Oxytocin

Two literatures under one name

Oxytocin injection has been an approved prescription medicine in the United States for decades. Drugs@FDA lists NDA 018261, NDA 018248, NDA 018243 and two abbreviated applications as currently marketed prescription products, and every one of them is an injectable. The labelled indications are narrow: medically indicated induction of labour, stimulation or reinforcement of labour in selected cases of uterine inertia, adjunctive therapy in incomplete or inevitable abortion, and production of uterine contractions in the third stage of labour to control postpartum bleeding. The route is intravenous infusion or intramuscular injection. Elective induction is excluded by the label itself, in a notice printed above the indications.

A second literature runs under the same name and shares almost nothing with the first. It administers oxytocin as a nasal spray to healthy volunteers and to psychiatric and neurodevelopmental populations, and measures trust, gaze, social withdrawal and amygdala perfusion. No intranasal oxytocin product is currently approved and marketed in the United States; Drugs@FDA lists one nasal solution, NDA 012285, with a marketing status of Discontinued. This page does not establish the regulatory position in other jurisdictions. Intranasal oxytocin — the formulation the behavioural literature studies, and the formulation supplied as a research preparation — is an unapproved presentation of an approved molecule.

The two also differ in scale. ClinicalTrials.gov returns 921 registered studies with oxytocin as an intervention, 549 of them recorded as completed, which is an unusually large registered footprint for anything covered on this site. Forty-five name autism as a condition. The obstetric arm long ago answered its question and moved on to comparators and heat-stable formulations. The behavioural arm has run for two decades without settling its foundational claim. Neither literature transfers to the other, and secondary pages routinely borrow the credibility of the first to describe the second.

Identity, and the two residues shared with vasopressin

The identity data are unambiguous and cross-check against two independent authorities. PubChem CID 439302 gives a molecular formula of C43H66N12O12S2, a molecular weight of 1007.2 g/mol, CAS 50-56-6, UNII 1JQS135EYN and ChEMBL395429. The sequence is Cys-Tyr-Ile-Gln-Asn-Cys-Pro-Leu-Gly-NH2 — nine residues, closed by a disulfide bridge between positions 1 and 6, terminating in a glycinamide. The approved product label independently states the same empirical formula and a molecular weight of 1007.19. This is one of the few compounds on this site where a regulatory document and a chemical database can be checked against each other and agree.

Among the synonyms PubChem carries is 3-isoleucine-8-leucine vasopressin. That name encodes the relationship: oxytocin and arginine vasopressin differ at two of nine residues. The approved label states the same relationship but counts the peptide as eight amino acids rather than nine — NDA 018261 attributes the drug's pressor and antidiuretic properties to the fact that oxytocin and vasopressin differ in regard to only two of the eight amino acids. This page reports both figures and does not reconcile them.

The two currently marketed approved labels then disagree about what follows from that similarity. NDA 018261 states that oxytocin even in its pure form has inherent pressor and antidiuretic properties which may become manifest when large doses are administered, and its adverse reactions section records severe water intoxication with convulsions and coma associated with a slow infusion over a 24-hour period, together with maternal death attributed to oxytocin-induced water intoxication. NDA 018248 states that synthetic oxytocin does not possess the cardiovascular effects, such as elevation of blood pressure, as exhibited by vasopressin found in posterior pituitary injection. Both labels carry an intrinsic antidiuretic statement in their precautions. This page does not reconcile them.

The trust finding, and what replication did to it

Kosfeld and colleagues reported in Nature in 2005 that intranasal oxytocin caused a substantial increase in the transfers investors made in a trust game, and that the effect was not a general increase in willingness to bear risk but was specific to social risk. PubMed types the paper as a controlled clinical trial. It is among the most cited results in behavioural neuroendocrinology, and nearly everything written about oxytocin outside obstetrics descends from it, directly or at one remove.

Nave, Camerer and McCullough reviewed the field in 2015 and concluded that the simplest promising finding — intranasal oxytocin associated with higher trust — had not replicated well, and that plasma oxytocin measurement was methodologically compromised. Five years later Declerck and colleagues, working with Fehr, an author of the original, ran a registered replication powered above 95 per cent that reproduced the original's minimal-social-contact condition. They found no effect of oxytocin on trusting behaviour in that condition. An exploratory post hoc analysis suggested an effect in participants with a low disposition to trust under the no-contact condition, which the authors explicitly flag as requiring confirmation.

