Metabolism · Clinical Literacy

    Hungry Brain, Hungry Gut: The Biology of When You Start and Stop Eating

    Homeostatic eating explained — hunger, satiation and satiety — and what the Mayo Clinic obesity phenotypes do and don't show

    Reviewed by Dr. Sanjeev GoelNovember 21, 202611 min read
    An empty ceramic bowl on a dark surface, rim-lit in deep harbour blue

    Two people with the same BMI sit down to the same buffet. One stops at about 700 calories. The other is still not full at 1,200. A few hours later, a different pair: both ate the same 320-calorie breakfast, and one of them is hungry again before the other has finished digesting it.

    Those are not differences in willpower. In a Mayo Clinic cohort of 450 adults with obesity, they were measurable, reproducible differences in physiology — and they were largely different people. The group that ate 62% more before feeling full was not, for the most part, the same group whose stomach emptied 31% faster.[1]

    That is the idea behind a handout now circulating at obesity meetings: obesity is not one disease with one mechanism. This post is about the half of appetite your body runs automatically — homeostatic eating — and what the evidence for "eating phenotypes" actually supports. The other half, eating driven by emotion and reward, gets its own post.

    The 60-second answer

    Homeostatic eating is the body's energy-balance system: the brain–gut signalling that tells you to start eating, when you've had enough, and how long that "enough" lasts. It has three separate parts — hunger (the drive to start), satiation (the signal that ends a meal) and satiety (how long fullness lasts before hunger returns). They can fail independently. In the Mayo Clinic phenotyping study, about a third of people with obesity had abnormal satiation (a "hungry brain") and about a third had abnormal satiety, measured as fast gastric emptying (a "hungry gut").[1] Choosing anti-obesity medication to match those phenotypes was associated with nearly double the weight loss in a real-world clinic study — but that study was not a blinded randomised trial, and the result has still not been confirmed in one.[1] The practical message is solid even where the trial is not: if you know which signal fails you, you can choose tools that target it.

    Three signals, not one

    Most people use "hunger" for everything. Physiologists split it three ways, and the split matters because each part has different machinery.[1]

    Hunger — the start signal. The desire to eat. Its best-known messenger is ghrelin, a hormone made mostly in the stomach that rises before meals and falls after them. Infusing ghrelin into healthy volunteers makes them eat more at a buffet. We covered the hormone in detail in our ghrelin article; the short version is that it is a genuine biological drive, and that dieting turns it up.

    Satiation — the stop signal. How many calories it takes to end a meal. As food arrives, the gut stretches and releases a cascade of signals — cholecystokinin (CCK), glucagon-like peptide-1 (GLP-1), peptide YY (PYY) — that the brainstem and hypothalamus integrate into "that's enough." Satiation is measured in the lab exactly the way it sounds: put someone in front of a buffet and record how much they take before they say they are full.

    Satiety — the duration signal. How long fullness lasts after a meal ends, before hunger returns. One of its main determinants is mechanical: how quickly the stomach empties into the small intestine. A stomach that empties fast stops sending "full" sooner.

    The distinction between satiation and satiety is the useful one. Someone with poor satiation eats large meals and then may go hours without thinking about food. Someone with poor satiety eats normal-sized meals and is hungry again in an hour or two. Same scale reading; opposite days.

    What the Mayo Clinic phenotype study measured

    The handout is adapted from Acosta and colleagues, published in Obesity in 2021.[1] It is worth knowing exactly what was measured, because the labels are catchier than the tests.

    The team studied 450 adults with obesity (BMI over 30; mean age 39, 72% women, 93% White) recruited from the community, and put each through a battery of tests after an eight-hour fast:

    • Satiation by ad libitum buffet: calories eaten to reach maximal fullness.
    • Satiety by gastric-emptying scintigraphy: a radiolabelled 320-calorie, 30%-fat meal, with the time taken for half of it to leave the stomach (the T½).
    • Hedonic eating by questionnaire: principally the anxiety subscale of the Hospital Anxiety and Depression Scale.
    • Energy expenditure by indirect calorimetry: measured resting energy expenditure compared against the Harris–Benedict prediction.

