Don’t take our word for it. Here’s every fact on this site, with its evidence grade, its caveats, and a link to the source, so you can check us.
We start you on the strong ones: findings that have been replicated or come from one large, well-controlled study. Moderate and weak facts are here too, labeled for what they are. Use “Link to this fact” to share a single card.
Hidden ingredients / Other evidence
On a US food label, 'spices' or 'natural flavor' can be listed as one word, but if a spice mix or flavor contains protein from one of the nine major allergens (milk, egg, fish, crustacean shellfish, tree nuts, peanuts, wheat, soybeans, sesame), FDA says the allergen must be named, in parentheses or in a 'Contains' line.
Strong evidence: Nonbinding guidance, but it restates the statutory duty in FD&C Act 403(w); applies only to protein-containing ingredients from the nine major allergens.
Source: FDA food allergen Q&A Ed. 5 (2025), FDA guidance for industry (Docket FDA-2022-D-0099) Link to this fact
Hidden ingredients / Other evidence
Even an allergen present only as an incidental additive inside a spice blend, such as wheat flour used as an anticaking agent, must be declared on the finished food: 'spices (wheat)' or 'Contains wheat'.
Strong evidence: Applies to intentionally added incidental additives, not to unintended cross-contact.
Source: FDA food allergen Q&A Ed. 5 (2025), FDA guidance for industry (Docket FDA-2022-D-0099) Link to this fact
Hidden ingredients / Government agency
Under US food labeling rules, "natural flavor" is defined by where the flavoring constituents come from (a spice, fruit, vegetable, yeast, herb, bark, bud, root, leaf or similar plant material, meat, seafood, poultry, eggs, dairy products, or fermentation products thereof) and by flavoring being the substance's significant function, not by safety, simplicity, or number of constituents (21 CFR 101.22(a)(3)).
Strong evidence: Text is the 1 April 2024 CFR edition; later amendments were not checked. The definition is quoted from the regulation, which is the law's own position, not a safety finding.
Source: 21 CFR 101 (4-1-24 ed.), Code of Federal Regulations, Title 21, Part 101 Link to this fact
Hidden ingredients / Government agency
A protein hydrolysate used in food for its flavor effect cannot be declared simply as "flavor", "natural flavor" or "flavoring"; onion, garlic, and celery powders must also be named (21 CFR 101.22(h)(7), (h)(3)). The collective term is therefore not unlimited.
Strong evidence: Only these named classes (plus salt and MSG) are carved out; other flavor constituents may stay under the collective term.
Source: 21 CFR 101 (4-1-24 ed.), Code of Federal Regulations, Title 21, Part 101 Link to this fact
Hidden ingredients / Other evidence
A Canadian extract or essence is a solution in ethyl alcohol, glycerol, propylene glycol, or any combination, which may contain water and a sweetening agent (9.1.1); a Canadian flavor may contain a sweetening agent, water, and edible vegetable oil (9.1.3).
Strong evidence: These are what the definitions name as permitted; the text does not say whether other substances are allowed or what must be declared on a label.
Source: CFIA Canadian Food Compositional Standards Vol. 9 (page dated 2025-05-01) Link to this fact
Hidden ingredients / Other evidence
Canada's vanilla standard (9.1.23) requires Vanilla Extract, Essence, or Flavor to be prepared from the vanilla bean (Vanilla planifolia or V. tahitensia) and to contain, per 100 mL, the soluble substances extractable from at least 10 g of beans (25% moisture or less) or 7.5 g moisture-free basis, by official method FO-17 (dated September 15, 1989).
Strong evidence: A product with vanilla flavor from non-bean sources would fall under the Artificial/Imitation sections by my reading of 9.1.2 and 9.1.4; no CFIA interpretation was captured.
Source: CFIA Canadian Food Compositional Standards Vol. 9 (page dated 2025-05-01) Link to this fact
Hidden ingredients / Other evidence
Canada's single-flavor standards set marker constituents for some flavors: cassia at least 80% cinnamic aldehyde (9.1.9), Ceylon cinnamon at least 65% cinnamic aldehyde and no more than 10% eugenol (9.1.10), lemon at least 0.2% citral from lemon oil (9.1.13) and peppermint oil at least 50% free and combined menthol (9.1.16).
