Hunger, Leptin & Why Diets Fight Back

Millions of people believe that regaining weight proves they lacked willpower. It doesn't. It proves the system worked. Deep in your brain, in a knot of cells called the arcuate nucleus, two populations of neurons fight a permanent tug-of-war: AgRP/NPY neurons are the accelerator (hunger on), POMC/CART neurons are the brake (hunger off). Hormones from your fat, your stomach and your gut arrive continuously and lean on that balance — and the whole circuit is wired to defend your body fat far harder than it defends you from gaining it. Press Play, then walk through the scenarios and watch the balance beam tip.

Try this: switch to Obesity + leptin resistance and look at two readouts at once — leptin is enormous (the fat is shouting) while the brain's perceived energy state reads STARVING. Then hit 🍽 Eat a meal and watch the satiety wave arrive… and fade, while the starvation signal stays exactly where it was.

The Science in Plain Language

The control room: one accelerator, one brake, and a receptor worth knowing

At the base of your brain, in the hypothalamus, sits a small cluster called the arcuate nucleus. It contains two populations of neurons that want opposite things and that actively silence each other.

AgRP/NPY neurons are the accelerator. When they fire, you want food. They are not subtle: in mice, switching these neurons on artificially makes an already-full animal start eating within seconds, and switching them off stops a starving animal mid-meal. Hunger, in other words, is not a passive report of an empty stomach — it is an output that these cells generate.

POMC/CART neurons are the brake. POMC is a big precursor protein that gets cut up into smaller pieces; one of those pieces is α-MSH. α-MSH travels to a downstream neuron and binds a receptor called MC4R, and MC4R activation is, as close as biology gets, the "stop eating" switch.

Here is the elegant, slightly diabolical part: AgRP is an inverse agonist at MC4R. It doesn't just compete with α-MSH — it actively pushes the receptor below its resting activity. The accelerator and the brake are wired to the same pedal.

This is not a theoretical circuit. Mutations in MC4R are the commonest single-gene cause of severe obesity in humans — found in a few percent of people with severe, early-onset obesity. Those individuals are relentlessly hungry from childhood, and it is not a character flaw; it is a broken switch. And the story has a real payoff: setmelanotide, a drug that stimulates MC4R directly, is approved for people with certain rare genetic defects upstream of the receptor (POMC deficiency, leptin-receptor deficiency, Bardet–Biedl syndrome). When you fix the actual lesion, the hunger goes away.

Leptin is a starvation alarm, not a diet pill

This single sentence reframes everything on this page, so read it slowly: leptin's evolutionary job is to scream when your fat stores fall, not to stop you eating dessert.

Leptin is made by fat cells, and how much you make is roughly proportional to how much fat you have. It crosses into the brain and does two things in the arcuate: it inhibits AgRP and activates POMC. Both pushes point the same way — toward "we have reserves, stand down."

When leptin was discovered in 1994 in Jeffrey Friedman's laboratory, it looked like the answer to obesity. Mice with a broken leptin gene are enormous and eat constantly; give them leptin and they normalise. The obvious next step was to give leptin to people with obesity. It didn't work — and understanding why it didn't work is the most useful thing in this article (next section).

What leptin is genuinely superb at is the opposite direction. Drop your body fat, and leptin falls — and it falls further and faster than your fat loss alone would predict, because an active energy deficit suppresses leptin secretion on top of the effect of having less fat. To the arcuate, that reads as an emergency. AgRP fires. POMC goes quiet. MC4R shuts down. You get hungry, and you stay hungry.

The system is asymmetric. It defends hard against fat loss and only weakly against fat gain. That asymmetry made perfect sense for an animal that periodically had to survive a bad winter. It makes no sense at all in a world with a supermarket, and it is the single biggest reason that "just eat less" fails as a strategy.

Why the scale fights back — and why it isn't willpower

Run the two-year chart above. Here is what it is showing you, and every part of it is documented.

You lose 10% of your body weight. Fat cells shrink. Leptin falls disproportionately. Ghrelin, which normally settles after weight loss, instead rises and stays risen. AgRP neurons fire hard. Hunger climbs — and, crucially, it does not habituate. It does not fade after three weeks while your new habits "lock in." It is still elevated a year later.