    Each phenotype was then defined by a cut-off taken from the first 100 participants — the top or bottom quarter, separately for women and men:

    How each of the four Mayo Clinic obesity phenotypes was defined
    PhenotypeDefined asCut-off
    Hungry brain (abnormal satiation)Eating a lot before feeling fullBuffet intake > 894 kcal (women), > 1,376 kcal (men)
    Hungry gut (abnormal satiety)Stomach emptying fastGastric T½ < 101 min (women), < 86 min (men)
    Emotional hunger (hedonic eating)High anxiety scoreHADS-Anxiety ≥ 7 (both sexes)
    Slow burn (low metabolic rate)Burning less at rest than predictedResting energy expenditure < 96% (women), < 94% (men) of predicted

    How different were the groups? Compared with participants who did not have each trait, the hungry-brain group ate 62% more calories before reaching fullness; the hungry-gut group emptied the stomach 31% faster; and the slow-burn group burned about 12% less at rest.[1] In absolute terms, the hungry-brain group's mean buffet intake was 1,224 kcal against a whole-cohort mean of 929; the hungry-gut group's mean gastric T½ was 83 minutes against 110; and the slow-burn group's measured resting energy expenditure averaged 89% of what the Harris–Benedict equation predicted, against 100% for the cohort as a whole.

    Reading the handout's Venn diagram correctly

    The handout shows 18% hungry gut, 16% hungry brain, 12% emotional hunger and 12% slow burn. Those are the people with only that phenotype. Counting everyone who had it, alone or in combination, the numbers roughly double: hungry brain 32%, hungry gut 32%, emotional hunger 21% and slow burn 21%.[1] About 27% of people had two or more phenotypes, and 15% had none of the four.

    So the two homeostatic phenotypes are the most common, by a clear margin — each one touches about a third of the cohort on its own. That is the finding I would emphasise to patients. It is also why "just eat less" lands so differently on different people: for many, the problem is precisely that the signal which is supposed to make eating less feel natural is the one that is miscalibrated.

    Figure 1

    How common each phenotype was

    This phenotype onlyAny, including overlaps

    % of cohort

    Percentage of 450 adults with obesity meeting each phenotype definition. The two series sum to more than 100% because 27% of participants met two or more phenotype definitions; the grey series counts only people with that phenotype and no other. For 'No phenotype' the two values are identical by definition. The slow-burn percentages are out of 400 participants rather than 450, because resting energy expenditure was not available for everyone.

    Source: Acosta et al., Obesity 2021 (ref. 1).

    The trial: what it found, and why I'd hold it loosely

    The second half of the paper is the part that gets quoted. Patients at Mayo's weight-management clinic received one of the anti-obesity medications available at the time, either chosen to match their phenotype or chosen the usual way — by side-effect profile, comorbidities, patient preference and insurance.[1] The matching rules were set in advance:

    • Hungry brain → phentermine–topiramate (or lorcaserin, since withdrawn from the market)
    • Hungry gut → liraglutide, a GLP-1 receptor agonist
    • Emotional hunger → naltrexone–bupropion
    • Slow burn → low-dose phentermine plus resistance training

    At 12 months, the phenotype-guided group had lost 15.9% of body weight against 9.0% — a difference of 6.9 percentage points (95% CI 4.5 to 9.4). Seventy-nine percent lost more than 10% of their weight, against roughly a third of the comparison group.[1]

    Figure 2

    What the pragmatic trial reported

    Phenotype-guidedUsual care

    Mean weight lost over time

    Share reaching each weight-loss threshold at 1 year

    Read down each panel, not across. 84 patients received phenotype-guided medication and 228 received usual care. This was a pragmatic, real-world comparison, not a randomised or blinded trial — patients were assigned to physician groups by appointment, and the authors list 'lack of blinded randomization' among their limitations. The groups differed at baseline in age, systolic blood pressure and fasting glucose. Error bars in panel A are standard errors of the mean, not confidence intervals. One note on panel B: the paper's results section reports 35% of usual-care patients losing more than 10%, while its abstract prints 34%; the figure uses the results-section value.

