If you have been taking a GLP-1 drug like Ozempic, Wegovy, or Zepbound correctly — the right dose, the right frequency, with appropriate dietary and lifestyle changes — and your results are substantially lower than what clinical trial averages suggest, science now has a biological explanation for a subset of those cases.
Researchers at Stanford Medicine, ETH Zurich, and the University of Adelaide published a study in Genome Medicine on April 10, 2026 finding that approximately 10% of the population carries genetic variants in the PAM gene that measurably reduce the biological effectiveness of GLP-1 receptor agonist drugs. The finding applies to drugs including semaglutide (Ozempic, Wegovy) and liraglutide — and provides a biological mechanism for what clinicians have long observed anecdotally: some patients who are clearly adherent to their regimen consistently experience lower-than-expected results.
The research does not change current clinical guidelines or create a new standard-of-care test. But it matters enormously for patients who have been told — implicitly or explicitly — that their poor GLP-1 response reflects inadequate effort.
Why This Matters
The clinical framing of GLP-1 "non-response" has, up to this point, largely attributed low results to compliance: patients taking the drug inconsistently, at insufficient doses, without adequate dietary changes, or stopping treatment too early.
But the Stanford and ETH Zurich research identifies a distinct population for whom the problem is not compliance — it is biology. Scientists have identified variants in the gene coding for the enzyme PAM (peptidyl-glycine alpha-amidating monooxygenase) that appear to cause GLP-1 resistance. This genetic change is carried by approximately 10% of the population.
For those individuals, GLP-1 drugs work through the same molecular pathway but produce a dampened signal. Paradoxically, people with PAM variants have higher circulating GLP-1 levels — but about 18% lower GLP-1 sensitivity, meaning the hormone is present but less biologically effective. The experience of this patient population is not "I didn't take the drug" — it is "I took the drug correctly and it did less than it should have."
What We Know So Far
The genetic research on GLP-1 response now encompasses two major complementary studies:
The Stanford/ETH Zurich PAM Study (April 10, 2026): Led by Anna Gloyn (Stanford Medicine) and Markus Stoffel (ETH Zurich), published in Genome Medicine with lead authors Mahesh Umapathysivam (University of Adelaide) and Elisa Araldi (University of Parma), the study found that two missense variants in the PAM gene (p.S539W and p.D563G) impair the amidation of peptide hormones — a biochemical step required for GLP-1 to reach its full biological effect. Approximately 10% of the population carries these variants.
The study's primary finding applies to glycemic control (HbA1c reduction): about 25% of non-carriers reached the recommended HbA1c target after six months of treatment, compared with 11.5% of participants with the p.S539W variant and 18.5% with the p.D563G variant. Importantly, as Dr. Gloyn herself noted, the study included only limited weight loss data — "only two of the clinical trials analyzed in the study provided weight data, which showed no difference in weight loss between those with and without PAM variants, but the data is too limited to be conclusive." The weight loss implication requires further study.
The Nature/23andMe Study (April 8, 2026): An international research team conducted a genome-wide association study of self-reported weight loss and treatment-related side effects in 27,885 people following GLP-1 receptor agonist therapy, published in Nature. The study identified a missense variant in the GLP1R gene itself associated with increased GLP-1 drug efficacy — specifically approximately 0.76 kg additional weight loss per copy of the effect allele. A separate GIPR variant was associated with increased nausea under tirzepatide specifically.
Reading the two studies together, there are at least two independent genetic layers that influence GLP-1 drug response: a ligand-processing layer (PAM) and a receptor layer (GLP1R, GIPR). A non-responder can have a PAM problem, a GLP1R problem, or both.
What the Studies Found and Did Not Find
MedicalDaily Evidence Check
- PAM study: Published in Genome Medicine (DOI: 10.1186/s13073-026-01630-0), April 10, 2026 ; Stanford Medicine and ETH Zurich; 1,119 participants analyzed; PAM variants found in ~10% of population; associated with reduced GLP-1 effectiveness for blood sugar control (HbA1c)
- GLP1R/GIPR study: Published in Nature (April 8, 2026); 23andMe and international collaborators; 27,885 GLP-1 users; identified GLP1R variant for increased efficacy and GIPR variant for tirzepatide-specific nausea
- What the research found: Genetic variants that measurably influence GLP-1 drug effectiveness and side effect profiles
- Key limitation on weight loss: The PAM study had insufficient weight loss data to draw conclusions about whether PAM variants affect weight loss; the glycemic control finding is well-powered, the weight loss implication is not yet established
- What it did not prove: That genetic testing should currently change clinical prescribing decisions; no standard clinical test exists yet
- What it does not mean: That non-responders should stop their GLP-1 drug — compliance-related reasons for poor response are still more common
Where the Practical Implications Are Greatest
Patients most likely to benefit from this research framework are:
- People who have been on a GLP-1 drug at the recommended dose for at least 3 months, with documented good compliance, and whose blood sugar control is significantly below trial averages
- Patients who have been told they are "non-responders" and for whom compliance has been incorrectly assumed as the explanation
- Prescribers managing patients with unexplained low GLP-1 response who want to counsel patients accurately rather than attributing poor results to behavioral factors
What Doctors and Experts Say
"When I treat patients in the diabetes clinic, I see a huge variation in response to these GLP-1-based medications and it is difficult to predict this response clinically," said Mahesh Umapathysivam, the lead author of the study.
"This marks a turning point in how we think about GLP-1 therapy response," said Anna Gloyn, professor of pediatric diabetes research at Stanford Medicine. "For years, poor response was attributed almost entirely to adherence. The genetics show that for a meaningful fraction of patients, there is a biological limitation that adherence cannot overcome" — at least for glycemic control, where the evidence is now well-established.
What You Can Do Now
- If you believe you are a GLP-1 non-responder, discuss this specifically with your prescribing clinician. Ask: "Is my dose at the recommended therapeutic level? Is there evidence of compliance? Have we ruled out other contributors to my response?"
- Do not stop your GLP-1 drug based on this research alone. Genetic variants reduce drug effectiveness — they do not eliminate it. Stopping a drug that is providing partial benefit may leave you without any treatment.
- Ask your prescriber about dose optimization. Some non-responders benefit from higher doses or longer titration periods before concluding the drug is ineffective.
- Note that pharmacogenetic testing for GLP-1 response is not yet a standard clinical service. Research-context testing exists but is not routinely available through clinical labs.
The Bottom Line
Approximately 10% of people carry genetic variants in the PAM gene that reduce the biological effectiveness of GLP-1 drugs — particularly for blood sugar control — offering a biology-based explanation for the subset of GLP-1 non-responders whose low results cannot be attributed to compliance. The weight loss implication requires further study to confirm. This research does not change current prescribing guidelines and does not mean non-responders should stop their medication. It does mean that the clinical conversation about poor GLP-1 response needs to include biology alongside behavior.