Strong evidence: These are composition minimums or maximums for the flavor, not label disclosure; several markers are also common fragrance allergens.
Source: CFIA Canadian Food Compositional Standards Vol. 9 (page dated 2025-05-01) Link to this fact
Hidden ingredients / Other evidence
OEKO-TEX Standard 100 treats flame retardants and biologically active finishes as 'no intentional use' in all product classes, with exceptions for treatments OEKO-TEX itself accepts, and bans named flame retardants including TCEP, TCPP, TDCPP, PBDEs, HBCDD, antimony trioxide, and borates.
Strong evidence: The 'accepted treatments' list is online and was not captured, so the exceptions cannot be checked; Edition 02.2025.
Source: OEKO-TEX Service GmbH, Standard 100 limit values Annexes 4-5, Ed. 02.2025 Industry-funded Link to this fact
Hidden ingredients / Other evidence
Canada's Food and Drug Regulations say the ingredients and components of flavoring preparations, artificial flavoring preparations, spice or herb mixtures, and flavor-enhancer preparations need not be shown on the label of the food they are added to (B.01.009(2), table items 2, 3, 4, and 10).
Strong evidence: The text is a legal exemption, not a safety finding; Division 10 and CFIA guidance were not read here.
Source: src-2026-canada-food-drug-regulations-b-01-009-to-b-01-010-4-allergens, Canadian regulation Link to this fact
Hidden ingredients / Other evidence
Even inside Canada's exempt flavor and spice preparations, salt, glutamic acid or its salts, hydrolyzed plant protein, aspartame, potassium chloride, any component that 'performs a function in, or has any effect on, that food', and any supplemental ingredient must be listed by common name (B.01.009(3)).
Strong evidence: No test is given for what counts as performing a function; how CFIA applies paragraph (f) to flavor constituents was not read.
Source: src-2026-canada-food-drug-regulations-b-01-009-to-b-01-010-4-allergens, Canadian regulation Link to this fact
Hidden ingredients / Other evidence
Canada forces peanut oil, fully hydrogenated peanut oil, and modified peanut oil to be listed even when they sit inside an ingredient on the exempt tables (B.01.009(4)).
Strong evidence: The text gives no reason for the oil-specific override.
Source: src-2026-canada-food-drug-regulations-b-01-009-to-b-01-010-4-allergens, Canadian regulation Link to this fact
Hidden ingredients / Other evidence
The one flavoring rule left in Canada's Division 10 (B.10.008) is about wording: for an artificial or imitation flavoring, the word "artificial" or "imitation" must be part of the name, in identical type and identically displayed, on labels and in advertisements.
Strong evidence: It says nothing about what makes a flavoring artificial or imitation, or whether its ingredients must be listed.
Source: Canada FDR Division 10 (2026 consolidation), Canadian regulation Link to this fact
Hidden ingredients / Journalism
EWG's own Skin Deep method says its hazard score "does not account for exposure or individual susceptibility": it ranks how hazardous an ingredient is flagged to be, not how much of it you are exposed to.
Strong evidence: EWG describing its own method; a primary statement of what the score is, not a test of how well it predicts exposure.
Source: EWG Skin Deep methodology page (2026 capture) Link to this fact
Hidden ingredients / Journalism
Skin Deep scores a product from the ingredients printed on its label or website: its worst ingredient plus the average of the others, so the amount of each ingredient is not part of the score.
Strong evidence: Method as published by EWG; the page does not say whether the live calculation has changed since the described version.
Source: EWG Skin Deep methodology page (2026 capture) Link to this fact
Hidden ingredients / Journalism
A low Skin Deep hazard score can simply mean little is known: missing toxicity data carries zero weight in the hazard score, and EWG advises choosing products with low hazard AND at least "fair" data availability.
Strong evidence: EWG's own advice and weighting table. The data-availability rating is built half from PubMed study counts, so heavily studied ingredients rate higher regardless of safety.
