Designing low-irritation formulations for sensitive skin

Sensitive skin is a sensory syndrome, not a disease

The expert group of the International Forum for the Study of Itch defines sensitive skin as a syndrome in which unpleasant sensations — stinging, burning, pain, pruritus and tingling — occur in response to stimuli that would not normally provoke them. The decisive clause is that these sensations cannot be explained by lesions attributable to any skin disease. Erythema may or may not accompany them; the skin can look entirely normal (Misery et al., Acta Dermato-Venereologica, 2017).

The implication for formulation is direct. Sensitive skin is a problem of sensation rather than visible pathology, so working on barrier structure alone is not enough — the neurosensory axis has to be addressed as well.

The scale is substantial. A meta-analysis of 51,783 people across 18 countries found that 71% report some degree of sensitive skin, and 40% even when narrowed to very or moderately sensitive (Chen et al., JEADV, 2020). A nationally representative Korean survey of 1,000 people put the figure at 56.8%, the highest among the countries compared (Kim et al., Skin Research and Technology, 2018).

And there is one finding manufacturers should meet head-on. A meta-analysis of triggering factors covering more than 12,000 people worldwide found the leading trigger was cosmetics themselves, at an odds ratio of 7.12 — roughly double the second-ranked factor, humid air at 3.83 (Brenaut et al., JEADV, 2020). Low-irritation design is a baseline responsibility, not a marketing concept.

Reducing irritants in ingredient selection

The starting point is not shortening the ingredient list but removing substances with a documented irritation history. Fragrance allergens are the clearest case. In Korea, 25 fragrance allergens designated by the MFDS must be named individually — rather than declared simply as 'fragrance' — when they exceed 0.01% in rinse-off products or 0.001% in leave-on products (Regulation on Precautions for Use and Labelling of Allergenic Substances in Cosmetics, in force since 1 January 2020).

Europe has gone considerably further. Regulation (EU) 2023/1545 added 56 fragrance allergens to the individual-declaration list. The deadline for placing new products on the market, 31 July 2026, has already passed; existing stock must comply by 31 July 2028.

Preservatives diverge more sharply still. Methylisothiazolinone is capped at 0.0015% in rinse-off products and prohibited in leave-on products in both Korea and the EU, and both jurisdictions forbid combining it with the methylchloroisothiazolinone mixture in the same product.

Parabens are a different story. The EU caps butylparaben and propylparaben at a combined 0.14% (as acid), prohibits them in leave-on products for the nappy area of children under three, and bans isopropyl- and isobutylparaben outright. Korea has none of those three restrictions, regulating the p-hydroxybenzoic acid family only at 0.4% single / 0.8% mixed. A formulation built to Korean limits alone will fail in the EU. Quaternium-15 is another split: permitted at 0.2% in Korea, prohibited in the EU.

Paraben limits may tighten further. In an April 2025 opinion, the EU Scientific Committee on Consumer Safety concluded that applying butylparaben at 0.14% across all product types simultaneously leaves safety concerns for children aged 0.5–10 that cannot be excluded. It found a tiered approach acceptable across all age groups instead: 0.14% retained for rinse-off, 0.002% for leave-on and 0.092% for oral care (SCCS/1674/25). It has not been written into the regulation yet, but for EU-bound formulations it is a signal worth preparing alternatives for now.

Barrier lipids — the ratio matters more than the inclusion

Lipidomics has been reshaping this area fast. A 2025 study of 95 women aged 18–25, split into sensitive dry (32), sensitive oily (31) and healthy skin (32), found transepidermal water loss of 10.88±2.79 g/m²h in the sensitive dry group against 9.26±2.48 in healthy skin, and hydration of 40.10±9.32 against 57.16±7.96. Between those two groups, 101 lipid constituents differed significantly (Xie et al., Metabolites, 2025).

The direction, though, is not settled. Another 2025 study reported the opposite: significantly lower stratum corneum ceramides and fatty acids in sensitive-skin subjects, even though their sting scores matched controls (Kim et al., Journal of Cosmetic Dermatology, 2025). The first study read its elevated ceramide share as a compensatory response to a deficit in the other lipid classes. In 2026, an LC-MS/MS study quantified tape-stripped stratum corneum to compare ceramide profiles across sensitive dry, non-sensitive dry and normal skin directly (Metabolites, 2026;16(4):260, 93 subjects).

So 'more ceramide' versus 'less ceramide' is still an open question. That is precisely why designing composition and ratio is a safer bet than competing on use level alone.

Adding ceramide is not by itself a basis for a barrier claim. Stratum corneum intercellular lipids consist of ceramides, cholesterol and free fatty acids. By weight ceramides make up roughly half, but on a molar basis the three are close to equimolar. The widely quoted '3:1:1' is not the physiological composition — it is a therapeutically optimised ratio.