    Source: Acosta et al., Obesity 2021 (ref. 1).

    That is a striking result, and I want to be straight about why it is not yet a settled one:

    1. It was not a randomised, blinded trial. The abstract says patients were "randomly assigned", but the methods describe patients being given appointments with physician groups that either did or did not use phenotyping, and the authors themselves list "lack of blinded randomization" as a limitation.[1]
    2. The groups were unequal and different. 84 phenotype-guided patients against 228 controls. The phenotype-guided patients were on average 7.2 years younger, with systolic blood pressure 6.2 mm Hg lower and fasting glucose 21 mg/dL lower. The authors adjusted for these and the gap held — 13.6% against 6.6% — but adjustment cannot fix what was not measured.
    3. Being tested may itself change behaviour. Patients who spent a day being phenotyped and then had their results explained to them may simply have been more engaged. The authors name this "testing bias" as a limitation.
    4. "12 months" is thinner than it sounds. Mean follow-up was 8.2 months in the phenotype-guided group and 8.1 in the comparison group, and only 59.4% and 54% of patients respectively had a physician visit between months 6 and 12. The 12-month figures come from those who stayed.
    5. Interests. Phenomix Sciences holds an exclusive licence to the senior authors' biomarker technology, and the lead author is a stockholder in that company. That does not make the data wrong; it is something a reader should know.

    The same group has since tested phenotype-tailored lifestyle programmes without drugs. In a 12-week study of 165 adults, people given a programme matched to their phenotype lost 7.4 kg against 4.3 kg on a standard programme — a difference of 3.1 kg (95% CI 1.1 to 5.1).[2] Again, though, participants were assigned by when they enrolled, not randomised, and the authors say plainly that a randomised trial is needed to confirm causality.

    That randomised trial now exists, and it is still running. Mayo Clinic is enrolling 228 adults in a properly randomised comparison of a phenotype-tailored lifestyle programme against a standard one; it opened in January 2025 and is not scheduled to report its primary outcome until December 2028.[9] Until it does, the honest summary is the one a 2026 review of precision obesity pharmacotherapy reaches independently: phenotype-guided treatment is the most practical precision approach available today, and the best-validated predictor of long-term response remains the simplest one — how much weight someone has lost in the first three or four months.[8]

    My reading: the classification is useful and biologically sensible. The claim that matching treatment to it nearly doubles weight loss is plausible and promising — and unproven.

    Hungry brain: when the stop signal is late

    If satiation is the problem, the meal is the battleground. The useful strategies are the ones that make the stop signal arrive earlier, or that cap the meal before it has to.

    Food structure and volume. The most instructive experiment here isn't about phenotypes at all. At the NIH, Kevin Hall's group housed 20 adults on a metabolic ward and fed them ultra-processed or unprocessed diets for two weeks each, matched for presented calories, energy density, macronutrients, sugar, sodium and fibre, and told them to eat as much or as little as they liked. On the ultra-processed diet they ate 508 kcal a day more, gained 0.9 kg, and lost 0.9 kg on the unprocessed diet.[3] With the obvious nutrition labels held equal, the form of the food still moved where people stopped. For a hungry brain, whole and minimally processed foods are not a moral choice; they are a satiation tool.

    What Mayo actually prescribed. In the lifestyle study, the satiation group received a time-restricted, volume-based low-calorie plan — more food by weight and volume for the calories, eaten within a fixed window.[2]

    Medication. In the original trial this phenotype was matched to phentermine–topiramate. Newer GLP-1-based drugs act heavily here too: in a 20-week trial of semaglutide 2.4 mg in 72 adults with obesity, energy intake at an all-you-can-eat lunch was 35% lower than on placebo.[4] That trial was run and authored largely by Novo Nordisk, the drug's manufacturer.

