DSIP clinical trials: what the human evidence actually shows

How many DSIP clinical trials are there?

Zero are registered. A ClinicalTrials.gov search under all four of the peptide's names returns no study at all. What exists is a scatter of small papers, with every human sleep study falling between 1981 and 1992, and across the intravenous studies FDA could identify in its 2026 review, the count came to 209 subjects.

DSIP stands for delta sleep-inducing peptide, and under its international non-proprietary name it is called emideltide. It was isolated from rabbit cerebral venous blood by a group in Basel in 1977 [15], and it was named for what it did to brainwave traces in rabbits given it by infusion directly into the brain, in an experiment that evaluated 58 rabbits including controls [1]. The name is therefore a description of a rabbit EEG result, not a finding about human sleep. Nearly fifty years later, the question people actually search for is whether the human trials ever followed.

The registry answer is blunt. A ClinicalTrials.gov search run for this article on 2026-09-19, covering "delta sleep-inducing peptide", "DSIP", "delta sleep inducing peptide" and "emideltide" across all fields, returned a total count of zero studies, and a second query on the non-proprietary name alone returned zero as well [17]. State that carefully, because it is not the same claim as "no trial was ever run". ClinicalTrials.gov is the United States registry, it carries many studies from elsewhere but it is not every national registry, and registration only became routine long after most of this work was published [17].

What exists instead is a small, old and almost entirely intravenous literature. Every human sleep study of DSIP was published between 1981 and 1992, most of them by one research group, and all of them are tiny. Human dosing did not stop there, it moved off sleep: two endocrine studies that found nothing, in 1993 and 1995 [21][22], an open opioid detoxification trial in 1998 [19], a report on intranasal administration in 2001 [20], and an anaesthesia study in 2009 that is covered further down [13]. When FDA reviewed the literature in 2026 it identified several studies that gave the peptide to humans by vein, and across those studies the count came to 209 subjects, at doses between 25 and 150 nmol/kg, for 1 to 15 days [16]. That is the human database FDA worked from, efficacy and safety together: a tally of the studies it identified, not a census of every person ever dosed. A review of the field in 1986, after nearly a decade of intense work, had already concluded that the physiological functions of DSIP and a possible mechanism of action remain to be established [14].

What the first human studies measured

The first human study gave six healthy volunteers a slow intravenous infusion of 25 nmol/kg in the morning and reported total sleep time up 59% inside a 130-minute window, with no classic sedation. A companion study in six chronic insomniacs found the effect only in the second hour after injection.

The first time anyone gave this peptide to a person was a pilot study in six healthy volunteers, four men and two women, run as a double-blind crossover, meaning each person received both the peptide and a dummy treatment in turn and nobody involved knew which was which at the time [2]. The dose was 25 nmol/kg, given as a slow infusion into a vein in the morning rather than at bedtime [2]. Sleep rose by 59% on median total sleep time inside a 130-minute window after treatment compared with placebo [2]. The authors then added the qualifier that almost never travels with that number: detailed behavioural and EEG analysis revealed no sedation in the classic pharmacologic way [2].

A companion paper the same year moved to patients, with six middle-aged chronic insomniacs and the same 25 nmol/kg by vein [3]. The result is more interesting than the summary usually suggests. Sleep-promoting effects appeared only in the second hour after injection, and in the first hour the authors recorded a slight arousing effect instead [3]. The sleep-enhancing capacity was visible for up to six hours of night sleep, with no daytime sedation reported [3].

Two more small studies fill out the positive side. An open study, meaning nobody was blinded and there was no control group at all, treated seven patients with severe insomnia with a series of ten injections and reported that sleep was normalised in all but one case, over follow-up periods of three to seven months [6]. And a summary paper by the group's lead author sets out the pattern they believed they were watching: single injections of 25 nmol/kg before sleep improving sleep in two separate studies, and repeated administration producing a build-up toward normalised sleep structure after four doses [5].

