Hepp et al. 2014 — FDA total-content survey of 150 U.S.-market cosmetics for As, Cd, Cr, Co, Pb, Hg, and Ni by ICP-MS and CVAFS
This U.S. Food and Drug Administration analytical survey, published in J. Cosmet. Sci., reports total-content measurements of seven inorganic elements (As, Cd, Cr, Co, Pb, Hg, Ni) in 150 cosmetic products spanning 12 product types purchased on the U.S. market in spring/summer 2011. The contract analytical work was conducted at Frontier Global Sciences (Bothell, WA) under FDA contract. The methodology used closed-vessel microwave-style oven HNO₃/HF/HCl digestion at 130 °C for 12 h (modified from the lipstick method validated in FDA’s 2007 and 2010 lipstick surveys), ICP-MS for As/Cd/Cr/Co/Pb/Ni, and cold-vapor atomic fluorescence spectrometry (CVAFS) for Hg. The paper is the FDA’s broad-survey companion to its earlier 20-lipstick (2009) and 400-lipstick (2012) Pb-only surveys, extending the analyte panel to seven elements and the product scope from lipsticks alone to 12 cosmetic product types.
Key numbers
All values are total-content concentrations in mg/kg (= ppm = µg/g) of finished cosmetic product as placed on market; mean and median are computed across the per-product replicate-mean values within each product-type group, with <LOD (NF) values set to zero and <LOQ but >LOD (TR) values included at the numerical mid-range reported. The 95th-percentile column in Table XXIX is computed across the full n=150 dataset. Where the source provides a parenthesised value (column “(value)”), that is the same statistic recomputed after Grubbs outlier exclusion at p = 0.05.
Limits of detection and quantitation (Table II, p. 126)
| Element | LOD (mg/kg) | LOQ (mg/kg) |
|---|---|---|
| Arsenic | 0.048 | 0.16 |
| Cadmium | 0.018 | 0.058 |
| Chromium | 0.052 | 0.18 |
| Cobalt | 0.0066 | 0.022 |
| Lead | 0.0084 | 0.028 |
| Mercury | 0.0010 | 0.0032 |
| Nickel | 0.032 | 0.10 |
LODs were calculated as three times the standard deviation, and LOQs as ten times the standard deviation, of ten replicate reagent blanks fortified with low-level standards.
Per-product-type summary statistics (Tables XVII-XXVIII)
Mean and median across the per-product replicate-mean values within each product-type group, in mg/kg. Values in parentheses are the same statistic recomputed after Grubbs outlier exclusion (p = 0.05) where the source reports an outlier-removed companion value.
| Product type | n | As mean (median) | Cd mean (median) | Cr mean (median) | Co mean (median) | Pb mean (median) | Hg mean (median) | Ni mean (median) |
|---|---|---|---|---|---|---|---|---|
| Eye shadow | 30 | 0.50 (0.43) | 0.032 (NF) | 940 (11) — three Cr-pigment outliers | 5.5 (2.6); 3.5 (2.4) after 64 mg/kg outlier removed | 4.2 (4.0) | 0.0068 (0.0025) | 16 (9.8); 5.2 (9.6) after 1600 mg/kg outlier removed |
| Blush | 30 | 0.51 (0.46) | 0.056 (NF); NF (NF) after 1.2 mg/kg Cd outlier removed | 23 (6.5) | 2.0 (1.5) | 3.9 (3.3) | 0.0027 (NF) | 8.0 (7.3) |
| Lipstick | 30 | 0.14 (0.13) | NF | 2.8 (1.5) | 0.86 (0.45) | 1.0 (0.82) | NF | 2.6 (1.8) |
| Lotion (medium/low priced) | 5 | 0.12 (0.10) | NF | TR (NF) | TR (NF) | 0.010 (NF) | NF | 0.11 (0.053) |
| Premium lotion | 5 | 0.14 (0.12) | NF | 0.084 (0.084) | 0.027 (NF) | 0.019 (NF) | NF | 0.062 (0.053) |
| Lotion with sunscreen | 5 | 0.12 (0.13) | NF | 0.11 (0.10) | 0.021 (0.019) | 0.19 (NF) | NF | 0.036 (0.044) |