At the same time, and independently, your resting energy expenditure drops below what your new, lighter body mass predicts. Some of the fall is simply because a smaller body costs less to run — that part is expected and unremarkable. But there is an extra, additional deficit on top of it. This is adaptive thermogenesis: the body running the same machinery more cheaply. Thyroid hormone conversion shifts, sympathetic nervous system output falls, muscles become more efficient per unit of work.

So you are simultaneously hungrier than you were and burning less than your body should. Both effects persist for years, not weeks. Sumithran and colleagues followed dieters for twelve months after a very-low-energy diet and found leptin still below baseline, ghrelin still above it, and subjective hunger still elevated. Fothergill and colleagues went back to competitors from a televised weight-loss competition six years later and found resting metabolic rate still suppressed below predicted — even in those who had regained most of the weight.

Read that last clause again. The metabolic defence was still running after the weight came back. That is a system defending a target, not a person failing a test.

If you have regained weight, this is what happened to you. You were not weak. You were outnumbered by a circuit that has had several hundred million years to get good at its job.

Leptin resistance: the signal is loud and the receiver is deaf

Now the second reveal, and the one that surprises almost everyone.

In obesity, leptin is not low. Leptin is high — often five to ten times a lean person's level. There is more fat, so there is more leptin. And yet the arcuate nucleus behaves exactly as if it were starving: AgRP fires, POMC is quiet, hunger persists. Switch the visualisation to Obesity + leptin resistance and look at the contradiction directly: the leptin meter is nearly full, the fat-mass readout says 45 kg, and the brain's perceived energy state reads STARVING.

This is leptin resistance, and it has at least three overlapping mechanisms:

1. Transport failure. Leptin cannot simply diffuse into the brain. It has to be carried across the blood–brain barrier by a saturable transport system. Saturable means it maxes out. Past a certain blood level, doubling leptin in the blood barely changes leptin in the brain. In the animation, watch the particles pile up against the barrier and bounce.
2. Negative feedback inside the neuron. Leptin binds its receptor (LepRb) and signals through JAK2 and STAT3. That same signalling switches on SOCS3, and chronic exposure raises PTP1B — two proteins whose job is to shut the pathway off. Shout at the receptor for long enough and it turns down its own volume.
3. Hypothalamic inflammation. Chronic overnutrition produces inflammation and gliosis in exactly this region of the hypothalamus, which further blunts leptin signalling.

And this is why the leptin trials failed. Recombinant leptin given to children with genuine, congenital leptin deficiency — an extraordinarily rare condition — is close to miraculous. Those children are ravenously, dangerously hungry, and leptin normalises them. But when recombinant leptin was given to adults with ordinary obesity, average weight loss was small, highly variable, and required uncomfortable doses. It never became a treatment for common obesity, and today leptin (as metreleptin) is approved only for the leptin-deficient: generalized lipodystrophy and congenital leptin deficiency.

It is one of the great disappointments in metabolic medicine, and it deserves to be told honestly rather than quietly dropped. The lesson is precise: adding more of a signal is useless when the problem is that the signal isn't being heard. That is not a failure of the leptin hypothesis — it is a confirmation of it.

Ghrelin: why hunger is partly scheduled, not purely metabolic

Ghrelin is the only well-established hormone in your bloodstream that increases appetite. It comes from X/A-like cells in the lining of the stomach, it rises when the stomach is empty, it activates AgRP neurons, and it falls after you eat.

But here is the detail that changes how you should think about your own hunger: ghrelin rises before your habitual meal times. Not after your stomach empties — before the meal you normally eat. It is entrained to your schedule. If you always eat at 12:30, your ghrelin starts climbing at around noon, whether or not your body needs food at noon.

That means a real part of what you experience as "I am starving" at 11:45 is a conditioned, anticipatory signal — a habit written in hormones. It is genuinely felt. It is not imaginary. But it is not a fuel gauge either.

Two practical consequences. First, a hunger pang at your usual mealtime carries much less information about your actual energy status than it feels like it does — and it will pass. Second, if you change your eating schedule, the ghrelin rhythm will eventually follow, which is a large part of why intermittent fasting feels brutal for about two weeks and then, for many people, stops feeling like anything at all. You have not "learned discipline." You have re-trained a hormone.