Source: EWG Skin Deep methodology page (2026 capture) Link to this fact
Hidden ingredients / Other evidence
The CIR draft says 1,4-dioxane, a known animal carcinogen, can contaminate ethoxylated ingredients such as polysorbates and that industry 'reported that it is aware' of it and uses 'additional purification steps' (polysorbates are steam-stripped), but it reports no measured 1,4-dioxane or ethylene oxide levels.
Strong evidence: The absence of numbers is a fact about this draft only; other sources may hold specifications or FDA measurements.
Source: CIR 2015 draft Industry-funded Link to this fact
Hidden ingredients / Other evidence
In IARC's 2018 list, titanium dioxide (Vols 47, 93, 2010), carbon black (2010), diethanolamine (2013) and coconut oil diethanolamine condensate, the cocamide DEA surfactant (Vol 101, 2013), were each printed as Group 2B, possibly carcinogenic to humans.
Strong evidence: 2018 snapshot; 2B is limited evidence; hazard not risk; the list does not give exposure from any product.
Source: IARC 2018, Agents Classified by the IARC Monographs, Vols 1-123 Link to this fact
Hidden ingredients / Other evidence
The 2018 list printed talc without asbestos or asbestiform fibers as Group 3 and perineal use of talc-based body powder as 2B (Vol 93, 2010), with a note that asbestos-containing talc should be regarded as carcinogenic to humans.
Strong evidence: Superseded in part: talc was reclassified in 2024 (Vol 136); this page records only the 2018 printed status.
Source: IARC 2018, Agents Classified by the IARC Monographs, Vols 1-123 Link to this fact
Hidden ingredients / Other evidence
Drug makers must tell FDA the name of every inactive ingredient in a listed drug, but they can mark individual names confidential, and FDA withholds those names from the public when the claim is valid.
Strong evidence: This is disclosure to FDA, not on the label, and concerns drugs only (not cosmetics or fragrance). Text is the April 1, 2022 CFR edition; later amendments are not shown.
Source: 21 CFR 207 (2022 ed.) Link to this fact
Hidden ingredients / Other evidence
A US food label does not list every substance that touched or is in the food: processing aids that are removed or end up at insignificant levels, and packaging or equipment migrants that meet the food-additive rules, can legally go unlisted, so absence from the list does not show absence from the food.
Strong evidence: The exemption requires insignificant levels and no technical or functional effect in the finished food; the regulation does not say how often these substances are present.
Source: 21 CFR 101.100 (eCFR, 2026-10-15) Link to this fact
Hidden ingredients / Other evidence
In US law, 'natural flavor' is defined by where the flavoring constituents come from (plant, meat, seafood, poultry, egg, dairy or fermentation sources) and by a flavoring rather than nutritional function; the definition sets no limit on how many constituents it contains and no safety test (21 CFR 101.22(a)(3)).
Strong evidence: Reading of the text; the definition is silent on constituent count and safety rather than stating either.
Source: 21 CFR 101.22 (eCFR 2026), Code of Federal Regulations, Title 21, Section 101.22 (eCFR) Link to this fact
Hidden ingredients / Government agency
EU rules cap the manufacturing impurities in polysorbate additives at 0.2 mg per kg ethylene oxide, 5 mg per kg 1,4-dioxane, 0.25% ethylene glycols, and 3, 2, 1 and 1 mg per kg of arsenic, lead, mercury and cadmium; EFSA recommended revising the four metal limits downward.
Strong evidence: Regulatory specification text quoted in the opinion; JECFA allows 10 mg per kg 1,4-dioxane for E 435 and E 436. Limits, not measured levels.
Source: EFSA ANS Panel (2015), full opinion, EFSA Journal Link to this fact
Indoor air / Review of other studies
A chemical released indoors is roughly 1,000 times more likely to be breathed in than the same amount released outdoors (exact when comparing the low end of each range; only ~100x at the high end, since the ranges overlap at 10⁻³), so cleaning products and air fresheners (about 1% of California's smog-forming emissions) may deliver an inhaled dose on the same scale as all outdoor sources combined.
Strong evidence.
