Why GLP-1 Drugs Don't Work for 10% of People: Stanford Identifies the PAM Gene Variant Behind GLP-1 Resistance

A decade-long, international research effort led by Stanford Medicine has identified a genetic explanation for why GLP-1 drugs fail to work adequately in a significant subset of patients. The study, published April 10, 2026 in Genome Medicine, found that variants in the PAM gene — carried by roughly 10% of the general population — reduce the biological effectiveness of GLP-1, both the body’s natural hormone and the synthetic drug versions.

The PAM gene encodes peptidyl-glycine alpha-amidating monooxygenase, an enzyme with a unique and critical function: it is the only enzyme in the human body capable of performing amidation. Amidation is a chemical modification that adds an amide group to the C-terminus of peptide hormones. This seemingly small modification has outsized biological importance — it increases the potency, receptor binding affinity, and half-life of dozens of peptide hormones, including GLP-1.

The two key variants identified are p.S539W and p.D563G. Both impair PAM’s enzymatic function, which means that in carriers, GLP-1 and other peptide hormones are produced but not properly amidated — resulting in reduced biological activity even when circulating levels appear normal or elevated.

This is a fundamental discovery for the peptide field. PAM does not only modify GLP-1 — it amidates a wide range of bioactive peptides throughout the body. The implications extend well beyond GLP-1 drugs, though that is where the clinical evidence is strongest.

One of the most counterintuitive findings in the study is that people with PAM variants actually have higher circulating GLP-1 levels than non-carriers — the opposite of what researchers initially expected.

The Stanford team recruited adult participants with and without the p.S539W variant, administered a standardized glucose challenge (a sugary drink), and measured blood biomarkers every five minutes for four hours. Carriers of the PAM variant showed elevated GLP-1 levels — yet their blood sugar response was impaired compared to non-carriers.

The explanation lies in the amidation mechanism. When PAM function is reduced, the body’s GLP-1 is produced in its non-amidated form. Non-amidated GLP-1 is less potent at the GLP-1 receptor — it binds less effectively and triggers a weaker intracellular signaling cascade. The body compensates by producing more GLP-1, which explains the elevated circulating levels. But quantity cannot fully compensate for reduced quality.

Quantitatively, p.S539W carriers showed an 18% reduction in endogenous GLP-1 sensitivity. This was confirmed in both human participants and PAM-knockout mouse models, providing strong cross-species validation.

Why this matters for drug therapy. If the body’s own GLP-1 is less effective in PAM variant carriers, it follows that synthetic GLP-1 receptor agonists (which work by activating the same receptors) may also be less effective. The clinical trial data confirmed this hypothesis.

The researchers validated their findings against clinical trial data from three GLP-1 drug trials involving a combined 1,119 participants. The results were clear and consistent.

HbA1c response. After six months of GLP-1 receptor agonist treatment, approximately 25% of non-carriers reached the recommended HbA1c target. Among p.S539W carriers, only 11.5% reached target. Among p.D563G carriers, 18.5% reached target. Both differences were statistically significant.

Specificity to GLP-1 drugs. Critically, participants with PAM variants did not respond differently to other common diabetes treatments — including metformin, sulfonylureas, and DPP-4 inhibitors. This specificity is important because it indicates the resistance mechanism is tied specifically to GLP-1 biology, not to a general metabolic or drug-processing difference.

Mouse model confirmation. PAM-knockout mice showed the same pattern: elevated GLP-1 levels but reduced GLP-1 sensitivity. When treated with GLP-1 receptor agonists, the knockout mice showed less glycemic improvement than wild-type controls, mirroring the human clinical trial findings.

The convergence of human clinical data, human physiological testing, and mouse genetic models makes this one of the more robust pharmacogenomic findings in the GLP-1 field. The 10% prevalence figure — if it holds across larger studies — means that millions of current GLP-1 drug users may be experiencing reduced efficacy due to their PAM genotype.

The practical implications of this discovery are significant for both patients currently on GLP-1 therapy and prescribers who manage GLP-1 drug regimens.

For patients experiencing suboptimal GLP-1 response. If you are taking semaglutide, tirzepatide, or another GLP-1 agonist and experiencing less blood sugar improvement or weight loss than expected, a PAM variant could be a contributing factor. This does not mean the drug is useless — 11.5% of carriers still reached target HbA1c — but it may mean that higher doses, combination therapy, or alternative approaches could be more appropriate.

For prescribers. This study supports the case for pharmacogenomic testing before or during GLP-1 therapy. Currently, GLP-1 drugs are prescribed empirically — patients start at a low dose, titrate up, and response is evaluated over months. If PAM variant testing were incorporated into prescribing workflows, physicians could identify likely poor responders earlier and adjust treatment plans accordingly, potentially saving 3–6 months of suboptimal treatment.

For the precision medicine field. GLP-1 drugs are now the most widely prescribed class of medications for obesity and type 2 diabetes. Identifying a genetic biomarker that predicts response in 10% of the population is exactly the kind of finding that can drive clinical pharmacogenomics forward. The question is how quickly testing will be incorporated into practice.

What this does NOT mean. PAM variants do not make GLP-1 drugs dangerous — they make them less effective. There is no safety concern specific to carriers. The finding also does not apply to all patients who fail to respond to GLP-1 drugs — there are many other reasons for poor response, including adherence, diet, comorbidities, and other genetic factors.

While the Stanford study focused on GLP-1 drugs, the broader implications extend to the entire field of peptide therapeutics. PAM is not specific to GLP-1 — it amidates more than 50% of all bioactive peptide hormones in the human body.

Peptides known to require amidation for full biological activity include GLP-1, GLP-2, GIP, oxytocin, vasopressin, calcitonin, gastrin, substance P, neuropeptide Y, and many others. If PAM variants impair amidation broadly, carriers may have subtly altered responses to a wide range of peptide-mediated biological processes — not just glucose metabolism.

For the peptide therapy community specifically, this raises questions about whether PAM genotype affects response to other peptide treatments — including BPC-157, thymosin alpha-1, and other research peptides that may require amidation for optimal activity. No studies have examined this directly, but the biological logic is compelling.

For drug development, the PAM finding suggests that future peptide drugs could be designed with pre-amidated or amidation-independent structures to ensure consistent efficacy across all genotypes. This is a solvable engineering problem once the biological target is understood.

The Stanford study demonstrates that our understanding of individual variation in peptide biology is still in early stages. As peptide therapeutics become an increasingly central part of medicine — from GLP-1 drugs to healing peptides to neuropeptides — understanding the genetic factors that influence peptide processing and activity will become essential.

The Stanford PAM variant study provides the first robust genetic explanation for GLP-1 drug resistance — a phenomenon that clinicians have observed anecdotally for years but had no biological basis to explain. Roughly 10% of the population carries variants that reduce GLP-1 effectiveness by impairing the amidation process that gives peptide hormones their full biological potency.

For patients, this means that poor response to GLP-1 drugs may not be a failure of willpower or adherence — it may be genetic. For prescribers, it offers a potential biomarker to guide treatment decisions. For the pharmaceutical industry, it highlights the importance of pharmacogenomic stratification in an era when GLP-1 drugs are being prescribed to tens of millions of patients.

The study was published in Genome Medicine on April 10, 2026, and represents a decade of international research combining human clinical trials, physiological testing, and mouse genetic models. The 10% prevalence figure, if confirmed in larger populations, has major implications for how GLP-1 drugs are prescribed and monitored.

Patients currently taking GLP-1 medications should not stop treatment based on this study. Those experiencing suboptimal results should discuss pharmacogenomic testing options with their healthcare provider.