The Year the Obesity Drugs Worked

For most of a working career, an obesity physician's pharmacological options amounted to a short shelf of compounds that shaved a few percent off a person's weight, a rather longer shelf of compounds that had been taken off the market, and a conversation about surgery. Then, in 2021, a trial of once-weekly semaglutide in adults with overweight or obesity reported a mean body-weight change of about −14.9% at sixty-eight weeks, against −2.4% in the placebo arm 2. That single line is the turning point, and what makes it a hinge rather than a good quarter for one company is the half-century standing behind it: a field with a documented habit of producing drugs that worked modestly and then harmed people had just produced a number belonging, until that moment, to the operating theatre.

The graveyard behind the number

You cannot read 2021 correctly without the decades in front of it, because the decades are what made the number legible. Obesity pharmacology had by then a reputation problem — not the ordinary sort, where a field is thought unproductive, but the sharper sort, where it is thought actively dangerous. Compounds reached the market on a plausible mechanism and a respectable trial, and came off it again when the longer exposure of real-world prescribing surfaced signals the trials had been too small or too short to see. The harms that ended those careers clustered around the heart and around the mind.

Any new entrant inherited the resulting expectations whether it deserved them or not. Regulators approached the category with the caution of an institution that has been embarrassed before, and clinicians who had once prescribed something later recalled had learned to wait. The arithmetic was brutal: if a drug delivers three or four percent of body weight, the harm that can be tolerated in exchange is close to none. A small benefit does not buy much risk.

That arithmetic propped up a framing which dogged the field for years. Obesity treatment was widely read — by payers, by parts of the profession, by the public — as essentially cosmetic, and so as a place where the usual willingness to trade side effects for benefit did not apply. Fair or not, it set the bar. Anything new had to clear both the safety record and the suspicion that it was not really medicine.

A number that invited a second look

Against that background, consider what the 2021 trial actually reported. Adults with overweight or obesity and no diabetes, randomly assigned to once-weekly semaglutide or to placebo alongside lifestyle support, followed for sixty-eight weeks. Mean body-weight change in the treated arm: approximately −14.9%. In the placebo arm: −2.4% 2.

Two things are worth noticing. The first is the placebo arm, which is not a null condition. Those participants received counselling, contact and the sustained attention that being in a clinical trial brings, and they lost 2.4%. That is roughly what determined lifestyle intervention delivers under favourable conditions, and a useful measure of the problem the drug was being asked to solve.

The second is the size of the gap. Weight-loss pharmacotherapy had generally operated in single-digit percentages: respectable in the paper, disappointing in the clinic, and not obviously worth the risk given the field's history. Fifteen percent was a different kind of number. It approached a range associated until then almost exclusively with bariatric surgery — not a comparison anyone expected to be making about an injection, and one that quietly changed what the drug was for. A few percent modifies risk factors at the margin. Fifteen percent, sustained, can move a person out of the clinical category they were in.

One caution, because figures like this travel badly once they leave the paper. The −14.9% is an average across an arm. Some in that arm lost a great deal more and some very little; the mean conceals a distribution and describes a group under study conditions, never an individual.

Why the safety file came first

The other reason 2021 was believed rather than merely noted had been established years earlier, in a different population and for a different purpose. Semaglutide's cardiovascular outcomes had already been examined in patients with type 2 diabetes at elevated cardiovascular risk 1 — a trial run to satisfy a regulatory requirement that an earlier generation of diabetes drugs had itself created. The requirement exists because the field was burned once and decided not to be burned again.

That study mattered to the obesity story out of all proportion to its subject matter. The historical failure mode here was never drugs that did nothing. It was drugs that worked and then harmed — the sequence in which an effective compound accumulates users, accumulates exposure, and eventually accumulates a signal. A large trial with adjudicated endpoints, meaning events reviewed by an independent committee kept blind to who received what, meant the question the field feared most had already been put in a related population before the weight-loss result landed.

The trial that changed the argument

Then came the study that settled the matter. SELECT enrolled people with obesity and established cardiovascular disease who did not have diabetes, and asked what the field had argued about for decades without evidence: does treating obesity pharmacologically prevent anything? It reported roughly a 20% reduction in major adverse cardiovascular events — a composite counting cardiovascular death, non-fatal heart attack and non-fatal stroke 3.

Until then, the case for treating obesity with drugs had rested on surrogate endpoints: weight itself, blood pressure, blood lipids, glycated haemoglobin. A surrogate is a measurement accepted as a stand-in for the thing you actually care about, because the thing you care about — a heart attack that does not happen — takes years and thousands of participants to count. It is how most of metabolic medicine is argued, and it is permanently open to the objection that the stand-in moved while the outcome did not.

