Only 5.1% of records carry a Cramer class, and Class III does not mean toxic

About a 4 minute read.

There is a field in this database I had never touched: structure_class. Its only values are I, II and III.

Just 2,174 of 42,225 records have one — 5.1%.

Why it is nearly empty

Compare records that have a class with records that do not:

	Has a class (2,174)	No class (40,051)
Has a FEMA number	68.0%	5.3%
Listed with JECFA	63.8%	2.2%
10+ suppliers	39.3%	5.7%

This field came in attached to food flavouring safety assessments. JECFA is the FAO/WHO Joint Expert Committee on Food Additives, and the Cramer classification is the first step of their evaluation procedure.

So the 94.9% of blanks is not missing data — those materials never went through the procedure. (I measured the same pattern in the storage field and the 145 refrigerate records: a field arrives with its source.)

What the three classes are

The Cramer decision tree, proposed in 1978, is a series of yes/no questions that sorts substances into three classes by structural features:

• Class I: simple structure, known metabolic route, low expected toxicity
• Class II: intermediate
• Class III: a structure offering no basis for presuming safety

The distribution:

Class	Records	Share
I	1,312	60.3%
II	525	24.1%
III	335	15.4%
(written as the digit 1)	2	0.1%

Six in ten land in Class I. And the two records written "1" instead of "I" are 2,3-butanediol (8 suppliers) and (±)-sec-butylamine (6) — another small error catchable by shape alone, the same kind I collected in checks that need no chemistry.

What each class looks like

Class I (median 7 suppliers): geraniol (198), vanillin (193), citral (135), linalool (135), geranyl acetate (130), benzyl alcohol (125)…

That is the industry's most basic shelf.

Class III (median 6 suppliers): buchu mercaptan (89), maple furanone (67), strawberry glycidate (63), 2-methyl-3-tetrahydrofuran thiol (61), dihydrocoumarin (55)…

The sulfur share climbs monotonically:

Class	Sulfur in the name
I	12.8%
II	25.0%
III	32.2%

And the odour types agree — sulfurous is the most common type in Class III (33 records) and drops to fourth in Class I.

But Class III does not mean toxic

This is the sentence that matters here.

The Cramer tree judges whether a structure can support a presumption of safety, not whether the thing will hurt you. A sulfur atom pushes a molecule into Class III because the tree is not confident enough about how sulfur heteroatoms are metabolised, not because there is evidence of harm.

Okamura and colleagues' 2015 paper is a good worked example. They evaluated five acetal flavourings used only in Japan through the JECFA procedure:

As no genotoxicity study data were available in the literature, all five substances had no chemical structural alerts predicting genotoxicity. Using Cramer's classification, acetoin dimethyl acetal and hexanal dibutyl acetal were categorised as class I, and acetaldehyde 2,3-butanediol acetal, hexanal glyceryl acetal and 4-methyl-2-pentanone propyleneglycol acetal as class III.

Five compounds, none of them with toxicity data, split across two classes — sorted by structure, not by evidence.

And what follows the sorting is the point: they estimated daily intakes of 1.45–6.53 µg/person/day by the production-based method, or 156–720 µg/person/day by the single-portion exposure technique.

The class only decides which threshold you compare that intake against.

What this procedure usually concludes

EFSA's 2022 FGE.07 Revision 6 evaluated 55 flavouring substances:

The substances were evaluated through a stepwise approach that integrates information on the structure-activity relationships, intake from current uses, toxicological threshold of concern (TTC) and available data on metabolism and toxicity. The Panel concluded that none of the 55 substances gives rise to safety concerns at their levels of dietary intake, when estimated on the basis of the 'Maximised Survey-derived Daily Intake' approach.

Fifty-five evaluations, fifty-five no-concerns.

This system manages exposure. It is not a banned list.

In practice

1. Do not read Class III as a warning. It marks "this one needs an intake number before anything can be said."
2. Its direct use in perfumery is limited — it is a classification for food use, and skin is an entirely different exposure route.
3. What it is useful for is as a signal: a material carrying a Cramer class and a JECFA listing has been through a formal evaluation. 94.9% of records have not.
4. What you should actually check is IFRA and the supplier specification. This field does not substitute for either.

What this doesn't establish

• I did not read the Cramer decision tree itself. "Class III means no basis for presuming safety" is the standard formulation of the method, not something I confirmed against the 1978 paper.
• "The 94.9% of blanks means those materials never went through the procedure" is an inference from the FEMA/JECFA coverage gap, not from any documentation.
• The sulfur share is judged from name strings (thiol, mercapt, sulf, thia, thio), which misses sulfur compounds not named with those roots.
• Both papers are food safety assessments, not fragrance research. I cite their account of how the classification is used.
• EFSA's "no safety concerns" applies to dietary intake and cannot be extrapolated to skin exposure from perfume.
• I did not check any material's IFRA status.