| Mascara | 10 | 0.20 (0.17) | NF | 1.8 (0.99) | 2.8 (1.7) | 0.13 (0.074) | 0.0065 (NF) | 4.0 (2.3) |
| Foundation | 10 | 0.18 (0.14) | 0.0072 (NF) | 2.3 (2.0) | 0.77 (0.47) | 0.64 (0.69) | NF | 1.8 (1.2) |
| Body powder | 5 | 0.32 (0.29) | NF | 1.6 (1.7) | 0.82 (1.1) | 0.48 (0.21) | NF | 2.1 (1.6) |
| Compact powder | 5 | 0.50 (0.59) | 0.014 (NF) | 7.6 (4.8) | 4.8 (3.0) | 4.6 (4.4) | 0.0010 (NF) | 12 (9.1) |
| Shaving cream | 5 | NF | NF | 0.21 (0.16) | NF | NF | NF | NF |
| Face paint | 10 | 0.42 (0.28) | 0.02 (NF) | 1600 (2.7) — one Cr-pigment outlier | 1.4 (0.36) | 1.0 (0.44) | NF | 4.7 (1.3) |
NF = Not Found, i.e. result below the LOD for that element. TR = Trace, i.e. result between LOD and LOQ; summary statistics in the source set <LOD values to zero and include TR values at their numerical mid-range, per the source’s stated convention (Survey of Cosmetic Samples paragraph, p. 141).
All cosmetic categories combined (Table XXIX, p. 142; n = 150)
Outlier-excluded companion values in parentheses where the source reports them.
| Statistic | As | Cd | Cr | Co | Pb | Hg | Ni |
|---|---|---|---|---|---|---|---|
| Maximum | 1.7 | 1.20 (0.36) | 22000 | 64 (7.5) | 14 | 0.06 | 1600 (42) |
| Minimum | NF | NF | NF | NF | NF | NF | NF |
| 95th percentile | 0.95 | 0.067 (0.062) | 46 | 7.7 (7.2) | 6.9 | 0.010 | 18 (18) |
| Mean | 0.33 | 0.020 (0.012) | 300 | 2.2 (1.8) | 2.1 | NF | 16 (5.5) |
| Median | 0.21 | NF | 3.1 | 0.91 (0.91) | 0.85 | NF | 2.7 (2.7) |
The maximum-cell entries in Table XXIX correspond to individual outlier samples whose attributes are described in the Discussion (pp. 141-143): the 22,000 mg/kg Cr in one eye shadow is attributed to chromium-pigment use (“chromium oxide greens” or “chromium hydroxide greens”); the 1,600 mg/kg Ni and 64 mg/kg Co in one eye shadow were identified by Grubbs test (p = 0.05) as outliers and were rejected from category-summary t-tests; a qualitative X-ray fluorescence scan of the eye-shadow holder did not detect Co or Ni, suggesting the cosmetic itself (and not the container) was the source of the metals.
Regulatory comparison points reported by the source
The Discussion notes (p. 143) that “FDA limits the level of mercury to less than 1 ppm (1 mg/kg) in most cosmetics and none of the products examined exceeded that limit.” The U.S. regulatory framework discussed (p. 125): U.S. regulations under 21 CFR §73, 74, 82, and 700.13 specify limits for elemental contaminants in color additives used in cosmetics, but there are no limiting specifications for these contaminants in cosmetics themselves, with mercury being the one exception. The paper does not propose new thresholds or compare its findings to other-jurisdiction limits.
Methods (brief)
Sampling and sample selection. 150 cosmetic samples purchased between April 22, 2011 and August 16, 2011 by Frontier Global Sciences (Bothell, WA) according to a detailed list provided by FDA: 30 eye shadows, 30 blushes, 30 lipsticks, 15 lotions (5 medium-and-low priced, 5 premium/high priced, 5 with sunscreen), 10 mascaras, 10 foundations, 5 body powders, 5 compact powders, 5 shaving creams, 10 face paints. FDA developed the product list with primary emphasis on market share. Powder products and face paints were selected from individual web sites to span high-, medium-, and low-priced and niche items.
