The Hormone That Was Supposed to Cure Autism

In October 1998 a television programme told American viewers about a boy with autism whose parents said he had begun to speak after receiving a hormone during a diagnostic procedure, and within days the hormone, secretin, was reported to be almost impossible to obtain. Within a year a controlled trial in the New England Journal of Medicine found no difference from placebo, and by 2012 a Cochrane review of the randomised evidence concluded that it did not work 23. The whole arc, from a rumour among parents to a closed question, took about fourteen years. The first and most dramatic half of it took a few months.

This is not a discussion of any therapy. It is a case history of how a claim moves: from a tiny report, through the media, into a market and a shortage, into the trials that tested it, and out the other side. It is among the best-documented examples of its kind, which is why it deserves to be told in order.

A hormone with a clear job

Secretin has an honourable place in the history of physiology. In 1902 William Bayliss and Ernest Starling, working in London, showed that when acid from the stomach reaches the upper intestine, the intestine releases a chemical messenger into the blood that tells the pancreas to secrete 4. A few years later Starling proposed the word hormone for substances of this kind. Secretin was the first one.

A century later it was a peptide with a narrow and unglamorous clinical role. Because it stimulates the pancreas, a dose can be used in a hospital to test pancreatic function or to help with certain diagnostic procedures. That is how it came to be given to children in a hospital in the 1990s, and why a parent was in a position to notice something afterwards.

A diagnostic procedure and a mother's observation

The account that circulated in 1998 began with a child with autism and severe gastrointestinal symptoms who underwent an endoscopy, during which secretin was given as part of the work-up. The family reported that in the weeks afterwards the child's behaviour, eye contact and speech changed. The mother pursued the observation, and similar accounts were gathered from other families.

There is nothing implausible about a parent noticing a change, and nothing dishonest about reporting it. The difficulty is epistemic. Children with autism change over time with or without any intervention, the weeks after a hospital procedure are not a controlled period, and a family that hopes for improvement is the least well placed to be blind to it. A single child's course, however striking, cannot distinguish an effect from a coincidence.

Three children in a 1998 case report

The observation reached the literature in 1998 as a short report in a small journal, the Journal of the Association for Academic Minority Physicians, describing three children with autistic spectrum disorders who had received secretin for gastrointestinal investigation and were reported to show improved social and language skills afterwards 1. The title carried the claim in its first words.

As evidence, a report on three children without a comparison group occupies the lowest rung of the hierarchy. It generates a hypothesis. It does not test one. Nobody could say from it whether children without secretin would have improved just as much, whether the observers knew which children had been given the hormone, or whether three out of how many had been selected. Those are not criticisms of its authors so much as a description of what the format can bear.

Dateline, and a hormone that vanished

The step that changed everything was not scientific. In October 1998 the American television news magazine Dateline NBC broadcast a segment on the story, featuring a family who said their child had improved. Reporting from the period describes a flood of telephone calls to physicians and to the small number of firms that supplied the hormone, and a rapid scarcity of the product for families who now wanted it for their children.

What followed is a pattern familiar from other episodes: a demand created by media outpaced the supply of a product made in small quantities for a specialist diagnostic market. Parents who could not obtain the pharmaceutical product looked for substitutes, and the conditions for harm, even in a drug with a good diagnostic safety record, were created by the sheer scale of improvised access. The dates and details of individual supply failures are not set out here, since a figure from memory would be worse than none.

1999: a single-dose trial in the New England Journal

The scientific response was fast. In 1999 a team led by Sandler published in the New England Journal of Medicine a randomised, double-blind, placebo-controlled trial of a single dose of synthetic human secretin in children with autism or pervasive developmental disorder, under the title Lack of benefit of a single dose of synthetic human secretin 2. Sixty children took part. The report found no difference between the hormone and placebo on the measures it used.

A negative result carries its own caveats, and the trial was a single-dose design. Advocates of the idea responded, reasonably, that a single dose was not what had been described by families. Further trials followed, with repeated doses and different designs, and the pattern across them did not change. That is how the evidence collapsed: not by one decisive study but by a series of properly controlled attempts, none of which found what the early accounts had implied.

Another lesson sits in the details of the trials. A trial of this kind has to decide in advance what it will measure, and measures of social and language skill in young children are noisy: raters disagree, children have good days and bad ones, and a modest improvement in one scale may mean little. Blinding, randomisation and a pre-specified outcome are what turn that noise into something an outsider can trust, and they are exactly what the early reports lacked. The negative trials did not show that families had imagined what they saw. They showed that, once the comparison was fair, the hormone did not explain it.

A company's bet

Where there is demand there is commerce. A biotechnology firm acquired rights to develop secretin for the new purpose, and its prospects, and its share price, moved with the news from the trials. Its development programme did not produce a result that could carry a marketing application, and the programme ended without one. The details of that corporate history are matters of public record in the company's filings and are not reproduced here from memory.

The episode illustrates something that applies well beyond this hormone. A commercial interest in a positive answer is not evidence of bad faith, but it is a reason for readers to ask who ran each trial, who paid for it, and what a negative outcome would have cost.

Hope also has a clock. Families facing a developmental condition in early childhood feel the weight of every month, and the argument that waiting for proof means losing time is emotionally powerful and, in a case like this, impossible to answer with statistics alone. The reply that careful trials were the fastest route to a real answer, because they spared families from pursuing a dead end, was correct, but it was a harder message to deliver than the promise it was meant to temper.

2012: the Cochrane verdict

The file was closed by synthesis. In 2012 the Cochrane Collaboration published a systematic review of intravenous secretin for autism spectrum disorders, pooling the randomised, placebo-controlled trials that existed 3. It concluded that the evidence did not demonstrate effectiveness. A systematic review is the slow, unglamorous end of a controversy, and it carries weight because it counts all the evidence, including the studies that did not make the news.

What the episode shows about hope, hormones and evidence

The secretin story is sometimes told as a story about gullible parents. That reading is unfair and wrong. The people most affected were those with the most to lose and the fewest options, and their reaction to an apparent breakthrough was entirely human. The failures were elsewhere: in a media format that presented a hypothesis as a discovery, in the speed with which anecdote travelled compared with the months that a trial needs, and in the absence of any brake between a promising observation and a mass movement.

A reader meeting a new claim about a peptide can use the ladder in the table above as a checklist. Ask what the design was, how many people took part, whether there was a comparison group, and whether the finding has been repeated by someone with no stake in it. The broader pattern is discussed in the piece on why claims outrun evidence; secretin is the single clearest, best-documented case study of it.