What that sequence establishes is narrower than either camp usually states. The 2005 result carries no retraction and no expression of concern. PubMed links two comment articles to it, neither a critique — a Nature News and Views, "Human behaviour: brain trust" (PMID 15931200), and a Revue Médicale Suisse essay (PMID 16044807). A well-powered replication run with an author of the original did not recover the effect under the conditions the original specified. That replication bounds the conditions under which the effect has been observed; it does not withdraw the original report, and both facts belong in the record.

Whether an intranasal dose reaches the brain

Striepens and colleagues sampled blood and cerebrospinal fluid in the same subjects after intranasal administration of 24 IU, with eleven receiving oxytocin and four placebo. Concentrations rose significantly in both compartments. Plasma peaked at fifteen minutes and decreased after seventy-five; cerebrospinal fluid took up to seventy-five minutes to reach a significant level. The correlation between the two was below 0.10. A rise in the cerebrospinal compartment is therefore demonstrated in humans, and the plasma measurement that most behavioural studies rely on does not predict it.

Leng and Ludwig worked through seven studies reporting cerebrospinal measurements after intranasal dosing and concluded that at most 0.005 per cent of an administered dose entered that compartment, while peripheral concentrations reached supraphysiological levels acting on multiple organ systems. Martins and colleagues then compared standard spray, nebuliser and intravenous routes against regional cerebral blood flow, and reported that oxytocin-induced decreases in amygdala perfusion were explained entirely by the rise in systemic circulation, while still finding evidence that the intranasal route can engage specific regions. That paper carries a 2022 author correction to its area-under-curve calculations, which the authors state did not change the pattern of results.

Sources disagree here, and averaging them would misrepresent both. Quintana and colleagues reviewed the same territory in 2021 and argued that converging human and animal evidence supports a functionally relevant nose-to-brain route underlying the behavioural effects. Leng, Leng and Ludwig returned in 2022 with a citation-network analysis of the social-peptide claim and concluded that if the most highly cited papers in each segment of the evidence represent the best available, then the evidence is weak. These are two review-level readings of an overlapping literature, and neither is treated here as a measurement.

Where the large trials landed

The largest randomised trial of intranasal oxytocin in autism was registered as NCT01944046 and reported by Sikich and colleagues in the New England Journal of Medicine in 2021. It enrolled 290 children and adolescents, ran twenty-four weeks, and used the Aberrant Behavior Checklist modified Social Withdrawal subscale as its primary outcome. Least-squares mean change from baseline was minus 3.7 in the oxytocin group and minus 3.5 with placebo, P equal to 0.61. Adverse-event profiles were similar between groups. The trial did not separate treatment from placebo on its primary endpoint.

Two meta-analyses published within a year of each other read the wider trial set differently. Audunsdottir and colleagues preregistered their analysis, found a summary effect of d equal to 0.22 for social outcomes at p below 0.001 and d equal to 0.14 for routinised behaviours at p equal to 0.22, and reported that frequentist and Bayesian assessments both indicated the summary effects might be inflated by publication bias. Zhang and colleagues pooled twelve randomised trials and 498 participants, found no significant overall effect on social impairment, and recovered significance only within a dose-response subgroup at the highest daily amounts. Zhang states that pooled set; Audunsdottir does not state a trial count in the abstract, so the degree of overlap between the two sets cannot be established from the published abstracts.

The pattern recurs outside autism. A phase 2 trial of an intranasal oxytocin product in chronic migraine, NCT05679908, randomised 88 patients across placebo and two active arms, completed in October 2023 and posted results. Least-squares mean change in monthly migraine headache days was minus 8.17 on placebo, minus 7.78 on the once-daily active arm and minus 5.77 on the twice-daily arm. The placebo arm produced the largest reduction of the three. The result was posted to the registry, and no journal publication of it was located.

The receptor turned out to be less decisive than assumed

Receptor genetics complicated the mechanism after the human trials were already running. Ferguson and colleagues reported in 2000 that male mice lacking the oxytocin gene failed to develop social memory, while olfactory detection of non-social stimuli, spatial memory and startle habituation remained intact, and that treatment with oxytocin but not vasopressin restored the deficit. An oxytocin antagonist produced a social-amnesia-like effect in wild-type animals. That mouse result, and pharmacological blockade work in prairie voles, supplied most of the mechanistic framing that the human intranasal programme inherited.

