The things called ‘collagen boosters’ — what has been verified, class by class

They sit on the same shelf and so they look alike, but the materials that actually go into the body are entirely different. They divide broadly into three branches.

① Synthetic polymers — PLLA · PDLLA · PCL. These are biodegradable polymers long used in surgical sutures. They break down slowly inside the body, and in the course of that the surrounding tissue responds.

② Mineral — CaHA. Calcium hydroxylapatite is a microsphere of the same substance that makes up human bone and teeth. It goes in suspended in a gel.

③ Biologically derived — PN · hADM. PN (polynucleotide) is DNA fragments purified from salmon testes, and hADM is the extracellular matrix left behind when the cells are removed from donated human skin.

Of these Miso Clinic keeps the CaHA · PCL · PDLLA · PN · hADM classes. The reason we do not keep only one is written further down.

We will start with the figures.

The clinical trial released in March 2026 when the US FDA approved the décolletage (upper chest wrinkles) indication is this — multicentre · randomised 3:1 · evaluator-blinded, 152 subjects (116 treated, 36 control), followed for 84 weeks.

Response rate at 24 weeks — treatment group 71.2% (95% confidence interval 61.4–79.4) versus control group 6.3% (1.5–22.9). And at the 60-week point more than 60% still had their improvement maintained.

Treatment-related adverse events were 11.7% (16/137) and the most common was bruising, 5.1% (7/137). In this trial no nodules and no granulomas were reported.

Gathered together

Amiri M et al., Journal of Clinical Medicine 2024;13(6):1686 gathered 13 controlled trials · 1,171 subjects of CaHA (8 facial, 5 dorsum of the hand, 93.3% women, mean age 50.2).

• Global improvement (GAIS) 99.1%
• Wrinkle improvement CaHA 85% versus control 3% (p<0.0001)
• Compared with hyaluronic acid, superior 47% versus inferior 5% (p<0.0001)
• Nodules — 1–3 cases each in four facial studies, 7/112 (6.2%) on the dorsum of the hand, all mild
• The authors' own words: “No granulomas, allergic manifestations, ulceration or other serious adverse events were reported.”

What was verified in human tissue

Zerbinati N & Calligaro A, Clinical, Cosmetic and Investigational Dermatology 2018 placed CaHA in skin that was due to be excised surgically and took that skin two months later and looked at it under the microscope (5 subjects, the non-injected site on the same person serving as control).

Under polarised light microscopy and picrosirius red staining, newly formed thin type III collagen was significantly increased between the existing fibres (p<0.01). That this is data obtained by actually taking tissue from the human body is what gives this study its worth.

What this study verified is type III collagen. Data quantifying an increase in type I collagen in humans we were not able to verify.

PLLA is a class with two randomised controlled trials registered with the US FDA. Because these were submitted to a regulator, the design and the figures are in the public domain.

The 2023 cheek wrinkle indication (PMA P030050/S039)

Randomised 2:1 · evaluator-blinded · multicentre, 149 subjects (97 treated, 52 control), 12 months plus a 12-month extension.

In this trial injection-site nodules were 1/97 (1.0%), and masses likewise 1/97 (1.0%).

Why the story about nodules changed

PLLA is a class in which the early data reported a high rate of nodules. But that number came down a great deal over time, and the course of it is recorded in the literature.

As the practice of diluting it sufficiently took hold, 17.2% became 1.0%. That said, over that period the site · the depth of injection · the technique · the era all changed as well, so it cannot be concluded that this is the effect of the dilution volume alone. In a trial that divided subjects at random for comparison (Palm 2021), no difference between the results of the two methods appeared at the 48-week point.

In 2021 the manufacturer also formally permitted reconstitution immediately before use. The practice of leaving it to stand for a long time is supported in animal experiments (150 mice, the 72-hour group superior to the control group, p=0.016), and in human clinical studies a difference was not verified. We take the course of diluting it sufficiently within the range written on the label.

