I see this pattern every week. A patient is 12 weeks into semaglutide therapy. They’ve lost 2%-3% of their initial body weight, and they look defeated. Before they’ve even had time to see the full effect of therapy, someone has labeled them a “nonresponder.” Maybe it was their prior clinician. Maybe it was a telehealth algorithm. Or maybe it was themselves.
So, what happens? They stop their prescription. The clinical relationship gets disrupted. And the patient walks out taking away the exact message that obesity medicine works so hard to prevent — that the failure is theirs, and theirs alone.
I want to push back on this word, nonresponder. We reach for it far too easily, and it’s a convenient scapegoat for when the clinical picture gets messy.
What True Resistance Looks Like
To be fair, real pharmacologic resistance to GLP-1 receptor agonists exists, and we have evidence of it for the first time. But it’s rare, and the patients we describe as “nonresponders” are almost never the ones we actually see in clinical practice.
Michael Albert, MD
A 2026 study in Genome Medicine zeroed in on loss-of-function variants in PAM, the enzyme that amidates GLP-1 into its active form. Carriers for the variant, which roughly encompass about 10% of people, paradoxically run higher GLP-1 levels that signal less effectively. Pooled across three trials, people with a loss-of-function variant were far less likely to hit their A1c target on a GLP-1, about 11%-19% vs a quarter of noncarriers. Yet most people responded normally to metformin, sulfonylureas, and DPP-4 inhibitors. So, this is highly genetic. It is pathway specific and very selective to a drug class. Resistance is real.
Antibodies are the other textbook way a peptide drug can fail. In the case of antibodies, the immune system recognizes this foreign agent and neutralizes it. But that barely bears any practical relevance to the GLP-1s we prescribe today. The older exendin-based agents, like exenatide, were far more immunogenic. But modern human analogues were built to mimic native GLP-1, and thus they avoid immune neutralization. The data support this: Antibodies appear in about half of tirzepatide patients and have no clinical significance on outcomes. Antibodies against drugs are real, but they have almost no bearing on the patient sitting in front of you.
The 12-Week Fallacy
We also need to keep the genetic finding in context. This is a signal around glycemic control from diabetes trials, and the weight-loss data in PAM carriers is still limited. Even with a 10% population prevalence, it only explains a small fraction of the nonresponders we see in the clinic. So, who are the rest of these people?
Mostly, it’s people who were judged and labeled far too early. Response varies substantially on these medicines, as observed in the STEP trials: 10%-17% of patients lose under 5%, while a third to 40% lose more than 20%. Yet we hand down a conclusive verdict by week 12, based on current guidance and payer policies which often recommend stopping obesity pharmacotherapy if a patient hasn’t achieved 5% weight loss after 3-6 months on a full dose. This standard was derived from older agents and expert consensus rather than long-term data on newer incretin-based therapies.
And in fact, we know that convention is wrong — at least, in the context of newer incretin-based therapies. When SURMOUNT-1 was re-analyzed in 2025, patients on tirzepatide who had not reached 5% by week 12 overwhelmingly caught up. About 90% of those patients obtained at least 5% weight loss by week 72, averaging nearly 25 weeks to get there. If you’re condemning these people to a lifetime label of nonresponse, then you aren’t representing the full picture.
Dose & Exposure
In my practice, nonresponder is not a conclusion. It is merely a hypothesis. And the explanation I see far more often is one about exposure. We know from modeling of the STEP data that circulating semaglutide levels largely drive the response curve, and that exposure drops as body weight climbs. The larger the patient, the lower the circulating drug level for a given dose.
A separate analysis in diabetes went even further: If you control for plasma level, the receptor sensitivity differences mostly wash out. The factor that matters usually isn’t someone’s background biology; it’s dose and exposure. And many nonresponders never get enough exposure, because side effects or affordability stall titration and continued use.
This is exactly where the field is going. The approval of high-dose semaglutide, 7.2 mg, in March 2026 — based off the phase 3 STEP UP trial — is proof. About 19% (placebo-subtracted) average weight loss was achieved with high-dose semaglutide, against roughly 15% on the standard 2.4 mg dose. And a third of people taking high-dose semaglutide lost greater than 25% of their body weight. That’s the field conceding, in writing, that a significant portion of nonresponse was just under-dosing.
We also have to talk about adherence, which we are awful at measuring. In one large commercial cohort, two-thirds of patients quit their GLP-1 medication within a year, and the average adherence was a mere 27%. Someone taking the drug a quarter of the time isn’t resistant. They’re barely taking it.
Reclaiming the Plateau
Even the weight-loss plateau is a trap. The flattening of the weight-loss response curve around 60-72 weeks isn’t the drug stopping. It’s the body defending a set point. And the tell is what happens when you stop: The weight recurs, proof that the disease was being managed, not cured. A plateau means the drug is working.
So why do we make a fuss over one word? Because the word carries real weight. It ends treatment for someone who would have responded. It reinstates the old narrative that obesity is merely a product of willpower. And it limits the next move: a higher dose of the medicine, a switch to a more effective therapy, the addition of an adjunctive agent, or even consideration of an alternative like bariatric surgery.
The volume-driven, pill-mill telehealth models exacerbate this. There’s no margin for a slow response, or for a nuanced conversation around resistance, in a high-volume, minimal-interaction clinical environment. So, the label of resistance or treatment failure becomes the easiest exit.
We need to be measured before we write it. We need to confirm that patients are on adequate dosing and are adhering to their medication. Patients need trials longer than 12 weeks. Comprehensive support and care need to be built around the injection, full stop. If you do that, your true nonresponders shrink to the tail of the genetics. Everyone else, written off too soon, is usually capable of responding to the drug, if only we give them the chance.