Imagine a molecule pipetted straight onto cultured human fibroblasts. Expression of a collagen-related gene increases. The result travels through a product-development cycle and returns as a claim: the ingredient has been shown to support collagen in human skin cells.
A change in gene expression has already become a broader statement about collagen. Put that statement with a serum’s promise of firmer skin, and the user is being invited to make another leap. The experiment has measured neither the delivery of that serum nor its effect on a face.
In the dish, the cells received a known dose for a known length of time. Without having to survive the formula, leave the vehicle, or cross the outer barrier. It met its target directly, with no obstacles in between.
This makes cell culture evidence suited to a narrower and often important question: does this molecule alter this cell under these conditions? Problems begin when biological capability is presented as if formulation, delivery and an outcome in your skin came free with it.
The route through the barrier deserves just as much attention as the pathway at the end of it. A result in cells starts an investigation. How much of the journey it establishes depends on what the experiment actually tested.
The Target Has Already Been Reached
Cell culture is powerful partly because it removes context. Researchers control the concentration, exposure time and surrounding medium, then compare treated and untreated cells without an entire tissue changing ten things at once. Isolating pathways that are nearly impossible to track on a living face is exactly where a well-controlled cell study excels.
The context removed from the experiment is also part of the biology. A fibroblast in a dish has no epidermis above it and none of the blood vessels or mechanical forces of living tissue around it. Its extracellular matrix, the protein scaffold cells normally live within, is arranged very differently on standard culture plastic. A keratinocyte grown alone lacks neighbouring cells whose signals alter its response. The cells are human, but that doesn’t restore the human skin around them.
The missing journey is considerable, particularly for a dermal target. A molecule applied at the surface has to leave its vehicle, pass through or around the stratum corneum and reach the dermis at an effective concentration. Degradation, retention in other layers and clearance all affect that exposure. The relevant journey depends on the target: an ingredient acting at the surface does not need to reach a fibroblast.
The conclusion from an isolated-cell experiment is therefore conditional. At the tested concentration and exposure time, the molecule produced the measured response. To connect that result to a topical product, the evidence must also address whether the intended tissue receives a sufficient exposure.
In vitro means that the experiment took place outside a living organism. It covers models with varying levels of structure. A well-designed cell study is strong evidence for establishing mechanism. Penetration, tissue exposure and a clinical change remain further questions unless the study measured them.
The cells are human, but that doesn’t restore the human skin around them.
In Vitro Is Not One Model
Researchers add context back in several ways. Two cell types are grown together. Keratinocytes are organised into reconstructed human epidermis. Fibroblasts are included in fuller skin equivalents. Excised human skin retains its native barrier and tissue arrangement for a limited experiment. Each model restores something, and each leaves something out. They do not form one tidy league table.
The protocol is part of the test. Exposure conditions, controls and the timing of the viability measurement form part of the validated irritation method. Tissue batches also have to meet defined quality criteria. The validation applies to that procedure and its endpoint. [1]
A model is validated for a defined method and endpoint, not awarded a general certificate for being sufficiently skin-like. The Organisation for Economic Co-operation and Development (OECD) publishes internationally agreed test methods for chemicals, and its 2025 Test Guideline 439 uses reconstructed human epidermis to identify irritant chemicals through loss of tissue viability. Cells convert the dye MTT into a coloured product that provides the viability readout. The method answers a question about irritation hazard, within its stated limits. It does not validate the same tissue for measuring the absorption of another molecule. [1]
The OECD’s skin-absorption guideline asks a different question. A skin sample separates donor and receptor chambers while researchers measure the movement of a surface-applied substance into and through it. [2] The model has to suit the claim, including the concentration and tissue context. Validation for irritation does not establish performance in a penetration experiment.
More tissue makes an experiment more representative in one respect and less controllable in another. Excised skin brings back the native permeability barrier, but both circulation and immune recruitment are absent. A human trial tests use on living skin, yet the mechanism becomes harder to isolate. The question is whether the model can carry the conclusion being asked of it.
