Material guide: costus root oil — the notes field says "do not use on any part of the body", and the two sensitising molecules each have their own record

About 5 minutes.

Short version

• Name: costus root oil, from Saussurea lappa, CAS 8023-88-9
• Odour: animal type — orris, green, hairy, woody, unripe melon
• Strength: medium; substantivity 316 hours
• Suppliers: 21; FEMA 2336, and regulatory still reads FEMA GRAS
• cosmetic_uses: not used anymore
• First line of the notes field: "Do not use on any part of the body because of potential sensitizing."
• Acute toxicity neither high nor low: oral rat LD50 3400 mg/kg, dermal rabbit > 5000 mg/kg

That note

Most records use notes for isolation history or chemical background. This record's first sentence is an instruction:

Do not use on any part of the body because of potential sensitizing.

The same cell then gives the reason:

"…used in perfumery & cosmetics; contains sesquiterpene lactones reported to cause contact hypersensitivity."

This is the bluntest field I have read in this series. Most materials keep their risk in the GHS or toxicity columns, where you have to interpret it. This one puts the conclusion on the first line of the notes.

The two sensitising molecules each have their own record

Which sesquiterpene lactones? The database can answer, because they are in it:

Record	CAS	Suppliers	Category
costus lactone (costunolide)	553-21-9	7	natural substances and extractives
dehydrocostus lactone	477-43-0	5	natural substances and extractives
dihydrodehydrocostus lactone	4955-03-7	0	natural substances and extractives

Those two are sold on their own. They are not theoretical constituents; they are things you can order.

And in the dermatology literature they have a named role. Paulsen reviewed the sesquiterpene lactone mix in Contact Dermatitis in 2023 — a screening agent developed in the 1980s for Compositae sensitisation, made of equimolar alantolactone, costunolide and dehydrocostus lactone.

The constituents that make costus oil sensitising are used as the standard patch-test reagent.

The prevalences given there: around 1% in Europe, 0.8% in North America, and between 0% and 10.7% elsewhere.

Paulsen and Andersen's fifteen-year screening data from 2011 is finer-grained: 157 of 7,163 people (2.19%) tested positive to the SL mix, and of the three constituents, costunolide elicited positive reactions most frequently, then dehydrocostus lactone, then alantolactone.

That 1982 experiment

Cheminat, Stampf and Benezra did something very clean in Acta Dermato-Venereologica in 1982.

They open by establishing the problem: costus oil, used in perfumery, is "responsible of numerous cases" of allergic contact dermatitis, and the haptens are costunolide and dehydrocostus lactone.

Then they took those two lactones out of the oil — with aminoethyl-polystyrene, a polymer the lactones bind to, so that the solution passing through is lactone-free.

Two groups of guinea pigs:

• The group given the original costus essential oil could be successfully sensitised.
• The group given the polymer-treated oil could not.

Same oil, two molecules removed, no longer sensitising.

This is a 1982 animal experiment, with the scale and methods of its time, and I have not checked whether it was replicated. But it establishes something: for this oil, sensitisation can be attributed to specific molecules rather than to "naturals are too complex to say".

(The problem with "naturals are gentler" is measured in Synthetics give me headaches, so I only use naturals. Costus oil is one of that article's counterexamples.)

Its regulatory status today is cut in half

The regulatory field: FEMA GRAS, FEMA number 2336. The cosmetic_uses field: not used anymore.

This is exactly pile one from last round's cut — of the 80 records flagged "not used anymore", 26 still carry a FEMA number, having left cosmetics rather than the flavour register. See Those 80 "not used anymore" materials.

The flavour side is live: the flavor field reads "costus", and the taste description carries "Fatty metallic like herbal. For creation in spirits tonalities and brown notes".

What that means for you is clear: do not put this in a perfume or in anything that touches skin.

It has a replacer with 14 suppliers

The same database holds costus root oil replacer, 14 suppliers, categorised as fragrance agents, with an empty cosmetic_uses field — no retirement flag.

Costus root oil is one of only three pairs, out of 87 original-versus-replacer matches, where the original is genuinely flagged as retired (the other two being musk ambrette and dihydrocoumarin). Most replacers are not replacing something banned, which is The database has 178 "replacers".

This one is in the small group where a replacer is the only option.

That odour field

The odour field reads: orris, green, hairy, woody, unripe melon.

"Hairy" is not a common word in this database, and pairing it with "orris" makes sense — orris's powderiness and this kind of dry animalic quality point the same way. "Unripe melon" is a separate thread: green, raw, not yet sweet.

This oil was prized precisely because that combination is hard to assemble from anything else. Fourteen suppliers for a replacer does not mean it smells the same.

What this doesn't establish

• I have not smelled it, and going by the notes field I will not. The whole article is fields, database-wide counts and three papers.
• I did not check the current IFRA standards. This article is about database fields and published literature. Establish the regulatory status from the IFRA standards and your own region's cosmetic regulation.
• The Cheminat paper is a 1982 guinea-pig experiment, with the scale and methods of its time. I did not check for replication.
• The SL mix prevalences are from patch-tested populations — people who were being tested because contact dermatitis was suspected — not from the general population.
• "Costunolide elicited the most positives" is from Paulsen and Andersen's Danish data. Compositae exposure differs by region, so the ranking need not hold elsewhere.
• The "do not use on any part of the body" line is the database's own wording, not any regulator's.
• This record has no physical data beyond substantivity — no melting point, no vapour pressure, no logP, and the appearance field is marked estimated. That is typical of natural extracts.
• I did not compare the replacer's smell against the original.