Fragrance and essential oils: the commonest cause of cosmetic contact allergy

Published 6 August 2026Zane Hitchcox, editor

The premise this page exists to dismantle is that natural means gentle. It is not an argument about chemistry in the abstract. It is a claim with a measurable prediction attached, and the prediction has been tested, repeatedly, in tens of thousands of patch-tested patients. It fails.

Two different injuries share one word

"Reacting" to a product covers two unrelated events. Irritant contact dermatitis is direct damage: the substance injures the skin and breaks the barrier, and given enough concentration, occlusion or repetition it will do that to anybody. DermNet, the New Zealand dermatology reference, puts the diagnostic problem plainly: there is no test for it, and patch testing may be needed to distinguish it from the other thing.¹⁰

The other thing is allergic contact dermatitis, a type IV delayed hypersensitivity reaction that typically appears 24 to 72 hours after exposure and requires the immune system to have been sensitised first. A person may use a substance for years without incident and then become sensitised to it, after which the response is reproducible and lasting.¹¹ That asymmetry is why years of uneventful use are not evidence of safety, and why a rash that arrives a day after application is a different kind of event from a product that stings on contact. Which of the two is happening is a clinical judgement, and only a clinician can order and read a patch test.

How often, and in whom

The patch-test literature is unusually good, because contact dermatitis clinics have been running standardised screening series for decades and publishing the totals. Three datasets carry most of the weight.

The following reporting period moved in the same direction. In 2021–2022, across 3,056 patients at 12 centres, hydroperoxides of linalool were positive in 10.1% and were again among the handful of commonest positives, and reactions to hydroperoxides of limonene increased significantly against the previous period. The authors' summary was that fragrance allergy remains very common while the composition of the markers is changing.² European surveillance says something similar with different instruments: across 22,581 patients tested with the European baseline series in 2019 and 2020 at 53 departments in 13 countries, fragrance mix I was positive in 6.80% and Myroxylon pereirae — balsam of Peru, a resin used as a fragrance marker — in 6.62%.³

None of that describes the general public, and the difference matters enough to be measured separately.

A 2020 review in Dermatitis reconciles the two worlds: up to about 4.5% of the general adult population may be allergic to some fragrance material, while among consecutive patients patch tested for suspected contact dermatitis the figure reaches 20% to 25%.⁶

What a patch test actually tests, and what it misses

Routine screening does not test fragrances. It tests three proxies. Fragrance mix I has been in most baseline series since the late 1970s and contains eight chemicals at 1% each: amyl cinnamal, cinnamal, cinnamyl alcohol, eugenol, oakmoss (Evernia prunastri) extract, geraniol, isoeugenol and hydroxycitronellal. Fragrance mix II, added to the European baseline series in 2008, holds six more: citral, citronellol, coumarin, farnesol, hexyl cinnamal and HICC. Myroxylon pereirae is the third marker.⁶ More than 150 individual fragrance materials have been reported to cause contact allergy; one tally used in the clinical literature puts the documented count at 162, of which 80 are essential oils.⁵ Fourteen chemicals and a resin are being asked to stand in for all of them. One English clinic tested nearly two thousand consecutive patients against both the markers and the individual substances, which is the only way to find out what the shortcut costs.⁷

Two consequences follow, and both are uncomfortable. A negative screening panel does not clear fragrance. And a positive one frequently fails to say which substance is responsible: when patients positive to fragrance mix I are retested against its eight ingredients individually, on average about a third react to none of them.⁶ A positive patch test is also not by itself a diagnosis; it means something only alongside a real exposure that fits.⁵

Limonene and linalool are weak allergens until the bottle is opened

This is the part of the subject that changes how a reader should think about an old jar. Linalool, a terpene alcohol abundant in lavender and many other plant oils, and limonene, from citrus peel, are only occasional sensitisers themselves. On exposure to the oxygen in air they oxidise into hydroperoxides, and those hydroperoxides are, in DermNet's words, "much more potent sensitisers" than the parent compounds — which is why a person sensitised to them may tolerate a freshly opened pack and react to the same pack months later.⁹ A half-used bottle that has been opened and closed in a warm bathroom for a year is, in the only sense that matters here, not the same product that was bought.