The Model Was Skin-Like Until the Molecule Changed
Fritz Schmook, Josef Meingassner and Andreas Billich made that dependence unusually visible in 2001. They compared human, pig and rat skin with two reconstructed human models using four topical drugs with different chemical properties: salicylic acid, hydrocortisone, clotrimazole and terbinafine. [3]
Excised human and animal skin and reconstructed human tissues. The researchers compared four drugs across human, pig and rat skin and two reconstructed models. The large clotrimazole differences concern Graftskin under the conditions tested in 2001. They do not describe every reconstructed model now in use. [3]
One reconstructed model, Graftskin LSE, provided an adequate barrier to salicylic acid. Through the model, measured flux from the more hydrophobic clotrimazole was about 900 times higher than through human skin, while the concentration retained in the model was about 50 times higher. [3] Flux describes movement through the tissue per unit area and time. Retention describes what remains within it. Both had changed, by different amounts.
The model had not become less human between assays. Its barrier performance depended on the compound. A result that looked adequate for one permeability question gave a very different answer for another. Pig skin came closest to human skin in the same experiments, within twofold on flux and fourfold on skin concentration. The authors’ own conclusion was that currently available reconstituted models cannot be regarded as generally useful for in-vitro penetration studies. [3]
Reconstructed human epidermis. Clotrimazole creams were compared using EPISKIN tissue, with different-strength controls to test the method’s ability to separate responses. The comparison concerned permeation profiles under a shared method, rather than an absolute prediction of delivery through human skin. [4]
Now, those figures describe the models and conditions tested twenty-five years ago. A newer clotrimazole study shows why that qualification is necessary. In 2022, Ana Sofia Agonia and colleagues used reconstructed epidermis to compare creams, including controls with different drug strengths. The model was more permeable than human skin, yet the method did distiguish the different drug strengths under the tested conditions. [4]
Their question concerned relative product performance under a shared method. It did not require the model’s absolute permeability to match human skin. The experiment supports the comparison without establishing the extent of permeation through human skin. The same model characteristic has different consequences depending on the conclusion being sought.
Sometimes the Missing Context Hides the Effect
If simplified models only exaggerated what ingredients could do, this would be an easier article to write. Sometimes the missing context contains the mechanism, and the effect disappears when the biology is taken apart.
Cultured cells and reconstructed epidermis. Pigment production and transfer were different experimental questions. The transfer result came from melanocytes grown with keratinocytes; the isolated-melanocyte experiment measured pigment production. The reconstructed tissue supplied another level of cellular context. [5]
Tomohiro Hakozaki and colleagues followed niacinamide through several models in 2002. When models were cultured alone, niacinamide did not inhibit purified mushroom tyrosinase, an enzyme that drives pigment production, or reduce pigment production in melanocytes, the pigment-making cells. In contrast, when melanocytes and keratinocytes were grown together niacinamide cut melanosome transfer by 35 to 68 percent. Melanosomes being the packets of pigment melanocytes pass to their neighbours. Interestingly, pigmentation as a whole also fell in reconstructed epidermis. [5]
Pigment synthesis within an isolated melanocyte and pigment transfer to a neighbouring keratinocyte are different processes. The first experiments had not tested the interaction where the researchers found an effect.
Human pigmentation study. Eighteen participants used 5 per cent niacinamide moisturiser and its vehicle in a paired comparison. Pigmentation outcomes were assessed from facial images. This tests benefit under the study conditions; it does not identify the cellular mechanism by itself. [5]
What the vehicle controls. The vehicle is the comparison formulation without added niacinamide. It helps test whether niacinamide contributes beyond the base under the study conditions. A benefit against vehicle does not establish that every niacinamide formulation will produce the same result. [5]
The paper also included a paired comparison in eighteen participants with hyperpigmentation, using a 5 per cent niacinamide moisturiser and its vehicle. The researchers reported a significant decrease in hyperpigmentation and an increase in skin lightness against vehicle after four weeks. [5] The clinical result supported a benefit under the tested conditions. The cell models supplied the mechanism. Both gave evidence from different angles.
This reverses the usual warning about cell evidence. Direct exposure bypasses delivery and risks overconfidence about a product. Isolation also removes interactions that a response requires, so a negative result has to stay attached to the process the assay tested.
Follow the Claim Backwards
When a product says an ingredient worked in human skin cells, start with the unglamorous questions. Which cells? Was the raw molecule tested or the finished formulation? At what concentration, for how long, and in what medium? Was the endpoint gene expression, secreted protein, cell survival or the outcome promised to the reader?
Then put the missing journey back. Does the formula preserve and release the ingredient? What tissue does it need to reach, and what evidence establishes exposure there? If the response depends on neighbouring cells or the state of the tissue, does the model include that relationship?
