Why GLP-1 Weight Loss Drugs Stop Working for So Many People
Science | May 27, 2026
GLP-1 weight loss drugs like semaglutide have reshaped how medicine treats obesity, but researchers have struggled to answer a basic question: why does weight loss eventually plateau in most patients, and why do two people on the same drug at the same dose lose such different amounts of weight? New research from the National Institutes of Health, published this week in Nature Metabolism, offers the most detailed cellular-level answer yet, and it points toward a future generation of treatments that could address both problems at once.
The study, led by Claire Gao, Ph.D., a postdoctoral fellow at NIH’s National Institute of General Medical Sciences, used fluorescence imaging to observe in real time how semaglutide, the active compound in Ozempic and Wegovy, triggers signaling activity inside individual brain cells in mice. What Gao’s team found was not the clean, uniform signal that prior research implied. It was something more variable, more interesting, and more clinically useful to understand.
What GLP-1 Weight Loss Drugs Are and Why They Took Over
To understand why the NIH findings matter, it helps to understand the scale of what GLP-1 treatments have become in a short time. Semaglutide, developed by Novo Nordisk, was originally approved for type 2 diabetes management under the brand name Ozempic. Its weight-loss formulation, Wegovy, received FDA approval in 2021. Tirzepatide, Eli Lilly’s dual GLP-1/GIP receptor agonist sold as Mounjaro and Zepbound, followed and in many trials produced even larger weight reductions.
The financial scale is staggering. Eli Lilly’s first nine months of 2025 saw Mounjaro and Zepbound generate $39.5 billion in combined revenue, surpassing Keytruda, Merck’s blockbuster cancer immunotherapy, as the world’s best-selling medicine. The global GLP-1 market was valued at roughly $94 billion in 2025 and is projected by some analysts to reach $216 billion by 2035. Approximately 10 million Americans were on GLP-1 treatment by 2025, with projections pointing to 25 million by 2030.
The drugs work, in their broadest outline, by mimicking glucagon-like peptide 1, a hormone the gut produces after eating that tells the brain to reduce appetite and slow gastric emptying. They bind to GLP-1 receptors on neurons in several brain regions involved in appetite regulation, triggering a signal that the body interprets as satiety. This suppresses the urge to eat, reduces caloric intake, and produces weight loss.
That mechanism was well understood. What was not understood, until this week’s publication, was what happens inside those neurons at the molecular level, and why the response is so inconsistent between cells, between patients, and over time.
What the New Research Found
The NIH team focused on the area postrema, a small region at the base of the brainstem that plays a central role in appetite regulation and nausea responses. Using fluorescence imaging that allowed them to observe individual cells in living brain tissue, Gao and her colleagues tracked what happened when semaglutide bound to GLP-1 receptors on neurons in this area.
The key molecule in their findings is cyclic adenosine monophosphate, known as cAMP. When a GLP-1 receptor is activated by semaglutide, one of the downstream effects is an increase in intracellular cAMP levels. This is the molecular signal that, in turn, produces the appetite-suppressing effect. The NIH team found that cAMP increases reliably when semaglutide is present. What they also found is that the magnitude and duration of that increase varies substantially from one neuron to the next.
“It was not an all or nothing phenomenon,” said Michael Krashes, Ph.D., a senior investigator at the National Institute of Diabetes and Digestive and Kidney Diseases and co-corresponding author on the paper. “We observed that cAMP responses across cells varied on a continuum.”
Some neurons maintained elevated cAMP levels for extended periods during semaglutide exposure. Others showed only brief, temporary increases before cAMP levels fell back toward baseline. The team traced part of this variability to an enzyme called PDE4, which breaks down cAMP. Cells that expressed higher levels of PDE4 activity showed shorter-lived responses, effectively dampening the signal that produces the appetite suppression. When the researchers used a drug called roflumilast, which inhibits PDE4, to block that enzyme, more neurons shifted toward the longer-lasting response pattern.
The plateau phenomenon that affects most patients, studies suggest roughly 23% hit a plateau within three to six months even without changing their behaviour, now has a plausible molecular explanation. As the body adapts to sustained semaglutide exposure, some neurons reduce their responsiveness by internalising or degrading GLP-1 receptors, and the PDE4-driven breakdown of cAMP shortens the signal that survives activation. The drug is still binding to its target. The downstream effect is simply being attenuated.
Why Responses Vary So Much Between Patients
The cellular variability the NIH team documented also offers a framework for understanding something that has puzzled clinicians since the first large semaglutide trials: the enormous spread in outcomes. In clinical trials for Wegovy, the average weight loss is approximately 15% of body weight over 68 weeks. But the distribution around that average is wide. Some patients lose 25% or more. Others lose less than 5%. A meaningful proportion stop losing weight entirely well before their physician expects them to.
