A 23andMe genome-wide study of 27,885 people links GLP1R and GIPR variants to GLP-1 weight-loss response and nausea or vomiting risk

The study identified a missense variant in GLP1R, rs10305420 (p.Pro7Leu), associated with GLP1 medication weight-loss efficacy (P = 2.9 × 10⁻¹⁰), conferring an additional 0.76 kg of weight loss per effect allele (95% CI [−1.27, −0.34] kg), with an additive mode of action (no evidence of dominance, P = 0.80). Earlier reports linking GLP1R variation to weight-loss response came from much smaller studies in specific disease contexts and pointed in the opposite direction; this study provides a directionally clear, quantified association in a 15,237-person European-ancestry GWAS and replicates the same direction in 4,855 people with electronic health record data from the All of Us cohort (P = 0.001). GWAS sample of 15,237 reaching genome-wide significance; within the 99% credible set rs10305420 is the only coding variant and carries the highest posterior probability (35%), and no expression quantitative trait loci colocalized with the signal, leading the authors to conclude it is probably the causal variant; external replication was directionally consistent, while UK Biobank did not replicate, which the authors attribute to low expected power in that cohort.

The study found signals at the GLP1R locus for nausea and vomiting side effects (vomiting index SNP rs11760106, P = 2.5 × 10⁻²⁷, OR = 1.57; nausea index SNP rs9357296, P = 2.6 × 10⁻²⁸, OR = 1.36), and colocalization analysis indicated these probably represent the same signal as the efficacy association, meaning more nausea and vomiting tracks with greater BMI loss. This supplies genetic-level evidence for the long-observed clinical pattern that people with more side effects lose more weight, framing efficacy and tolerability as linked at a single locus rather than as two independent problems. Based on GWAS of 11 side-effect phenotypes, with both signals reaching genome-wide significance; colocalization posterior probabilities were H4 = 96.6% for ΔBMI% versus nausea, H4 = 88.5% for ΔBMI% versus vomiting, and H4 = 92.1% for nausea versus vomiting, with a 72.6% posterior probability from multi-trait colocalization.

Among tirzepatide users, the study separately identified a missense variant in GIPR, rs1800437 (p.Glu354Gln), associated with vomiting (index SNP rs71338792, P = 4.2 × 10⁻⁹, OR = 1.84; rs1800437 P = 5.1 × 10⁻⁹), an association not observed in semaglutide users; people homozygous for risk alleles at both GLP1R and GIPR had 14.8-fold increased odds of tirzepatide-related vomiting (95% CI [6.2, 35.8]). This is the first localization of tolerability differences for the dual agonist tirzepatide to its second target, GIPR, suggesting the GIP component may buffer the nausea-inducing effects of the GLP1 component and offering a target-level lead for drug design and treatment choice. Based on GWAS within the tirzepatide-treated population; rs1800437 is in near-perfect linkage disequilibrium with the index SNP (r² = 0.99) and was concluded to be the causal variant; the direction was consistent in the Latino population (P = 0.03, OR = 0.49) and a fixed-effect meta-analysis increased significance (P = 1.1 × 10⁻¹⁰, OR = 0.54); evidence for interaction between the GLP1R and GIPR variants was weak (P = 0.018).

The study integrated genetic and non-genetic variables into models of efficacy and side effects: the efficacy model explained 25% of the variance in self-report data (consistent between training and held-out test sets) and stratified predicted weight loss in 642 people with HealthKit electronic health record data; the nausea and vomiting models achieved receiver operating characteristic areas under the curve of 65.4% and 68.0%, though the contribution from genetics was relatively modest compared with non-genetic factors. This demonstrates a path from association discovery toward a stratification model that could inform treatment planning, while showing that at current data scale genetic information is only one part of the overall prediction. Models were evaluated in a held-out test set and an independent electronic health record sample, with calibration plots confirming good calibration in the test set; the authors explicitly note that the genetic contribution is relatively modest compared with non-genetic factors and that effect sizes themselves are small.

This work addresses people with overweight or obesity who are using or about to start GLP1 receptor agonists (semaglutide, tirzepatide, and their compounded versions), and its value lies in moving "who will lose more weight and who is more prone to nausea and vomiting" from clinical intuition toward computable probability stratification. For clinicians, it points toward estimating a treatment trajectory before initiation by combining genotype, diabetes status, drug type, and dose; for drug developers, the signal peptide variant in GLP1R and the p.Glu354Gln variant in GIPR offer mechanistic entry points for further validation. The study also places self-report and electronic health record data side by side, indicating that self-report surveys can serve as a complementary instrument for capturing side-effect information, a methodological path relevant where tolerability data are hard to extract from medical records.

One caveat is that the underlying text here is the abstract and main body, so details in figures and supplementary materials are not fully visible and some subgroup results and sensitivity analyses can only be understood through the main-text narrative. The genetic effect sizes are modest, most explained variance in the models comes from non-genetic factors, and the incremental value of genetic information awaits larger samples and longitudinal data. Associations in non-European populations did not reach significance on their own and were only directionally consistent, leaving cross-population applicability an open question. There is a systematic gap between self-reported weight and electronic health records (self-report shows greater weight loss), and its effect on model calibration and clinical interpretation deserves continued attention. In addition, the GLP1R variant points in the opposite direction from earlier small studies, which the authors explain by sample size and disease context; final resolution of that discrepancy still requires independent data. Future work incorporating dose escalation, fuller treatment trajectories, and more ancestral populations would help clarify where these genetic markers belong in real clinical decisions.