Oxytocin Is Approved as an Injection for the Uterus and Sold as a Spray for the Mind

01What it is

Oxytocin is nine amino acids. The American label for Pitocin calls it "a nonapeptide found in pituitary extracts from mammals" and standardises each millilitre to 10 units. It is dosed in international units, and the standard behind that unit is one freeze dried ampoule: the WHO fourth standard, code 76/575, holds about 21 micrograms and is assigned 12.5 units, or roughly 1.68 micrograms per unit. Every microgram figure below that started life as a unit figure is my arithmetic on that ratio.

It is also a near twin of the hormone that makes the kidneys hold on to water, which is where the label's worst warning comes from. CLINICAL PHARMACOLOGY says the two "differ in regard to only two of the eight amino acids", while the nonapeptide sentence, nine amino acids, is in DESCRIPTION. I report both as printed rather than reconciling them.

02What is approved, and by which route

Obstetrics, and nothing else: starting labour where there is a medical reason for it, reinforcing labour that has stalled, helping the uterus empty in incomplete or inevitable abortion, and controlling bleeding after delivery. Under a heading reading IMPORTANT NOTICE the same section rules out elective induction, where there is no medical indication, because the available data are inadequate. That is not a boxed warning; the version I read has no boxed warning section at all.

The route is narrower than the indication, and the route is the whole of this entry. "Intravenous infusion (drip method) is the only acceptable method of parenteral administration of Pitocin for the induction or stimulation of labor." After delivery the label also permits ten units into a muscle once the placenta is out. Nothing in it goes under the skin and nothing goes up the nose. Oxytocin does not last, either: a half-life in the blood of about 1 to 6 minutes, the uterus responding almost at once into a vein and settling inside an hour, or within 3 to 5 minutes into a muscle and persisting 2 to 3 hours.

I queried FDA's product database for the generic name oxytocin on 10 August 2026 and got three application records, all injectable, the marketed Pitocin application first approved on 19 November 1980. A query for a nasal route with oxytocin as the active ingredient returns one record, the discontinued Syntocinon at 40 units per millilitre. I could not retrieve its package insert, so I am not attributing an indication, a dose or a reason for discontinuation to it.

03What the autism trials found

The largest trial I found is SOARS-B, in the New England Journal of Medicine in 2021 (PMID 34644471). Of 355 screened, 290 were enrolled, 146 assigned to oxytocin and 144 to placebo, children and adolescents aged 3 to 17 with autism, 24 weeks, target 48 units a day by nose. The prespecified primary endpoint was a 13 item caregiver questionnaire on social withdrawal, higher scores meaning less social interaction. Both groups improved by almost the same amount, minus 3.7 against minus 3.5, a difference of minus 0.2 with uncertainty running from minus 1.5 to 1.0, and secondary outcomes generally did not differ.

The second largest I found ran in Japan across several centres, double blind, in 106 adult men with autism aged 18 to 48, 53 assigned to six weeks at 48 units a day and 53 to placebo (PMID 29955161). Its primary endpoint, a clinician rated measure of social reciprocity, fell from 8.5 to 7.7 on the drug and from 8.3 to 7.2 on the dummy spray, so the placebo group moved further and the two did not separate. Two secondary endpoints did separate, repetitive behaviour and gaze at socially relevant parts of a scene, and secondary results after a null primary are a hypothesis. The authors say they cannot recommend the treatment at that dose and duration for the core social symptoms in adult men.

Then the placebo problem. In an Australian trial's lead-in, 87 autistic children aged 3 to 12 all received a dummy spray for three weeks before anyone was randomised, and 42 of them, 48.3 percent, improved by 10 points or more on a social responsiveness questionnaire, the authors' own threshold for a meaningful response (PMID 41550040). Nearly half got better on nothing. It is also why the meta-analyses split. A 2021 pooling of 28 studies and 726 patients found a benefit while admitting single arm and uncontrolled work. A 2025 one restricted to randomised controlled trials, 12 of them covering 498 patients, found no significant effect on social impairment in its main analysis, with a benefit appearing only in the subset dosed at 48 units a day and in a model fitted across doses. That surviving signal sits at exactly the dose both large trials used, and it should not be left out of the summary.

