Peptide Glow Journal

COSRX The Peptide Collagen Hydrogel Eye Patch: Four Peptides at Positions 31, 33, 34 and 39 of 41

The four peptides in the product name sit in the last quarter of a 41-ingredient list, below niacinamide at 4, adenosine at 19 and caffeine at 22. Two have no PubMed record, and one is a 99,369-dalton collagen.

Fiona G · Edited by Caroline S · Published 2026-09-18

Illustration: A shimmering hydrogel eye patch on a rose-beige ceramic surface in soft morning light.
Illustration

The name promises peptides and collagen. The declaration puts both behind a vitamin, a nucleoside and a stimulant. Niacinamide is fourth. The first peptide is thirty-first of forty-one.

Where each active sits

Pos. Ingredient (INCI) What it is PubMed, 2026-09-18
4 Niacinamide vitamin B3 45 RCTs (topical)
19 Adenosine nucleoside 44 records vs wrinkles
22 Caffeine methylxanthine 11 records vs eye-area endpoints
23 Collagen native protein, not hydrolysed
25 Cyanocobalamin vitamin B12
31 Acetyl Hexapeptide-8 Argireline 47 records
33 sh-Polypeptide-121 recombinant collagen XXI, 99,369 Da 0 records
34 Palmitoyl Tripeptide-5 Syn-Coll, 611.9 g/mol 4 records
39 Acetyl Tetrapeptide-5 492.5 g/mol 0 records

Full list read on the brand's own product page 2026-09-18; counts via NCBI E-utilities the same day; weights from PubChem. Positions 35 to 38 are four hyaluronic-acid forms, and citric acid and disodium EDTA close the list at 40 and 41.

The four peptides are the last actives named before the pH adjuster and the chelator. Our guide to reading a peptide ingredient list sets out what position can and cannot establish, and the short version applies cleanly here: everything below the 1% line is in discretionary order, so positions 31 and 39 do not mean one peptide outweighs the other — but the three actives listed above them are a fact about the formula, not an artefact of the rule.

The peptide with no literature

Acetyl tetrapeptide-5 is the second reason to read this list slowly. It is the peptide in the formula sold specifically into the puffy-under-eye problem, and on 2026-09-18 a PubMed title-or-abstract search returned zero records for it under its INCI name or its trade name.

That is not the same as a peptide with weak evidence. It is a defined molecule — PubChem holds it as CID 11620163, C₂₀H₂₈N₈O₇, 492.5 g/mol — with a registry entry, a supplier and no indexed published literature at all. Everything a reader might be told about what it does traces to material that is not in the indexed record.

There is a second observation in the same number. At 492.5 g/mol it is the only one of the four under the 500-dalton line that our peptides for skin census explains. The peptide most likely to cross the stratum corneum on size alone is the one with nothing published about what happens if it does.

The three ingredients above the peptides

Niacinamide, at position 4, carries 45 randomized controlled trials of topical use. It is the single largest published file in the patch and it is not in the product name.

Adenosine, at 19, returned 44 records joining it to wrinkles — more than any one peptide here. It is a nucleoside rather than a peptide, and in Korean formulation practice it is a long-standing wrinkle-claim ingredient.

Caffeine, at 22, is the ingredient with the most direct connection to the visible problem a patch is bought for. Searching caffeine against periorbital, dark-circle, eye-bag or puffiness terms returned 11 records on 2026-09-18, of which one carries the randomized-controlled-trial tag. That is a thin file in absolute terms, and still thicker than three of the four peptides.

Our peptide eye cream explainer counted the eye-area trial literature across the whole category and found seven randomized trials of a peptide around the eye against six for retinoids — an unusually even split by this site's standards, and a reminder that the eye area is genuinely under-studied for everything, not only for peptides.

One of the four peptides is a collagen

The product counts four peptides, and the count is arithmetically correct. One of the four is not a peptide in any sense a reader would recognise.

sh-Polypeptide-121 is sold under the trade name HumaColl21, and it is recombinant human collagen alpha-1(XXI): UniProt Q96P44, gene COL21A1, 957 amino acids, 99,369 daltons. The sh- prefix in an INCI name means synthesized human — a protein produced by recombinant expression rather than chemical synthesis — and it covers molecules of any size, including ones like this that are proteins rather than short chains.

It returns no PubMed records under its INCI name.

So the patch contains two collagen-scale molecules, one of them declared as collagen and one counted among the peptides. At 99,369 daltons, sh-Polypeptide-121 is roughly 200 times the 500-dalton line and about 265 times the weight of the smallest peptide in the same list. Listing a tripeptide and a 957-residue protein under one heading is what our ingredient-list guide means when it says the word peptide on a label is a category, not a size.

What "collagen" means on a patch

Position 23 reads Collagen, with no qualifier. Not hydrolysed collagen, not a collagen amino-acid blend, not a peptide fragment.

Native type I collagen is a triple helix of roughly 300,000 daltons — about 600 times the 500-dalton figure usually quoted as the ceiling for passive movement through intact skin. On the surface, a protein that size holds water and forms a film, which changes how skin looks and feels while the patch is on. That is a real cosmetic effect and it is not the same claim as the molecule reaching the dermis.

The distinction matters more here than on a cream, because a hydrogel patch is occlusive by design: it holds a water phase against skin for the wear time. Occlusion alone reduces transepidermal water loss and temporarily plumps the stratum corneum. A patch is an unusually good delivery format for hydration and an unusually hard format in which to attribute anything to one of forty-one ingredients.