One more counterintuitive fact, because it matters: ghrelin is generally lower in obesity, not higher. People with obesity are not being driven by excess ghrelin. Their problem is on the leptin side of the ledger. But after weight loss, ghrelin goes up and stays up — which is exactly the wrong direction, and exactly what the two-year chart shows.

Why protein, fibre and volume genuinely blunt hunger — and liquid calories barely register

Not all satiety advice is folklore. Some of it maps directly onto the circuit above.

When food arrives, your gut sends up a wave of signals, each with a real job:

• Stomach stretch — mechanoreceptors in the stomach wall fire along the vagus nerve to the brainstem (the NTS) and from there into the hypothalamus. This is a pure volume signal. It does not care about calories.
• CCK — released by I-cells in the duodenum in response to fat and protein. It acts on vagal afferents and is a fast, short-acting meal-terminator.
• PYY — released by L-cells further down the intestine, in proportion to the calories that arrive. Longer-acting than CCK.
• GLP-1 — also from L-cells. Slows gastric emptying, reduces intake, and (remember this for the next section) acts directly on the arcuate circuit.
• Insulin — besides its blood-sugar job, insulin is itself an adiposity signal in the hypothalamus, where it also inhibits AgRP and stimulates POMC.

Now the practical translation:

Protein is the most satiating macronutrient, and the effect is large and reproducible. In controlled feeding studies, raising protein from roughly 15% to 30% of calories — with calories initially held constant — reduced appetite so much that people spontaneously ate several hundred fewer calories a day when allowed to eat freely, and lost weight without being asked to restrict anything. Protein triggers CCK and PYY strongly and blunts ghrelin.

Fibre and volume work through a different door: they slow gastric emptying (keeping the stretch signal alive longer) and increase PYY. A large salad and a small brownie can carry the same calories and produce completely different signals to the arcuate nucleus. This is not moral. It is mechanical.

Liquid calories barely register. A soda empties from the stomach fast, so the stretch signal is brief; there is no protein to speak of, so the CCK response is small. Studies comparing energy given as a drink versus the same energy as solid food consistently find that people fail to compensate for liquid calories at later meals — the calories go in, and appetite is not adjusted downward. If you drink 300 calories, you will very likely still eat your normal meal.

Sleep loss turns both dials the wrong way

This is one of the most robust and most ignored findings in the whole field.

In a controlled laboratory study, healthy young men were restricted to about four hours in bed for two nights and compared with the same men after ten hours in bed. Short sleep lowered leptin by roughly 18% and raised ghrelin by roughly 28%, and their hunger and appetite ratings rose — most sharply for calorie-dense, high-carbohydrate foods. A large population study (the Wisconsin Sleep Cohort) found the same signature in the real world: habitual short sleepers had lower leptin, higher ghrelin, and higher BMI.

Look at what that does to the circuit on this page. Leptin down means less inhibition of AgRP. Ghrelin up means more activation of AgRP. Both dials, wrong way, at the same time. You wake up short of sleep and your arcuate nucleus has been told, quite falsely, that you are running out of fuel.

Which means: if you are trying to lose weight and you are sleeping five hours a night, you are dieting against a headwind that you could simply switch off. Sleep is not a soft, nice-to-have adjunct to weight management. It is an upstream input to the hunger circuit. Fix it first — it is the cheapest intervention on this entire page.

How GLP-1 drugs actually work — and what they honestly cost you

Semaglutide, liraglutide and tirzepatide are not appetite suppressants in the old, stimulant sense, and they do not "give you willpower." They work on exactly the circuit drawn above.

Native GLP-1 is a gut hormone with a half-life of a couple of minutes — it is destroyed almost immediately by the enzyme DPP-4. The drugs are engineered versions that survive for days. They act on GLP-1 receptors in the brainstem and the hypothalamus, and the net effect in the arcuate is to turn down AgRP drive and support POMC. They also slow gastric emptying, which keeps the stretch signal alive for longer.

That is why the thing patients describe is not "I forced myself to stop." It is "the food noise went quiet." The constant, intrusive, background negotiation about food — which is precisely what a hyperactive AgRP population feels like from the inside — simply turns down. Switch the visualisation to On a GLP-1 agonist and watch the AgRP meter collapse while the balance beam tips toward FULL.

