Secretin for Autism: A Closed Question

The measured set

The number of randomized trials of secretin in autism is usually reported as a handful. It is not. A scoped census run against PubMed on 2026-09-02 — the query pairing secretin with autism, autistic and pervasive developmental — returns 132 indexed records, 22 under the clinical-trial publication type and 19 under the randomized-controlled-trial type ^[1]. The filter was proved live in the same run: the control term aspirin returned 6,115 records under the identical randomized filter and 8,906 under the clinical-trial filter, the invalid tag clinicaltrial[pt] returned zero for that same control term, and a deliberately invalid record identifier errored ^[1].

All 19 randomized-type records were retrieved and opened, because a count is not a reading. Sixteen are randomized trials of secretin in autism or pervasive developmental disorder. Of the remaining three, one is a published Comment on another paper, one is a trial in refractory schizophrenia rather than autism, and one is a dietary-intake study that carries the tag without testing the peptide ^[1]. Sixteen is therefore the figure this review argues from, and it is a measured figure rather than a selected one.

What the sixteen found

The sequence opens in the New England Journal of Medicine in 1999. Sixty children were randomized to a single dose of synthetic human secretin or placebo and 56 completed; secretin was not associated with significant improvement in any outcome measure, and the authors concluded that a single dose is not an effective treatment for autism or pervasive developmental disorder ^[2].

The obvious objections were then tested one at a time. Wrong dose: a randomized placebo-controlled trial of repeated doses of porcine secretin found no group differences on formal measures of language, cognition or autistic symptomatology, and no effect in subgroups defined by cognitive level, diarrhea or a history of regression ^[3]. A separate controlled study of multiple doses reported that they did not produce any symptomatic improvement ^[4]. Wrong preparation: a three-arm trial randomized 85 children to biologic porcine secretin, synthetic porcine secretin or placebo and found no amelioration beyond placebo from either preparation, whether or not the children had gastrointestinal problems ^[5]. Wrong route: a crossover trial of transdermal secretin in 15 children found no significant differences in speech, sociability, sensory or health scores ^[6].

The larger parallel-group trials agree. A 2000 trial randomized 95 children and found no significant differences in language or autistic behavior at three weeks ^[7]. A 2001 multisite crossover trial in 56 children reported a change in the primary Autism Diagnostic Observation Schedule social-communication score of -0.8 on placebo against -0.6 on secretin, t54 = 0.346, p < .73, and concluded that there was no evidence for efficacy ^[8]. A 42-child crossover trial of intravenous synthetic human secretin is titled for its result ^[9]. A 61-child trial concluded that a single intravenous dose is not effective in changing behavior and communication ^[10]. A trial of single-dose intravenous secretin in children found no significant differences at six weeks ^[11].

The most quietly devastating entry tested the mechanism of the belief rather than the drug. Under double-blind conditions, parents of young children with autism were unable to distinguish secretin from placebo: 27 families guessed their child's assignment correctly and 27 guessed incorrectly ^[12].

The two positive signals, stated in full

A negative literature is only trustworthy if it reports its own exceptions, and this one has two. In the single-dose intravenous trial, a marginally statistically significant improvement in autistic behaviors appeared in the treatment group at three weeks after injection, p = .051, which had disappeared by the six-week primary assessment; the authors describe the effect as transient and marginal ^[11]. In the transdermal crossover trial, improvement in speech reached p = 0.0479 among the subset of children not taking other medications ^[6].

Both are unadjusted, both are secondary to a primary endpoint that was missed, and one of them sits inside a fifteen-child study. Neither author team claimed more. Reported honestly, they are what a well-powered negative literature looks like when it is not suppressed, and they do not disturb the conclusion the other fifteen trials reached.

What secretin is actually for

The approved use is diagnostic and narrow. Given intravenously, secretin provokes a measurable pancreatic response, and controlled work established that synthetic porcine secretin performs accurately in pancreatic function testing in people with chronic pancreatitis ^[13]. Two products remain marketed on that basis: ChiRhoStim, synthetic human secretin, approved April 9, 2004, and Porcine Secretin, approved April 4, 2002; two earlier products are recorded as discontinued ^[14].

The distance between that indication and the one the trials tested is the ordinary distance between a diagnostic agent and a therapeutic claim. It is worth naming because the compound's fame ran entirely the other way.

Why this belongs in a regulatory index

The record has a live consequence. Secretin is not only an approved product; the bulk substance is separately nominated for pharmacy compounding, and the section 503A list published May 14, 2026 files "Secretin, human 99%" in Category 3, the category for substances nominated without adequate support to evaluate ^[15].

A reader who meets that line without the trial record behind it may reasonably infer that something is pending and promising. What is actually on file is a nomination the agency could not evaluate, attached to a molecule whose best-known therapeutic claim was tested sixteen times and did not survive once. The monograph carries the dated ledger, and Category 3 explains what that filing is and is not.

Elsewhere in this index