As Head of R&D at a Swiss cosmetics laboratory, I read a lot of the evidence behind the cosmetic actives that brands ask us to put in their formulas. The same pattern keeps turning up. A well-cited in-vitro mechanism gets marketed as if it were a proven clinical effect. Consumers now read ingredient lists closely, and regulators are starting to ask what dermatologists have asked for years: does it work on human skin, at the concentration in the product?
Why in-vitro data falls short
Cosmetic claims have mostly rested on two things: in-vitro assays, usually on cultured dermal fibroblasts, and data from raw material suppliers, who have a commercial stake in the results. Neither is dishonest. Cell culture shows that a molecule can act under laboratory conditions. It does not show that the molecule reaches the target cells at a useful concentration after topical application to intact skin. Among cosmetic actives, peptides, growth factors, and microbiome-targeting actives each sit at a different point on the evidence scale, so I will take them one at a time.
Peptides
A 2025 review found that most cosmetic peptides are backed only by in-vitro and ex-vivo data, and that there is no standard model or protocol for testing peptide activity, which makes it hard to compare one peptide’s evidence with another [1]. A systematic review of skincare peptides described strong mechanistic data for carrier peptides such as GHK-Cu, but weak human outcome data for cosmetic use, and several classes with no completed human trials [2]. Another review on skin senescence noted that most bioactivity studies use in-vitro assays because they are cheap, and that in-vivo trials are scarce [3]. Palmitoyl pentapeptide-4 is one of the few exceptions. It has a double-blind, placebo-controlled, split-face trial in 93 women behind it [4]. That one study gets cited across the industry mainly because so few peptides have anything comparable.
Growth factors
I expected this class to be weaker than peptides, but it has the better clinical record, including randomised, double-blind, split-face trials of topical formulations [5,6]. The effects, though, are modest. A 2023 systematic review reported median investigator-assessed improvement below 50% for skin texture, below 35% for fine lines and wrinkles, and below 20% for overall facial appearance. Participant-reported improvement was generally higher than what investigators graded [7]. Growth factors do work to a degree, but copy built on participant testimonials instead of blinded grading tends to describe a bigger result than the trials support.
Microbiome-targeting actives
This is the newest category and the evidence is the thinnest. A narrative review of 60 studies concluded that few have examined efficacy, safety, and mechanism specifically for skin [8]. A review of 14 clinical trials on well-characterised probiotic and postbiotic strains did report significant gains in skin moisture, elasticity, and wrinkle depth [9]. A 2026 review was more careful. It pointed to varied study designs, small samples, and no agreed criteria for what counts as a “microbiome-friendly” claim, and described the evidence as promising but exploratory [10]. That last gap is a regulatory problem as much as a scientific one, because a substantiation standard cannot be applied evenly to a claim with no fixed meaning.
What good substantiation looks like
For cosmetic actives, the practical step between cell culture and a full clinical trial is the ex-vivo human skin model. Excised skin keeps its barrier structure and metabolic activity, stays viable for up to two weeks, and has given results consistent with clinical data [11]. It does not replace a human trial, but it predicts real-world performance better than a cell monolayer.
The randomised, double-blind, split-face trial is the closest thing cosmetic dermatology has to a gold standard. Test and control products go on opposite sides of the same face, so differences in age, skin type, and environment cancel out within each subject. Instrumental measures make the readout less subjective than a questionnaire: transepidermal water loss, elasticity, and hydration can be quantified with calibrated devices under controlled conditions.
What the rules currently require
In the European Union, cosmetic claims fall under Article 20 of Regulation (EC) No 1223/2009 [12] and Commission Regulation (EU) No 655/2013 [13]. Six common criteria apply: legal compliance, truthfulness, evidential support, honesty, fairness and informed decision-making. Claims must be backed by adequate and verifiable evidence that is relevant to the finished product. The Responsible Person holds this evidence in the Product Information File.
The United Kingdom retained the same framework after Brexit. Great Britain uses its own notification system (SCPN) and requires a UK-based Responsible Person, but the six common criteria and the need for product-relevant evidence remain in force.
Switzerland aligns closely with the EU through the FDHA Ordinance on Cosmetics. A Product Information File and a safety assessment are required, yet there is no pre-market notification portal. Manufacturers and importers carry primary responsibility under a self-monitoring system.
In the United States the Modernization of Cosmetics Regulation Act (MoCRA) strengthened post-market duties. Facility registration, product listing and safety substantiation are now mandatory. There is still no EU-style pre-market safety report or common criteria regulation, but the direction is clear: regulators in all four markets expect claims to rest on evidence that matches the product as sold, not only on ingredient supplier data sheets.
Extrapolation and smaller laboratories
The failure I see most often is extrapolation, not fraud. A fibroblast assay shows higher collagen gene expression, and the product is then sold as “clinically proven anti-ageing,” even though culture concentrations, exposure times, and delivery conditions look nothing like a cream applied once or twice a day to intact skin.
Smaller formulation laboratories cannot fund a pharmaceutical-grade clinical programme, and that is a fair constraint. What I see changing is the middle tier. More labs are commissioning ex-vivo testing, and more are running small, randomised, instrumented split-face pilots instead of open-label studies where participants say their skin felt softer. None of this needs pharmaceutical-grade infrastructure. It needs substantiation to be planned at the start of formulation, not assembled afterwards to defend a claim that was written first.
Conclusion
Science itself is not the problem. The rigour behind cosmetic actives is unevenly spread. It is reasonable for growth factors, thin for most peptides, and early for microbiome actives, and marketing language rarely shows the difference. Consumers, and increasingly regulators, will want to know which of those categories a claim belongs to.
Author Bio