Regulators have accepted the chemistry and written it into the ingredient schedules. New Zealand's Schedule 5 restricts D-limonene, L-limonene and DL-limonene to a peroxide value below 20 mmoles per litre, and caps a long run of pine, fir and terpene materials at 10 — the Abies and Pinus oils, steam-distilled turpentine, alpha-terpinene, terpinolene. Then the schedule's own footnote says exactly what those numbers do not cover: "This limit applies to the individual ingredient and not to the finished cosmetic product."¹⁵ A peroxide value is a specification for the raw material as supplied to the formulator. Nothing in the instrument measures what the product is doing in month nine, on a shelf, in the steam.

The network's response was not to stop testing but to narrow it: monitor the preparations chemically to reduce their irritancy, interpret readings carefully, and test lower concentrations, aimed at the patients in whom a fragrance exposure is actually suspected.⁸ That sits awkwardly beside North American practice, which keeps linalool hydroperoxides on the screening tray and finds them among the five commonest positives in 4,121 patients.¹ The disagreement is genuine: one group is reading a share of those reactions as irritation, the other is counting them. It is the clearest available illustration of why a percentage in a patch-test table is not a count of people harmed.

What a label is compelled to name

Since 2005 the European Union has required 26 named fragrance substances to appear individually in the ingredient list once a leave-on product exceeds 0.001% (10 ppm) or a rinse-off product exceeds 0.01% (100 ppm), rather than hiding inside the single word parfum.^{6, 12} In 2023 it added 45 further entries to Annex III, numbered 327 to 371, on the same thresholds and a three-year runway: products placed on the market from 31 July 2026 must comply, and stock already in shops may be sold until 31 July 2028.¹³ That is why limonene and linalool turn up by name at the tail of lists that otherwise disclose nothing about their scent, a pattern the guide to reading an ingredient list works through position by position.

The list is not only expanding. Where the evidence became bad enough, the European Union removed substances outright: Commission Regulation (EU) 2017/1410 added HICC — the fragrance sold as Lyral — to Annex II at entry 1380, and atranol and chloroatranol at 1381 and 1382, its recitals naming the last two as components of the oakmoss (Evernia prunastri) and treemoss (Evernia furfuracea) extracts that Annex III still permits. Cosmetics containing any of the three could not be placed on the Union market after 23 August 2019, or sold there after 23 August 2021.¹⁴ HICC was the most frequent sensitiser in fragrance mix II and, tested alone at 5%, produced one of the higher positive rates in the whole European baseline series before it went.⁶

New Zealand named its allergens on 1 January 2026

This is a live change and it is not widely reported. Cosmetics in New Zealand are controlled by the Environmental Protection Authority under the Cosmetic Products Group Standard 2020, approval HSR002552, a point covered in full in the article on who regulates cosmetics here. That standard borrows the European ingredient schedules almost wholesale — the 26 allergens appear as Schedule 5 entries 67 to 92, each carrying the same 0.001% and 0.01% thresholds — but until recently a single clause cancelled the effect. Schedule 1, condition 1(6) read: "Despite subclause (4) a cosmetic product containing a component at reference numbers 67-92 of Schedule 5 may not be labelled with the name of that component, provided the label lists the flavours or fragrances which must be described by the words, 'fragrance' or 'fragrances' or 'parfum' or 'parfums'."²⁵ Subclause (4) is the one that makes the Schedule 5 labelling entries binding. In the version consolidated and effective 1 January 2026, condition 1(6) reads, in full, [Revoked].¹⁵

Two limits on that good news. The retail tail means shelves will carry older labelling into 2027.¹⁶ And New Zealand's Schedule 5 stops at entry 324, methyl salicylate. It has not taken in the entries the European Union inserted at 327 to 371 in 2023 — no vanillin, no carvone, no menthol, no terpineol.¹⁵ A product sold in both markets from 2027 will legally carry a longer allergen list in Europe than in Auckland.

Citrus oils and sunlight: a third mechanism

Phototoxicity is neither irritancy nor allergy. Furocoumarins — psoralens — absorb ultraviolet A and damage skin only in combination with light, which is why the reaction appears as a burn in the shape of whatever touched the skin. Expressed citrus peel oils carry them. The European Union's Scientific Committee on Consumer Products, in the opinion it adopted at its sixth plenary on 13 December 2005, sets out an earlier committee's list of eleven such oils — angelica root, bergamot, expressed grapefruit, lemon, lime, bitter orange peel and rue among them — and then reaches its own conclusion: because "the consumer is exposed to sunlight after using various types of cosmetic products, not only bronzing or sun protection products", total furocoumarin concentrations above 1 ppm in any finished cosmetic product "will be of concern with regard to consumer safety".¹⁷

