The decision question. When someone with obesity loses weight and regains it, what does the evidence say is happening — a lapse of discipline that more willpower would fix, or a regulated biological system defending a higher body weight against the loss? The answer changes what an advisor recommends and what the person expects: whether the target is “try harder” or “counter a physiological defense that does not relent on its own.” The held evidence points to the second, with a symmetric guard against the opposite over-swing (biology-as-destiny).
Body weight is centrally regulated, not a running calorie tally [EXTRACTED]
Body weight is a defended variable, not a passive ledger of intake minus expenditure. Sumithran describes the architecture: «Body weight is centrally regulated, with peripheral hormonal signals released from the gastrointestinal tract, pancreas, and adipose tissue integrated, primarily in the hypothalamus, to regulate food intake and energy expenditure…» (Sumithran et al., 2011). Hall states the same in the framing of an energy-balance model: «…the brain is the primary organ responsible for body weight regulation operating mainly below our conscious awareness via complex endocrine, metabolic, and nervous system signals to control food intake…» (Hall et al., 2022). The control loop sits below conscious awareness — which is precisely why a model that locates the lever at conscious choice mislocates it.
The setpoint is defended after weight loss — the hormonal signals do not reset [EXTRACTED]
The central claim of the biological model, and the one with the sharpest evidence, is that diet-induced weight loss provokes a persistent hormonal defense of the higher weight. Sumithran measured the appetite-regulating hormones one year after a supervised reduction: «One year after initial weight reduction, levels of the circulating mediators of appetite that encourage weight regain after diet-induced weight loss do not revert to the levels recorded before weight loss…» (Sumithran et al., 2011). The signals pushing toward regain are still elevated a year out — not a transient adjustment but a sustained counter-force. This is the mechanism catalogued in detail on Weight-Loss Maintenance and Metabolic Adaptation; the point here is what it means for the model: the difficulty of maintenance is a property of the regulated system, not a measure of the person’s resolve.
Pharmacology is consistent with the mechanism — and with a defense still there underneath [EXTRACTED]
A drug acting on appetite biology produces effect sizes that structured lifestyle programmes do not. In SURMOUNT-1, tirzepatide (a GIP/GLP-1 agonist) drove: «The mean percentage change in weight at week 72 was -15.0%… -19.5%… and -20.9%… with 15-mg doses and -3.1%… with placebo (P<0.001…)» (Jastreboff et al., 2022). Both arms received lifestyle counselling with a 500 kcal/day deficit and >=150 min/week of activity, so the comparator was a structured lifestyle programme, not exhortation. (Jastreboff et al., 2022) A pharmacological handle on the endocrine control loop moves weight roughly 5-7x as much as the placebo-plus-structured-lifestyle arm (15.0-20.9% vs 3.1%) — consistent with the loop being the governing variable, though an appetite-lowering drug would also help under a restraint-limited reading (see Evidence status). (corrected 2026-10-06: proves / order of magnitude more than exhortation / direct evidence -> consistent with, ~5-7x vs structured lifestyle; self-critique)
The withdrawal result is the strongest supporting signal, because it is what a defended setpoint that was suppressed rather than abolished would predict. In the STEP-1 off-treatment extension — a non-randomized follow-up of a subset of the randomized trial, in which drug and lifestyle support were stopped together in both arms — «Following treatment withdrawal, semaglutide and placebo participants regained 11.6 (SD: 7.7) and 1.9 (SD: 4.8) percentage points of lost weight, respectively, by week 120…» (Wilding et al., 2022). About two-thirds of the drug-induced loss returned within a year, leaving a net loss of 5.6% at week 120 (Wilding et al., 2022) — the same direction as the signature Sumithran measured hormonally, read out here as regain. This is the regulated-defense reading of the GLP-1 evidence catalogued on GLP-1 Drugs.
The symmetric guard — regulated does NOT mean fated [EXTRACTED]
The disease model corrects a willpower over-attribution; it must not be inverted into a biology-is-destiny fatalism, which is the same error with the sign flipped (the symmetric-standards trap — replacing an overclaim with its mirror image, see Model Error vs Sampling Error). Two held facts hold the line:
- The population trend is environmental, not genetic. Hall: «…changes in the food environment are likely the primary driver of the increased obesity prevalence in recent decades (4).» (Hall et al., 2022). Genes did not change over the decades the prevalence rose; the environment did. So the food environment and the behaviors it shapes remain real, modifiable levers — the model relocates difficulty, it does not remove agency.
