Start with a fact that does not get said often enough: the response to these medications is remarkably uneven. In one analysis of semaglutide, the average weight loss was around 10% of body weight, but roughly 1 in 20 people lost more than a quarter of their body weight, while nearly a third lost less than 5% or gained. Same molecule, same dose, very different results.
It is tempting to read that spread as a measure of effort, and it is nothing of the kind. A large part of it comes down to a point that the single word obesity hides: the weight was not put there by one thing. It is the end point of several different drivers, which act alone in some people and combine in most. And each of those drivers plausibly shapes not only whether the medication works, but how much of it a person needs. At Anova we think that is the more useful way to look at the question of response, so this piece takes the drivers apart one at a time before narrowing on the one that is least understood and arguably most interesting.
Obesity is not one condition
It helps to see the drivers separately first, even though almost no one is a pure single type.
Researchers at the Mayo Clinic have spent over a decade building exactly this kind of map, and their framework sorts the drivers of overeating and excess weight into recognisable groups. There is the “hungry brain,” the person who can eat a full meal and not feel full, whose satiety signal arrives late or faint. There is the “hungry gut,” the person who feels satisfied at the table but is genuinely hungry again an hour or two later, whose fullness does not last. There is emotional hunger, eating driven by mood and stress rather than physical need. And there is the “slow burn,” the person whose problem is less about intake and more about a metabolism that runs economically, burning fewer calories than expected. Notably, this group has reported that which of these patterns dominates can be measured, and that it predicts who responds well to semaglutide, the early shape of a precision approach to obesity rather than the one-size dose everyone currently starts on.
Sitting alongside those internal drivers is a whole second set that has little to do with appetite wiring at all. Age and the metabolic slowing that comes with it. The composition of the diet, not just its quantity, with highly processed, energy-dense food driving weight through routes that have nothing to do with willpower. And the contributors we have written about at length elsewhere, because they are so often the uncorrected reason a person is heavier than they want to be.
Sleep is the clearest example. Poor or insufficient sleep raises the hunger hormone ghrelin, lowers the satiety hormone leptin, worsens insulin resistance, and shifts food preference toward calorie-dense options, and the effect is large enough that we have argued treating weight without addressing it is fighting with one hand behind your back. A person whose weight is substantially sleep-driven, and whose sleep is never corrected, has a blunting factor working against the medication the entire time. The same logic applies to circadian disruption in shift workers, to untreated mood, and to several other contributors. Each of these can dampen the result, and each, in principle, changes the calculus of how the medication should be dosed and for how long.
The practical point of laying them out this way is simple. If the weight came from several places, the medication is only acting cleanly on some of them, and the honest expectation depends on which drivers are in play. A person whose weight is mostly slow-burn metabolism and uncorrected sleep is asking the drug to do something different from the person whose weight is mostly a hungry brain, and it is unsurprising they get different results from the same dose.
Narrowing on one driver: the volume of the food noise
Hold all of that aside for a moment, the way a clinician mentally controls for the other contributors, and one driver in particular is worth examining on its own, because it is the one most directly in the medication’s line of fire and the one where the open questions are most interesting.
Most people who have struggled with weight will recognise it long before they have a word for it: the constant, intrusive background hum of thoughts about food, the snack half-planned an hour out, the pull of the cupboard at 9pm, the meal you are already thinking about while finishing the last one. The popular term is food noise. The clinical construct closest to it is hedonic hunger, the preoccupation with and desire to eat for pleasure rather than physical need, and it can actually be measured, with a validated questionnaire called the Power of Food Scale. We have written separately about what happens when that noise goes quiet, including the unexpected ways the quiet can spread beyond food.
One honest caveat belongs here, because it sharpens rather than weakens the point. Hedonic hunger is a strong predictor of cravings and loss-of-control eating, but on its own it does not reliably predict how heavy a person is. That fits the theme of this piece exactly: food noise volume is not a measure of weight, it is a marker of which engine was driving the eating. Two people at the same weight can have completely different food-noise volumes, because they got there by different routes.