The brand's numbers, as published

COSRX states "-25% reduction in puffiness" and "+17% improvement in firming around eyes", and attributes the testing to P&K Skin Research Center Co., Ltd, a contract testing laboratory.

Read on 2026-09-18, the page gives no participant count, no duration, no control arm and no measurement method. Two percentages with none of those four attached can be read for direction and not for size: a 25% reduction measured on ten people against their own baseline and a 25% reduction against a vehicle patch in a hundred people are different statements that look identical in a bullet point.

The attribution problem sits on top of it. Even a well-designed single-arm study of this patch could not separate niacinamide, adenosine, caffeine, four peptides, four hyaluronic-acid forms and an occlusive hydrogel from each other. That is a structural limit on any finished-product study, and it is the reason our own standard is a split-face design with a named comparator rather than a before-and-after.

Alongside the other COSRX formula on this site

Read next to our analysis of COSRX's 6 Peptide Skin Booster Serum, the two products use the same house vocabulary for different architectures. The serum counts its peptides in the name and puts them in a water-and-humectant base; the patch counts four peptides but leads its list with a B3 vitamin and delivers everything under an occlusive hydrogel. Neither page places one above the other, and an ingredient list is not the instrument that could.

What forty-one names in order actually show

They establish four ingredients counted as peptides in the final quarter of the declaration — one of them a 99,369-dalton recombinant collagen, and two of them with no indexed literature under any name; a vitamin, a nucleoside and a stimulant listed above all four, each with a larger published file; an unhydrolysed structural protein at position 23 that cannot pass intact skin and does not need to in order to hold water; and a marketing claim expressed in two percentages with no headcount behind either.

They do not establish what the patch does to an under-eye. We have not put one on a face, and a list read in order is a statement about a formula, never about a result.

Sources

  • COSRX, The Peptide Collagen Hydrogel Eye Patch — published ingredient list, product claims and testing attribution, cosrx.com, read 2026-09-18.
  • PubChem compound records read 2026-09-18: CID 11620163 (acetyl tetrapeptide-5), CID 11950477 (palmitoyl tripeptide-5).
  • UniProt entry Q96P44, collagen alpha-1(XXI) chain, gene COL21A1, 957 aa, 99,369 Da, read 2026-09-18.
  • PubMed record counts via NCBI E-utilities, run 2026-09-18.

Frequently asked questions

What peptides are in the COSRX Peptide Collagen Hydrogel Eye Patch?

Four: acetyl hexapeptide-8 at position 31, sh-polypeptide-121 at 33, palmitoyl tripeptide-5 at 34 and acetyl tetrapeptide-5 at 39, in a declared list of 41 ingredients. All four sit in the final quarter of the list, below niacinamide at 4, adenosine at 19 and caffeine at 22.

Does the ingredient order mean the peptides are at low concentration?

US and Korean declarations list ingredients in descending order of weight down to 1%, after which the order is at the manufacturer's discretion. Four peptides at positions 31, 33, 34 and 39 of 41 are therefore below 1% — as essentially every cosmetic peptide is, because that is the use level the raw materials are sold at. What the order does show is the relative position of the peptides against the other actives in the same formula, and here three ingredients with much larger research files are listed above all four.

What is acetyl tetrapeptide-5?

A four-amino-acid peptide, 492.5 g/mol, registered in PubChem as CID 11620163 and sold into the eye-care category on a de-puffing story. On 2026-09-18 a PubMed title-or-abstract search returned zero records for it under its INCI name or its trade name. It is a defined molecule with a registry entry and no indexed published literature, which is a different situation from a peptide whose trials are small or mixed.

What is sh-polypeptide-121?

A collagen, despite being counted as one of the four peptides. The 'sh-' prefix designates a synthesized-human protein made by recombinant expression, and sh-polypeptide-121 — sold under the trade name HumaColl21 — is human collagen alpha-1(XXI): UniProt Q96P44, gene COL21A1, 957 amino acids, 99,369 daltons. It returns no PubMed records under its INCI name. Neither the label nor the brand's page says which protein the number designates.

Is the collagen in these patches doing anything to skin?

The list names 'Collagen', not a hydrolysate. Native type I collagen is a triple-helical protein of roughly 300,000 daltons, around 600 times the 500-dalton figure usually quoted as the practical ceiling for passive penetration of intact skin. On the surface a protein of that size behaves as a film-forming humectant, which is a real effect on how skin looks and feels while it is in place. Whether the molecule itself reaches living tissue from a patch is not something the ingredient list can answer.

What does the -25% puffiness figure mean?

COSRX attributes it, with a +17% firming figure, to testing by P&K Skin Research Center Co., Ltd, a contract testing laboratory. As read on 2026-09-18 the page states no participant count, no duration, no control arm and no measurement method. Without a headcount and a comparison the direction of the number can be read but its size cannot, and with niacinamide, adenosine, caffeine and four peptides in one patch no figure could be attributed to any single ingredient.

Have you worn these patches?

No — nobody here has had one of these on their face. This page is a reading of COSRX's own declared list against the published record for each ingredient in it. A product assessment on this site means six weeks of split-face use by two testers under a fixed photographic standard, and that has not happened for these patches. What a hydrogel patch does to an actual under-eye is left open here.