- Heritability explains individual variation, not individual fate. Hall: «…BMI is highly heritable and genetic differences explain ~75% of BMI variability among individuals (1, 2).» (Hall et al., 2022). Heritability partitions variance across people in a given environment — it says why two people in the same food environment differ, not that any one person’s weight is unalterable. Biology sets the difficulty and the individual starting point; it does not close the door.
The correct synthesis is both: a strong biological regulation of body weight, defended against loss, operating within — and provoked by — an obesogenic environment. Neither pole alone is the evidence.
Evidence status — what is settled, what is emerging
The page must not launder an ascendant framing into a settled one — the disease model is currently in fashion, and being in fashion is an anti-signal, not a warrant (the same bar the willpower model got, applied symmetrically). The pieces do not all carry the same grade, so separate them:
- Well-evidenced (the load-bearing core). That body weight is centrally/hormonally regulated is textbook physiology. That the appetite-hormone shift persists about a year after diet-induced loss rests on a single, small cohort (Sumithran, n~50) — corroborated by the broader metabolic-adaptation literature but not itself high-GRADE (Sumithran enrolled 50, analysed 34 completers). That the GLP-1/GIP drugs produce large losses is a high-GRADE RCT finding (SURMOUNT-1, STEP-1); regain after stopping comes from a non-randomized off-treatment extension of a randomized STEP-1 subset, with lifestyle support stopped in both arms — lower-grade than the efficacy result. (corrected 2026-10-06: regain was filed as high-GRADE RCT; self-critique)
- An interpretation, not a direct test. The step from “drug works, regain on withdrawal” to “therefore a defended setpoint” is an inference the evidence is consistent with — not a claim any single trial was designed to identify. Regain on withdrawal is what a defended-setpoint model predicts, but it does not uniquely rule out weaker readings (a settling-point that tracks the food environment, behavioral relapse once appetite suppression lifts).
- Emerging / contested / framing-grade. The unifying paradigm — “obesity is a biologically defended disease” — and the disease label itself are not GRADE-able empirical facts; they are a framing on which bodies take positions (see the source note below). Whether the strong-setpoint version or a weaker regulation-plus-environment version is right is actively debated in the very sources here (Hall’s paper sits inside the energy-balance-vs-carbohydrate-insulin debate).
So the honest status is: the regulation-and-defense mechanism is real and partly high-GRADE; the
disease paradigm built on top of it is an emerging, still-consolidating framing that has not
reached settled consensus. This page asserts the mechanism (which the evidence carries) and holds
the paradigm at confidence: medium — it does not present the label as established.
Synthesis
This is a worked instance of Model Error vs Sampling Error. The willpower model is not imprecise — it is aimed at the wrong variable. No sample size corrects it, because the error is in what the estimate is of: a study that measures “compliance” or “calories chosen” as though these were the free inputs is estimating a downstream readout of a regulated system as if it were the control knob. That is a model error — out-of-CI by construction — and it is why decades of advice-based interventions produced small, non-durable effects that looked like patient failure rather than model failure.
The mechanism-vs-outcome structure. The individual worked sections above are each modest on their own (one cohort’s hormone panel; one RCT’s efficacy result plus a non-randomized off-treatment extension of another; a narrative-review framing). Their value is configurative: the hormonal-defense mechanism (Sumithran), the pharmacological effect size (Jastreboff), the withdrawal-regain (Wilding), and the regulation-plus-environment framing (Hall) triangulate on one claim from independent angles — mechanism, intervention, de-intervention, and epidemiology — that no single one of them carries alone. The defended-setpoint reading is what makes sense of all four at once.
The compensation link. The willpower model’s cousin on the expenditure side is the naive “calories out” model — burn 300 kcal in cardio, lose 300 kcal of fat. The organism compensates: appetite rises, non-exercise activity falls, and resting expenditure may drop beyond what the lost tissue predicts (adaptive thermogenesis) — though that limb is the contested and smaller one: the held energetics review flags its persistence as controversial and ranks the intake rise as >3-fold larger than the expenditure adaptation -> Weight-Loss Maintenance and Metabolic Adaptation . That is the same regulated-defense logic applied to the output side, worked on Net Effect vs Intended Effect and Exercise Energy Compensation — “you can’t outrun a bad diet” is this compensation, not a slogan. Both the intake-willpower model and the expenditure-tally model share the defect: they treat a regulated quantity as a free one.