The results are genuinely large. In the STEP 1 trial, semaglutide 2.4 mg produced a mean weight loss of about 15% over 68 weeks, against about 2.4% on placebo. In SURMOUNT-1, tirzepatide (which hits both GIP and GLP-1 receptors) reached roughly 21% at its highest dose over 72 weeks. Nothing in the history of obesity pharmacotherapy short of surgery has come close.

Now the honest half, because you deserve it:

• The weight comes back if you stop. In the STEP 1 extension, participants who came off semaglutide regained roughly two-thirds of the weight they had lost within a year. This is not a scandal — it is the whole thesis of this page. The drug suppresses the defence; it does not delete it. Stop the drug and the defence is still there, waiting, exactly as it was.
• Muscle loss is real. As with any large energy deficit, a substantial share of the weight lost is lean tissue, not just fat. This is not unique to these drugs, but the losses are big and fast, which magnifies it. Resistance training and adequate protein are not optional if you are on one of these — they are the difference between losing fat and losing your body.
• Gastrointestinal side effects are the norm, not the exception. Nausea, vomiting, diarrhoea and constipation are the commonest adverse events, are dose-related, and are the main reason people stop.
• Cost. At US list price, without insurance, these run on the order of a thousand dollars a month. For a chronic medication that must be continued indefinitely, that is a serious and unequal barrier — and it is a legitimate part of the clinical picture, not a footnote.

None of this makes them bad drugs. It makes them drugs — with a mechanism, a magnitude, and a price. Read the fuller picture on our GLP-1 receptor agonists page.

Set point or settling point? — and what actually helps

Scientists genuinely disagree about how the body picks the weight it defends, and it would be dishonest to pretend otherwise.

The set-point model says the brain has a target body-fat level and actively steers back to it, like a thermostat. Its strongest evidence is everything on this page: the hunger surge, the metabolic suppression, the persistence for years.

The settling-point model says there is no target at all — weight simply settles where the forces balance (food availability, palatability, activity, sleep, stress, drugs), and changing those forces moves the settling point without any thermostat being involved. Its strongest evidence is that populations have got heavier over decades far faster than genes could change, and that moving people into a different food environment moves their weight.

The honest synthesis, and the one most researchers now hold, is somewhere in between — often called a dual-intervention-point model: the defence against loss is fierce and clearly biological, the defence against gain is weak and easily overwhelmed, and in the wide gap between the two, environment does most of the deciding. That single asymmetry explains both why weight creeps up easily and why it is so brutally hard to get back down.

So what actually helps, given a defended system?

• Protein at every meal. The most reliable non-drug lever on the satiety side of the circuit — more CCK, more PYY, less ghrelin.
• Resistance training. It does not burn many calories. That is not why it is here. It protects lean mass, which protects resting energy expenditure — the exact quantity the defence is trying to lower.
• Sleep. Free, upstream, and it moves both leptin and ghrelin at once.
• Slower losses. The size of the counterattack scales with the size and speed of the deficit. Aggressive crash dieting recruits a bigger defence, and the defence outlasts the diet.
• Volume and fibre. Cheap mechanical satiety — stretch receptors and PYY, working exactly as designed.
• Medication, when it is warranted. If the biology is defending a weight that is harming you, treating the biology is a reasonable, evidence-based choice — not a shortcut, and not cheating.
• And never, ever framing regain as a moral failure. This is not sentiment; it is accuracy. The regain is the predicted output of a working system. A person who regains weight has learned nothing about their character. They have simply met their arcuate nucleus.

You cannot out-argue a circuit that has been optimised since before there were primates. You can, however, understand it, work with the levers it actually responds to, and stop blaming yourself for losing a fight that was never fair.

↑ Back to the animation

Connections

• All Interactive Visualizations
• Blood Sugar & Insulin — the other half of the fuel system
• Digestion — where CCK, PYY and GLP-1 are released
• The Gut Barrier & Microbiome
• Sleep & Circadian Rhythm — why short sleep raises ghrelin
• Stress & the Cortisol Axis
• GLP-1 Receptor Agonists — how they work, and their limits
• Obesity — a defended system, not a character flaw
• Insulin Resistance
• Metabolic Syndrome
• Fasting — and what ghrelin does during one
• Caloric Restriction
• Exercise — why resistance training protects your metabolic rate
• Sleep Hygiene — the cheapest lever on the hunger circuit
• Fasting Insulin
• HOMA-IR