Caveats
Modeled population intake fractions (not measurements), population-average estimate. Traced to the primary source, Lai, Thatcher & Nazaroff 2000 (JAWMA 50:1688-1699), which derives outdoor PITF 10⁻⁶-10⁻³ vs indoor/in-vehicle PITF 10⁻³-10⁻¹ from Gaussian-plume and mass-balance modeling (Nazaroff co-author of both papers, so this is not independent replication). The ranges overlap at 10⁻³, so '1000x' is a floor-to-floor comparison, not a fixed multiplier under all conditions.
Source: Nazaroff & Weschler 2004, Atmospheric Environment Link to this fact
Indoor air / Review of other studies
Scent chemicals such as limonene and pine oil react with ozone indoors to make new pollutants, including formaldehyde and fine particles, that keep forming for hours after cleaning ends.
Strong evidence: Size of the effect depends on dose and indoor ozone; confirmed by later chamber and field studies.
Source: Nazaroff & Weschler 2004, Atmospheric Environment Link to this fact
Indoor air / Measured in people or real products
Limonene was detected in every childcare room studied and was more than 1,000 times higher indoors than outdoors, so it came from products used inside.
Strong evidence: Outdoor detected in 1/20 sites.
Indoor air / Measured in people or real products
Questionnaires about habits and routines correctly sorted only 20% of people by their measured exposure to cleaning and fragrance chemicals, vs 61% for traffic fumes; exposure depends on individual behavior and product choice.
Strong evidence.
Caveats
'Product choice' is the authors' explanation, not directly tested in this paper; independently corroborated by a larger, more direct 2021 UK study (Heeley-Hill 2021, 60 homes, daily diary-logged — not recalled — product-use frequency vs measured concentrations) that found cumulative product-use frequency has essentially no predictive power for measured indoor VOC concentrations (R²<0.001 for total VOC; no individual VOC significant).
Source: Edwards 2006, Atmospheric Environment Industry-funded Link to this fact
Indoor air / Review of other studies
Scent chemicals released by perfumes and scented products can react with ozone in indoor air to form new pollutants, including formaldehyde and fine particles.
Strong evidence.
Caveats
Formation is well established (Fiedler 2005: 40 ppb O3 + a limonene/alpha-pinene-containing VOC mix -> HCHO 13 -> 40 µg/m³, UFP 2,500 -> 46,000 cm-3; Singer 2006; Rossignol 2013; Destaillats 2006 gives the underlying per-ozone yield). Amounts depend on dose and ozone; the harm at real indoor levels is not shown. A far more rigorous, independent toxicologist review (Wolkoff & Nielsen 2017) reaches the same qualitative chemistry conclusion via its own literature synthesis and adds a quantitative caution: in the two largest human tests to date (Fiedler 2005, n=130; Fadeyi 2015, n=71), the resulting reaction mixture caused no significant airway/eye effects, and its own threshold estimate is that >200 µg/m³ ozone at high limonene would be needed before sensory effects are expected. Cite primaries for numbers.
Source: Alblooshi 2025, Frontiers in Toxicology Link to this fact
Indoor air / Measured in people or real products
Across decades of US monitoring where people carried personal air samplers, the air they breathed held more of almost every hazardous chemical than outdoor air did: exposure happens indoors, not at the outdoor monitoring station.
Strong evidence: Older studies include smokers' homes; compound-specific.
Source: Weisel 2002, Environmental Health Perspectives Link to this fact
Indoor air / Measured in people or real products
Air fresheners, deodorizers, and moth crystals were identified as the household source of p-dichlorobenzene, which reached about 300 µg/m³ indoors while being nearly absent outdoors.
Strong evidence: P-DCB is a deodorizing active, not a fragrance ingredient.
Source: Weisel 2002, Environmental Health Perspectives Link to this fact
Indoor air / Measured in people or real products
Formaldehyde in New Jersey homes averaged 55 parts per billion indoors versus 13 outdoors; compared home by home, indoor levels averaged about seven times the outdoor level (mean indoor/outdoor ratio 7.2).
Strong evidence.