Why the framing is easy to get wrong in both directions (kept peripheral). The attribution has a stigma cost — a willpower model licenses blame, and blame is a documented barrier to care — but the wiki weights framings by their evidence and decision consequences, not by their social valence. The disease model earns its place here because the held mechanistic and interventional evidence supports it, not because it is kinder; and it is bounded by the symmetric guard above so it does not become an excuse-model. The label “disease” is itself contested in the literature and is largely a framing/guidance question rather than an evidence one — this page makes the biological-regulation claim, which the evidence carries, and does not rest on the disease label, which position statements assert but do not evidence (see Which Questions Get Studied on why the framing layer and the evidence layer must be kept apart).
Intervention consequence — the model flips the recommendation (Layer 3)
This is where the model pays off. If the higher weight is defended — not merely the residue of past choices — then the reflex advice “reverse what got you here” is mis-specified: removing the inputs that raised the weight does not lower the defended level, so the person diets back into the teeth of the defense and regain is the modal outcome -> Weight-Loss Maintenance and Metabolic Adaptation (where regain is “the norm, not the exception”). The willpower-and-misery prescription is not merely unkind: under adherence-is-part-of-the-effect, a maximal push that is abandoned has ~0 net effect, and the repeated loss-regain it produces carries its own cost -> Weight Cycling and Cardiometabolic Risk.
So the levers that counter the biology are structurally favored (the structural-leverage rule): they change the regulated system rather than demand a sustained fight against it. The strongest held evidence is pharmacological — SELECT is the first demonstration that a GLP-1 agonist cuts hard cardiovascular events, not just weight -> Semaglutide for Cardiovascular Risk in Obesity — moving the drug route from cosmetic to outcome-indicated in the population SELECT studied: established cardiovascular disease (secondary prevention). It carries no hard-outcome evidence in primary prevention, so “outcome-indicated” does not yet extend to the whole obese population. Three guards keep this from becoming “a drug for everyone”:
- Continued, not cured. The drug works while taken; after stopping (drug and lifestyle support stopped together), about two-thirds of the loss returned within a year, net -5.6% at week 120 (Wilding withdrawal, above). “Pharma indicated” means a chronic medication with its own cost / side-effect / reversibility ledger — a blood-pressure-style commitment, not a course.
- Lifestyle is not null — it is the comparator and the substrate. SELECT’s drug arm sat on top of lifestyle, not instead of it, and the food environment remains the population driver (Hall, above). The claim is that willpower is the wrong sole lever against an engaged defense, not that behavior is irrelevant.
- Stratum: prevention likely easier than reversal — on mechanism, not interaction evidence. If the defense is provoked by weight gain and then guards the higher level, acting before the setpoint moves should avoid engaging the defense at all — so the not-yet-obese and established-obesity strata plausibly warrant different emphasis (protect environment/behavior vs structural leverage). But this is a route-(b)-flavored effect-modification claim resting on the unvalidated mechanistic defense model — no positive interaction evidence is held that the same lever works better applied earlier — so it is marked as mechanism (directional, discounted), not a clean baseline-risk (a) stratification that would need none. The magnitude (how much cheaper) is not something any held source demonstrates; the direction is the claim, not the size.
What this does NOT claim — the vicious-cycle version is contested, not held. The stronger reading — that obesity self-perpetuates through positive-feedback hormonal loops (leptin resistance; hyperinsulinaemia driving further fat storage) — is a feed-forward claim beyond the defended setpoint the held evidence carries, and its leading form (the carbohydrate-insulin model) is explicitly rebutted by a source this page holds (Hall 2022 is a CIM rebuttal; “leptin resistance” does not occur in it). The intervention flip above does not depend on it: a defended setpoint (negative feedback around a raised level) already makes reverse-the-inputs insufficient and structural leverage indicated. Whether a true self-reinforcing spiral operates is an open mechanistic question — the insulin-resistance feed-forward loops the fabric does hold sit in the T2D / ectopic-fat pathway -> Ectopic Fat and Depot-Specific Risk, not in a shown obesity-begets-obesity setpoint drift.
The open loop. None of this is validated against realized long-term outcomes here — the wiki grades coherence and source-fidelity, not validity; the durability of drug-based defense-countering beyond trial horizons is itself an open question.