04The trust result, and two attempts to reproduce it

The finding that made oxytocin famous was published in Nature in 2005 and was about trust in an economic game. I could not open its full text and am not reporting a sample size or a dose from it.

In 2020 a team including one of the original authors ran a registered replication, meaning the design and the analysis were locked in and accepted by the journal before any data existed, powered above 95 percent, reproducing the original's minimal social contact condition, in which participants see each other briefly without interacting (PMID 32514040). No effect on trusting behaviour. In 2026 a second registered report ran 211 more people, pooled them with 321 from the 2020 study for a combined 532, and found no evidence of an effect in either (PMID 41880977). A 2016 analysis of the field's statistics concluded these studies are generally underpowered, too small to find the effect they seek even when it is real.

05What it is sold for, and what was studied for those uses

For social anxiety, the closest I found is 25 people given 24 units alongside exposure therapy, randomised and double blind, who rated their own appearance and speech better as sessions went on without that changing how they ended up (PMID 19246160). For the sexual claim, the study most often pointed at is 29 couples at home. Twenty-four units changed none of the standard measures of sexual function. A statistical model did find more intense orgasm and more contentment afterwards, and the authors themselves describe those effect sizes as small to moderate (PMID 24503174).

Milk is the one with a proper trial behind it. Fewtrell and colleagues randomised 51 mothers of babies born before 35 weeks, 27 to spray and 24 to placebo, weighed the milk, and found total production no different, with most mothers believing they were on the active spray (PMID 16223754). The National Institutes of Health lactation database describes an earlier trial reporting 1964 mL against 510 mL over days 2 to 5, and that is a subgroup figure: it is among first-time mothers, the trial was stopped after only 8 first-time mothers had been studied, and among the 4 mothers in it who had given birth before there was no significant difference. The later, larger study that same record reports as finding no difference is the Fewtrell trial, not a third one.

Whose evidence this is matters as much as what it says. A 2011 systematic review counted 38 randomised trials between 1990 and 2010 and 1,529 participants across them, which is participants and not recipients of the drug, 79 percent male, and 8 percent of them people with developmental or mental health difficulties. What that record does not contain is the person buying a compounded spray for bonding or for sleep.

06Where 24 units came from

The standard research dose is a physical object rather than a calculation. In the methods of the study that measured it in spinal fluid, 24 units is six squirts of a commercial spray, three into each nostril, metered at 4 units each, made by Novartis in Switzerland. Nothing I opened establishes that the Swiss product and the discontinued American application are the same formulation, and I am not treating them as such.

I could not find the study that chose 24. The papers using it describe it as what everybody else uses, which is a convention and not a derivation, and a 2025 systematic review of dose against effect says the same from the other side: across studies using 1 to 48 units in healthy adults most administered 24, evidence either side of it is mixed, and the idea of a sweet spot in the middle rests largely on studies that never compared two doses in one experiment.

One dosing page carries a second route entirely, injection under the skin, in micrograms: 100 micrograms low, 200 to 300 moderate and 400 to 500 upper, once daily, with the page itself saying those ranges come from community and research-use protocols rather than human trial efficacy data. Reported, not recommended. Converting at 1.68 micrograms per unit, which is my arithmetic, that is roughly 60 to 300 units a day; the page's own conversion of about 2 micrograms per unit would make it 50 to 250. Against either, a 1 mL single-dose vial of the approved American injection holds 10 units. I could not find a human trial of oxytocin under the skin at those amounts.

07Whether the nose reaches the brain

This is the premise everything above rests on, and it is unresolved in the primary literature.