This class has one piece of data that no other class has. It is a biopsy in which the same person's skin was taken again four years later.

Kim JS, Aesthetic Surgery Journal 2019;39(12):NP484 — 13 Asian women (mean age 38.7), a PCL preparation diluted and injected into the dermis with a 31G automatic injector, with the untreated temple on the opposite side as the control.

Dermal thickness measured by ultrasound was also +21.3% ± 7.7%. Immunostaining for type I and type III collagen, Masson's trichrome and Herovici staining were carried out alongside.

That the dermis was still in a thickened state four years later is the heart of this data. A four-year follow-up in people, and one done by taking tissue at that, is rare in this field.

GOURI is a little different

Even within PCL, GOURI is a liquid form with no microspheres, while Ellansé is a suspension in which microspheres are held in a gel. Their physical properties differ, so the layer they go into and the way they spread differ.

The four-year biopsy above is a study carried out with the suspension form. Published human data on liquid PCL are still at the scale of case reports — in 2 cases reported from Australia (women aged 28 and 32, 2 sessions 4 weeks apart, 12 weeks of follow-up), the mean patient assessment was 4.0 and the operator assessment 3.0.

The liquid form and the suspension form are the same ingredient but not the same product. That is why in consultation we tell you first which form we are using.

PDLLA is of the same polylactide class as PLLA, but its molecular arrangement differs, making it a form that breaks down faster. The Korean product also carries non-cross-linked hyaluronic acid with it.

The human data have come out in the form of pilot studies.

• Seo et al. 2024 — 20 subjects, face
• Ma et al. 2024 — 10 subjects, dorsum of the hand, 3 sessions at 3-month intervals
• Lin et al. 2022 — 10 subjects, lower eyelid
• Seo et al. 2024 — 20 subjects, combined with CO₂ laser, stretch marks

In the animal data the direction is consistent. In a mouse model the gene expression of type I collagen (COL1A1) and type III collagen (COL3A1) increased, and an antioxidant factor (NRF2) and an anti-inflammatory factor (IL-10) rose alongside them.

In 2026 a study comparing PDLLA and PLLA on the left and right sides of the same person's face was published in Skin Research and Technology. We have not written the figures because we were not able to access the full text — we will amend this article as soon as we can verify them.

There is already a column dealing with Juvelook (PDLLA) on its own. If you would like to look at the ingredient · the approval · the evidence in more detail, see How far Juvelook (PDLLA) has been verified as a product.

PN is polynucleotide purified from salmon testes. That it is a molecule with the same backbone as human DNA, so that the body does not recognise it as foreign, is the starting point of this material.

Among the published human data, the one whose design is clear is this — Kim JH et al., Aesthetic Plastic Surgery 2022;46(4):1902. PN was injected into crow's feet and measured objectively with an Antera 3D camera (30 enrolled, 28 completed, 18 weeks).

• Crow's feet grade p<0.001
• On the 3D measurements, wrinkles · texture · pores · indentation · haemoglobin all p<0.05
• No significant change in melanin

What this study showed is that not only wrinkles but texture and pores as well, measured by machine, improved together. However, since the design compared before and after the procedure with no control group, it is not a trial divided against a placebo for comparison.

The authors of a review gathering the PN · PDRN class as a whole (Cureus 2026, 7 randomised trials · 183 subjects) wrote that “the samples are small and the protocols are inconsistent, so further randomised trials of adequate size are needed”.

The paper on Rejuran's phase 3 trial published in 2014 was retracted by the journal in 2016. The ground of retraction was not a problem with the data but an incorrect statement of the funding source and the conflicts of interest. That is why we have not used the figures from that paper in this article.

The five classes above are on the side of stimulating the body to make collagen. hADM goes the opposite way — it is on the side of putting the material that has been lost back in directly.

The cells and antigens are removed from donated human skin and what remains, a matrix made up of collagen · elastin · glycosaminoglycans, is made into fine particles and used. Material of the same class has long been used in burn treatment and in breast reconstruction surgery.