    Hungry gut: when fullness doesn't last

    If satiety is the problem, the gap between meals is the battleground. The aim is to slow emptying and lengthen the "full" signal.

    Protein. Higher-protein preloads raise fullness more than lower-protein ones; a meta-analysis found the effect consistent, while warning in its own title and conclusions that it was measured under tightly defined conditions and should not be over-extrapolated to body weight.[5] Mayo's lifestyle programme gave this phenotype protein before meals.[2] Our protein guide covers how much people actually need.

    Medication. This is where GLP-1 drugs earned their place in the phenotype model. In a Mayo randomised, placebo-controlled pilot of 40 adults with obesity, liraglutide delayed gastric emptying by a median of 70 minutes at five weeks, against 4 minutes on placebo — and in a post-hoc analysis, the size of that delay at week 5 correlated with weight lost at week 16 (Rs 0.567, p = 0.018).[6] The authors suggested gastric emptying might help predict who keeps responding.

    One honest complication: in the semaglutide trial above, there was no evidence of slowed gastric emptying by week 20, measured indirectly, even as intake fell by a third.[4] Gastric slowing seems most pronounced early in treatment and to fade; the drugs keep working through other routes, including the brain. "GLP-1 drugs fix a fast stomach" is too simple.

    Slow burn, briefly

    The fourth phenotype isn't about eating at all — it's the expenditure side of the ledger. People in the slow-burn group had less muscle (46.7% versus 50.7% lean mass), reported being active less often and did less structured exercise.[1] Mayo paired this phenotype with resistance training and, in the lifestyle study, post-workout protein and interval training.[2] It deserves its own article; for now, our training piece is the place to start.

    Why dieting makes all of this harder

    The homeostatic system defends weight. After a 10-week very-low-energy diet, 50 adults who had lost an average of 13.5 kg still had elevated ghrelin and still-altered levels of the satiety hormones PYY and CCK 62 weeks after starting — more than a year later — and still reported more hunger than at baseline.[7] Weight loss turns hunger up and satiation down. That is not a character flaw; it is the system doing exactly what it evolved to do, and it is the strongest argument for choosing tools that work with your particular weak signal rather than against all of them at once.

    Can you find out your phenotype?

    Not precisely, outside a research setting. The Mayo tests — a standardised buffet, gastric-emptying scintigraphy, indirect calorimetry — are specialised clinical procedures, performed over most of a day, and they are not cheap. Commercial tests derived from this work exist; I'd treat them as early.

    But you can notice the pattern, and that is clinically useful:

    • Do you struggle to stop once you start, but are fine between meals? That leans hungry brain.
    • Are your portions normal, but you're hungry again within an hour or two? That leans hungry gut.
    • Is your eating driven more by stress, mood or boredom than by your stomach? That's the hedonic side, and it's the subject of the companion post to this one — see our blog index for it.

    Bring that observation to your doctor. It is a far better starting point for choosing a treatment than "I need to lose weight."

    The practical bottom line

    • Appetite has at least three separately breakable parts: the start signal, the stop signal and the stay-full signal. Knowing which one fails you is more useful than any general rule.
    • For a late stop signal: whole foods, high-volume meals, defined meal windows.
    • For short-lived fullness: protein at meals, especially early in the meal.
    • Phenotype-matched medication is a promising idea resting on one non-randomised clinic study. Ask about it; don't treat it as proven. The randomised trial reports in 2028.
    • If you track food, track when you get hungry as well as what you eat — the pattern is the diagnosis. Peak Nutrition can help with the logging, and if weight is the goal our weight-loss guide is the wider picture.

    Frequently asked questions

    What is homeostatic eating?

    Eating driven by the body's energy needs, regulated by brain–gut signalling. It has three components: hunger (desire to eat), satiation (calories needed to feel full and end a meal) and satiety (how long fullness lasts).

    What's the difference between satiation and satiety?