The bonding reputation starts with an infusion into a rodent's brain: in ovariectomised female prairie voles, oxytocin released into the brain ventricles produced a preference for the partner, while the same drug under the skin did nothing (PMID 7920590). So the question was always how to get it in. The founding human paper is not what it is usually taken for. Born and colleagues, Nature Neuroscience 2002, is the study the nose to brain case is built on, described in a later spinal fluid paper as the only previous report on the effects of intranasal administration of other neuropeptides on spinal fluid and plasma concentrations in humans, and it did not test oxytocin. It gave a melanocortin fragment, vasopressin and insulin, and found they reached the fluid around the brain and spinal cord within 30 minutes, bypassing the bloodstream.

Oxytocin itself was measured eleven years later, in a study its authors describe as the first to measure it in human spinal fluid after a nasal dose (PMID 24310737). They gave 24 units to 11 people and saline to 4. All 15 were men, and none was a healthy volunteer: they were neurology inpatients already having a needle put into the lower spine for diagnostic reasons. Two further patients were excluded, one on a diagnosis of multiple sclerosis and one over problems with the spinal fluid sampling, and the paper does not say whether its diagnosis breakdown describes the 15 who remained. Each gave one sample and one only, so the sampling times are three small groups rather than a curve through the same people. In blood the dose arrived fast and left fast, raised against baseline from 15 to 60 minutes, lower at 75 and 90 than at the peak, and indistinguishable from the dummy group by 90. In spinal fluid nothing measurable had happened at 45 or 60 minutes, and only the 75 minute samples rose above the dummy group, though not above the 60 minute samples. That rise was more than 60 percent above the placebo group's level, in three people, and there was no before-and-after baseline in spinal fluid at all, because each person gave one sample. Blood and spinal fluid did not track each other, which matters because blood oxytocin is what nearly every study and every sales page measures. Measurement stopped at 90 minutes in blood and 75 in spinal fluid.

The work since runs both ways. Six rhesus macaques given 80 units of labelled oxytocin, tagged so it could be told apart from the animal's own, showed no advantage of the nose over a vein (PMID 28289281). A human imaging study comparing a spray, a nebuliser built for the nose to brain region, and a drip found the change in blood flow to the amygdala explained entirely by the rise in the bloodstream, while saying in the same paper that the nasal route can target specific brain regions (PMID 32127545). That paper carries an Author Correction from 2022 which I could not read, so what it changes is unknown to me. A 2023 study closed the small blood vessels in the nose first, so a 24 unit dose could not raise blood levels, and most of the brainwave change went away (PMID 37185959). Pushing the other way, fluid sampled straight out of the hippocampus and amygdala of rats and mice rose 30 to 60 minutes after a nasal dose, read by the authors as a central route of transport (PMID 23579082). And the reviews split: Leng and Ludwig, both at the University of Edinburgh's Centre for Integrative Physiology, argue in Biological Psychiatry that very little of what is sprayed reaches spinal fluid while blood levels go above anything the body makes naturally, while a 2021 review in Molecular Psychiatry reads the same literature and reports converging evidence for direct nose to brain delivery.

08What could go wrong

The serious harms on the approved label are about giving it to a woman in labour: maternal deaths from high blood pressure episodes, bleeding around the brain and rupture of the womb, and fetal deaths, with oxytocic drugs injected for induction or augmentation. The harm that belongs to the molecule rather than the womb is water intoxication, and it has killed, with convulsions and coma during a slow infusion over 24 hours. The label attaches a number to that, 40 to 50 thousandths of a unit per minute infused for long periods, which is 2.4 to 3 units an hour sustained into a vein by my arithmetic, a different kind of exposure from six squirts of a spray. On long term risk the label's own answer is that the studies were not done: none on carcinogenicity or mutagenicity, and no information on fertility.

For the nasal route in research settings the record is unremarkable, and the 2011 review says so in three numbered points: no detectable subjective changes, no reliable side effects, and no association with adverse outcomes at 18 to 40 units for short term use in controlled research settings. It also notes three published case reports of adverse reactions from misuse and longer term use, which I did not retrieve and cannot describe.