The reason the committee is insistent is the toxicology. The International Agency for Research on Cancer classifies methoxsalen (8-methoxypsoralen) plus ultraviolet A radiation in Group 1, carcinogenic to humans,¹⁸ and 5-methoxypsoralen — bergapten, the furocoumarin most associated with bergamot — in Group 2A, probably carcinogenic to humans.¹⁹ Those classifications come from therapeutic PUVA dosing and animal studies, not from cosmetic exposure at parts per million, and it would be dishonest to present them as a risk estimate for a scented moisturiser. They are the reason the limit is 1 ppm rather than something looser.

The oil in the jar is usually there for the smell

An essential oil listed high on an ingredient list is doing a job. The question is which job. Sabroe and colleagues open their 2016 paper with the flat statement that essential oils are fragrance substances, labelled on cosmetic products by their INCI names, and that this potentially confuses consumers.²² Their finding is the practical one. Thirty-four patients reacted to at least one of the oils tested; eleven of those reacted to none of the six marker substances in the baseline series. Counted oil by oil: 4 of 11 reactions to tea tree, 3 of 4 to sandalwood, 2 of 7 to lemongrass, 2 of 3 to peppermint and 1 of 5 to ylang-ylang would have been missed without testing the oil itself.²²

Tea tree oil has produced more published allergic reactions than any other essential oil since the first cases in 1991. A 2016 review found prevalences of positive patch test reactions in routine testing ranging from 0.1% to 3.5%, described fresh tea tree oil as a weak to moderate sensitiser whose allergenic potency increases on oxidation, and noted that most reported cases followed application of the pure oil rather than use of a formulated cosmetic.²⁰ Lavender shows the same fashion-sensitivity: a Japanese series reported a 3.7% positivity rate to lavender oil across 1990–1998, but with an annual range of 0% to 13.9% and a sharp spike in 1997 that the authors attributed to a vogue for dried lavender in household goods rather than to cosmetics.²¹ Contact allergy tracks what people are actually putting on themselves that decade, which is a fact about culture as much as chemistry.

The commercial point is separate and worth stating bluntly. An essential oil present at a fraction of a percent, near the bottom of a list, is at a perfuming concentration. Nothing on the label distinguishes an oil put there to make the product smell of something from an oil put there because a brand believes it treats skin, and no ingredient list is required to. The evidence for the second proposition is examined ingredient by ingredient in the survey of New Zealand botanicals; for most of them it is thin.

"Preservative-free", with essential oils, is a claim about a word

The New Zealand group standard defines preservatives as substances "which are exclusively or mainly intended to inhibit the development of micro-organisms in the cosmetic product", and permits only those listed in Schedule 7 — a positive list with maximum concentrations attached. The definition turns on intention. The preamble to that schedule then says something a shopper never sees: "Other substances used in the formulation of cosmetic products may also have anti-microbial properties and thus help in the preservation of the products, as, for instance, many essential oils and some alcohols. These substances are not included in this Schedule."¹⁵

So the phrase is accurate and uninformative at the same time. It reports a formulator's stated intention, not the microbiology of what is in the jar, and the substances it substitutes in are the ones this entire page is about. That is the sense in which "essential-oil-based, preservative-free" can be the worst of both worlds: it trades a preservative with a legal maximum for a fragrance material without one. The wider problem with the vocabulary is set out in the article on what "natural" and "clean" mean.

And in the United States, still one word

American cosmetic labelling has permitted a single collective term for more than half a century. Under 21 CFR 701.3(a), ingredients are declared in descending order of predominance "except that fragrance or flavor may be listed as fragrance or flavor".²³ The Modernization of Cosmetics Regulation Act of 2022 changed the statute: section 609(b) of the Federal Food, Drug, and Cosmetic Act now requires the responsible person to identify each fragrance allergen on the label — but which substances count as fragrance allergens is left to a regulation the Secretary must write, and the Act's deadline for even a proposed rule was 18 months from 29 December 2022.²⁴ Until that regulation is in force, a United States label may still say only "fragrance". Terms such as "fragrance free" and "unscented" have no legal definition there either, and products described as unscented can contain a masking fragrance added precisely to cover an odour.⁵ Those words are dealt with alongside their relatives in the piece on label terms with no legal meaning.