The Hunger Signal: GLP-1 Drugs and Autistic Interoception

The question

Weight-loss drugs like semaglutide (a.k.a., GLP-1’s) work by changing how the body senses hunger and fullness. Many autistic people sense hunger and fullness differently. Does the drug therefore work differently for us - better, worse, or simply strangely?

As far as I can find, nobody has asked.

Where the question came from

In Chapter 3 of Loving Variant Minds I described interoception as the most important sense you have never heard of. It is the sense that reports on the inside of your body: hunger, thirst, temperature, needing the toilet, the racing heart that tells you that you are anxious. Many autistic people describe this sense as unreliable, and it fails in both directions. Some do not notice hunger until they are shaking and irritable. They forget to drink. They push through illness without registering how unwell they are. Partners end up asking “have you eaten today?” like a parent, and both people find it frustrating. Others have the opposite experience: hunger that never quite switches off, a meal that ends without the body registering that it has ended, food occupying far more mental space than it seems to for anyone else. The single autistic case report in this literature is a young man of exactly that second kind.

Now consider what a GLP-1 drug actually does.

GLP-1 is a hormone your gut releases when food arrives. It is one of the messengers that tells your brain the meal is over. Semaglutide - sold as Ozempic and Wegovy - is a manufactured copy of that messenger, engineered to last far longer in the body than the natural version. Liraglutide is an earlier drug of the same family. The advertised effect is weight loss. The actual mechanism is a change to the hunger signal itself.

So we have a class of drug that acts directly on a body signal, and a population known for reading that signal differently. That is a question. It might even be a good one.

What the evidence actually says

I want to be honest about how little there is.

Mice. Several teams have given semaglutide to BTBR mice, a strain used as a model of autism. One 2024 study reported that repeated treatment reduced autism-like behaviours without affecting motor performance.¹ Another, published in 2026, credited changes in gut bacteria and metabolism.² Both are real papers in real journals. But notice what neither of them proposes as the mechanism: interoception. The first is about DNA repair and oxidative stress. The second is about the gut-brain axis. Neither is testing my question, and mouse behaviour is a long way from human experience.

One human being. In 2019 a Finnish team published a case report on a twenty-year-old autistic man with moderate intellectual impairment.³ He had severe food obsessions, binge eating and aggressive behaviour. On liraglutide, the food obsessions dropped away quickly, his weight fell by twelve to thirteen percent, and - this is the interesting part - the aggression and the non-food compulsions reduced too. No side effects were recorded. The authors were careful: one patient is one patient, and they called for proper trials.

There is a detail in that case worth holding onto. His weight gain had been caused by the antipsychotic medication he was given for behavioural problems in the first place. That is not a footnote. It is a common autistic story, and it is one reason autistic people end up as candidates for weight-loss drugs at all.

Trials. I have not been able to find a single registered trial of a GLP-1 drug in an autistic population.⁴ I would be glad to be corrected. There are active semaglutide trials for alcohol dependence, opioid dependence, Alzheimer’s, asthma, arthritis and adolescent bone health. There is not one for autism. Not recruiting, not planned, not completed.

So the entire human evidence base for my question is a single case report from 2019.

The comparison that makes me nervous

If you want to know whether a drug that fixes satiety signalling helps people whose satiety signalling is unusual, there is an obvious place to look. Prader-Willi syndrome is a genetic condition in which the hunger switch does not turn off. People with it experience relentless hunger, and obesity is the main cause of early death.

Liraglutide has been properly tested there. A multicentre trial ran for fifty-two weeks in children and adolescents with Prader-Willi syndrome, with a sixteen-week blinded period against placebo.⁵ The two main endpoints - change in body mass index score at sixteen weeks and at fifty-two weeks - were not met. Hyperphagia scores did shift somewhat in the adolescents, enough that the authors said further study was warranted, but the trial did not show what it set out to show.

I find this sobering, and I think anyone excited about my question should sit with it. In the one population where the satiety signal is definitively atypical, the drug performed worse than it does in the general population, not better. That is the opposite of the hopeful reading.

The part where I argue against myself

There is a bigger problem with my question, and it is a problem with the premise rather than the conclusion.

The idea that autistic people have atypical interoception is not settled science. It is a live argument.

The strong version of the claim came from Quattrocki and Friston in 2014, who proposed that a failure of interoception gives rise to the autistic phenotype itself.⁶ Within a year, Brewer, Happé, Cook and Bird published a direct rebuttal.⁷ Their argument was that the interoceptive difficulties belong not to autism but to alexithymia - the difficulty identifying and describing your own emotions - which happens to co-occur with autism in around half of cases. In 2016 Shah and colleagues published a study whose title states the position plainly: alexithymia, not autism, is associated with impaired interoception.⁸ A 2018 study went further and argued that interoceptive impairment does not sit at the heart of either condition.⁹

The evidence runs the other way too. A 2023 study found reduced interoceptive accuracy and sensibility in autistic participants, with alexithymia acting as a mediator rather than an explanation.¹⁰ A 2016 review found enough to justify serious attention to interoception in autism.¹¹ The honest summary is that this is contested, and the contest is not close to over.