Caveats
Regional sample; building materials are a major source alongside products The 7.2 is the mean of per-home ratios, not 55 ÷ 13 (about 4.2); corrected 2026-09-28 after a site-team check.
Source: Weisel 2002, Environmental Health Perspectives Link to this fact
Indoor air / Measured in people or real products
For p-dichlorobenzene, an often unlabeled deodorizer in air fresheners and moth repellents, 97% of people's exposure came from sources inside the home, not outdoor air.
Strong evidence: Deodorizer/pesticide active, not a fragrance ingredient.
Source: Wallace 1991, Environmental Health Perspectives Link to this fact
Indoor air / Measured in people or real products
In week-long air measurements across 11 European cities, the scent chemicals limonene and alpha-pinene were the compounds most concentrated in homes compared with workplaces (about 4x), attributed to perfumes, cleaning and fragranced products.
Strong evidence: Source attribution by the authors, not measured per product.
Source: Geiss 2011, Atmospheric Environment Link to this fact
Indoor air / Measured in people or real products
Limonene was 13x higher indoors than outdoors in European offices and classrooms, one of the two highest indoor/outdoor ratios of 23 chemicals measured.
Strong evidence: Ratio of medians.
Source: Geiss 2011, Atmospheric Environment Link to this fact
Indoor air / Measured in people or real products
Most European homes had low limonene (median 9.5 ug/m3 over a week), but the heaviest-use homes reached weekly averages up to 493 ug/m3, the range used in lab studies of indoor ozone chemistry.
Strong evidence: Upper tail only; median home 20-60x below chamber levels.
Source: Geiss 2011, Atmospheric Environment Link to this fact
Indoor air / Measured in people or real products
In a real test house, using a limonene-scented foam cleaner exactly as the label directed produced a burst of 45,000-150,000 tiny new particles per cubic centimeter within minutes, but only when ozone was in the air.
Strong evidence.
Caveats
One product, one house; particle number not mass. Independently replicated in a second real building (a working primary school; Morawska 2009), with a different detergent and country, at a similar order of magnitude (5.89-6.85×10^4 cm-3 in controlled tests) and the same water-only-no-burst, ozone-dependent pattern.
Source: Rossignol 2013, Atmospheric Environment Link to this fact
Indoor air / Measured in people or real products
Scent-type chemicals typical of fragranced and cleaning products (limonene, pinene, carene) were the signature of home air: close to zero outdoors but tens to hundreds of times higher inside homes.
Strong evidence.
Caveats
Small original study (n=7); the indoor-source signature (not the specific product-category attribution) is now independently replicated by a much larger 2021 UK study with real paired outdoor sampling (Heeley-Hill 2021, 60 homes: limonene indoor:outdoor ~19x, alpha-pinene ~10x), though that larger study also found cumulative product-use frequency does not predict concentration (only insecticide/plug-in-air-freshener use showed weak covariance with limonene) — so the indoor-source signature is strong, the specific 'cleaning products' attribution stays inferential. Givaudan-funded (COI); analysis independent.
Source: Gokhale 2008, Science of the Total Environment Link to this fact
Indoor air / Lab study (cells or chemistry)
What you breathe after using a scented cleaner is not only what was in the bottle: the scent chemicals keep reacting in the air to make new compounds.
Strong evidence: Well-established chemistry; magnitude indoors depends on ozone levels.
Indoor air / Review of other studies
Scented cleaners react with ozone indoors and release bursts of tiny new particles; every study able to separate them found this.
Strong evidence.
Caveats
Particle number, not mass; magnitude depends on instrument size cutoff and dose. Now confirmed across chamber, real test-house, real-office, and real-classroom settings (Rossignol 2013, Rosales 2022, Wu 2024, Morawska 2009).
Source: Salonen 2024, Environment International Link to this fact
Indoor air / Measured in people or real products
In a controlled chamber, the people themselves were the main thing reacting with indoor ozone: four seated adults removed about 60–65% of it, in two separate campaigns with men and with women, and their skin oils turned it into carbonyls such as 4-oxopentanal and 6-MHO. The same DTU chamber's protocol paper independently reproduced the ~60–69% removal in three more groups, teenagers and seniors as well as adults, across 26 experiments, and showed the sink is dermal (a breath-only control lost almost no ozone).