The two independent replications came out flat

Two groups outside Basel repeated the design and neither confirmed it. A Uruguayan crossover study at 25 nmol/kg concluded the sleep improvement was of little clinical significance. A Dutch parallel-group study in sixteen insomniacs found shorter sleep latency but called the effects weak and possibly an artefact of the placebo group.

Everything in the section above comes from one group of researchers. The test of any finding is whether somebody else can get it, and two teams tried.

In Montevideo, a team recorded polysomnograms, the full overnight sleep study with electrodes measuring brain, eye and muscle activity, in chronic insomniacs given 25 nmol/kg or placebo by vein across four nights in a double-blind crossover design [8]. Several measures moved in the hoped-for direction: fewer awakenings in the night, shorter latency to non-REM sleep, less total waking time [8]. None of it reached significance against baseline or against the double-blind placebo nights [8]. Where the DSIP-versus-placebo difference was significant, for non-REM sleep time and stage 2 sleep, the authors noted that the same differences already existed in the baseline values [8]. Their conclusion was that sleep improvement under DSIP treatment is of little clinical significance [8].

In the Netherlands, sixteen chronic insomniacs were split into matched pairs and assigned to parallel groups, half receiving 25 nmol/kg by vein and half a glucose solution, on the afternoons before the third, fourth and fifth nights in the laboratory [9]. This study did find something: higher sleep efficiency and shorter sleep latency on the objective recordings compared with placebo [9]. Then the authors looked harder at it and reported that the statistically significant effects were weak and in part could be due to an incidental change in the placebo group, that no other measure including subjective sleep quality changed at all, and that short-term treatment of chronic insomnia with DSIP is not likely to be of major therapeutic benefit [9].

FDA's 2026 review summarised the split in a single sentence: the studies authored by Schneider-Helmert reported improved sleep outcomes after short-term intravenous dosing, whereas the two double-blind placebo-controlled studies by Bes and by Monti, running similarly designed investigations, did not reach the same conclusions [16].

The most-quoted trial had no placebo arm

The 1987 study of fourteen chronic insomniacs is the one the vendor pages lean on, and it describes itself as placebo-controlled. FDA read the full paper in 2026 and found there was no placebo arm in it at all. The comparison was each patient's own baseline plus an outside group of healthy sleepers.

If one paper is carrying the weight on every page that sells this peptide, it is this one. Fourteen middle-aged chronic insomniacs received DSIP for seven successive nights under what the paper calls placebo-controlled, double-blind conditions, with polysomnograms at a placebo baseline, at the beginning and end of treatment, and on one post-treatment night [7]. Night sleep improved with the first dose and further with repeated doses, the gains held into the first post-treatment night, daytime alertness and performance rose, and the author concluded that the study demonstrates the efficacy of DSIP for impaired sleep and daytime function [7].

FDA obtained the full article for its 2026 evaluation and reported figures the abstract does not carry. Mean sleep onset latency, which is how long it takes to fall asleep once the lights go out, fell from 58.4 minutes at baseline to 38.9 minutes after the first injection and 27.6 minutes after the final one, across seven successive intravenous injections of 30 nmol/kg [16].

Then FDA said the thing that changes how the whole paper reads. Even though the authors described the study as placebo-controlled, there was no placebo arm in the study, so no placebo results were reported [16]. What the study actually compares is each patient against their own pre-treatment baseline, plus the group against an age and sex-matched external control group of healthy subjects who were given placebo injections to establish normal values [16]. That is a real design with real uses, and it is not a placebo comparison. FDA's own objection is that an externally controlled design carries inherent limitations, including imbalances in measured and unmeasured confounders [16].

Two claims are worth keeping apart here. The 1987 paper is not dishonest, and its numbers are very likely an accurate description of what happened to those fourteen people. What it cannot tell anyone is how much of the improvement was the peptide, because in a condition where expectation alone moves the measurements, nothing in this design isolates that.

