One in Ten People May Be Genetically Resistant to GLP-1 Diabetes Drugs Like Ozempic

The body makes plenty of the hormone. Arguably too much of it. And yet the hormone does almost nothing. This is the paradox at the heart of a new study that could reshape how doctors prescribe GLP-1 drugs, the blockbuster class of medications that includes Ozempic and Wegovy, to people with type 2 diabetes. About 10 percent of the general population, it turns out, carry genetic variants that leave them partially deaf to GLP-1, the very molecule these drugs are designed to mimic.

The finding, published in Genome Medicine, is the product of a decade-long international collaboration spanning experiments in humans and mice, clinical trial data from over a thousand participants, and biochemistry that kept delivering results nobody anticipated. It is also the first in-depth investigation of what the researchers call GLP-1 resistance.

GLP-1, or glucagon-like peptide-1, is a gut hormone released after you eat. It nudges the pancreas to produce insulin, slows the rate at which your stomach empties (which is partly why the drugs curb appetite), and helps keep blood sugar in check. GLP-1 receptor agonists work by impersonating this hormone, flooding the system with a synthetic stand-in. More than a quarter of people with type 2 diabetes now take them. The trouble is that for some patients, the drugs don’t work nearly as well as expected, and until now clinicians have had little way of predicting who those patients would be.

The story begins with an enzyme called PAM, peptidyl-glycine alpha-amidating monooxygenase. It is, as far as biologists can tell, unique: the only enzyme in the human body capable of performing a chemical tweak called amidation, which boosts the potency and lifespan of many hormones, GLP-1 among them.

Two genetic variants in the PAM gene were already known to raise the risk of type 2 diabetes. Anna Gloyn, a professor of pediatrics and genetics at Stanford Medicine who co-led the study, had previously shown that these variants impair insulin release from pancreatic beta cells. The logical next question was whether the same genetic glitch also mucks up GLP-1. Perhaps, the team reckoned, carriers of the variants would have lower circulating levels of the hormone, since the unamidated version might be less stable. They recruited healthy volunteers from the Oxford Biobank, had them drink a sugary solution, and drew blood every five minutes for four hours.

“What we actually saw was they had increased levels of GLP-1,” Gloyn said. “This was the opposite of what we imagined we would find.”

More hormone, but no extra benefit. The carriers’ blood sugar did not drop any faster than non-carriers’. Their incretin response, the boost in insulin secretion you get from eating versus receiving the same glucose intravenously, was essentially the same. When the researchers calculated a ratio of GLP-1 levels to biological effect, carriers showed an 18% reduction in GLP-1 sensitivity. The hormone was there, sloshing around in abundance, but the body could not seem to hear it properly. Gloyn likens the situation to insulin resistance, a phenomenon clinicians have grappled with for decades without fully understanding its mechanism. “We have ticked off this enormous list of all the ways in which we thought GLP-1 resistance might come about. No matter what we’ve done, we’ve not been able to nail precisely why they are resistant.”

That result was so surprising the team spent years confirming it. Collaborators in Zurich built knockout mice lacking the PAM gene entirely. Those mice, too, had elevated GLP-1 but showed no corresponding improvement in blood sugar control. Their stomachs emptied faster than normal, and treating them with a GLP-1 receptor agonist did nothing to slow it down. Researchers in Copenhagen tested whether the problem lay at the receptor itself, checking binding affinity, signaling through cAMP, beta-arrestin recruitment and receptor internalization. Nothing differed between the amidated and non-amidated forms of GLP-1. Whatever was going wrong, it was happening further downstream, somewhere in the signaling cascade after the hormone docks with its receptor. In the pylorus (the muscular valve between stomach and small intestine), cAMP levels in knockout mice were lower after GLP-1 stimulation, even though the receptor was expressed at normal levels.

The clinical stakes became clear when the team examined data from drug trials. In a meta-analysis of 1,119 participants with type 2 diabetes treated with GLP-1 receptor agonists across three cohorts, carriers of the more severe PAM variant saw their average blood sugar marker, HbA1c, drop by just 0.69 percentage points after six months, compared with 1.24 points in non-carriers. That amounts to a 44 percent relative loss of the drug’s glycemic benefit. Roughly 11.5% of carriers hit the recommended HbA1c target of below 7%, versus about a quarter of non-carriers. Carriers of the milder variant fared somewhat better but still lagged behind. And crucially, neither variant affected how well patients responded to metformin, sulfonylureas, or DPP-4 inhibitors. “We can see very clearly that this is specific to medications that are working through GLP-1 receptor pharmacology,” Gloyn said.

There are wrinkles, though. Two pharmaceutical company trials using longer-acting GLP-1 receptor agonists did not show the same disparity between carriers and non-carriers. One possible explanation is that continuous receptor stimulation from these longer-acting formulations might sort of override the resistance, though that remains speculative. Weight loss data, which everyone wants to know about given the drugs’ popularity as obesity treatments, were available from only two of the five trials and showed no significant difference, but the dataset is too small to draw conclusions. “For the newer GLP-1 medications, it would be useful to look at whether there are genetic variants, like the variants in PAM, that explain poor responders to their medications,” Gloyn said, noting that pharmaceutical companies routinely collect genetic data on trial participants but have been slow to share it.

Mahesh Umapathysivam, an endocrinologist at Adelaide University and one of the study’s lead authors, sees the work as a step toward something clinicians have long wanted: a way to match patients with the drugs most likely to help them. “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,” he said. “This is the first step in being able to use someone’s genetic make-up to help us improve that decision-making process.”

The mechanism remains, as Gloyn puts it, “the million-dollar question.” But there is a useful precedent. Insulin resistance was poorly understood for years, and still is in many respects. That did not stop researchers from developing insulin sensitizers, drugs that help the body respond to insulin more effectively. Perhaps something similar could be done for GLP-1. “There are a whole class of medications that are insulin sensitizers, so perhaps we can develop medications that will allow people to be sensitized to GLP-1s or find formulations of GLP-1, like the longer-acting versions, that avoid the GLP-1 resistance,” Gloyn said. For roughly one in 10 people, that possibility is not abstract pharmacology. It is the difference between a drug that works and one that doesn’t.

Journal reference: Genome Medicine, DOI: 10.1186/s13073-026-01630-0