That distinction matters because of what the experiments show. Applying only one or two of the three lipids to a disrupted barrier actually delayed recovery; normal recovery appeared only with equimolar mixtures; and increasing any one component up to threefold accelerated repair further (Man, Elias et al., Journal of Investigative Dermatology, 1996).

Getting the ratio wrong is demonstrably harmful. In aged skin, a cholesterol-dominant mixture performed best and the equimolar mixture came second, while a fatty-acid-dominant mixture significantly delayed barrier recovery at 3, 6 and 24 hours (Zettersten et al., Journal of the American Academy of Dermatology, 1997).

In a low-irritation formulation, lipid design is a question of proportion and lamellar structure rather than inclusion. Ceramides have high melting points and are difficult to solubilise in the oil phase, so the outcome usually turns on solubilisation and lamellar structuring technique rather than on use level.

A 2025 human study then showed that topical supplementation with physiological lipids rebalances the stratum corneum ceramide profile and strengthens barrier function (Andrew et al., British Journal of Dermatology, 2025). The ratio principle established thirty years ago is being confirmed again by current lipidomics.

pH — one number separates two opposing enzyme families

Healthy skin surface is mildly acidic, with a measured range of roughly pH 4 to 6 once individual variation is included. In a multicentre study (N=330) that measured the volar forearm before and after 24 hours without washing or cosmetic application, the mean fell from 5.12 to 4.93, giving an estimated 'natural' surface pH averaging about 4.7 (Lambers et al., International Journal of Cosmetic Science, 2006). The industry-standard 'pH 4.5–5.5' is not wrong, but it states the range too narrowly and the centre too high.

The number is decisive because enzymes working in opposite directions are separated by it. β-glucocerebrosidase and acid sphingomyelinase, which process lipids into a functioning barrier, have pH optima of 5.6 and 4.5 respectively; β-glucocerebrosidase activity is roughly ten times lower at pH 7.4 than at 5.5. The serine proteases that break down corneodesmosomes to shed corneocytes — kallikrein 5 and kallikrein 7 — have neutral pH optima (Ali and Yosipovitch, Acta Dermato-Venereologica, 2013).

This is causal, not correlational. Raising surface pH experimentally delayed barrier recovery through reduced β-glucocerebrosidase activity while simultaneously activating serine proteases within 20 minutes, degrading corneodesmosomes and undermining stratum corneum integrity. Inhibiting the proteases reversed it (Hachem et al., Journal of Investigative Dermatology, 2003).

Preservation — the trap in the phrase 'preservative-free'

Start with the legal position. Korea regulates preservatives on a positive list: any ingredient not listed in Annex 2 of the Regulation on Safety Standards for Cosmetics may not be used as a preservative. 1,2-hexanediol is not on that list, nor is it in the EU's preservative annex (Annex V). Legally it is therefore a polyol humectant and solvent, not a preservative, and its antimicrobial contribution should be described as preservative support. Labelling or advertising it as a preservative amounts to declaring the use of an unlisted preservative.

The mechanism itself is real. Alkanediols such as 1,2-hexanediol and caprylyl glycol are amphiphilic, inserting into microbial membranes and disrupting them while also lowering water activity. Glyceryl caprylate, a caprylic acid monoglyceride, destabilises microbial membranes, and ethylhexylglycerin acts as a classic booster, lowering interfacial tension so that other preservatives penetrate and perform better.

A 'preservative-free' claim is permitted, but conditionally. The MFDS does not accept manufacturing records or raw material test data showing the substance was never added; it requires analytical testing showing the substance is absent from the finished product. For prohibited ingredients, a 'free-from' claim cannot be used at all.

The line that actually has to be held is the distribution standard. Total aerobic microbial count must not exceed 500 CFU/g(mL) for infant and eye-area products and 1,000 CFU/g(mL) for other cosmetics, while wet wipes are limited to 100 CFU/g(mL) each for bacteria and fungi. E. coli, P. aeruginosa and S. aureus must not be detected (Regulation on Safety Standards for Cosmetics, Notification 2026-19, Article 6).

Preservative challenge testing is not legally mandatory in Korea. It is nonetheless effectively required to justify a shelf life and to support export, and the international reference is ISO 11930:2019 including Amendment 1:2022. Criterion A requires a 3 log reduction in bacteria and a 1 log reduction in yeast by day 7, and for mould, no growth at day 14 plus a 1 log reduction at day 28. The MFDS preservative efficacy testing guideline (Guidance 1236-01, 2022) sets substantially the same bar. USP <51> Category 2, by comparison, asks only for a 2 log bacterial reduction at day 14, so passing ISO Criterion A satisfies it automatically.