Berendzen and colleagues used CRISPR mutagenesis to generate three independent oxytocin-receptor-null prairie vole lines and reported in Neuron in 2023 that the mutants still showed a behavioural preference for a mating partner over a stranger across different experimental setups, that mothers lacking the receptor delivered viable pups, and that parents cared for their young to weaning. Pharmacological antagonism in the same species had supported the opposite conclusion for years. Genetic ablation and acute pharmacological blockade produced opposite answers to the same question in the same species.

Measurement, and the cold chain

Every peripheral oxytocin figure in the older literature depends on an assay that Szeto and colleagues examined head-on. Measuring 39 human plasma samples, they found that unextracted plasma read by enzyme immunoassay was more than a hundredfold higher than extracted plasma from the same samples, with a Spearman correlation of minus 0.10 at p equal to 0.54. Radioimmunoassay left more than ninety per cent of samples below the limit of detection with or without extraction. Chromatographic fractionation of the extracts recovered multiple immunoreactive products besides oxytocin. Their conclusion was that changes in degradation products probably contribute to previously reported responses to social and behavioural challenge.

Saliva carries an artefact of its own. Van IJzendoorn and colleagues gave 46 women 16 IU, 24 IU or placebo intranasally and sampled saliva hourly for seven hours. Salivary concentrations peaked around one hour and were still significantly elevated at seven, and the two amounts did not differ from each other. MacLean and colleagues validated salivary assays in domestic dogs and reported canine salivary concentrations much higher than those typically detected in humans, with results varying by swab type and salivary stimulation. What a salivary figure indexes is taken up in the unsourced record below.

The published chemistry is specific about handling. Collins and colleagues describe oxytocin as having limited shelf life in aqueous solution, particularly above 25 degrees Celsius, with injectable aqueous formulations requiring refrigeration below 8 degrees. The CHAMPION trial, which randomised 29,645 women across 23 sites in ten countries to intramuscular heat-stable carbetocin at 100 micrograms or oxytocin at 10 IU after vaginal birth, kept both drugs at 2 to 8 degrees purely to preserve blinding — the trial existed because oxytocin's cold-chain requirement is a practical obstacle in hot climates. Blood loss of at least 500 ml or use of additional uterotonics occurred in 14.4 per cent of the oxytocin group.

What is not known

The approved evidence covers uterine contraction and nothing else. No labelled indication, and no adequately powered trial, addresses social cognition, mood, bonding, attachment, anxiety or sexual function, and the largest randomised trial in autism did not separate from placebo on its primary endpoint. Whether intranasal administration produces functionally relevant central concentrations remains contested at review level, with Leng and Ludwig reading the published cerebrospinal studies as showing only a negligible fraction of an administered dose reaching that compartment and Quintana and colleagues reading the same literature as supporting a nose-to-brain route; Striepens showed a real cerebrospinal rise but no correlation with the plasma measurement most studies use as a proxy, so a given study's plasma or salivary number cannot be interpreted as a central exposure. Dose-response is unresolved. 16 IU and 24 IU produced indistinguishable salivary curves, the largest autism trial used 48 IU daily without benefit, and one meta-analysis recovers significance only at the highest amounts while a preregistered one reports that its own positive social estimate might be inflated by publication bias. Human pharmacokinetics exist only for the intravenous route in small samples, where the label states a plasma half-life of about 1 to 6 minutes and a 1980 infusion study in 8 men measured 10.3 minutes with a range from 5.3 to 17.3; these are not reconciled anywhere located. The two currently marketed approved labels also disagree with each other about whether the synthetic product carries vasopressin-like cardiovascular effects, and nothing located resolves that. Peripheral assay validity is unresolved, and it contaminates every observational association built on immunoassay before extraction became standard. Receptor genetics do not support the mechanistic story the human programme inherited: three independent oxytocin-receptor-null prairie vole lines still formed partner preferences, delivered viable pups and reared them. PubMed lists 23 papers with oxytocin in the title carrying a retracted-publication or expression-of-concern type, none of them among the papers cited here, which indicates a field with an integrity tail that any reader working outward from this page will meet. Nothing in the published record addresses the safety, purity or handling of unregulated preparations supplied outside the approved supply chain.

Questions