-intestinal transit, mediated through peripheral vagal afferent and central area postrema pathways — explains the class-wide tendency toward GI upset, but correlates only weakly with how bad any individual’s symptoms become. Something else must determine who tolerates the drugs well and who does not.

The first candidate system is genetics. In 2026, a genome-wide association study drew on self-reported side effect and outcome data from 27,885 GLP-1 receptor agonist users in the 23andMe research cohort, of whom 15,237 formed the efficacy analysis after quality control. Two loci reached genome-wide significance. At GLP1R, the gene encoding the drug’s own target, the missense variant rs10305420 was associated with greater weight-loss efficacy, while distinct index variants predicted medication-related nausea and vomiting. At GIPR, a variant predicted vomiting, but only among users of tirzepatide, the dual GIP/GLP-1 co-agonist — suggesting that intolerance genetics for incretin therapy may be partly receptor-target specific rather than uniformly class-wide.

Those genetic findings come with important caveats. A combined genetic and non-genetic model explained only about 25 percent of weight-loss efficacy in held-out test sets, with sex, drug type, dose and treatment duration accounting for more variance than any single variant. The discovery cohort was 78.3 percent European ancestry and predominantly female, and self-reported weight change was roughly double the record-derived estimate. Independent replication of GLP1R candidate variants has been mixed: in a 4,571-participant pharmacogenomic study of glycemic response, rs10305420 showed no effect, and in two Chinese cohorts a different variant was associated with GI adverse reactions in one but not the other. Meanwhile, the candidate genes most often proposed on mechanistic grounds — the serotonin receptor genes HTR3A and HTR3B, the bile acid receptor gene GPBAR1, and the motility gene SCN5A — have never been tested in a GLP-1 RA-treated cohort with GI outcomes, and the one genome-wide scan available reported no signal at any of them.

The second system is the gut microbiome, which sits on both sides of the relationship. Microbial fermentation of dietary fiber produces short-chain fatty acids that stimulate L-cell GLP-1 release, and bile acids raise GLP-1 secretion through the GPBAR1/TGR5 receptor — a mechanism that is receptor-specific, since signaling through the nuclear receptor FXR runs the other way. In humans, a 24-week randomized trial of inulin-propionate ester increased postprandial GLP-1 and peptide YY and improved insulin sensitivity, while a galacto-oligosaccharide trial multiplied Bifidobacterium levels fivefold without moving GLP-1 at all, showing that a microbiota shift does not guarantee a hormonal one.

The reverse direction — the drug reshaping the microbiome — rests on a smaller and more heterogeneous base. The best-designed human study, a 12-week double-blind placebo-controlled trial of liraglutide in 51 participants, found no change in alpha or beta diversity despite clear glycemic and bile acid effects. Of 38 studies in a 2025 systematic review, only nine were in humans, and no human semaglutide study was identified. Positive findings are frequently confounded: increased Akkermansia with liraglutide appeared to track weight loss in a comparable mouse study, and in the MIGHTY trial the liraglutide arm carried metformin throughout, a drug that itself causally alters the microbiome and produced a larger Faecalibacterium increase in a head-to-head comparison. Most tellingly, the authors found no human study testing whether baseline or drug-induced microbiome composition predicts GI adverse events — the literature is oriented toward efficacy, not tolerability.

When the authors surveyed across drug classes — including metformin-induced GI intolerance, levodopa metabolism, digoxin inactivation and chemotherapy-induced mucositis — they could not find a single study fitting a multiplicative host-genotype-by-microbiome interaction model against a drug tolerability or adverse-event outcome. The closest attempts fall short: the only human study to fit an explicit genotype-by-microbiome interaction term analyzed intestinal mucosal gene expression rather than drug response, and the nearest attempt at drug GI intolerance used blood mRNA rather than germline genotype in just 18 samples without an interaction term. The largest genotype-plus-microbiome resources, including MiBioGen with 18,340 participants and a Swedish-Norwegian GWAS with more than 28,000, are general-population cohorts with no GLP-1 RA exposure arm.

Why has nobody done this study? The review points to methodological arithmetic rather than neglect. Detecting a multiplicative interaction requires at least four times the sample size needed for a main effect of the same magnitude, and measurement error in the interacting exposure inflates the requirement sharply: under one illustrative parameter set, detecting an interaction with exposure and outcome reliability coefficients of 0.3 and 0.4 would require nearly 151,000 participants, while repeated measurement brings the figure below 5,000. Gut microbiome features sit squarely in the low-reliability regime — daily profiling of 713 stool samples from 20 women found that for 78 percent of genera, day-to-day abundance varies more within individuals than between them, with roughly five samples per person the optimal trade-off. Combining these parameters, the authors estimate that single-timepoint taxon-level testing would call for on the order of 10,000 to 100,000 treated participants, while five repeat samples with pathway-level features would lower this to roughly 5,000 to 10,000 — an illustrative calculation, they stress, not a formal power analysis.

The review also tempers expectations about what genetics and the microbiome can deliver clinically. A randomized open-label trial found that a 16-week flexible semaglutide escalation reduced GI-driven withdrawal to 2 percent, versus 19 percent on the standard 8-week schedule, with equivalent final HbA1c — a protocol-driven effect larger than any between-drug or between-genotype comparison in this literature. No genetics- or microbiome-guided program has been shown to alter a tolerability outcome, and the authors conclude that the rationale for precision tolerability management rests on mechanistic plausibility rather than demonstrated clinical benefit. What is needed, they argue, is a specific and nameable study: prospective enrollment of GLP-1 RA initiators with baseline genotyping at GLP1R and GIPR, repeat stool metagenomics at about five samples per person, structured prospective symptom capture, and powering designed for interaction detection. Until such work is reported, the two strongest biological explanations for who gets sick on these drugs will remain strangers to each other.

Subject of Research: How host genetics and the gut microbiome independently shape gastrointestinal tolerability of GLP-1 receptor agonist drugs

Article Title: Pharmacomicrobiomics of GLP-1 receptor agonist gastrointestinal tolerability: integrating host genetics and the gut microbiome

Article References: Kumbhare, S. V., Pedroso, I., Sinha, R., & Almonacid, D. E. (2026). Pharmacomicrobiomics of GLP-1 receptor agonist gastrointestinal tolerability: integrating host genetics and the gut microbiome. Frontiers in Microbiomes, 5, 2016564. https://doi.org/10.3389/frmbi.2026.2016564

Image Credits: AI Generated

DOI: 10.3389/frmbi.2026.2016564

Keywords: GLP-1 receptor agonists, pharmacomicrobiomics, gut microbiome, pharmacogenomics, GLP1R, GIPR, tirzepatide, semaglutide, nausea, gastrointestinal adverse events, gene-environment interaction, microbiome GWAS