The best powered posted numbers I found are SOARS-B's. Over 24 weeks, serious adverse events in 2 of the 146 on oxytocin against 1 of the 144 on placebo, other adverse events in 120 of 146 and 120 of 144, no deaths in either blinded arm. The commonest posted event in either arm was infection, 53 of 146 on oxytocin against 67 of 144 on placebo, then cough at 26 against 18; the behavioural events landed close to even, negativism 24 against 23 and anger or irritability 20 against 20, while the widest gaps were increased appetite at 23 against 14 and increased energy at 14 against 5. One death is posted, in the open label extension, 1 of 127 in the group titled "Oxytocin Nasal Spray: Open Label Phase", with no cause in the record I opened. What that participant received is not clear from the registry, because the intervention description says everyone in the open label phase received oxytocin while the posted table splits that phase into an oxytocin group and a placebo group.

09What nobody knows

Whether the nasal route delivers anything to the brain that a vein would not. The primate comparison and the human imaging study point one way, the rodent sampling and one review point the other, and the direct human measurement is eleven men with one sample each. Nothing I read connects the size of a spinal fluid change in one person to anything that person then did.

What a dose under the skin does. The circulating amounts convert to roughly 60 to 300 units a day, and I could not find a human trial at those amounts by that route.

What your own oxytocin level is. A 2013 evaluation found the commercial kits unreliable on unprocessed human samples, tagging molecules other than oxytocin.

What the discontinued American nasal product was approved for. FDA's database gives its route, strength, sponsor and marketing status and nothing else, and the monograph that might carry the original indications returned an access error to me.

10What the selling rests on

Almost every individual piece of oxytocin is real: a pharmacopoeial identity, a potency standard in an ampoule, an American approval in force since 1980, and a registered literature running to 920 records, including a 290 person double blinded trial in children that ran 24 weeks and reported its null straight. What does not transfer is the route. The approval is a drip, or a shot into a muscle, aimed at the uterus, and the drug is gone from the blood in minutes. The product is a spray aimed at the mind. Between the two sits a question the measurement papers have not closed, and the honest summary of the human side of it is three men at 75 minutes.

The sharper point is that the nasal route lost its approval and kept its uses. One compounding pharmacy lists a 40 unit per millilitre nasal spray under a five item string of uses that includes postpartum haemorrhage, a genuine approved indication for oxytocin by a vein or a muscle. Another lists 100 units per millilitre, describes one 0.1 millilitre metered spray per nostril as "approximately 10 IU total", and states on the same page that safety and effectiveness for the formulation have not been evaluated by FDA. Work that page's own arithmetic. At 100 units per millilitre a 0.1 millilitre spray is 10 units, and one spray per nostril is two sprays, which is 20.

Go and open the Pitocin label on DailyMed and read the section headed DOSAGE AND ADMINISTRATION. It names a vein, and it names a muscle. If you have bought a compounded oxytocin nasal spray, what strength is printed on your own bottle, and does the dose the pharmacy gave you divide into it?

Frequently asked

Is oxytocin nasal spray approved by the FDA?+

Not currently. Every oxytocin product I found listed under the generic name oxytocin in FDA's product database on 10 August 2026 is an injection: three application records, all dosage form injectable and route injection. There was once an approved American nasal oxytocin, application NDA012285, brand name Syntocinon at 40 USP units per millilitre, and its record is still in the database with marketing status Discontinued. What is sold now is compounded, meaning mixed to order by a pharmacy, and the compounding pages themselves say so: one states that safety and efficacy for its formulation have not been evaluated by FDA, and another that no compounded medications are reviewed by the FDA for safety or efficacy.

