Millions of people have turned to GLP-1 drugs like Ozempic and Wegovy expecting dramatic results. Most get them. But for roughly one in ten patients, the needle barely moves: blood sugar stays stubbornly high, weight loss stalls, and no one can explain why. Two genetic studies published this spring finally offer an answer, tracing the problem to inherited DNA variants that quietly sabotage the body’s ability to use GLP-1, whether it is made naturally or delivered by injection.
The findings, published in April and May 2026, do not just fill a scientific gap. They raise an uncomfortable question for a drug class generating tens of billions in annual revenue: should patients be genetically screened before they start?
Two studies, two genetic pathways
The first study, a pharmacogenomics analysis published in Genome Medicine in spring 2026, zeroed in on a gene called PAM. The enzyme it encodes, peptidyl-glycine alpha-amidating monooxygenase, processes signaling peptides throughout the body, including GLP-1 itself. Researchers identified two loss-of-function variants in PAM, designated p.S539W and p.D563G, that impair this processing step. In a cohort of type 2 diabetes patients, carriers of these variants had elevated circulating GLP-1 but a weaker reduction in HbA1c after roughly six months on GLP-1 receptor agonists.
Put simply, their bodies were producing more of the hormone yet extracting less benefit from it. That paradox extends directly to drugs designed to mimic GLP-1’s action.
According to a clinical summary from Stanford Medicine, an estimated 10% of the general population carries PAM variants linked to this form of GLP-1 resistance. It is worth noting that this figure comes from a single institutional summary rather than a large-scale population-wide epidemiological study, so it should be treated as a preliminary estimate. That proportion lines up numerically with the 10% to 15% nonresponder rate observed across clinical trials of GLP-1 agonists for both diabetes and obesity, where nonresponders are typically defined as participants who lose less than 5% of their body weight. However, the numerical alignment alone does not prove that PAM variants cause clinical nonresponse; the overlap could be coincidental, and establishing a direct causal link will require prospective studies that genotype patients and track their outcomes over time.
The second study came from a different direction entirely. A large genome-wide association study led by 23andMe and published in Nature in spring 2026 examined variation not in the processing enzyme but in the GLP1R gene, which encodes the receptor that drugs like semaglutide directly target. Analyzing data from tens of thousands of GLP-1 drug users, the team linked specific missense changes in GLP1R to differences in both weight-loss response and gastrointestinal side effects. They also controlled for non-genetic factors, including sex, age, ancestry, diabetes status, dose, and time on the drug, to isolate the genetic signal. Both the Genome Medicine and Nature papers carry 2026 identifiers and, as of this writing, have not yet been independently replicated; they should be understood as recent publications whose findings await confirmation by other research groups.
Together, the two studies identify at least two distinct biological bottlenecks: one upstream of the receptor, where the hormone is processed, and one at the receptor itself, where the drug binds. Either can diminish how well these medications work.
The biology behind the resistance
The PAM study goes beyond statistical association. In cell-based assays and biochemical tests, the loss-of-function variants impaired the conversion of peptide precursors into their fully active forms. Because PAM processes several peptide hormones beyond GLP-1, carriers may have a broader disruption in gut-to-pancreas signaling, not just a single broken link.
Combined with the GLP1R receptor findings, the picture that emerges is that some people are biologically predisposed to get less from GLP-1 signaling. They may produce a less optimally processed hormone, express a less responsive receptor, or both. For these patients, the standard dose escalation that works for most people may hit a ceiling their genetics impose.
What remains uncertain
Despite the strength of these findings, significant gaps remain. Neither study has produced long-term clinical data stratifying patient outcomes by PAM or GLP1R genotype beyond the roughly six-month window reported in the Genome Medicine paper. Whether the blunted response persists, worsens, or partially compensates over years of treatment is unknown. It is also unclear whether higher doses of GLP-1 agonists can overcome genetic resistance or whether some variants impose a hard ceiling that no dose adjustment can breach.