Standard pharmacological explanations, adherence, diet during treatment, baseline metabolic rate, account for some of this variance. They do not account for all of it. The NIH findings suggest that differences in the baseline ratio of responsive to less-responsive neurons, or in individuals’ expression levels of PDE4 and related enzymes, may predispose certain patients to weaker or more rapidly attenuating responses before the first injection is administered.
This has direct implications for how these drugs might be prescribed in the future. If a biomarker could identify patients whose neural signaling architecture is likely to produce attenuated responses, physicians could adjust dosing strategies earlier, combine treatments at the outset rather than after a plateau has been reached, or investigate whether PDE4 inhibitors like roflumilast might serve as useful adjuncts.
It also raises harder questions about long-term treatment design. Most patients who stop taking GLP-1 drugs regain the weight they lost, often rapidly. Research from Cambridge published earlier this year found that patients regain weight quickly after stopping but typically retain about a quarter of their total loss long-term. If the mechanism producing weight loss gradually attenuates during treatment, the therapeutic window, the period of meaningful benefit, may be more limited than the current generation of prescribing guidelines assumes.
What Remains Contested and Unknown
The NIH study was conducted in mice, not humans, and the specific neural circuits in the murine area postrema are not perfectly analogous to human neuroanatomy. The researchers were careful to frame their findings as a mechanistic blueprint rather than a confirmed account of how the drugs work in human patients, and the paper notes that human studies are needed to confirm whether the same variability in cAMP response is present in people.
There is also active disagreement among researchers about how much weight to assign neurological mechanisms versus peripheral ones. GLP-1 receptors are expressed not only in the brain but in the gut, the pancreas, the heart, and the kidneys. Some researchers argue that the gut’s contribution to the appetite-suppressing effect is at least as important as the brain’s, and that focusing on brain signaling provides an incomplete picture of why and when the drugs stop working.
The role of lean muscle mass loss adds another complication. Between 20% and 40% of the weight lost on semaglutide comes from lean mass rather than fat, according to published trial data. Muscle is metabolically active, and losing it reduces resting calorie expenditure. Some portion of the plateau effect almost certainly reflects the body reaching a new homeostatic equilibrium at a lower metabolic rate, not just attenuated neural signaling. Untangling these two contributions is not straightforward.
There is also unresolved debate about whether plateaus are primarily a neural phenomenon at all, or whether they reflect the body defending a biologically determined set point through multiple redundant mechanisms that no single drug intervention can fully override. Research into set-point biology has been running for decades without producing a clear therapeutic target, and some obesity researchers are sceptical that addressing any one piece of the neural signaling puzzle will produce meaningfully better outcomes at the population level.
What Comes Next
The most immediate clinical implication of the NIH findings is the potential for combination approaches. If PDE4 inhibition can extend and intensify the cAMP response to GLP-1 activation, roflumilast or next-generation PDE4 inhibitors might extend the duration of effective weight loss on semaglutide before a plateau is reached. Roflumilast is already an approved drug, used for severe chronic obstructive pulmonary disease, which means its safety profile is well characterised and clinical trials for obesity indications could move relatively quickly.
The broader significance is what the NIH team’s imaging approach represents for the field. The ability to watch individual neurons respond to a drug in real time, tracking intracellular signaling at the level of molecules rather than observing only behavioural outcomes, provides a new resolution of observation that prior obesity research lacked. The question of why GLP-1 weight loss drugs stop working for so many patients is not fully answered by this week’s paper. But the tools to answer it, in much finer detail than was previously possible, are now clearly in view.
For the roughly 10 million Americans currently taking GLP-1 treatments, and the many more who will start over the next several years as oral formulations reduce cost barriers and Medicare coverage expands access, that matters more than it might initially seem. A drug that works for 68 weeks is not the same as a drug that works indefinitely. Understanding, at the cellular level, why the difference exists is the first step toward closing it.
Sources: NIH Researchers Identify Avenue for Enhanced GLP-1-Induced Weight Loss, NIH | Scientists Discover Why Ozempic and Wegovy Weight Loss Eventually Plateaus, ScienceDaily | Semaglutide Effects on Energy Balance Are Mediated by Adcyap1+ Neurons in the Dorsal Vagal Complex, Cell Metabolism00256-6) | Patients Regain Weight Rapidly After Stopping Weight Loss Drugs, University of Cambridge | How Supply and Demand for Weight Loss Drugs Is Playing Out in 2026, J.P. Morgan | Obesity GLP-1 Market Size to Hit USD 66.57 Billion by 2035, Precedence Research