Strong evidence.
Caveats
Same DTU 22.5 m³ chamber and fan setup in all campaigns (Wang 2024: 4 men, ~37 ppb; Zannoni 2021: 2 groups of 2 women + 2 men, ~35 ppb; Bekö 2020: 5 groups — 3 adult, teenagers, seniors — same chamber/fans, ~35 ppb); mixing fans put deposition velocities at the top of the literature range, and a furnished home also loses ozone to surfaces, so the occupant share there will be lower. Chemistry only, no health endpoint; ICHEAR 6-MHO values may be inflated by an inlet artifact.
Source: Wang 2024, Environmental Science & Technology Link to this fact
Indoor air / Measured in people or real products
Four people in a clean, well-ventilated test room removed about two-thirds of the ozone coming in, confirmed in a second campaign that included women as well as men.
Strong evidence: Same DTU chamber as Wang 2024; ~65% is evidence base arithmetic from a ~100 ppb target inlet; furnished homes also lose ozone to surfaces.
Source: Zannoni 2021, Environmental Science & Technology Link to this fact
Indoor air / Measured in people or real products
When scent chemicals from a pine-oil cleaner, an orange-oil degreaser, and a plug-in air freshener met ozone in a lab chamber, about one formaldehyde molecule formed for every 3-5 ozone molecules they used up (yields of 20-30% in most tests).
Strong evidence.
Caveats
Same three products as Singer 2006, same research group/CARB contract, not independent replication; the highest single value (90%) occurred at air-freshener vapor loads 30-45x a real room's plug-in levels.
Source: Destaillats 2006, Environmental Science & Technology Link to this fact
Indoor air / Measured in people or real products
The moment ozone was added to pine-oil, citrus, and air-freshener scent vapors in a lab chamber, a burst of new ultrafine particles formed immediately and grew within minutes into fine particles, reaching hundreds of micrograms per cubic meter.
Strong evidence.
Caveats
Bench chamber at cleaning-pulse-scale vapor loads; ultrafine particles were only 1-4% of the mass at steady state, not the 4-16% the paper itself mislabels as an 'ultrafine' yield (that figure is total particle mass, all sizes).
Source: Destaillats 2006, Environmental Science & Technology Link to this fact
Indoor air / Measured in people or real products
In a controlled test room, people removed roughly 60–69% of the indoor ozone in every one of five very different groups tested, teenagers, young adults, and people in their 70s alike, showing this is a general property of occupied rooms with ozone present, not a quirk of one age group.
Strong evidence: Chamber study (22.5 m³, DTU, mixing fans push ozone deposition to the top of the literature range), fragrance-free hygiene products only, no health outcome measured.
Indoor air / Measured in people or real products
Limonene was the most prevalent and highest-concentration of 88 volatile organic compounds measured in a national survey of Canadian homes: detected in 99% of homes, with an arithmetic mean of about 45 ppb (roughly 250 ug/m3) and a geometric mean of about 25 ppb (roughly 140 ug/m3), levels well above limonene concentrations reported in Australian, European, and Californian indoor-air studies.
Strong evidence.
Caveats
The ppb-to-ug/m3 conversion is evidence base arithmetic (limonene MW 136.23, 25C/1atm), not stated by the authors, who reported concentrations only in ppb. Comparator studies used different sampling durations, locations, and methods, so the size of the gap should be read cautiously even though the direction (Canada higher) is clear.
Source: Dales & Cakmak 2019, Environmental Pollution Industry-funded Link to this fact
Indoor air / Other evidence
The mouse Alarie test (RD50, the airborne concentration that cuts breathing rate in half) predicts the no-effect concentration for irritation in human chamber studies well, with humans about 40 times more sensitive than the RD50 value would suggest.
Strong evidence: Cannot be used alone for a full guide value because the Alarie test only captures sensory irritation, not systemic effects that could occur at lower doses.
