Show the work

Methodology & Evidence Limits

A rating is useful only when the source, test model, uncertainty, and matching rules stay attached. Dataset 2026.07.10.1 · Last reviewed 2026-07-10

01 · Primary source

What the launch data reproduces

The 161 launch records reproduce factual ingredient names and comedogenicity grades from Table I of James E. Fulton Jr.’s 1989 paper, Comedogenicity and irritancy of commonly used ingredients in skin care products.

The paper is a historical ingredient screen. This site does not copy its narrative prose or photographs. Each dataset record stores the reported grade or range and a page-level table locator.

02 · Test conditions

How the historical screen worked

Ingredients were generally mixed in propylene glycol at a 10% concentration. The study used three New Zealand albino rabbits per assay, applied material on weekdays for two weeks, and compared the treated ear with the untreated ear.

Follicular keratinisation was measured visually and microscopically. The source emphasizes that rabbit-ear skin is unusually sensitive and that results can change with solvent, vehicle, refinement, and raw-material source.

03 · Rating scale

Why the site uses signal labels

4–5Stronger historical signal

The paper describes this range as reproducibly positive in its model.

2–3Borderline historical signal

The source explicitly calls this range borderline.

0–1Low historical signal

The source considered this range not significant under its conditions.

These labels deliberately say “historical signal.” They do not translate the scale into human severity, a finished-formula grade, or a purchase recommendation.

04 · Matching pipeline

How pasted labels are compared

  1. 01Parse locally. The browser splits commas, semicolons, line breaks, and bullets while preserving slash-containing INCI names.
  2. 02Normalize typography. Case, Unicode dash variants, explanatory parentheses, and written concentrations are normalized.
  3. 03Match exactly. Only canonical names and reviewed CosIng aliases can match. No substring, family, or fuzzy inference is used.
  4. 04Keep unknowns visible. Unmatched entries remain in the result and are never assigned a default low signal.
  5. 05Attach evidence. Every match retains the source record and table locator.

05 · Limitations

What this method cannot establish

Exact percentages or ingredient purity from label order.
Interactions across the complete formula or delivery vehicle.
Differences between leave-on, rinse-off, face, scalp, lip, or body exposure.
Whether an individual reaction is acne, irritation, allergy, or another condition.
Current product composition after a brand reformulates.
A finished-formula or human evidence conclusion from an ingredient-level animal screen.

The 2025 JAAD Reviews clinical review is used to frame these limits. It reports inconsistent translation from rabbit-ear grades to human outcomes and emphasizes concentration, purity, vehicle, and complete formulation.

06 · Updates & corrections

A versioned dataset, not a timeless list

Every release publishes a review date, record count, and version. A correction must identify the disputed record, provide a traceable source, preserve the evidence model, and pass the repository’s integrity and language tests.

Dataset version
2026.07.10.1
Last reviewed
2026-07-10
Record count
161 ingredient-level records
Rating source
fulton-1989 · Table I

Source registry

Every source, with its limits

011989

Comedogenicity and irritancy of commonly used ingredients in skin care products

Journal of the Society of Cosmetic Chemists, 40, 321-333 · Primary rabbit-ear screening study

Ingredients were generally tested at 10% in a rabbit-ear model. The paper calls the assay extremely sensitive, reports source and vehicle effects, and says the survey is not definitive or a substitute for finished-formula and human evidence.

Open source ↗
022025

Comedogenicity in cosmeceuticals: A review of clinical relevance, regulatory gaps, and future directions

JAAD Reviews, 6, 78-83 · Clinical review covering animal and human testing

The review describes inconsistent translation from rabbit-ear ratings to human skin and emphasizes that concentration, purity, vehicle, complete formulation, and individual response can change the practical outcome.

Open source ↗
032025

CosIng cosmetic ingredient database and glossary

European Commission cosmetic ingredient database · Regulatory nomenclature and ingredient-name reference

CosIng is used here only to reconcile common label names and INCI nomenclature. The Commission states that inclusion does not mean an ingredient is authorised, effective, or appropriate for a particular finished product.

Open source ↗
042026

Adult acne: Treatment and self-care guidance

AAD public guidance · Dermatology patient guidance

This guidance supports general consumer education and professional consultation. It does not validate this site's ingredient-level ratings or establish how any specific finished formulation will affect an individual.

Open source ↗