Vitals Knowledge Vault

GLP-1 Agonist Pharmacogenomics Non-Response

10 min read

GLP-1 Agonist Pharmacogenomics Non-Response

TL;DR

GLP-1 pharmacogenomics is a genuine scientific field with biologically plausible mechanisms, but no finding has demonstrated clinical utility. The most-studied variant, GLP1R rs6923761 (Gly168Ser), shows a directional signal in severe obesity (n=112) but has failed to replicate in two subsequent cohorts. The largest GWAS to date (Nature 2026, n=27,885) identified a GLP1R missense variant associated with −0.76 kg/copy additional weight loss; a separate multi-cohort GWAS (DIRECT, n=4,571) found ARRB1 — not GLP1R — as the strongest HbA1c-response signal. Critically, MC4R deficiency — the most common monogenic obesity — does NOT cause GLP-1 non-response; tirzepatide is equally effective in MC4R mutation carriers. The best currently available predictor of GLP-1 non-response is early response trajectory at weeks 8–16, not genetic testing.

⚠️ Evidence boundary: No validated genetic test exists for GLP-1 response prediction. No guideline body (FDA, EMA, ADA, AACE) recommends pharmacogenomic testing before GLP-1 prescribing. Do not use current evidence for clinical decision-making without human review.

Why it matters for Vitals

  • Coaches: Genetic variants explain only a small fraction of individual response variation. Adherence, dose titration, diet, and baseline metabolic health remain far more important levers than any known variant.
  • Early response monitoring is actionable now: A patient’s weight trajectory at weeks 8–12 is the most evidence-backed predictor of long-term GLP-1 response — no genetic data required.
  • MC4R carriers can use GLP-1s: The most common monogenic obesity cause does NOT contraindicate GLP-1 therapy. Tirzepatide works equally well in MC4R mutation carriers.
  • CYP2D6 variants do not affect GLP-1 levels: If a client is on CYP2D6-metabolized medications, those levels may need monitoring — but the GLP-1 itself is unaffected.
  • Clinical boundaries: Vitals coaches should not recommend or discourage GLP-1 therapy based on genetic variants from DTC panels. Escalate genetic-data questions to the prescribing clinician.

Key Facts

Variant / FactorEffectEvidence Grade
GLP1R rs6923761 (Gly168Ser) — severe obesity, n=112AA homozygotes: 1.64%/month vs G carriers: 1.04%/month (p=0.03)Supported — single small cohort
GLP1R rs6923761 — oral semaglutide, n=210No significant association with HbA1c, BMI, BPNull / Reported
ARRB1 Thr370Met — DIRECT GWAS, n=4,5710.25%/copy greater HbA1c reduction (p=5.2×10⁻⁶)Supported — multi-cohort
MC4R mutation carriers — SURMOUNT-1Tirzepatide: 18.3% vs 19.9% WL (P=0.79); equally effectiveConfirmed
Early response trajectory (week 16)Best predictor of long-term response; no genetic predictor significantSupported
Low C-peptide / insulin deficiencyPredicts non-responseReported
GLP1R missense variant rs1030559 — Nature 2026, n=27,885−0.76 kg/copy additional weight loss (P=2.9×10⁻¹⁰)High (statistically); single source; self-reported cohort
Real-world discontinuation30–50% at 1 year — non-adherence is the dominant non-response causeConfirmed
Genetic contribution to treatment response<10% of total effect — far too small to explain clinical non-responseConfirmed
Clinical utility trial for GLP-1 pharmacogenomicsNoneGap
Any guideline recommends genetic testing for GLP-1No — FDA, EMA, ADA, AACEConfirmed

Mechanism Summary

GLP1R rs6923761 (Gly168Ser)

Located at the exendin(9-39) binding site in the third extracellular loop of the GLP-1 receptor — a region critical for ligand binding and receptor activation. In vitro functional assays show enhanced downstream signaling per unit ligand. The effect is pharmacodynamic, not pharmacokinetic — receptor sensitivity is altered, not drug clearance. Evidence grade: Supported (functional assays; specific mechanism not fully disclosed). Replication status: failed in two subsequent cohorts.

ARRB1 β-Arrestin-1 Variants

β-arrestin-1 (ARRB1) is a scaffolding protein involved in GLP1R desensitization and β-arrestin-dependent signaling. The low-frequency missense variant Thr370Met was associated with greater HbA1c reduction in DIRECT GWAS, suggesting that β-arrestin recruitment efficiency — not just ligand binding affinity — may be a key determinant of GLP-1 response. Evidence grade: Supported (β-arrestin bias is established GPCR pharmacology; multi-cohort GWAS).