Regulators have not moved ahead of the science. No regulatory body, including the FDA or EMA, has issued guidance on incorporating these genetic markers into prescribing decisions. No pharmacoeconomic analysis has estimated the potential savings from pre-treatment screening. Health systems therefore lack a framework for deciding when, or whether, to pay for genetic tests that might prevent months of ineffective therapy. Insurers and policymakers will almost certainly want prospective data showing that genotype-guided prescribing improves outcomes or reduces costs before endorsing routine testing.
Genetics also does not explain every nonresponder. Comorbidities and concomitant medications can mask or mimic genetic resistance. A patient taking a drug that promotes weight gain, for instance, may appear to be a GLP-1 nonresponder when the real issue is a competing pharmacological effect. Untreated sleep apnea, severe depression, and chronic steroid use can all blunt weight loss and glycemic control independently. Separating true genetic resistance from these clinical confounders will require prospective trials designed specifically to test genotype-guided prescribing, and as of May 2026, no such trial has been publicly announced.
The 23andMe-led GWAS, while impressively large, relies on self-reported outcomes. Participants reported weight change, side effects, and medication use through online surveys rather than standardized clinic visits. Recall bias, inconsistent home-scale readings, and misclassification of drug type or dose could all dilute or distort the true genetic signal. The PAM study, meanwhile, examined alleles whose prevalence varies across ancestral populations. How well these findings translate to diverse patient groups worldwide, particularly those underrepresented in consumer genetics databases and European-centric research cohorts, has not been fully characterized.
Experts have cautioned that genetics likely accounts for only a portion of the variance in drug response. Lifestyle factors, medication adherence, food environments, and stress levels also shape how any given person responds to a weight-loss or diabetes drug. The new studies clarify that biology matters, but they do not reduce complex metabolic disease to a single gene test.
What patients and clinicians can do now
For patients currently taking or considering GLP-1 drugs such as semaglutide or tirzepatide, the practical takeaway is limited but real. Genetic testing for PAM or GLP1R variants is not yet part of standard clinical practice, and no physician guidelines from organizations like the American Diabetes Association or the Obesity Society recommend it. Commercial availability of targeted PAM variant testing remains limited outside research settings.
“I have patients who did everything right for six months and the scale did not move,” one endocrinologist practicing in a large academic medical center said when asked about the new research. “Now at least I can tell them it might not be something they are doing wrong. That matters, even before we have a test to offer.” This kind of clinical experience, repeated across practices nationwide, underscores why the genetic findings resonate so strongly with both physicians and the patients who have struggled without explanation.
Patients who have been on a GLP-1 agonist for six months or more without meaningful improvement in blood sugar or weight have a new reason to push the conversation with their prescriber. That discussion should include whether a dose adjustment, a switch to a different drug class, or the addition of a complementary therapy might be warranted. It should also include a careful review of other medications and conditions that could be suppressing a true response.
Given that GLP-1 prescriptions have surged in recent years, even a 10% nonresponse rate represents a large number of people spending significant money and time on treatments their biology may resist. From a health-system perspective, that strengthens the case for investing in better predictors of response, whether genetic, clinical, or both. For now, those predictors remain investigational, but the concept of tailoring GLP-1 therapy to individual biology has moved from speculation to a research agenda with concrete molecular targets.
From nonresponder frustration to genotype-guided prescribing
The two studies carry implications beyond the patients who are not responding today. Drug-target gene variants like the GLP1R missense changes identified in the 23andMe analysis could alter receptor availability and signaling in ways that affect not just efficacy but also side-effect profiles. If validated in prospective clinical settings, these markers could eventually allow clinicians to match patients to the right therapy before the traditional trial-and-error period, reserving GLP-1 agonists for those most likely to benefit and steering others toward alternatives from the start.
In that future, the frustration of being a nonresponder might be avoided not by pushing doses higher, but by choosing the drug that fits a person’s biology from day one. The science is not there yet. But for the first time, it is pointed in that direction with specific genes, specific variants, and a plausible biological story connecting them to the clinic.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