Sample digestion (Sample Digestion Procedure, pp. 135-137). 0.5 ± 0.1 g of analytical portion transferred to a 60 mL Teflon digestion vessel (Savillex 100% high purity). 25 mL of 3:1 HNO₃:HF mixture plus 3 mL of HCl added. Vessels sealed and heated in an oven at 130 °C for 12 h. Solutions then transferred to centrifuge tubes and diluted to 50 mL. For mercury analyses, a 4 mL aliquot was transferred to a separate tube and diluted to 40 mL with 5% (v/v) bromine monochloride. For ICP-MS analytes, the original digest was transferred to Teflon beakers, heated on a hot plate at ≤200 °C until volume reduced to 2-3 mL, then 25 mL of concentrated HNO₃ was added and the evaporation repeated three further times (so HF was largely eliminated and no boric-acid neutralisation was applied — distinguishing this method from prior lipstick methods that added boric acid to redissolve precipitated metal fluorides). Two method blanks were included in each digestion batch of ≤20 samples. One organic oil reference material (Conostan for non-mercury metals, SPEX for mercury) and one inorganic reference material (NIST 2702) were included in each digestion batch. Each sample was prepared in duplicate.
ICP-MS determination of As, Cd, Cr, Co, Pb, Ni (pp. 137-138). Perkin Elmer Nexion 300X ICP-MS. Isotopes used for quantitation (bolded in source list): ⁵³Cr, ⁶⁰Ni, ⁷⁵As, ⁵⁹Co, ¹¹¹·¹¹⁴Cd, ²⁰⁶·²⁰⁷·²⁰⁸Pb (sum of three Pb isotopes used to account for isotopic variation between samples and standards). Internal standards: ⁴⁵Sc, ¹¹⁵In, ¹⁹⁵Pt. Initial calibration with 8 calibration standards (Table I, p. 126); continuing calibration verification (CCV) standards analysed every 15 analytical solutions. Standard-curve correlation coefficient > 0.997 in all cases; CCV recoveries within 90-110%. Samples analysed with a 1/10 dilution of the diluting solution (1% HNO₃ + 0.5% HCl v/v). Absence of matrix effects verified by comparing one analytical solution from each digestion batch with the same solution at a 1/20 dilution; relative percent differences (RPDs) < 20%. Method-blank QC: analytical solutions with values above the highest standard were additionally diluted to bring them on-curve.
CVAFS determination of Hg (pp. 138-139). Tekran 2600 mercury cold vapor gold-amalgamation atomic fluorescence spectrometer. Mercury was analysed prior to the evaporation steps used to prepare samples for ICP-MS (to avoid loss of Hg through volatilisation). Solutions were first oxidised (1% v/v bromine monochloride solution prepared from 1.5% potassium bromide and 1.1% potassium bromate in concentrated HCl) to ensure all Hg was in a stabilised oxidation state, then treated with hydroxylamine hydrochloride (25% w/v in 1% HCl) to neutralise remaining oxidant, then reduced with stannous chloride (25% w/v in 15% HCl) through an online mixing system to form volatile elemental Hg vapour. A gas-liquid separator removed the Hg vapour, which was deposited on a gold bed through amalgamation; the gold bed was then heated to release the Hg vapour, carried by argon gas into the atomic fluorescence spectrometer at 253.7 nm. Calibration: blank plus 5 calibration standards (0.5, 1, 5, 2, 40 ng/kg) prepared daily from dilution of the primary stock standard (1000 mg/kg, High Purity Standards). CCV recoveries within 90-110%. QC reference materials: NIST 1641e (mercury in water).
Reference materials and recovery (Tables III and IV, pp. 127-128). Four matrix-matched reference materials were analysed: Conostan custom oil (1 µg/g As, Pb, Cd; 5 µg/g Cr, Co, Ni — no certified Hg), Spex custom oil (5 µg/g Hg), NIST 695 trace elements in multi-nutrient fertilizer, NIST 2702 inorganics in marine sediment, NIST 1573a tomato leaves. Recovery limits 70-130% for certified values < 10× the LOQ; 90-110% for certified values ≥ 10× the LOQ. Reported mean recoveries (Table IV): Conostan 93-106% across the seven elements; NIST 695 96-102%; NIST 2702 90-104%; NIST 1573a 90-96% (Pb not certified for NIST 1573a). One exception: Hg recovery from Spex custom oil ran 71-79% across triplicate weight tests (0.25, 0.50, 0.75 g) — lower than the 75-104% Hg recoveries on the inorganic and biological matrices. The authors note the cause is not understood and report the values uncorrected.
