94.5% of the dermal toxicity field is a rabbit's lethal dose, and sensitisation appears once

Last time I counted the records carrying human data. This article steps back to a more basic question: does the toxicity field measure the route you actually use? Of the 3,156 materials with ten or more suppliers, 51.7% have oral toxicity data, 31.2% dermal, and just 6.6% inhalation — the route a perfumer never uses has the best coverage. And opening those 985 dermal entries does not give you what you need: 94.5% are LD50 lethal doses, 888 of them in rabbits, while "sensitisation" appears in one record and "irritation" in four. The field that answers dermal questions is ghs_classification, not toxicity_dermal.

About 4 minutes.

Last time I counted the 156 records carrying human data and noted that those figures describe ingestion.

This article steps back and asks: which routes does the toxicity field measure at all?

Coverage runs opposite to need

There are 3,156 materials with ten or more suppliers. The three toxicity fields cover:

Route	Records	Coverage
Oral	1,632	51.7%
Dermal	985	31.2%
Inhalation	207	6.6%

The route a perfumer never uses has the best coverage.

And inhalation — smelling, the reason this industry exists — sits at 6.6%.

A further 658 materials have oral data but no dermal data. For those you can look up what happens if you swallow it and not what happens if you put it on skin.

Even when there is dermal data, it is not the right kind

Breaking down the content of those 985 dermal entries:

Kind of datum	Records	Share
LD50 (lethal dose)	931	94.5%
LDLo (lowest lethal dose)	128	13.0%
Irritation	4	0.4%
TDLo (lowest toxic dose)	3	0.3%
Sensitisation	1	0.1%

Nearly 95% is "how much applied will kill".

And the test animals:

Animal	Records
Rabbit	888
Mouse	181
Rat	151
Guinea pig	111

What the dermal toxicity field actually contains is acute lethality in rabbits.

But a perfumer's questions are different

Putting a material into a perfume, what you worry about is:

Will it cause an allergy? (sensitisation)
Will it make skin red and sting? (irritation)
Will it react in sunlight? (phototoxicity)

The toxicity_dermal field answers the first once, the second four times, and the third never.

The field's name suggests it answers dermal questions. It answers acute toxicology questions.

The answer lives in another field

ghs_classification is where dermal questions get answered.

Among GHS's classification entries:

H317 — May cause an allergic skin reaction (skin sensitisation)
H315 — Causes skin irritation
H319 — Causes serious eye irritation

I measured earlier that 63 records database-wide carry skin sensitisation in the GHS field, including geraniol at 198 suppliers, citral at 135, eucalyptus oil at 129, citronellol at 119, eugenol at 97 — all materials in daily use.

A concrete comparison: tea tree oil has 101 suppliers and has caused more published allergic reactions than any other essential oil. Its toxicity_dermal field reads, in full:

skin-rabbit LDLo 5000 mg/kg

And its GHS field carries both H315 (skin irritation) and H317 (skin sensitisation, Category 1).

One material, two fields, entirely different amounts of information. (That material has its own guide.)

But GHS coverage is poor

To be fair about it: only 359 records database-wide (0.9%) have a GHS entry, rising to 337 of 3,156 (10.7%) among well-supplied materials.

So the right statement is not "just read the GHS field", but:

toxicity_dermal has data, but it is not the answer to your question.
ghs_classification is the right field, and nine tenths of materials leave it empty.
When neither is available, the real answers are in the IFRA standards and the supplier's SDS.

What to do

Do not use toxicity_dermal to judge whether a material can touch skin. It measures something else.
Check ghs_classification for H315 / H317 / H319 first. Where present, that is a direct answer.
Both fields empty does not mean safe. At 0.9% coverage, "not written" carries almost no information.
For formula doses, read IFRA. That is the document built for this question.
Phototoxicity is the third question, and it lives in notes. Forty-six records say keep it off your skin outright.

What this doesn't establish

This classifies field content; it is not a toxicological assessment. I matched keywords (LD50, sensit, irritat) against field text rather than reading the underlying studies.
"Sensitisation appears once" is how often that string occurs in toxicity_dermal, not a claim that only one material has ever been sensitisation-tested. That data mostly lives elsewhere — the GHS field, IFRA documents, REACH dossiers.
Ten or more suppliers is my threshold.
The three coverage figures count "content that is neither empty nor Not determined", with no assessment of quality or vintage.
I consulted no SDS and no IFRA text for any material.
GHS classifications are supplier-declared and can disagree — beta-caryophyllene's GHS field contradicts its own toxicity field.