Does sniffing oxytocin actually reach the brain?+

That is still argued in the papers that measured it, and the direct human measurement is small. Striepens and colleagues gave 24 units by nose to 11 men and saline to 4, all of them neurology inpatients already having a needle put into the lower spine, and took one spinal fluid sample from each. Nothing measurable had happened at 45 or 60 minutes; only the three men sampled at 75 minutes were above the placebo group, by more than 60 percent, and blood and spinal fluid did not track each other. In six rhesus macaques given 80 units of labelled oxytocin, the nasal route offered no advantage over a vein. A human imaging study found amygdala blood flow changes explained entirely by the rise in the bloodstream, while stating in the same paper that the nasal route can target specific brain regions. Rodent work sampling fluid from brain tissue argues for a direct route, and two review teams reading the same literature reach opposite conclusions in print.

Does oxytocin help autism?+

The two largest randomised trials I found both missed their primary endpoint. SOARS-B randomised 290 children and adolescents aged 3 to 17, of whom 146 received nasal oxytocin at a target of 48 units a day for 24 weeks and 144 received placebo; on the prespecified caregiver questionnaire of social withdrawal both groups improved, minus 3.7 against minus 3.5, a difference of minus 0.2. A Japanese multicentre trial enrolled 106 adult men, of whom 53 received 48 units a day for six weeks, and its primary measure of social reciprocity improved slightly more on placebo. Two secondary endpoints in that trial did favour oxytocin, which after a null primary endpoint is a hypothesis rather than a finding. In a separate Australian trial, 42 of 87 autistic children improved by a clinically meaningful amount on a dummy spray during a three week lead-in, before anyone was randomised.

Does oxytocin increase trust?+

The 2005 Nature experiment that made the claim famous has been tested twice by preregistered replications, meaning the design and analysis were locked in and accepted by a journal before any data existed. A 2020 study including one of the original authors, powered above 95 percent, found no effect on trusting behaviour, with an after-the-fact analysis hinting at something in low trust individuals that the authors say needs confirming. A 2026 registered report ran 211 more people, pooled them with 321 from that 2020 study for a combined 532, and reported no evidence of an effect in either dataset. A 2015 review of the whole trust literature concluded the cumulative evidence is not robust, and a 2016 analysis concluded that intranasal oxytocin studies are generally underpowered and that most published findings probably do not represent true effects.

Where does the 24 unit dose come from?+

It is a spray bottle rather than a calculation, and it is reported here as the convention the studies used and not as a recommendation. The methods of the study that measured oxytocin in human spinal fluid describe 24 units as three puffs into each nostril, six in total, metered at 4 units each, from a commercial spray made by Novartis in Switzerland. I could not find the study that chose the number; papers using it describe it as the dose given in the majority of behavioural and neuroimaging studies, which is a description of a convention rather than a derivation. A 2025 systematic review of dose against effect reports that across studies using 1 to 48 units in healthy adults most administered 24, and that evidence for anything lower or higher is mixed.

What is known about the safety of intranasal oxytocin?+

In research settings the record is unremarkable and should be reported as such. A 2011 systematic review of 38 randomised trials concluded that intranasal oxytocin produces no detectable subjective changes, no reliable side effects and no association with adverse outcomes at 18 to 40 units for short term use in controlled research settings, while noting three published case reports of adverse reactions from misuse and longer term use that I did not retrieve. The best powered posted numbers I found are from SOARS-B: over 24 weeks, serious adverse events in 2 of the 146 children on oxytocin against 1 of the 144 on placebo, other adverse events in 120 of 146 and 120 of 144, and no deaths in either blinded arm. One death is posted in the open label extension that followed, 1 of 127, with no cause given in the record. The approved injectable label carries much heavier warnings, but they concern infusion into a vein during labour rather than a spray.

---

This content is for educational and informational purposes only and is not medical advice. Peptides discussed are research compounds and may not be approved for human use. Nothing here should be used to diagnose, treat, cure, or prevent any disease. Always consult a qualified healthcare professional before starting any peptide, supplement, or protocol. Individual responses vary. Do not self-administer compounds without proper medical supervision.