Every trial used a slow drip, not a subcutaneous shot

Every human sleep study of DSIP delivered it into a vein, by slow infusion. FDA found no effectiveness or safety data at all for the subcutaneous route it evaluated in 2026. The published effect also depends on infusion speed: over one minute the peptide was completely ineffective, and optimal effects needed four to eight minutes.

This is the detail that matters most to anyone reading a vendor page, and it almost never appears on one. Every human sleep study described above used an intravenous infusion. FDA looked for anything on the subcutaneous route, an injection into the fat under the skin, which is the form it evaluated and the form these peptides are sold in, and reported that there was no study evaluating effectiveness for that route, and that clinical studies carrying safety information for it were not identified either [16]. It was only able to find data for the intravenous route [16].

Route is not a technicality here. The peptide has a plasma half-life of 8 minutes [16], so how fast it goes in largely decides what concentration anything downstream ever sees. The original group knew this and said so plainly: slow injection proved essential [4]. FDA's reading of their dose-response work puts numbers on that sentence. At an infusion time of 2.5 minutes, total sleep time rose by 30 minutes; at 7.5 minutes it rose by an hour; the peptide was completely ineffective if infused over 1 minute; and the authors reported optimal effects when the infusion was given over a range of 4 to 8 minutes [16].

Read that curve again, because it is strange. A compound that does nothing when pushed in over a minute, does something at two and a half, and does more at seven and a half is a compound whose reported effect is tied to a narrow intravenous delivery window. Nothing about that window survives the move to a subcutaneous injection, where absorption is slow and uncontrolled by design. The same group also described a latency to sleep induction of about an hour but a duration of effect of up to 20 hours [4], an unusual shape for a molecule cleared from blood in minutes, and one that nobody has explained since.

FDA flagged the route change as a safety question in its own right. Subcutaneous administration is generally associated with increased immunogenicity compared with intravenous, immunogenicity being the chance that your immune system starts producing antibodies against the injected peptide [16]. A nine-amino-acid peptide given by that route may therefore pose a significant risk of it, potentially amplified by aggregation and by peptide-related impurities [16]. One paper in the record does describe a different route, a 2001 report on intranasal administration [20], but PubMed carries no abstract for it, so what it measured cannot be checked from the record here.

The trials outside sleep, and what they were like

Sleep is not the only place DSIP was tested. Two unblinded withdrawal series treated 67 and then 107 inpatients, a pain pilot treated seven patients, narcolepsy rests on a single case report, and an anaesthesia trial found the peptide lightened anaesthetic depth instead of deepening it.

The largest human series in the DSIP literature is not about sleep at all. Working from the hypothesis that the peptide acts on opioid receptors, a Geneva group gave 25 nmol/kg by vein to 67 patients in withdrawal, 28 from alcohol and 39 from opiates, and reported a beneficial effect in 48 of the 49 patients they judged evaluable, with 27% of the original group lost or unsuitable for evaluation [10]. A follow-up extended the work to 107 inpatients, 47 in alcohol withdrawal and 60 in opiate withdrawal, with clinical symptoms disappearing or improving markedly and rapidly in 97% of the opiate patients and 87% of the alcohol patients [11].

Those are the biggest numbers anywhere in this literature and they come from its weakest design. Assessment of effect was based on the clinical evaluation by the physician and the nursing staff, with no blinding, no randomisation and no control group [11]. Without a control group, nothing in that design separates what the peptide did from what would have happened over the same days without it. FDA reached the same verdict, and recorded the safety side of it in full. No significant adverse events in 95 of the 107 subjects; nine with minor transient side effects such as perspiration, headaches, nausea and vertigo; and three subjects with serious adverse effects, two of whom experienced hypotension at the beginning of the first injection while the third had repetitive episodes of general discomfort with perspiration and nausea lasting 15 minutes [16]. One of those three received a second injection and experienced progressive hypotension, and FDA records no further information about them [16].