    Satiation ends a meal; satiety is how long you stay full afterwards. Poor satiation means large meals; poor satiety means hunger returning soon after normal ones.

    What is the "hungry brain" phenotype?

    A Mayo Clinic label for abnormal satiation — needing to eat substantially more before feeling full. In their study it was defined as eating more than 894 kcal (women) or 1,376 kcal (men) at a buffet, and about a third of adults with obesity met it.

    What is the "hungry gut" phenotype?

    A label for abnormal satiety, measured as unusually fast gastric emptying, so fullness fades quickly. About a third of the study cohort had it.

    Does matching weight-loss drugs to phenotype work?

    One real-world study found phenotype-guided prescribing was associated with 15.9% versus 9.0% weight loss at a year, but it was not a blinded randomised trial and the groups differed at baseline. A randomised trial of the lifestyle version is enrolling now and reports in 2028. It's promising, not proven.

    Can I get tested for my obesity phenotype?

    The research tests are specialised and costly. Paying attention to your own pattern — trouble stopping versus trouble staying full versus eating for mood — and discussing it with your physician is a sensible, free first step.

    References

    1. [1]Acosta A, Camilleri M, Abu Dayyeh B, et al. Selection of antiobesity medications based on phenotypes enhances weight loss: a pragmatic trial in an obesity clinic. Obesity (Silver Spring). 2021;29(4):662–671. doi:10.1002/oby.23120. PMID 33759389. (ClinicalTrials.gov NCT03374956. Correction: Obesity. 2021;29(9):1565–1566. doi:10.1002/oby.23236.)
    2. [2]Cifuentes L, Ghusn W, Feris F, et al. Phenotype tailored lifestyle intervention on weight loss and cardiometabolic risk factors in adults with obesity: a single-centre, non-randomised, proof-of-concept study. EClinicalMedicine. 2023;58:101923. doi:10.1016/j.eclinm.2023.101923. PMID 37007741.
    3. [3]Hall KD, Ayuketah A, Brychta R, et al. Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake. Cell Metab. 2019;30(1):67–77.e3. doi:10.1016/j.cmet.2019.05.008. PMID 31105044.
    4. [4]Friedrichsen M, Breitschaft A, Tadayon S, et al. The effect of semaglutide 2.4 mg once weekly on energy intake, appetite, control of eating, and gastric emptying in adults with obesity. Diabetes Obes Metab. 2021;23(3):754–762. doi:10.1111/dom.14280. PMID 33269530.
    5. [5]Dhillon J, Craig BA, Leidy HJ, et al. The effects of increased protein intake on fullness: a meta-analysis and its limitations. J Acad Nutr Diet. 2016;116(6):968–983. doi:10.1016/j.jand.2016.01.003. PMID 26947338.
    6. [6]Halawi H, Khemani D, Eckert D, et al. Effects of liraglutide on weight, satiation, and gastric functions in obesity: a randomised, placebo-controlled pilot trial. Lancet Gastroenterol Hepatol. 2017;2(12):890–899. doi:10.1016/S2468-1253(17)30285-6. PMID 28958851.
    7. [7]Sumithran P, Prendergast LA, Delbridge E, et al. Long-term persistence of hormonal adaptations to weight loss. N Engl J Med. 2011;365(17):1597–1604. doi:10.1056/NEJMoa1105816. PMID 22029981.
    8. [8]Lopez Delgado DS, Reyes-Zermeño MG, Sanjuanelo Lemus ED, et al. Phenotypic and genomic frameworks for precision pharmacotherapy in obesity: a narrative review. Front Med (Lausanne). 2026;13:1803474. doi:10.3389/fmed.2026.1803474. PMID 42519802.
    9. [9]ClinicalTrials.gov. Phenotype-Tailored Lifestyle Intervention for Obesity: A Randomized Trial. NCT06770049. Mayo Clinic; enrolment 228; primary completion December 2028.
    Disclaimer: This article is for education and is not medical advice. Anti-obesity medications have contraindications and side effects; decisions about them belong with a physician who knows your history.