Why this engine matters for the medication comes down to how eating is organised in the brain. There are roughly two systems. The homeostatic system is the body’s genuine energy-need signalling, hunger when fuel is low, fullness when it is restored. The hedonic system is reward-driven, the wanting and pleasure of food regardless of need, and it runs through dopamine pathways. These medications act on both, but the hedonic effect is striking: in animal work published in 2025, semaglutide directly suppressed the dopamine neurons that sustain the consumption of palatable food — though a separate 2023 study found semaglutide increased rather than decreased VTA dopamine activity during reward consumption, suggesting the mechanism alters reward circuitry in context-dependent ways rather than simply dampening it. Loud food noise is, in large part, that reward system running hot. So a person who arrives with very loud food noise is arriving with a great deal of exactly the signal the drug is built to turn down. There is more of it to quiet, and quieting it produces a larger felt change.
Two open questions worth a clinician’s attention
This is where the research thins out and the honest thing to do is mark the edge of what is known rather than paper over it. Two questions follow naturally from everything above, and neither has been directly answered. We raise them not as Anova’s conclusions but as considerations a prescriber may want to hold in mind, drawing on the wider evidence as they go.
The first is about dose. If food noise volume reflects how strong the hedonic drive is, it is reasonable to ask whether it should also inform the dose. The intuitive read points one way. Loud, entrenched food noise is a powerful reward signal, and a powerful signal may take more medication to subdue, so the person with the loudest noise might need to climb higher up the dose ladder before the appetite drive genuinely quiets. By the same logic, a person whose food noise was never especially loud, whose problem was a faint satiety signal or a slower-burning metabolism rather than a roaring reward drive, has less of that signal to overcome and might reach most of their available benefit on a lower dose, or with a much slower, gentler titration. On this reading, sensitivity to the body’s own fullness cues is precisely what lets a smaller nudge do the work.
Two things support taking that direction seriously rather than treating it as a hunch. The medication’s action on the reward system appears dose-sensitive, and it acts on the homeostatic and hedonic systems to different degrees, so different drivers plausibly meet their limiting factor at different doses. And it runs with rather than against the population data: the trial dose-response curves show that, on average, more medication produces more weight loss, which is exactly what you would expect if the heavier reward drives need the larger doses. None of this has been tested directly by eating phenotype, so it remains an open question rather than a finding. It is a matter for the clinician to assess during titration, weighing response against tolerability in the individual in front of them, rather than a rule that can be read off a questionnaire. We raise it so that it is a conscious part of that assessment.
The second is about what happens later, on a maintenance dose or after stopping. Here the gap in the evidence is near total. Every study of what happens when people stop these medications finds the same thing: weight returns, by roughly two-thirds within a year, driven by a defended set-point reasserting itself as appetite hormones climb, satiety signals fall, and energy expenditure stays suppressed. What no study has yet done is ask whether that regain differs by the driver that put the weight there in the first place. It is plausible, and only plausible, that a high-food-noise person does well while the drug is quieting the noise but is more exposed to relapse when the noise returns to its old volume, the loud reward signal coming back online. It is equally plausible that a person whose weight was metabolic or age-related follows a different trajectory both on the drug and after it. We do not know. The data to answer it has not been collected.
The reason this is worth a clinician’s attention despite being unstudied is that the down-titration and maintenance decision is being made for real patients now, without it. A prescriber thinking about how, and how slowly, to taper a particular person may reasonably draw on what is known about that person’s drivers, their food noise volume, their sleep, their metabolic picture, their established habits, among other findings, even while acknowledging that no trial has validated doing so. Naming the gap is not the same as ignoring it. It is the responsible version of treating in the absence of complete evidence, which is most of medicine most of the time.
What this is really saying
For anyone who has watched the weight come off easily for someone they know and slowly for themselves, this is the point that matters. Response to these medications is not a measure of effort or character. It tracks the particular mix of things that drove the weight, how loud the food noise was, how the body burns, how well it sleeps, how clearly its fullness signals came through, and the medication acts cleanly on some of those and barely on others. Those are facts about biology and history, not verdicts on willpower.
The work of good care is to take that mix apart, treat the parts that can be treated, correct the contributors that are correctable, and set an expectation that fits the individual rather than the average. The medication is a powerful lever. How much it moves depends on what it is levering against, and being honest about that, including about the questions the research has not yet answered, is the difference between a prescription and a plan.
If you want the companion pieces, the reward-pathway research behind shifting exercise motivation covers the same brain circuitry from a different angle, sleep as the missing lever covers the most common correctable contributor, and life after the maintenance dose covers the stopping question in full.