The Shortage Years

In the second half of 2022, a particular kind of absence started appearing on pharmacy systems across several wealthy countries. Not a discontinued line, not a recall, not a supplier switching wholesalers — an approved, actively marketed, extremely profitable medicine that the manufacturer simply could not produce in the quantities being asked for. Over the following three years that absence became one of the largest sustained supply failures in modern pharmaceutical history, and it happened not to an old generic injectable running on a worn-out line, but to the best-selling drugs on earth.

The obvious explanation — that the peptide itself was hard to make — was wrong, and the fact that it was wrong is the most interesting thing about the whole episode.

Demand that arrived after the decisions

Capacity in this industry is planned years ahead against a forecast. The forecast for GLP-1 receptor agonists had been built on their original indication — type 2 diabetes — where they were one option among several and grew at a respectable, predictable rate. That forecast survived until the obesity trial results landed. When a large randomised trial of once-weekly semaglutide in adults with overweight or obesity reported mean weight loss of around fifteen per cent of body weight, an effect size with no precedent in pharmacological weight management, the addressable population changed overnight 1.

The word overnight is doing real work there. Trial results are published in a single week. A sterile manufacturing line takes years. Every decision that determined how much product could exist in 2023 had been taken before anyone knew what the 2021 results would say, and there was no mechanism by which those decisions could be revisited in time.

There is a further asymmetry that made the mismatch worse than a simple forecasting miss. The diabetes indication and the obesity indication do not consume the same amount of drug. Weight-management regimens escalate to higher maintenance doses, so a single patient moving from one indication to the other can represent several times the quantity of active ingredient and, more importantly, the same number of devices at a higher fill. Capacity planned in patients turned out to have been planned in the wrong unit.

What followed was compounded by the ordinary machinery of attention. Prescribing spread from specialists to primary care, then to telehealth, then into a cultural conversation that generated demand from people who had not previously been counted in any forecast at all. By the time the shortages were formally recognised, the gap between what existed and what was being asked for was not a matter of percentages.

The bottleneck was not the peptide

Here is the counter-intuitive part. Synthesising these molecules at scale is genuinely demanding — they are long for a synthetic peptide, they carry modifications that extend their circulating half-life, and the processes involved are proprietary and hard-won. But active ingredient was not what ran out first, and it was not what took longest to add. Peptide manufacturing capacity had been expanding for two decades in response to a therapeutic pipeline that had been growing steadily since the 1990s 4, and adding a synthesis train, while expensive, is a solved industrial problem.

The narrow point was downstream. These products are injectables, most of them delivered in pre-filled multi-dose pen devices, which means the manufacturing chain ends with aseptic fill-finish: filling a measured volume of sterile solution into a cartridge under conditions that guarantee sterility, then assembling that cartridge into a device with a dose-setting mechanism, then packaging and labelling it. Every one of those steps is capacity-constrained, capital-intensive and, crucially, not fungible. A line qualified for one product cannot simply be pointed at another.

This was not a new discovery. A decade earlier, an analysis of why generic sterile injectables kept running short had made the same structural argument: this segment of manufacturing has few producers, high quality-related fixed costs, little spare capacity, and a market that cannot observe quality and therefore will not pay for redundancy 2. The 2022 shortages ran the identical mechanism at the opposite end of the price spectrum, which should have settled the question of whether the problem was about cheap drugs. It was never about cheap drugs.

What a shortage designation actually does

It is worth separating two things that get conflated. There is a shortage, meaning that people who want a product cannot get it. And there is a shortage designation, meaning that a regulator has formally listed the product, on the basis of manufacturer reporting and its own assessment of supply against demand. The first is a fact about the world. The second is an administrative status, and it is the status rather than the fact that has legal consequences.

Those consequences are specific: enforcement discretion around importation, flexibility on certain labelling and packaging requirements, prioritised review of applications and supplements that would add supply, and — the one that mattered most here — a statutory opening for pharmacies and outsourcing facilities to prepare copies of a listed drug, which is otherwise prohibited. The designation is the switch. When a regulator later determines that supply meets demand and removes the listing, the switch flips back, and it does so on a defined timetable rather than gradually.

Regulators had spent years documenting how badly this whole apparatus works as a response rather than a prevention. A task-force report published in 2019, examining more than a hundred and sixty shortages across a five-year period, concluded that shortages were fundamentally driven by economics — that the market did not reward the investment in reliability and quality that would prevent them, and that the tools available once a shortage begins are palliative 3. The GLP-1 episode was an unusually well-lit demonstration of that conclusion.

The markets that opened and closed

What happened in the space created by those designations is a matter of record and belongs in this history as a description of events. Between 2022 and 2025 a substantial secondary supply layer grew up around the listed products, ranging from regulated compounding operations working within the statutory exemption, through telehealth services that bundled prescribing with supply, out to sellers with no plausible relationship to any regulatory framework at all.

It closed in stages. The tirzepatide shortage was declared resolved in December 2024 and the semaglutide shortage in February 2025, each followed by a short wind-down period after which the statutory exemption no longer applied. What the regulator then found at the far edge of that layer is documented in its own public statements: products labelled for research or explicitly not for human consumption but sold with dosing instructions, compounds that had never been approved for anything anywhere, and shipments arriving warm because whoever sent them had not accounted for temperature at all 5.

Recording that sequence is not the same as commenting on it. The point of interest here is structural: a supply gap in a heavily regulated product does not stay empty. It fills from whatever direction has the fewest obstacles, and the composition of what fills it is determined by how long the gap stays open and how wide the legal opening is at the time.

How long a line takes

The manufacturers did respond, at a scale that is easy to underestimate: multi-billion-currency-unit commitments to new sites, expansions of existing ones, and — most tellingly — the outright acquisition of established fill-finish businesses rather than building from nothing. That last choice is the clearest possible statement about lead times. Buying an operating, inspected, qualified facility was judged faster than constructing an equivalent one.

The reason is that construction is the short part. A new sterile line has to be built, then commissioned, then qualified — proving each system does what it is specified to do — then validated for the specific process, including media fills that demonstrate sterility can be maintained across a production run, then written into a regulatory submission, then inspected, then approved. Three to five years from decision to approved commercial output is normal, and the majority of that clock is qualification and regulatory review rather than concrete and steel.

Which produces the perverse rhythm every shortage of this kind has. Capacity arrives after the crisis it was ordered for, and often after demand has settled, which is exactly the pattern that makes companies reluctant to order it in the first place. The industry has known this for decades and has no good answer, because nobody is willing to pay to keep expensive sterile capacity idle against a shortage that may not come.

What the episode exposed

Strip away the specific molecules and a general picture of peptide manufacturing is left standing, and it is not the picture most people carry. Making the peptide is the well-solved, well-capitalised, geographically diversified part of the chain. Everything after it — sterile filling, device assembly, cold storage, distribution — is narrower, more concentrated, harder to expand and less visible, and it is shared across an entire industry that spent thirty years deciding it was more efficient not to own any of it.

That is the durable finding of the shortage years, and it is worth holding on to precisely because the crisis has passed and the temptation is to file it as a story about one unusually popular drug class. It was a stress test. What it measured was not whether a molecule could be synthesised, which was never in doubt, but how much of the world's capacity to put a sterile liquid into a device could be redirected in an emergency. The answer, for three years, was: not very much, and not very quickly.