MC4R Deficiency Does NOT Predict Non-Response

MC4R deficiency is the most common monogenic obesity cause. Despite mechanistic expectation that GLP-1 response would be reduced, tirzepatide is equally effective in MC4R mutation carriers (18.3% vs 19.9% weight loss, P=0.79). This suggests GIP receptor agonism may compensate for or bypass GLP1R pharmacogenomic variation.

GLP1R Missense Variant (rs1030559, Nature 2026)

Distinct from rs6923761. Sits at the exendin(9-39) binding site. n=27,885 (23andMe self-reported). Effect: −0.76 kg/copy additional weight loss. Self-reported outcomes; all authors 23andMe employees. Critical contradiction: DIRECT GWAS found ARRB1, not GLP1R, as the primary HbA1c signal. The two largest GWAS point to different primary genes for different outcomes.

Pharmacokinetics

GLP-1 agonists are large peptides cleared primarily by proteolytic degradation — not by CYP450 enzymes. CYP2D6 polymorphisms have no direct effect on GLP-1 RA pharmacokinetics. SLC22A1 (OCT1) and SLC47A1 (MATE1) transporter variants showed no significant effect on semaglutide response.

What the current evidence suggests

Most-studied variant: rs6923761 (Gly168Ser) — Failed to replicate

  • Positive signal (n=112, severe obesity, PMID 40384505): AA homozygotes lost weight at 1.64%/month vs G carriers at 1.04%/month (p=0.03)
  • Null signal (n=210, T2D, oral semaglutide, PMID 41307691): No significant association with HbA1c, BMI, or BP changes
  • Nominal trend (n=10, oral semaglutide, PMID 40996853): Did not survive FDR correction
  • Verdict: Mechanistically plausible but clinically unconfirmed across three independent cohorts. Do not use for prescribing decisions. This is the most-studied GLP-1 pharmacogenomic variant and its best-supported signal comes from the smallest cohort (n=112).

MC4R Deficiency — Tirzepatide Equally Effective

SURMOUNT-1 pharmacogenomic subgroup (n=32 carriers / 2,259 non-carriers): 18.3% vs 19.9% weight loss at 72 weeks, P=0.79. setmelanotide remains the approved targeted therapy for MC4R deficiency. Evidence grade: Confirmed.

Early Response Trajectory — The Best Current Predictor

Early weight loss at week 16 is the strongest predictor of liraglutide response at 56 weeks (n=125, adolescent obesity, PMID 37264767). No baseline demographic or genetic predictor reached significance. This is actionable without genetic data.

Non-Adherence Is the Dominant Non-Response Cause

Real-world GLP-1 discontinuation rates reach 30–50% at 1 year. Apparent non-response in clinical practice is far more likely to reflect non-adherence than any genetic variant.

Clinical Implications

Actionable Now

  1. Use early weight trajectory (≥4% loss at week 8–16) as the primary practical predictor of long-term GLP-1 response
  2. Assess baseline C-peptide and glycemic status — low C-peptide predicts non-response per PMID 26802434
  3. Conduct adherence assessment as first-line intervention before attributing poor response to genetics
  4. Do not deny GLP-1 therapy based on genetic variants from DTC panels

Extrapolation (with caveat)

  • In patients with strong family history of early-onset severe obesity, be aware that polygenic burden may predict lower response — inferred from general obesity genetics, not GLP-1-specific data
  • Review concomitant medications for CYP2D6 interactions — GLP-1 RAs are not CYP2D6 substrates but co-medications may be affected

Do Not Implement Without Validation

  • Commercial pharmacogenomic panels (23andMe, AncestryDNA) include some GLP1R variants — there is NO evidence these should guide GLP-1 prescribing
  • Polygenic risk scores for obesity or T2D could theoretically correlate with GLP-1 response — no validated PRS for GLP-1 response has been developed

Skeptic / Limitations Summary

The source corpus surfaces several high-severity limitations worth foregrounding:

  • rs6923761 replication failure: The most-studied GLP-1 pharmacogenomic variant shows a directional signal in the smallest cohort (n=112) and null results in two subsequent studies (n=10, n=210). The field’s most-cited finding remains unvalidated.
  • All studies are post-hoc: No pre-registered primary pharmacogenomic endpoint RCTs exist for any GLP-1 agent. This creates substantial false-positive risk across the entire literature.
  • “10% genetic non-response” figure has no confirmed primary PMID: Widely cited in secondary press reports. No primary evidence trail confirmed. Real-world non-response is far more likely explained by the 30–50% discontinuation rate at 1 year.
  • Population diversity gap: Most cohorts are European ancestry. Data in African, East Asian, South Asian, and Hispanic populations are largely absent.
  • No clinical utility trial exists: No study has demonstrated that using genetic information to guide GLP-1 prescribing improves outcomes.