Source: Mangelsdorf et al. 2021, Int J Hyg Environ Health, International Journal of Hygiene and Environmental Health Link to this fact
Indoor air / Measured in people or real products
Tobacco smoke is the largest known source of styrene exposure for smokers, whose blood/exposure levels run roughly 6 times higher than non-smokers'; cigarette smoke has been estimated to contain 0.5-48 micrograms of styrene per cigarette depending on the assessment.
Strong evidence: Range spans an older (WHO 1983, 18-48 ug/cigarette) and a more recent (CDC 2010, 0.5-10 ug/cigarette) assessment using different methods; not a single agreed figure.
Source: IARC 2019 (Vol. 121) Link to this fact
Indoor air / Other evidence
Of 109 air fresheners bought in the Netherlands in 2009, 20 of the 24 regulated fragrance allergens turned up; linalool was in 87% of products and limonene in 69%, and one product was 15% linalool by weight.
Strong evidence: One survey of one national market in January 2009; weight fractions only, not what is emitted into air.
Source: ter Burg 2014, Inhalation Toxicology Link to this fact
Indoor air / Government agency
When limonene and linalool sit in air they oxidize into allergens: the EU committee said oxidized limonene and oxidized linalool rank among the most common contact allergens in patch testing (about 2-3% of consecutive patients for limonene at 3-5%, and 0.8-7.2% for linalool depending on test strength).
Strong evidence.
Caveats
Clinic patients, not the general population; test concentrations differ between studies (limonene 3-5%, linalool 2-11%); no published data on peroxide content of finished cosmetic products existed when written; transformation chain supported by multiple independent clinics but patch-test groups overlap (several Swedish centers).
Indoor air / Other evidence
The 2018 list printed d-limonene as Group 3 (Vols 56, 73, 1999), acetaldehyde as 2B, and styrene as 2A (Vol 121, listed 'In prep.'), alongside household combustion of biomass fuel (2A) and coal (1).
Strong evidence: 2018 snapshot; Group 3 means not classifiable, not safe; hazard not risk.
Source: IARC 2018, Agents Classified by the IARC Monographs, Vols 1-123 Link to this fact
Indoor air / Other evidence
Indoor ozone is typically 10-50% of the outdoor level (about 0.10 or less in air-conditioned homes, 0.2-0.4 in measured homes and schools), because surfaces and air exchange remove it.
Strong evidence.
Caveats
Values come from several independent field studies in California and Mexico City, summarized by the review author; they depend on air exchange rate and assume ozone penetrates fully.
Source: Weschler 2006, Environmental Health Perspectives Link to this fact
Indoor air / Other evidence
Terpene ingredients in scented cleaners, air fresheners, perfumes, colognes and essential oils (limonene, pinene, linalool, linalyl acetate) are listed as ozone-reactive indoor sources that make formaldehyde and tiny particles when ozone is present.
Strong evidence.
Caveats
Table 1 is a literature summary; perfumes and essential oils rest on three chamber citations (Chao 2005, Karamalegos 2005, Shu 1997). No product was measured in this paper and it gives no product-specific numbers.
Source: Weschler 2006, Environmental Health Perspectives Link to this fact
The industry / Lab study (cells or chemistry)
The lab tests that replaced animal testing for setting cosmetic and fragrance allergy safety limits explain only about 62% of how strongly a chemical causes allergy.
Strong evidence.
Caveats
Regression R2 against mouse LLNA, not human outcomes; industry-authored. Natsch 2023 (same author) retrains against human-anchored data: the 62% figure survives on a blended human+LLNA composite (63.4%, n=139), but against strictly human data (n=62) accuracy drops to 32-45%, and even the LLNA itself only reaches 43% predicting human potency — so '62%' should not be read as human-relevant accuracy. Api, Basketter & Lalko 2014 (RIFM's own earlier LLNA-vs-human validation, 57 fragrance chemicals) independently corroborates, by a different metric (categorical concordance, not R2) and seven years earlier, that the LLNA itself is an imperfect human proxy: 76% agreement within half an order of magnitude, but up to ~790-fold mispredicted for salicylates and ~26-42-fold for trans-2-hexenal/methyl-2-nonanoate. Roberts & Api 2018 (same fragrance dataset, direct LLNA-vs-human regression) reports a much higher R2=0.784, but only for a 35-chemical subset that already excludes those 14 known outliers a priori — demonstrating that domain-restricted R2 figures like this one can substantially overstate accuracy on unselected chemistry, a caution that applies to curated NAM training/case-study sets generally, not only to this specific figure.