Speciation: total content only for all metals. ICP-MS without speciation chromatography; CVAFS quantifies total mercury after BrCl oxidation to convert all Hg species to Hg(II). Per CLAUDE.md Part 14, this paper’s As enters the wiki as tAs (not iAs), Hg as tHg (not MeHg), Cr as Cr (not Cr-VI) — the underlying measurements do not separate inorganic from organic species or oxidation states.
Implications
- Certification (HMTc). This paper contributes A-tier total-content occurrence data on the U.S. cosmetic market for seven of the HMTc-relevant elements (Pb, Cr, Ni, tAs, Cd, tHg, Co; Co is not on the HMTc 10-analyte panel per Part 14 but is in scope for category-fitness questions). Direct-evidence data are available for HMTc Cat 13 (Cosmetics and Personal Care — Leave-on) Rows 5 (foundation/powders/blush), 7 (eye makeup), and 10 (sun/suntan products); HMTc Cat 14 (Rinse-off) Row 7 (shaving cream/gel/foam); HMTc Cat 2 Row 12 (children’s face paint, to the extent the FDA’s “face paints” category includes products marketed to or used by children — the paper does not separate kids’ vs adult costume face paints). Per-product-type median values in the table above are the direct inputs HMTc would pool against literature-baseline percentiles in the Standards Workbench. The 22,000 mg/kg Cr, 1,600 mg/kg Ni, and 64 mg/kg Co outliers are pigment-based product-design choices (chromium-oxide green pigment, nickel/cobalt-rich mineral pigments) and would be relevant to a clean/dirty subcategory split if HMTc were considering pigment-based formulations as a separate subrow. The outliers were rejected by the authors’ Grubbs test for category-mean statistics but they are real measured values on real U.S.-market products and should not be silently dropped during HMTc pooling.
- Courses. Useful as a teaching case for (a) the FDA’s regulatory posture on cosmetic elemental contaminants (no limiting specifications for the cosmetic itself except Hg; limits only on color additives via 21 CFR §73/74/82), (b) the inadequacy of acid-only digestion (without HF) when cosmetic samples contain refractory mineral pigments such as titanium dioxide and mica — the paper’s modification of the prior lipstick method to add HF and remove boric-acid neutralisation is the methodological pivot, and (c) the use of isotope-sum quantitation (Pb at ²⁰⁶+²⁰⁷+²⁰⁸) to handle isotopic variation between cosmetic samples and calibration standards.
- App. Not directly relevant to ingredient
contamination_profile(the paper does not measure metals in cosmetic ingredients individually). The product-type-level medians and 95th-percentiles are however a useful prior for any future app feature that estimates contamination likelihood in cosmetic product categories from product type alone. - Microbiome: not addressed.
Wiki pages this source may touch
- lead
- cadmium
- arsenic-total
- mercury-total
- chromium
- nickel
- cobalt
- eye-makeup
- makeup-foundation-powders-blush
- lipstick
- body-hand-leave-on-skin-care
- sun-suntan-products
- dusting-talc-powders-adult
- shaving-cream-gel-foam
- face-paint
Verification notes
- 2026-05-18 fresh ingest (Claude Opus 4.7, autonomous v2.0 manual-fetch skill, daemon tick): NEW path. Three identity checks against
wiki/sources/returned no hits: no DOI assigned by J. Cosmet. Sci. for this article (paper carries no DOI text on first page or final page; J. Cosmet. Sci. historically did not register DOIs across all issues); raw_handleMFK_survey-of-cosmetics-for-arsenic-cadmium-chromium-c(52-char Kimi handle truncation) not present in any existing source page; cite-key stemhepp2014not present (hepponly appeared as an in-body author citation inopss2023-feasibility-action-limits-cosmetics-uk.mdandjitareanu2025-heavy-metals-cosmetics-overview.md— both reference this paper but neither is a wiki source page for it). PDF SHA-256b99fe3d498f139649b36f90725b5cad4a3aed2d73b51794fed375b86b85406dc. evidence_tier: A: FDA-authored, peer-reviewed J. Cosmet. Sci. paper reporting primary analytical data generated under FDA contract. Government regulator + peer review + primary data = A-tier per Part 