A third withdrawal study followed in 1998, and it is the one that shows what the idea looked like when somebody wrote it up more strictly [19]. FDA describes it as an open-label study in seven subjects meeting the DSM-4 diagnosis of opioid dependence, dosed at 35 nmol/kg intravenously on a fixed schedule, with mean withdrawal scores falling from 34.4 one hour before the first injection to 17.8 after it [16]. The same account then records what the headline leaves out: only two subjects received all the scheduled injections and detoxified on the regimen, and the recurrence of withdrawal symptoms was not suppressed after the second and following injections except in one subject [16]. The authors' own wording, as FDA quotes it, is that it remains an open question whether DSIP could be an effective alternative detoxification approach [16].

A pain pilot treated seven patients with migraine episodes, vasomotor headaches, chronic tinnitus or psychogenic pain attacks, comparing each patient's own history against a follow-up period, and reported significantly lower pain levels in six of the seven after infusions on five consecutive days followed by five more every 48 to 72 hours [12]. Narcolepsy, a neurological condition causing sudden overwhelming sleep attacks, is thinner still. FDA searched for evidence supporting that use, found that the nominators supplied no clinical references for it, and identified a single case report [16].

The one genuinely modern human study points the other way entirely. Anaesthetists gave twenty-four women either saline or one of three intravenous bolus doses of DSIP, at 25, 50 or 100 nmol/kg, before and during isoflurane anaesthesia, on the hypothesis that a natural hypnotic would deepen it [13]. Instead the peptide significantly reduced delta rhythm, the exact brainwave it is named after, while raising the bispectral index, a processed EEG number used to estimate how deeply anaesthetised somebody is [13]. It also raised heart rate and reduced heart rate variability [13]. The authors concluded that DSIP probably lightened the depth of anaesthesia [13].

What FDA concluded, and what the panel decided

FDA's 2026 briefing document concluded there is insufficient evidence on effectiveness for chronic insomnia, narcolepsy and opioid withdrawal, and that the substance is not well characterized. At the July advisory committee meeting, emideltide was the only one of seven peptides the panel voted down.

FDA received nominations for emideltide-related bulk drug substances for inclusion on the list of substances that may be used in compounding under section 503A of the Federal Food, Drug and Cosmetic Act [16]. Those nominations were later withdrawn, and FDA continued the evaluation at its own discretion [16]. What it evaluated was a compounded product for subcutaneous injection, for opioid withdrawal and two sleep disorders, chronic insomnia and narcolepsy [16]. The result was the first full regulatory reading this literature has ever had, published as a briefing document for the Pharmacy Compounding Advisory Committee meeting of July 23 and 24, 2026 [16].

Its conclusions are better quoted than paraphrased. On effectiveness, there is insufficient evidence to reach a conclusion for chronic insomnia, for narcolepsy or for opioid withdrawal by the intravenous route, and no data at all for the subcutaneous route under evaluation [16]. On the substance itself, emideltide free base is not well characterized from the physical and chemical standpoint [16]. Balancing every criterion together, FDA concluded that they weigh against the substances being placed on the 503A list [16].

One finding in that document cuts both ways and deserves saying plainly. A search of FAERS, the FDA adverse event reporting system, for emideltide through March 3, 2024 retrieved zero reports [16]. That is not a clean safety record. It is what a substance looks like when almost nobody is prescribing it through channels that report to FDA at all.

The committee declined to follow FDA on most of the agenda. Of seven peptides considered over the two days, it voted by narrow margins in favour of all but one, and emideltide was the only one to receive a down vote [18]. Reporting from the meeting noted that while the level of evidence varied by compound, all of them lack the kind of large, well-controlled trials in humans that establishing safety and effectiveness requires [18]. On DSIP specifically, a panel inclined to say yes said no.

Twenty years earlier, two researchers who had spent much of their careers on this molecule reached a compatible judgement from the science side rather than the regulatory one. The link between DSIP and sleep has never been further characterized, partly because no DSIP gene, protein or receptor has been isolated, and the hypothesis of DSIP as a sleep factor is extremely poorly documented and still weak [15]. Their own early experiments had found slow-wave sleep promotion from certain artificial structural analogues of DSIP, but not from DSIP itself [15].