How MySkinIQ verifies cross-reactivity

The claim, in one sentence

MySkinIQ’s cross-reactivity database is constructed from three classes of public, dermatologist-authored evidence: the North American Contact Dermatitis Group’s biennial patch test surveillance series, the American Contact Dermatitis Society’s Allergen of the Year designations, and EU Regulation 1223/2009 Annex III. Where surveillance and regulatory data disagree, we record both. Where neither is available for a putative cross-reaction, we do not record one.

That sentence is the entire moat. The remainder of this page is the work behind it.

The four source classes

Every family in our database is anchored to at least one of the following. Citations are recorded against the underlying allergen records (a research-grade Firestore collection internal to the app); families inherit traceability from their members.

How a family becomes a family

A “family” in MySkinIQ is a set of contact allergens for which sensitization to one member predicts an elevated risk of reaction to other members of the set. The construction rule has four steps:

1. Anchor compound. A single compound must be either (a) listed in EU Reg 1223/2009 Annex III, (b) on the NACDG screening series at prevalence ≥ 1.0% in the most recent reported cycle, or (c) carry an ACDS Allergen of the Year designation. If none of these hold, the compound enters the database as a standalone allergen, not a family.
2. Cross-reactivity rationale. A peer-reviewed source must describe the mechanism — shared structural backbone, shared metabolite, shared sensitizing impurity, or population co-positivity above a published threshold. We record the rationale verbatim from the source. We do not paraphrase mechanisms.
3. Member compounds. A compound is added to the family only if the cross-reactivity rationale names it explicitly or describes a structural rule it satisfies. Compounds added by structural rule are tagged for review at the next surveillance update.
4. Limitation note. Each family records what its rationale does not claim. The Lanolin family, for example, notes that “highly purified lanolin may be better tolerated but is not guaranteed safe” — the source caveat travels with the family record so the app cannot misuse it.

Compounds and families that fail any of the four steps are not added. We do not have an “AI suggestion” ingestion path. The model that runs on user scans never writes to the family taxonomy.

Three worked examples

Three examples from different parts of the database. Each shows the same audit trail: anchor → rationale → members → limitation.

About our cross-reaction families

Our database organizes cross-reactive allergens into families grouped by chemical class and sensitization mechanism. Each family record includes an anchor source (NACDG screening series, ACDS designation, EU Regulation 1223/2009 Annex III, or SCCS opinion), member compounds, and a limitation note. The three examples above represent the depth and transparency standard applied across the database. Family records are updated as surveillance data is published; the maintenance schedule is below.

What the data does not claim

• This is not a substitute for patch testing. Patch testing under dermatologist supervision is the only definitive way to identify your personal contact allergens. Our database tells you which compounds are likely related to a known allergen; it cannot tell you which compounds you will react to.
• We do not predict reaction severity. Cross-reactivity is a probability, not a guarantee. A compound flagged as cross-reactive may or may not produce a clinically meaningful reaction in any individual patient.
• We do not score “safety” on a 0–10 scale. Composite numerical scores require methodology disclosure that consumer-facing scoring systems usually omit. Our verdicts (CLEAR / CAUTION / HIGH RISK) reflect whether your declared allergen profile intersects the product’s ingredient list, not a global safety judgment.
• We do not invent cross-reactivity. The model that runs on user scans cannot write to the family taxonomy. New families are added by the editorial process described above, not by AI inference at scan time.
• Our family records contain known duplicates from prior migrations. They are taxonomy artifacts, not evidence of breadth. We are consolidating them and the next methodology update will reflect the deduplicated count.
• We do not certify products as “hypoallergenic.” The term has no regulatory definition in the United States. A product that returns CLEAR against your profile is one that does not contain the ingredients you flagged or their cross-reactors — nothing more.

How this page is maintained

This page is updated when (a) a new NACDG biennial surveillance report is published in Dermatitis, (b) a new ACDS Allergen of the Year designation is announced, (c) the EU Cosmetics Regulation Annex III is revised, or (d) we consolidate migration-era duplicate family records. Each update is recorded with the editorial date in the page footer below.

If you spot a factual error or want to point us to surveillance data we should be citing, write to [email protected]. We respond within 48 hours and source-correction reports take priority over feature requests.

Methodology page first published 2026-05-10. Allergen counts reflect the production Firestore allergens collection as queried on 2026-05-16: 237 total allergen records, 233 active contact allergens.