The human data

Lee JH et al., International Journal of Molecular Sciences 2026;27(5):2193 — randomised · split-face · double-blind, 20 subjects (mean age 54.7), 20 weeks. Particulate hADM was placed on one side and hyaluronic acid on the other, for comparison.

Skin density · volume · transepidermal water loss · wrinkle depth · pores · elasticity · pigmentation were all better than on the hyaluronic acid side (p<0.05). There were no serious adverse events.

In the laboratory data from the same study type I collagen · type III collagen · elastin all increased (p<0.05), and in a rat model fibroblast migration and an increase in density, together with new collagen on Herovici staining, were verified.

In China a 5-hospital · double-blind · multicentre randomised non-inferiority trial (175 completed) was carried out in the nasolabial folds, and against cross-linked collagen it was verified as not inferior — 88.4% versus 85.4% at 3 months and 70.9% versus 69.7% at 6 months.

The two trials above are data on the hADM class. Published clinical trial data on the individual products used in Korea (CellREDM · RITUO) themselves we were not able to verify. Nor is there any data comparing the two products directly. So we do not tell you that one is better than the other; we divide them by site and skin thickness — in thin places such as around the eyes we use the one with the finer particles.

One often hears it said that “boosters make collagen and fillers only fill”. But the material for which an increase in type I · type III collagen has been quantified most clearly by human biopsy is hyaluronic acid.

Wang F, Fisher GJ, Voorhees JJ et al., Archives of Dermatology 2007;143(2):155 — cross-linked hyaluronic acid was placed in photoaged skin and biopsies taken at 4 weeks · 13 weeks (11 subjects, mean age 74, normal saline as control).

• Increased collagen deposition (P<.05) around the filler
• Gene expression of both type I · type III procollagen increased (P<.05)
• Prolyl-4-hydroxylase increased (P<.05)
• Fibroblasts showed a mechanically stretched form

The mechanism is interesting. It is that as the gel physically pushed and stretched the dermis, the fibroblasts responded to that stimulus and began to make collagen. It touches on the point made earlier that the Korean approved wording is “physical restoration” — filling physically and collagen increasing are not opposites but continuous with each other.

So what we say in the consulting room is this — it is not a question to be divided into “booster or filler”, but a question of deciding first whether what has been lost in this face is thickness, volume or texture.

There is as yet no large randomised trial pitting the classes against one another with a clinical result as the primary endpoint. So “which is best” cannot be answered with data. Instead we look at what has been lost in the face and decide the class from that.

The reason we keep several classes is not superiority. It is that if you have only one thing, you end up explaining the face to fit that one thing.

If you are wondering whether a booster or lifting should come first, see Should I have a booster, or lifting, and for the number of sessions and how long it lasts, see How many sessions do I need and how long does it last.

We will add just three things.

First, the Korean approved wording records these products as bringing about “temporary improvement of facial wrinkles in adults through physical restoration”. Collagen production is not the scope the approval guarantees but the territory the research deals with. That is why we do not say things like “collagen increases by such and such a percentage”.

Second, “no adverse effects were reported” and “there are no adverse effects” are different. That a study of 20–30 people had zero adverse events also means that an uncommon reaction would not be caught at that size.

Third, the weight of the evidence for the number of sessions and the interval differs from class to class. For PLLA, “one to four sessions, typically three, at intervals of at least three weeks” is written in the approved instructions for use. The “three sessions” of the other classes is a number that came from each study's protocol, and a trial setting it against a different number of sessions for comparison we were not able to verify.

What we do in the consulting room is look at this table together. Once you have seen which cells are thick and which are empty, the decision becomes a great deal easier.

Written and reviewed by Lee Chi-Hak, MD, medical director of Miso Clinic in Daegu, South Korea. Every study cited above is given together with its design, its size and the limitations the authors themselves recorded, and where we could not find data, we have said that we could not find any.