Soothing actives — grade the evidence

The uncomfortable fact here is that the evidence for soothing actives is thinner than the marketing. A review that surveyed the ingredients actually used in pharmacy and dermocosmetic products for sensitive skin concluded that clinical studies on them remain sparse and lack methodological quality, and that only a handful of the most-used ingredients had ever been tested on sensitive-skin subjects specifically (Ferreira et al., International Journal of Cosmetic Science, 2022).

The ingredient with the clearest human evidence against sensory hyperreactivity itself is trans-4-t-butylcyclohexanol. It selectively inhibits TRPV1 (IC50 34 μM), and in 30 female subjects challenged with capsaicin at 31.6 ppm, a formulation containing 0.4% significantly reduced burning (P<0.0001, Kueper et al., Experimental Dermatology, 2010).

Panthenol's standard clinical concentration is 5%, at which it accelerated transepidermal water loss recovery and reduced erythema in sodium lauryl sulfate irritation models. Only the D-form (dexpanthenol) is active, so a DL racemic mixture delivers half the activity at the same use level — and since both carry the same INCI name, the form has to be confirmed on the raw material certificate.

Niacinamide increased ceramide synthesis four- to fivefold in cultured keratinocytes (Tanno et al., British Journal of Dermatology, 2000), and in a left-right comparison in 28 people with atopic dry skin, a 2% formulation reduced transepidermal water loss significantly more than white petrolatum (Soma et al., International Journal of Dermatology, 2005). The common belief that above 5% it becomes irritating does not match the CIR safety assessment, which found no stinging at concentrations up to 10%. The more plausible cause of flushing is nicotinic acid formed as an impurity at low pH and elevated temperature — which is why holding formulation pH near 6 is the practical answer.

Some ingredients deserve a colder look. Allantoin holds a secure regulatory position in the US OTC skin protectant monograph at 0.5–2% (21 CFR 347.10), yet a 6% cream failed to beat vehicle in a phase 3 trial in 169 epidermolysis bullosa patients (P=0.985, Paller et al., Orphanet Journal of Rare Diseases, 2020).

Madecassoside has split-face randomised evidence showing it performs comparably to a topical steroid in post-laser recovery (Lueangarun et al., Journal of Cosmetic Dermatology, 2019), but no confirmed human evidence against sensitive-skin stinging itself. The two are not competitors; they have different targets. Neurosensory hyperreactivity calls for trans-4-t-butylcyclohexanol; recovery from physical damage after a procedure calls for madecassoside and panthenol.

Substantiation — what actually backs a low-irritation claim

There is a common misconception here. The annex of substantiation-required claims in Korea's Regulation on Substantiation of Cosmetic Labelling and Advertising lists only eight items — suitability for acne-prone skin, antibacterial action, reduction of skin ageing and so on. 'Low irritation' is not among them.

That does not make it free to use. Even for claims outside the annex, the general substantiation duty under Article 14 of the Cosmetics Act applies: documentation must be submitted within 15 days of an MFDS request, and failure to do so triggers an order to cease the labelling or advertising. What does not exist is any MFDS-designated test method or subject count for 'low irritation' specifically.

The two numbers quoted in practice come from different sources. The human patch test in Annex 1 of the Regulation on Review of Functional Cosmetics requires at least 30 subjects, occlusive patching on the upper back or forearm, removal at 24 hours, and assessment by a dermatologist. Human studies used to substantiate labelling and advertising, by contrast, require at least 20 subjects with valid data so that statistical comparison is possible (Guideline on Test Methods for Substantiation of Cosmetic Labelling and Advertising, 2018). The two figures should not be used interchangeably. Grading follows the ICDRG scale (0–3: none, doubtful, weak, strong, extreme).

The wording boundaries are equally clear. An 'anti-allergy effect' is a drug claim and cannot be used. 'Allergy tested' is permitted where supported by human study data, provided it cannot be read as meaning the product does not cause allergy.

For stability, the guideline sets accelerated testing at 40±2°C and 75±5% RH and long-term testing at 25±2°C and 60±5% RH, each for at least six months across at least three lots. That guideline is advisory and carries no legal force. What is legally binding is Article 12(11) of the Enforcement Rule of the Cosmetics Act, requiring stability test records to be retained for one year past the expiry date for products containing retinol, ascorbic acid, tocopherol, peroxides or enzymes at 0.5% or more.

What changes from 2028

The amendment to the Cosmetics Act passed by the National Assembly in December 2025 introduces a phased cosmetic safety assessment regime: functional cosmetics from 2028, infant and children's products from 2029, newly launched items from 2030, and full application from 2031.

It marks a shift to a system in which 'low irritation' must be backed by a documented safety assessment rather than by wording. HBMIC builds that documentation structure into development from the ingredient selection stage onward.

Labelling is moving the same way. The MFDS began an e-label pilot with volunteer companies in March 2026 and has announced guidelines for October 2026, and it plans to designate a Cosmetics Safety Information Centre in December 2026 to run a raw material safety database.