Two further complications, both of which a dissertation would have to face directly.

The first is measurement. Much of this literature rests on a single task, unchanged since 1981: sit still, do not take your pulse, and count how many heartbeats you can feel. The closer your count to your actual heart rate, the better your interoception is judged to be.¹² That task has been criticised heavily, including by researchers who had used it themselves, on the grounds that people may be estimating from what they believe about their own heart rate rather than actually sensing it.¹³ A field resting on a shaky instrument produces shaky findings in both directions.

The second is more specific to my question, and I think it is the sharpest objection. Interoception is not one thing. It varies by domain, and the evidence that accuracy in one domain predicts accuracy in another is mixed. Almost all of the autism research has studied the heart. GLP-1 drugs act on the stomach. Even if the autism-interoception link were rock solid for cardiac signals, it would not automatically carry across to hunger.

I raise all of this because I would rather state the weakness in my own idea than have a supervisor state it for me. A dissertation on this topic could not assume the premise. It would have to test it.

The risk nobody is measuring

If I had to justify this research on public health grounds rather than curiosity, I would not lead with weight loss. I would lead with harm.

Avoidant/restrictive food intake disorder - ARFID - is not about body image. It is restriction driven by sensory aversion, by fear of choking or vomiting, or simply by a lack of interest in eating. It overlaps heavily with autism. A 2025 meta-analysis found autism in around sixteen percent of people with ARFID, and ARFID in around eleven percent of autistic people.¹⁴ A genetically informed study of a large autism cohort put the figure higher, estimating ARFID in around a fifth of autistic participants, and suggested it is routinely under-recognised.¹⁵

Now put a drug on top of that. A drug whose core action is to suppress interest in eating, which commonly causes nausea, and which changes how appealing food feels.

I am not claiming this causes harm. I am claiming that nobody has looked, that the mechanism for harm is obvious once you say it out loud, and that autistic people are already being prescribed these drugs. The absence of a trial is not the absence of a risk. It is only the absence of information.

What I would actually do

A study to answer this properly is not exotic. It is ordinary clinical research that happens not to have been done.

• Recruit autistic and non-autistic adults with obesity who are starting a GLP-1 drug for the usual clinical reasons. No new exposure, no new risk.
• Measure gastric interoception at baseline, not cardiac. If the drug acts on the stomach, measure the stomach.
• Measure alexithymia separately, as its own variable. This is the step that would let the study contribute to the autism-versus-alexithymia argument rather than sidestep it.
• Screen for ARFID at baseline and track it throughout.
• Record two outcomes, not one: the weight change, and the subjective experience. How does satiety feel? Is the nausea tolerable or intolerable? Does food become less interesting in a helpful way or a distressing way?
• Design it with autistic people rather than about them.

My honest prediction is that weight outcomes would look broadly similar between groups, because the drug probably acts below the level of conscious body awareness, but that the lived experience would differ sharply - and that a subgroup would do badly in a way that a weight-only trial would never detect.

I could be wrong. That is rather the point of doing the study.

Why this one is on the list

Because the drugs are already in autistic bodies. It is happening at scale, in every country where these medications are prescribed, and it will carry on happening. Asking whether we should allow it is not a serious question. The market has already answered it, and no dissertation is going to change that.

What a study like this could change is smaller, and more achievable, and it comes in two parts. The first is clinical. If the direction of someone’s interoceptive difference predicts how they respond, then prescribers have a reason to ask about autism before they write the prescription rather than after. They also have a reason to monitor for effects that would never appear in a neurotypical patient - because nobody looks for a side effect they have never been told to expect. A patient who quietly stops eating is not a mild outcome, and it is exactly the sort of thing that gets recorded as treatment success if the only number anyone is watching is the weight. Screening costs a conversation. Missing it could cost somebody their relationship with food.

The second is psychological. Therapists and psychologists supporting autistic clients cannot prepare someone for an effect that nobody has documented. If we knew how these drugs tend to land on a particular kind of nervous system, that knowledge could be handed to the client in advance: you may find that food stops interesting you altogether, and given how you already relate to eating, that might not be a small thing - so let us agree now what we do if it happens. That is a very different conversation from having it six weeks later, after the fact, with someone who has concluded that something is wrong with them.

Neither of those requires new law, or a change in what pharmaceutical companies do, or anyone’s permission. Both require somebody to have done the study.

---