Genetic predictors of GLP1 receptor agonist weight loss and side effects
Author style: Punchy — this is a high-impact pharmacogenetics study with clear clinical implications.
Summary
This large study (23andMe survey + GWAS; replication in All of Us) analysed self-reported and EHR-derived responses to GLP1 receptor agonists (mainly semaglutide and tirzepatide) to identify genetic predictors of weight-loss efficacy and treatment-related side effects. The authors detected a missense variant in GLP1R (rs10305420, p.Pro7Leu) associated with greater percentage BMI loss on GLP1 therapy, and a missense variant in GIPR (rs1800437, p.Glu354Gln) associated with increased risk of vomiting on tirzepatide. Models combining clinical and genetic features explained ~25% of variance in BMI loss and modestly predicted nausea/vomiting (AUCs ~0.65–0.68). The findings support a pharmacogenetic signal at the drug targets and point to potential usefulness of combined phenotypic/genetic prediction for precision prescribing.
Key Points
- Large cohort: >27,000 23andMe survey respondents; GWAS of ΔBMI% in n=15,237 (European ancestry) with replication in All of Us (n≈4,855).
- GLP1R missense variant rs10305420 (p.Pro7Leu): T allele associated with additional ≈0.64% BMI loss per allele (≈0.76 kg per allele) and replicated in All of Us.
- GLP1R locus also linked to nausea and vomiting; those side-effect signals co-localise with the efficacy signal, suggesting shared biology.
- GIPR missense variant rs1800437 (p.Glu354Gln) associates with higher odds of vomiting specifically in tirzepatide-treated patients (odds ratio ≈1.84 for the risk allele); protective allele frequency varies by ancestry.
- Non-genetic predictors (sex, drug type, dose, time on drug, T2D status, ancestry) account for most explained variance; combined models reached ~25% R2 for BMI-loss prediction and gave modest discrimination for nausea/vomiting.
Content summary
Background: GLP1 and GIP incretins are targeted by widely used weight-loss drugs (semaglutide, tirzepatide). Response varies substantially between individuals; genetics may explain part of that variability.
Methods: 23andMe launched a GLP1-use survey (Aug 2024) and analysed self-reported pre/post weights, drug type, dose, duration and side effects alongside genotypes imputed to a reference panel. The primary efficacy phenotype was percentage BMI change (ΔBMI%). GWAS was performed in Europeans; leading variants were tested across other ancestries and in external cohorts (All of Us, UK Biobank attempted).
Findings: A GLP1R signal (rs10305420, p.Pro7Leu in the signal peptide) reached genome-wide significance for ΔBMI% and is likely causal; allele frequencies vary by ancestry. Side-effect GWAS identified separate but co-localising signals for nausea and vomiting in the GLP1R region. In tirzepatide-only analyses, a GIPR missense variant (rs1800437, p.Glu354Gln) strongly associated with vomiting. Co-localisation suggests the efficacy and side-effect signals at GLP1R are related: increased nausea/vomiting tends to accompany greater weight loss. Predictive models combining clinical and genetic variables achieved ~25% explained variance for efficacy and modest AUCs (~0.65–0.68) for nausea/vomiting; genetics contributed less than clinical covariates but improved discrimination slightly.
Context and relevance
Why it matters: the loci map directly to the drug targets (GLP1R, GIPR), giving biological plausibility. This is one of the largest real-world efforts to link common coding variants in therapeutic receptors to differential drug response and adverse effects for obesity medicines now used widely in the population. The work shows pharmacogenetic signals that could — with more data and clinical validation — help stratify patients for likely efficacy versus side-effect risk and guide drug or dose choice (for example, deciding between semaglutide and tirzepatide or anticipating nausea management).
Limitations: primary GWAS in Europeans, self-report vs EHR discrepancies (self-reported weight loss larger than EHR), modest effect sizes for individual variants, and current models explain a minority of total variance. Further longitudinal and diverse-cohort work is needed before clinical implementation.
Why should I read this?
Quick take: if you care about who will actually lose weight on semaglutide/tirzepatide — and who might get sick from it — this paper is worth your time. It nails down real genetic hits at the drug targets, shows the trickiness of self-report vs medical records, and gives a practical starting point for prediction models. No jargon-heavy fluff — just big-sample genetics that could matter for precision prescribing as these drugs scale up.