DSIP: A 1977 Discovery With a Thin and Contradictory Evidence Record

The evidence for delta sleep-inducing peptide (DSIP) is thin and contradictory. It was isolated in 1977 as a nonapeptide that increased slow-wave EEG activity when infused into the brain ventricles of rabbits 1. Since then, no gene encoding it, no precursor protein and no receptor has been identified. A 2006 review concluded that its standing as a physiological sleep factor is poorly documented and still weak 2. The human data are a handful of small insomnia studies from the 1980s and early 1990s, and they do not agree.

The question here is narrow: what has been shown, in what system, and what has failed to replicate. DSIP makes a useful case study because its history is so long and so little has become settled. A compound studied for nearly fifty years without a molecular target is unusual. The reasons for that are more informative than any single result.

How it was found

The Basel group of Schoenenberger and Monnier took cerebral venous blood from rabbits kept asleep by electrical stimulation of the thalamus. They fractionated it and infused fractions into the brain ventricles of recipient rabbits. One fraction increased delta-wave and spindle activity on the EEG. The active peptide was sequenced as Trp-Ala-Gly-Gly-Asp-Ala-Ser-Gly-Glu. Synthetic material and several related peptides were then tested in 58 rabbits under double-blind conditions, including controls 1.

The design was careful for its time, and the result was a real observation in that model. Two features limit it. The evidence was electrophysiological: an EEG pattern, not a behavioural sleep measure. And it came from a central route that bypasses every barrier a peripherally administered peptide must cross. Both points matter later, because the human studies used intravenous administration.

The missing gene, precursor and receptor

Endogenous peptides normally have a traceable origin. They are encoded in a gene, cut from a larger precursor and recognised by a specific receptor. For DSIP none of these has been found. Kovalzon and Strekalova, reviewing three decades of work in 2006, identified that absence as the reason the link between DSIP and sleep was never characterised further. They called the sleep-factor hypothesis extremely poorly documented 2.

The same review makes a second point that is often missed. Sleep-promoting activity in later animal work was seen with structural analogues, not with the native peptide. The authors suggest that a DSIP-like molecule, not DSIP itself, may explain the effects attributed to it 2. If that is right, much of the literature has been measuring a related substance, or cross-reacting with one in immunoassays, and calling it DSIP.

What happens to it in blood

The molecule's stability is the second obstacle. In vitro, DSIP was degraded rapidly in human and rat plasma and serum, producing tryptophan-containing breakdown products at rates that depended on temperature and time. A phosphorylated analogue broke down more slowly and formed complexes, which the authors suggested could explain prolonged effects 3. For a nine-residue linear peptide with an unprotected N-terminal tryptophan, rapid clearance is expected.

This creates a chain of uncertainty. An intravenous study administers the native peptide. Most of it is likely broken down before it can reach the brain in intact form. Whether the fraction that does reach it, or a degradation product, or a complex, is responsible for any observed effect is not known. Unlike most modern peptide programmes, the DSIP literature contains no human pharmacokinetic study that resolves this.

The human studies disagree

The main human work came from Schneider-Helmert and colleagues in Switzerland. A 1983 report summarised five double-blind studies of intravenous DSIP. It described sleep onset after about an hour, normalised sleep in insomniacs after four consecutive injections, and improved alertness and stress tolerance while awake 4. A 1987 study in 14 middle-aged patients with severe chronic insomnia, placebo-controlled and double-blind over seven nights, reported that night-time sleep efficiency and daytime rest reached the levels of normal controls, with gains persisting after treatment 5.

Two other groups did not reproduce that picture. A 1987 double-blind crossover study from Monti and colleagues found fewer awakenings and more total and stage 2 sleep with DSIP. However, the differences from placebo were not statistically significant, baseline differences complicated the comparison, and the authors judged the changes clinically insignificant 6. A 1992 double-blind parallel-group study from Bes and colleagues in 16 chronic insomniacs found modest objective improvements in sleep efficiency and latency. The statistical support was weak, part of the difference may have come from an incidental change in the placebo group, and subjective sleep quality did not change 7.