Source: Natsch & Gerberick 2022, ALTEX Industry-funded Link to this fact
The industry / Lab study (cells or chemistry)
Accuracy has not improved: a 2015 model predicted allergy potency about as well as the 2022 one.
Strong evidence.
Caveats
Confirmed by reading the 2015 primary directly, not just inferred from one comparison line in the 2022 paper: Natsch 2015 Equation 7 gives R²(adj)=62.3% (Results, not the abstract's rounded 60%), matching 2022's 62% and 2023's 59-65%.
Source: Natsch & Gerberick 2022, ALTEX Industry-funded Link to this fact
The industry / Review of other studies
The cosmetics industry's own safety panel judged PGME 'safe' only for nail products and calculated only skin absorption; it did not estimate how much users or nail-salon workers breathe in, and relied on the smell to limit inhalation.
Strong evidence: Verdict probably right for occasional home users; the weakness is in the reasoning, not a demonstrated harm.
Source: CIR 2008, International Journal of Toxicology Industry-funded Link to this fact
The industry / Review of other studies
The Cosmetic Ingredient Review is funded by the cosmetics industry trade association, and the use data, exposure survey, and safety-margin calculation behind its 2008 PGME verdict were unpublished industry submissions.
Strong evidence: Much of the underlying toxicology is peer-reviewed; industry funding does not by itself make the conclusion wrong.
Source: CIR 2008, International Journal of Toxicology Industry-funded Link to this fact
The industry / Review of other studies
The fragrance industry's own standard caps this one ingredient at 0.003% in lip products and 0.02-0.05% in typical leave-on skin products – about 30 parts per million in the strictest category.
Strong evidence: IFRA Standards are voluntary; the limits apply to rose ketones individually or in combination.
Source: Lalko 2007 (RIFM), Food and Chemical Toxicology Industry-funded Link to this fact
The industry / Review of other studies
For this fragrance ingredient there was no data at all on absorption, repeat-dose toxicity, developmental effects, genetic damage, or cancer – the whole safety file is skin testing, and 8 of its 11 studies are unpublished company reports that outsiders cannot check.
Strong evidence: As of the 2006 literature search; later RIFM re-evaluations may have added data.
Source: Lalko 2007 (RIFM), Food and Chemical Toxicology Industry-funded Link to this fact
The industry / Review of other studies
Much of the evidence behind the fragrance industry's 'safe as used' verdict on salicylates is unpublished: about 65 of the cited studies are internal RIFM reports, several submitted by fragrance companies, and no modern cancer bioassay exists for any of the 17 chemicals.
Strong evidence: Counted from the paper's reference list; 'no modern bioassay' means untested, not shown to be carcinogenic.
Source: RIFM Expert Panel / Belsito 2007, Food and Chemical Toxicology Industry-funded Link to this fact
The industry / Self-reported survey
The study most often cited to call lavender and tea tree oils 'safe' for children was part-funded by the Australian Tea Tree Industry Association, the Tisserand Institute, and an aromatherapy association, while declaring no conflicts of interest.
Strong evidence: Funding is documented in the paper itself; funders are stated to have had no role. The paper's weakness is its power and design, not only its funding.
Source: Hawkins 2021, International Journal of Pediatrics and Adolescent Medicine Industry-funded Link to this fact
The industry / Review of other studies
The tea tree oil industry's trade association (ATTIA) challenged the lavender/tea tree case reports in a medical journal, and the same association later helped fund a survey that called the oils safe for children.
Strong evidence: Documents who argued what; says nothing about whether the oils cause breast growth.
Source: Ramsey et al. 2020 (JCEM reply); Hawkins et al. 2021, The Journal of Clinical Endocrinology & Metabolism Link to this fact
Information, not medical advice. See also: myths we won’t tell you.