The Elixir That Wrote the Drug Law

The bottles left Bristol, Tennessee, in the first week of September 1937. By the middle of October the American Medical Association was receiving letters from physicians in Tulsa describing something they could not explain: patients who had taken a raspberry-flavoured syrup for a streptococcal sore throat, felt better for a day or two, and then stopped producing urine and died over the following week. The syrup was Elixir Sulfanilamide. It carried a genuinely effective new antibacterial dissolved in diethylene glycol, the sweet and viscous solvent better known as a component of antifreeze. More than a hundred people died, a great many of them children.

Nobody had broken the law by selling it. That is the sentence to sit with, because it is the hinge on which the whole of modern drug regulation turns. In 1937 an American manufacturer could compound a new preparation, name it, flavour it, ship it to every state in the union, and never once be required to demonstrate to anybody that it would not kill the person who swallowed it. The law that existed had nothing to say about the question, and the law that replaced it exists because of these bottles.

A solvent chosen for taste

Sulfanilamide in 1937 was a legitimate marvel. It worked against streptococcal infection at a time when streptococcal infection routinely killed people, and it was already being sold as tablets and powders without incident. The problem the S. E. Massengill Company set out to solve was mundane and commercial: physicians in the American South were asking for a liquid version, easier to give to children and to patients who could not manage a tablet. Sulfanilamide is stubbornly insoluble in water and in alcohol. The company's chief chemist tried diethylene glycol, found that it dissolved the drug beautifully, added raspberry flavouring and pink colouring, tested the mixture for appearance, taste and fragrance, and sent it out.

It was not tested on an animal. There was no reason it should have been, in the sense that no rule and no professional expectation required it. The toxicity of diethylene glycol had in fact been described in the scientific literature before 1937, but nothing in the arrangement of the industry connected a published toxicology finding to a chemist choosing a solvent in a manufacturing plant. The clinical picture that followed — progressive kidney failure, anuria, death over roughly a week — was reconstructed afterwards by pathologists who matched the human cases against deliberately poisoned animals and confirmed that the solvent, not the drug, had done it 2.

The recovery operation is the part that best conveys how little the state actually had. The federal food and drug agency dispatched very nearly its entire field staff to chase down the shipment, bottle by bottle, from wholesalers and pharmacies and individual patients' kitchen shelves, because there was no recall power and no distribution record to consult. And when the case came to be prosecuted, the charge was misbranding — the product was called an elixir, and an elixir is by definition an alcoholic solution, which this was not 1. Had it been named a solution rather than an elixir, there would have been no violation at all. A hundred deaths, and the legal handle was a word on a label.

What 1938 actually changed

A federal food and drugs bill had been drifting through the United States Congress since 1933, opposed and amended and left to stall. The sulfanilamide deaths ended the stalling. The Federal Food, Drug, and Cosmetic Act was signed in June 1938, and it is the direct legal ancestor of nearly everything a pharmaceutical regulator does today 3.

Its central innovation was procedural rather than scientific. A manufacturer of a new drug now had to file an application with the agency before marketing, and that application had to contain evidence — reports of investigations, the composition, the manufacturing methods — showing the product was safe for its proposed use. If the agency did not object within a set period, the application took effect. It was a thin, permissive version of what would later become an approval process, but it inverted the burden of proof, and that inversion is the whole point. Before 1938, the state had to prove a marketed product harmful. After 1938, the maker had to offer evidence that it was safe before marketing at all.

Note what the 1938 Act did not require. It said nothing whatever about whether a drug worked. A manufacturer could satisfy the new law completely with a preparation that was demonstrably harmless and entirely useless, and for twenty-four years many did. The two halves of the modern question — is it safe, does it do anything — were separated by a generation, and the second half needed its own catastrophe.

The second catastrophe

Thalidomide reached the West German market in 1957 as a sedative, sold without prescription, and promoted as unusually safe — genuinely difficult to overdose on, which was true and, for a sedative in that era, a real selling point. It was taken for morning sickness. By 1961 obstetricians in Australia and paediatricians in Germany were independently noticing an unprecedented pattern of limb malformations in newborns, and the connection was made public in a short letter to a medical journal asking whether other readers had seen the same thing 4. The eventual count of affected children ran into the thousands across dozens of countries.

The United States was largely spared, and not through any strength of the 1938 framework. A medical officer at the agency, reviewing the American application, declined to let it through on the grounds that the chronic toxicity data were inadequate and that the absence of information about effects in pregnancy troubled her. She held the application through repeated resubmissions and considerable pressure from the sponsor until the malformation reports broke. The framework did not stop thalidomide in America. A person exercising judgement inside the framework did, and everybody involved understood the difference.

The amendments that followed in 1962 completed the sentence the 1938 Act had left half-finished. Manufacturers now had to show substantial evidence of effectiveness, drawn from adequate and well-controlled investigations, and not merely evidence of safety. Clinical testing was brought under a formal investigational scheme with reporting obligations. Consent from research subjects was written into the statute. And the effectiveness standard was applied backwards, to drugs already approved between 1938 and 1962, which triggered a review that eventually pulled hundreds of marketed products off the shelves. The historical literature on this period treats it as the clearest available case of a regulatory system being built out of its own failures rather than out of principle 5.

Where the research category sits in all this

Everything described above operates on a single pivot: intended use. The apparatus does not regulate molecules. It regulates what a molecule is offered for. A compound becomes a drug, in the legal sense, because somebody presents it as one — through a claim, a label, an indication, a route of administration meant for a person. That is why the 1937 prosecution turned on the word elixir, and it is why a regulator today can act against a product on the strength of its marketing alone, without any new chemistry.

The consequence is that the framework has an outside as well as an inside. Material supplied for laboratory investigation, offered with no therapeutic claim and no human indication, is not inside the approval structure, because the structure is triggered by the claim rather than by the substance. This is not a gap somebody discovered. It is a straightforward implication of building the whole edifice on intended use, and every jurisdiction that adopted the same pivot inherited the same boundary.

What that boundary does not confer is any of the assurance the inside of the structure provides. A compound outside the approval process has not been reviewed for safety, has not been reviewed for effectiveness, and carries none of the evidence that took eighty-odd years and two catastrophes to make mandatory. The whole point of the 1938 and 1962 machinery is that those assurances only exist where somebody has been compelled to produce them. Where nobody has been compelled, they simply are not there — which is a statement about the architecture, not an accusation about anyone operating within it.

The rest of this cluster follows individual rules back to their origins in the same spirit. Scheduling, compounding, warning notices, the research-use label itself: none of them was designed from first principles by somebody thinking clearly about the future. Each was a repair, made quickly, under pressure, to a specific thing that had already gone wrong. Knowing which accident produced which rule is the most reliable way to understand what that rule is actually for — and, just as usefully, what it was never built to do.