Every few weeks another condition joins the list of things these drugs are said to help. Some of those claims will turn out to be right — the knee osteoarthritis result came out of a randomized trial of 407 people. The useful question is not which claims sound plausible but what kind of study produced them, and there is a small study of a painful skin disease that makes an unusually clean example of the problem.
The study
Hidradenitis suppurativa is a chronic inflammatory skin condition, strongly associated with obesity, that produces recurrent painful lesions. A prospective open-label cohort followed 20 adults with the condition and obesity for six months on semaglutide above 1 mg a week, measuring weight, body composition, skin severity, pain, mood, and metabolic and inflammatory markers at baseline and at the end. [1]
The results are striking. Body weight fell 21.4 kg. Hurley stage, the standard severity grading, decreased in every single patient. The dermatology quality-of-life index improved by 8.8 points, a pain visual analog scale by 4.1, and a depression inventory by 12.9, all at p<0.001. Blood pressure, waist circumference and fat measures all fell as well. If you read only that paragraph you would conclude the drug treats the disease.
What is missing
There was no control group. Every participant got semaglutide, and each one’s six-month numbers were compared with their own baseline numbers. That design cannot separate the drug from everything else that happens in six months — regression to the mean in a relapsing-remitting disease, the attention of a research team, seasonal variation, other treatments, and the simple fact that people enroll in studies when their condition is bad.
It was also open-label. Participants knew they were taking the drug. Three of the four headline outcomes — quality of life, pain, and mood — are self-reported, which is where that matters most. A person who has lost 21 kg and knows why will answer a pain questionnaire differently.
And there were 20 people. Not 20 per arm. Twenty in total, which is roughly the size at which a single unusual responder moves a mean. None of this is misconduct; it is what a pilot study is. The paper is honest about its design in its own methods section.
Why this shape of claim keeps appearing
Because it is cheap to produce and genuinely interesting. Twenty patients, six months, no randomization and no placebo is a study a single clinic can run. It is the correct first step, and the correct thing to do with its result is to design the trial that would test it. The failure is downstream, when the pilot gets summarized as a finding and the finding gets summarized as an indication.
There is a legitimate version of the weight-independent claim, and it looks different. A prespecified analysis of 1,145 randomized participants found symptom benefit and weight loss moving apart inside a controlled trial, which is the kind of comparison that can support the inference. Same shape of conclusion, completely different grade of evidence.
How to read the next one
Three questions settle most of it. Was there a group that did not get the drug? Did anyone not know what they were taking? How many people, in total, not per headline? Those take ten seconds against an abstract and they sort nearly every claim you will meet. Effect sizes and intervals matter too, and what a comparator can tell you is the longer version of the first question.
Nobody on this roster sells semaglutide for hidradenitis suppurativa, and no seller claims it. What sellers do carry is a general implication that the benefits keep extending outward, which is where a twenty-person pilot ends up doing commercial work it was never built for. The ordinary weight-loss evidence is strong and worth reading on its own terms. It does not need help from studies like this one, and what a seller declines to say is usually more informative than what it implies.