Set point theory makes two predictions about a drug that takes weight off, and the incretin class breaks both. If these drugs reset the defended weight, then removing them should not much matter — yet a pooled analysis of 8 randomized trials in 2,372 participants found weight regain proportional to the original loss, at 2.20 kg (95% CI, 1.69 to 2.70) after liraglutide and 9.69 kg (95% CI, 5.78 to 13.60) after semaglutide or tirzepatide.[1] If instead the defended weight fights back harder the longer it is held down, then a fixed weekly dose should be progressively overpowered — yet in 17,604 adults, semaglutide produced loss that ran on for about 65 weeks and then held at −10.2% through 208 weeks, against −1.5% on placebo.[2] Four years, no escape, and no reset.
What a set point is supposed to be, and what was measured
The idea is not folklore; it has a number attached. Eighteen adults with obesity and 23 who had never been obese were held on a liquid formula diet in an inpatient setting and studied at their usual weight, after losing 10 to 20% of it by underfeeding, and after gaining 10% by overfeeding. Holding a weight 10% below baseline lowered total energy expenditure by 6 ± 3 kcal per kilogram of fat-free mass per day in the never-obese group and 8 ± 5 in the group with obesity (P < 0.001 for both). Holding a weight 10% above baseline raised it by 9 ± 7 and 8 ± 4 kcal/kg/day.[3]
That symmetry is the part that makes it look like a regulated variable rather than a ratchet. The body resisted being lighter and resisted being heavier by comparable amounts, in people who had never been obese and in people who had. It is also a small inpatient experiment, and the effect sizes are per kilogram of lean tissue, not per person.
The obvious objection — that the slowdown is a transient response to undereating — was tested and did not hold. Seven trios matched for sex and weight were studied on identical weight-maintaining formula diets: someone at usual weight, someone 5 to 8 weeks past a loss of at least 10%, and someone who had maintained a documented loss of more than 10% for over a year. Total and non-resting energy expenditure were lower in both weight-reduced groups, and the differences from the usual-weight group were qualitatively and quantitatively similar whether the loss was recent or a year old.[4] Whatever the reduction is, it does not time out.
The hormones do not time out either
The endocrine half was measured on 50 adults with overweight or obesity put through a 10-week very-low-energy diet, who lost a mean of 13.5 ± 0.5 kg. At the end of the program, leptin, peptide YY, cholecystokinin, insulin and amylin had fallen and ghrelin, gastric inhibitory polypeptide and pancreatic polypeptide had risen, with subjective appetite significantly increased. One year after the initial loss, leptin, peptide YY, cholecystokinin, insulin, ghrelin, gastric inhibitory polypeptide and pancreatic polypeptide were all still different from baseline, and so was hunger (P < 0.001).[5]
That is dieting, not drug treatment, and the distinction is the whole question. A model built on those findings says the body maintains a standing bid to return to its previous weight, transmitted partly through appetite. The drugs act on exactly that channel, which is what makes the comparison with dieting worth drawing at all, and what puts the horizon covered in the long-term article at the center of this question.
The withdrawal data say the set point was never moved
The proportionality in that pooled regain figure is the tell. A drug that had lowered the defended weight would produce regain unrelated to how much came off; what the trials show instead is that the further down the drug pushed, the further back up the body came.[1] The single cleanest demonstration is the semaglutide withdrawal design: after a 20-week run-in producing 10.6% loss, 803 adults were randomized to continue or switch to placebo, and over the next 48 weeks the arms moved −7.9% against +6.9%, a difference of 14.8 percentage points (95% CI, −16.0 to −13.5).[6] The arm-by-arm reading of that trial and its siblings belongs to the stopping article; what it contributes here is that the counterforce was waiting, undiminished, after nearly a year and a half of suppression.
Cycling on and off has been tested only in animals. Aged obese UM-HET3 mice put through three cycles of liraglutide treatment and withdrawal developed hyperleptinemia and visceral fat expansion and ended in worse metabolic health than mice held on continuous treatment; they regained fat mass after each cycle but failed to restore lean mass, which the authors read as a shift toward age-related sarcopenia.[7] That is a mouse result in a specific aged strain and it is not a human finding. It is, however, the only direct test of intermittent use anywhere in the literature, and it points the same way as the body-composition concerns in the muscle article.
And the four-year data say the defense never won
The other prediction fails just as cleanly. If a defended weight recruited progressively stronger counter-regulation, a constant dose should be gradually overwhelmed. In the cardiovascular outcomes population, weight loss continued through roughly week 65 and was then sustained: at week 208, mean reduction was 10.2% in weight, 7.7 cm in waist circumference and 6.9% in waist-to-height ratio, against 1.5%, 1.3 cm and 1.0% on placebo (P < 0.0001 for all).[2]
Two caveats sit under that figure and neither erases it. The population had pre-existing cardiovascular disease, so it is not the population buying a subscription; and discontinuation of trial product increased as BMI class decreased, which means the people still being measured at four years are not a random sample of those who started.[2] What survives is the shape: a curve that flattens and then stays flat for three more years, rather than drifting upward. The flattening itself has its own explanations in the plateau article.
The reading that fits both halves
A review of the competing frameworks — static, settling point and set point — concluded that the set-point model is the one most commensurate with current data, while noting that the components of energy balance are dynamically interrelated rather than independently controlled.[8] Stack that against the drug data and a consistent picture appears: the defense is fully intact, the drug supplies a continuous opposing force on the intake side, and neither exhausts the other. Nothing gets reset. Something gets held.
There is a measurement consistent with this. In a crossover trial of semaglutide at 1.0 mg over 12 weeks, resting metabolic rate adjusted for lean body mass did not differ from placebo despite a 5.0 kg loss.[9] The drug neither blunts the metabolic response to being lighter nor adds to it. It changes how much food is enough, and that is all it changes.
What nobody has measured
The obvious follow-up experiment does not yet have results. No published trial has tracked the hormonal panel that characterized diet-induced adaptation — leptin, ghrelin, peptide YY — across withdrawal from a GLP-1 drug, in the way it was tracked for a year after a very-low-energy diet. A registered randomized trial comparing gradual dose reduction against abrupt cessation, with fasting ghrelin among its secondary outcomes, has published its protocol and not its findings.[10] Until that reports, whether the counterforce is the same object after drug-induced loss as after diet-induced loss is an assumption, and the taper question it addresses is covered, with the evidence that does exist, in the tapering article.
What this makes the product
If the drug is a counterweight rather than a correction, then the relevant price is not the price of losing weight but the price of continuing to hold it — a different arithmetic, worked through in the maintenance cost article. It also means an advertisement promising to reset a metabolism or retrain a set point is describing something no trial has demonstrated on any product in this class.
Every figure above was produced on FDA-approved drug at labeled doses in randomized trials. Most sellers covered here dispense compounded semaglutide or tirzepatide, which is not FDA-approved and is not reviewed by the FDA for safety, efficacy or quality before it is dispensed, and no withdrawal or maintenance trial has ever been run on a compounded preparation. The four-year durability figure belongs to a branded product taken continuously inside a trial, and it does not transfer to a vial bought month to month.