The wording of a claim deserves the same scrutiny. Great Britain’s cosmetic claims guidance requires evidence relevant to the product and benefit claimed. Properties of an ingredient must not imply a finished-product benefit the product does not have. [6] The Committee of Advertising Practice, which writes the UK advertising codes, also says in its cosmetics guidance that efficacy claims generally need support from tests on people. [7] An accurate description of a cell result does not settle the broader promise made by the surrounding advertisement.
An accurate description of a cell result does not settle the broader promise made by the surrounding advertisement.
A rise in collagen-related gene expression in fibroblasts does not establish more collagen in the dermis. Even a measured increase in dermal collagen would leave the promised visible change to be tested. The steps are biologically connected, but the evidence must establish the connections the claim relies on.
So, does an ingredient that worked in cells work on skin? The result gives us a reason to investigate. The next experiment must address what the claim still needs: delivery to the relevant site, the response in tissue or a benefit during use. The pipette delivered a molecule straight to a cell, and everything between that dish and your face needs evidence of its own. Keep the cell result at the size of the experiment that produced it.
Fibroblasts. Connective-tissue cells that produce and maintain extracellular-matrix material. Dermal fibroblasts contribute to the skin’s collagen-rich support tissue.
Vehicle. The formulation base that carries an ingredient. In a vehicle-controlled comparison, the base without the tested ingredient serves as the comparator.
Keratinocyte. The main cell type of the epidermis. Keratinocytes form the living epidermal layers and develop into corneocytes that form the stratum corneum.
In vitro. Outside a living organism. The term covers isolated-cell experiments and more structured laboratory models, including reconstructed tissues.
Endpoint. The specified outcome measured in an experiment, such as gene expression, tissue viability or a change in pigmentation.
Flux. The amount of a substance passing through a defined area per unit time. It measures passage through the tissue, rather than the amount retained within it.
Melanosome. A pigment-containing organelle made in a melanocyte. Melanosomes are transferred to neighbouring keratinocytes as part of skin pigmentation.
References
1. OECD (2025). Test No. 439: In Vitro Skin Irritation: Reconstructed Human Epidermis Test Method. OECD Guidelines for the Testing of Chemicals, Section 4. Paris: OECD Publishing. https://doi.org/10.1787/9789264242845-en
2. OECD (2004). Test No. 428: Skin Absorption: In Vitro Method. OECD Guidelines for the Testing of Chemicals, Section 4. Paris: OECD Publishing. https://doi.org/10.1787/9789264071087-en
3. Schmook, F.P., Meingassner, J.G. and Billich, A. (2001). Comparison of human skin or epidermis models with human and animal skin in in-vitro percutaneous absorption. International Journal of Pharmaceutics, 215(1-2), pp. 51-56. https://doi.org/10.1016/S0378-5173(00)00665-7
4. Agonia, A.S., Palmeira-de-Oliveira, A., Cardoso, C., et al. (2022). Reconstructed Human Epidermis: An Alternative Approach for In Vitro Bioequivalence Testing of Topical Products. Pharmaceutics, 14(8), 1554. https://doi.org/10.3390/pharmaceutics14081554
5. Hakozaki, T., Minwalla, L., Zhuang, J., et al. (2002). The effect of niacinamide on reducing cutaneous pigmentation and suppression of melanosome transfer. British Journal of Dermatology, 147(1), pp. 20-31. https://doi.org/10.1046/j.1365-2133.2002.04834.x
6. Office for Product Safety and Standards. (n.d.). Regulation 1223/2009 and the Cosmetic Products Enforcement Regulations 2013: Great Britain. Product Claims, paragraphs 114-121. Accessed 5 October 2026. https://www.gov.uk/government/publications/cosmetic-products-enforcement-regulations-2013/regulation-20091223-and-the-cosmetic-products-enforcement-regulations-2013-great-britain
7. Committee of Advertising Practice (2026). Beauty and Cosmetics: General. AdviceOnline, 2 February 2026. Accessed 5 October 2026. https://www.asa.org.uk/advice-online/beauty-and-cosmetics-general.html
A short note
This article contains no affiliate links, sponsored content or product recommendations. The Schmook study came from Novartis; the Hakozaki paper included Procter & Gamble and University of Cincinnati researchers. The 2022 comparison received support from the tissue manufacturer EPISKIN, drug manufacturer Laboratórios Basi and testing company Labfit, with author affiliations disclosed. Their involvement belongs in the evaluation alongside the controls, endpoints and limits of each study. These papers provide examples of model performance, not independent confirmation of every current ingredient or product claim.