The pattern matters more than any single row. The strongest positive results come from the originating group, and the independent studies are weaker or null. Every study is small, short and published before trial registration or consistent reporting standards existed. None was large enough to rule out a modest effect, and none was designed in a way that could establish one convincingly. That is a record which cannot support a conclusion in either direction.

A tension inside the positive results

The positive human reports contain an internal problem that deserves attention. The 1983 summary describes sleep beginning about an hour after a slow intravenous injection, with effects on the sleep-wake pattern lasting up to twenty hours 4. The in-vitro plasma data describe a native peptide that is broken down quickly in human blood 3. A short-lived molecule producing a delayed and prolonged effect is not impossible. It could act through a longer-lived intermediate, trigger a downstream process that outlasts it, or reach a protected compartment. But each of those explanations is itself a hypothesis, and none has been tested for DSIP.

There is a simpler explanation that the study designs cannot exclude. Chronic insomnia fluctuates from night to night, and patients recruited at a bad moment tend to improve on later nights regardless of treatment. That is regression to the mean. A small study that compares treatment nights with an unusually poor baseline will find improvement. A study that compares treatment with contemporaneous placebo nights will find a smaller effect or none. That is roughly the split in this literature: the reports that emphasise change from baseline are the most positive, and the placebo comparisons are the least 567. It does not prove that DSIP is inert. It shows why the positive reports cannot establish that it is not.

Why it is still studied

DSIP has survived as a research subject for several reasons unrelated to how strong its evidence is. It was an early candidate in the search for endogenous sleep factors, a field that did later produce firm results with other molecules. Immunoreactivity described as DSIP-like has been reported in many tissues, which kept the idea of a physiological role alive, even though, without a gene, such reports are hard to interpret 2. And the peptide is short, cheap to synthesise and unpatentable, so it has circulated widely outside any sponsored development programme. That last point explains why the gap was never closed. No one had a commercial reason to fund the pharmacokinetic and receptor work that would settle the question.

Regulatory position in 2026

DSIP, under the international non-proprietary name emideltide, is not approved as a medicine by the regulators of the United States, the United Kingdom or the European Union. In the United States the FDA sorts bulk substances nominated for compounding under section 503A into categories. Category 2 holds substances for which it has identified significant safety risks 8. On 15 April 2026 the agency announced that twelve peptides, including emideltide, would come off Category 2 after their nominations were withdrawn. Removal did not place them in Category 1 and does not in itself permit compounding 9.

Emideltide was one of seven peptides the Pharmacy Compounding Advisory Committee reviewed on 23 and 24 July 2026. The committee voted 6–7–1 against recommending it for the 503A bulks list, the only negative vote of the seven. The vote is advisory, and the FDA must still complete rulemaking 10. It is an assessment of compounding eligibility, not a finding on efficacy. But the one peptide the committee declined is also the one whose evidence record is least able to bear weight.

What would resolve it

1. Identification of a gene and precursor, or a demonstration that DSIP is a degradation fragment of a known protein and not a dedicated signalling peptide.
2. Identification of a receptor or binding site, with an antagonist that abolishes an EEG or sleep effect in animals.
3. Human pharmacokinetics showing how much intact peptide survives in plasma and whether any reaches the brain.
4. An assay specific enough to distinguish native DSIP from the DSIP-like immunoreactivity that fills the older literature.
5. A registered, adequately powered, placebo-controlled polysomnography trial run independently of the originating group.

Until then the accurate statement is the one the 2006 review made. DSIP produced a real EEG observation in rabbits in 1977 1. The molecular basis for that observation was never found 2, and the human studies that followed conflict with each other 567. It is an open question. Nothing in the record has turned it into an answer.