Risks and Uncertainty

High-severity limitations

  • rs6923761 replication failure: Most-studied variant shows signal in one small cohort, null in two subsequent studies
  • All pharmacogenomic studies are post-hoc analyses: No pre-registered primary pharmacogenomic endpoint RCTs exist
  • No clinical utility trial: No study has demonstrated that genetic information improves GLP-1 prescribing outcomes
  • Population diversity gap: Most cohorts are European ancestry; African, East Asian, South Asian, Hispanic populations largely unstudied
  • Self-reported outcomes in primary GWAS: Nature 2026 cohort used self-reported weight — recall and social desirability bias concerns
  • ARRB1 vs. GLP1R contradiction: Two largest GWAS point to different primary genes for different outcomes — the pharmacogenomic landscape is not resolved
  • The “10% genetic non-response” figure has no confirmed primary PMID

Evidence quality by claim

ClaimGrade
MC4R mutation carriers: tirzepatide equally effectiveConfirmed
CYP2D6 variants have no direct effect on GLP-1 PKConfirmed
Genetic contribution <10% of total treatment responseConfirmed
No guideline recommends genetic testing for GLP-1Confirmed
ARRB1 Thr370Met → strongest HbA1c pharmacogenetic signalSupported (multi-cohort)
rs6923761 → differential response in severe obesitySupported (single small cohort)
Early response trajectory → best current predictorSupported
rs6923761 as clinical decision toolInsufficient replication
”10% genetic non-response” figureUnsupported — no confirmed primary PMID

Wearable / Vitals Monitoring Implication

Genetic non-response detection via wearables: not actionable. No wearable-derived metric (HRV, RHR, sleep architecture, CGM) has been validated to detect or predict GLP-1 genetic non-response.

Early weight trajectory is implementable now:

  • Patients losing <2–3% body weight at 12 weeks are likely poor long-term responders
  • Implementable with connected scales — no genetic data required
  • If trajectory suggests non-response, consider dose escalation or agent switching in consultation with prescribing clinician
def assess_glp1_response(weight_baseline, weight_12wk):
    pct_loss_12wk = ((weight_baseline - weight_12wk) / weight_baseline) * 100
    return "possible_non_response_signal" if pct_loss_12wk < 2.5 else "on_track"

What Vitals Can and Cannot Say

✅ Can say:

  • “Early weight trajectory is the best available predictor of long-term GLP-1 response”
  • “MC4R deficiency does NOT contraindicate GLP-1 therapy — tirzepatide works equally well in these patients”
  • “Genetic variants from DTC panels are not validated for GLP-1 prescribing decisions”
  • “If GLP-1 therapy isn’t producing results, check adherence and early response trajectory first”
  • “CYP2D6 variants don’t affect your GLP-1 levels — they only matter for some other medications you might be taking”

❌ Cannot say:

  • “Genetics explains why you may not respond to GLP-1 therapy”
  • “This genetic test can guide your GLP-1 treatment choice”
  • “Your wearable data shows you are a genetic non-responder”
  • “10% of GLP-1 users are genetic non-responders” (no confirmed primary PMID)
  • “rs6923761 testing is recommended before starting GLP-1 therapy”

Sources

  • PMID 40384505 — Phan et al., 2025: GLP1R rs6923761 directional signal in severe obesity (n=112)
  • PMID 40996853 — Tourtourikov et al.: GLP1R rs6923761 oral semaglutide nominal trend, n=10
  • PMID 41307691 — Candido et al., 2026: GLP1R rs6923761 null in T2D oral semaglutide (n=210)
  • PMID 40858971 — SURMOUNT-1 pharmacogenomics: MC4R carriers tirzepatide equally effective
  • PMID 41238444 — ARRB1/GLP1R systematic review: β-arrestin mechanism
  • PMID 36528349 — DIRECT GWAS, Lancet Diabetes Endocrinol 2023: ARRB1 Thr370Met strongest HbA1c signal (n=4,571)
  • PMID 37264767 — SCALE Teens liraglutide: early response trajectory predictor
  • PMID 26802434 — C-peptide / β-cell function: low C-peptide predicts non-response
  • PMID 41951734 — Nature 2026: GLP1R missense variant −0.76 kg/copy (n=27,885)
  • PMID 35894080 — TCF7L2 associated with liraglutide weight response (n=136)
  • PMID 30883264 — Chinese exenatide cohort: rs10305420 T allele opposite direction
  • NCT04839237 / NCT05762744 / NCT05071898 — actively recruiting GLP-1 pharmacogenomics trials

Last updated: 2026-04-23 (Batch 101, vault promotion) Evidence boundary: clinical applicability not established; no validated genetic test exists; no clinical utility trial completed

Graph View

Backlinks