SELECT removed the objection by counting the events. In doing so it achieved something against the cosmetic framing that no amount of advocacy had managed: it made obesity treatment a cardiovascular intervention, assessed on the same terms as the drugs given for blood pressure or cholesterol. Payers argue differently about a medicine that prevents strokes. So do guideline committees. So, for that matter, do patients.

The ceiling moved again

If 2021 was a surprise, what followed was closer to a recalibration. Tirzepatide — one engineered peptide acting at two gut-hormone receptors rather than one, adding GIP to the familiar GLP-1 target — was tested in adults with obesity in a phase 3 programme, and the mean weight reductions in its treated arms went further again than those seen with semaglutide 4.

The interesting part is not which compound came out ahead, but what the result implied about the shape of the problem. For thirty years the assumption had been that pharmacological weight loss had a low ceiling: that body weight is defended by so many redundant systems that any single intervention would be compensated for. Two compounds in a few years showed the ceiling to be an artefact of the molecules tried, not a property of the biology.

That reorganised an entire pipeline. Agonists reaching for three receptors at once, oral formulations, agents meant to protect muscle while fat is lost — a field with almost nothing worth developing now has more candidates than it can run trials for. Failure had been the organising fact of obesity pharmacology. After 2021 it stopped being.

What success does to supply

Effective drugs create demand, and demand for these arrived at a size the industry had no precedent for. The eligible population is not a rare-disease cohort; it is a substantial fraction of the adults in every high-income country. Manufacturing capacity for a modified peptide delivered by injection cannot be conjured at that scale in a hurry: synthesis, purification, sterile filling and the injection devices are each a constraint carrying a lead time measured in years.

Shortages followed, and shortage plus high price plus enormous demand is one of the more reliable recipes in commerce. Compounded preparations appeared, made outside the original process under the national rules that let pharmacies prepare a medicine when an approved product cannot be had. Beyond that, outside any regulatory framework whatever, a market emerged in vials labelled with the same molecule names and sold as research material.

For a readership interested in peptide science, this is the part worth sitting with, because the causation runs in the direction you would least like it to. The unregulated supply exists precisely because the trials succeeded. Nobody was selling grey-market obesity peptides during the decades when the drugs did not work. The evidence created the demand, the demand outran the legitimate supply, and the shortfall filled with material about which nothing is known.

That is the distinction the whole thing turns on. Trial evidence attaches to a specific manufactured product: a defined molecule at a defined purity, made by a documented process, released against tests for identity, sterility and stability, and given under supervision in a study that counted what happened. It does not attach to a name. A vial labelled with a molecule asserts its contents; it does not demonstrate them, and a figure such as −14.9% has nothing to say about what is inside it.

• It covers a defined product: a specified molecule at specified purity, made under a documented and inspected process.
• It covers administration under medical supervision, with monitoring and a route by which adverse effects get recorded and acted upon.
• It does not cover material of unknown synthesis. Peptide manufacture generates related-substance impurities, and which are present, in what quantity, only analysis can answer.
• It does not cover sterility or endotoxin control, which are properties of a facility rather than of a molecule.
• It does not cover storage. Peptides in solution degrade at rates set by formulation, temperature and time, none of them knowable for material of uncertain origin.
• A name on a label is a claim about contents, not a chain of evidence.

The questions nobody can answer yet

None of which is to suggest the approved medicines are a closed book. They are not, and the open questions are substantial enough that anyone claiming to know how this story ends is guessing.

The first is what happens when treatment stops. The mechanism is a sustained signal at receptors governing appetite and food intake, and it works while it is present 5. Extension studies have observed weight regain after discontinuation, which is what a mechanism of that description predicts: the drug suppresses appetite rather than resetting whatever the body is defending. That turns a treatment decision into something closer to a permanent arrangement — a materially different proposition for a patient, a prescriber and a health system than a course of therapy with an end date.

The second is composition. Weight lost is not only fat. A fraction is lean mass — muscle, and the tissue that keeps it company — and both the size of that fraction and what it means across decades remain unsettled. The question presses hardest in older people, for whom muscle is the difference between independence and a fall.

The third is the situation itself. A medicine taken indefinitely by a large share of a population is not new in principle — that is what the drugs for blood pressure and cholesterol have been for a generation — but those accumulated their record slowly, over decades of ordinary use, and it took that long to know what they did. Here the exposure is accumulating now, in real time, in a group far larger and rather younger than any trial enrolled. The long-run answers do not exist, for the unglamorous reason that the years have not passed.

Which leaves the field somewhere odd and rather healthy. The question that defeated it for fifty years — can a drug meaningfully and safely reduce body weight? — has been answered in the only currency that counts: randomised arms, adjudicated events, published in full. The questions that replaced it are slower and harder, and they are about consequences rather than possibility. That is what success looks like from the inside. It does not feel like an ending.