13 evidence grading. Promoted above the typical-B-tier ICP-MS academic survey because the FDA authorship gives the data direct regulatory standing and because the methodology was validated against multiple matrix-matched certified reference materials with recovery documented in Table IV.metals: [tAs, Cd, Cr, Co, Pb, tHg, Ni]uses non-speciated abbreviations because ICP-MS and CVAFS as deployed here are total-elemental techniques.tAsnotiAs;tHgnotMeHg;CrnotCr-VI. Co retained even though it is not on the HMTc 10-analyte certification panel (per Part 14 the panel is Pb, tAs, Cd, MeHg, tHg, iAs, Ni, Al, Cr-VI, Sn) because the paper measures Co directly and Co exposure from cosmetics is a recognised allergic-contact concern discussed in the source’s introduction (citing prior eye-pencil dermatitis case reports).products:routing rationale:eye-makeup(Cat 13 Row 7) — direct evidence; 30 eye shadows (Table V, p. 129; summary Table XVII, p. 136) plus 10 mascaras (Table XI, p. 133; summary Table XXIII, p. 139), n=40 of the 150-sample dataset.makeup-foundation-powders-blush(Cat 13 Row 5) — direct evidence; 30 blushes (Table VI, p. 130; summary Table XVIII, p. 137), 10 foundations (Table XII, p. 134; summary Table XXIV, p. 140), 5 compact powders (Table XIV, p. 135; summary Table XXVI, p. 141), n=45.lipstick(Cat 13 — currently RETIRED in wiki/products/, redirect targetchildrens-lip-balm-mineral-bearing; this paper is adult lipstick data) — direct evidence; 30 lipsticks (Table VII, p. 131; summary Table XIX, p. 137), n=30. Routed despite retirement status following the precedent set today (2026-05-18) byalmukainzi2022-saudi-topical-cosmetics-icpms.md, which also routes adult lipstick samples through the retiredproducts/lipstickslug pending Karen’s Step 0 Lock of a Cat 13 adult-lipstick row. The retired-page redirect points to children’s lip balm, which is NOT the appropriate destination for adult lipstick occurrence values; this routing is provisional until Cat 13 adult lipstick is built.body-hand-leave-on-skin-care(Cat 13 Row 4) — direct evidence; 5 medium-and-low priced lotions (Table VIII, p. 132; summary Table XX, p. 138) and 5 premium lotions (Table IX, p. 132; summary Table XXI, p. 138), n=10. The paper does not specify whether each lotion is face-targeted or body-targeted; routing to the body/hand row is the closest broad slug. If a sub-set of these were marketed as face moisturisers, a future routing pass could reroute those toface-neck-leave-on-skin-care(Cat 13 Row 3); the source does not provide product-level subcategorisation to support that split now.sun-suntan-products(Cat 13 Row 10) — direct evidence; 5 lotions with sunscreen (Table X, p. 133; summary Table XXII, p. 139), n=5.dusting-talc-powders-adult(Cat 13 Row 2) — direct evidence; 5 body powders (Table XIII, p. 134; summary Table XXV, p. 140), n=5. “Body powder” in the FDA’s category list maps to adult dusting/talc rather than baby talcum (the paper does not say “baby powder” or describe pediatric labelling).shaving-cream-gel-foam(Cat 14 Row 7) — direct evidence; 5 shaving creams (Table XV, p. 135; summary Table XXVII, p. 141), n=5.face-paint(Cat 2 Row 12 in current wiki taxonomy, even though FDA’s category includes both kid-marketed and adult costume face paint) — direct evidence; 10 face paints (Table XVI, p. 136; summary Table XXVIII, p. 142), n=10. The source motivation for including face paints (p. 128, citing references 34, 35, 36) was “Adverse reactions compelling recall of some face paints” plus Campaign for Safe Cosmetics 2009 Halloween-face-paint advocacy — both of which are dominated by kid-marketed costume cosmetics. Routing to the kid-positioned Cat 2 face-paint slug therefore matches the source’s predominant product motivation; if a future Cat 13 adult costume-cosmetics row is built, this routing can be reviewed.
matrices: [cosmetic-personal-care]follows the established bare-string matrix vocabulary in use across cosmetic-occurrence sources (almukainzi2022, arshad2020, attard2022, bashir2025, jitareanu2025, opss2023, poulsen2007, rbeida2023, sccs2023, li2021).jurisdictions: [US]: 150 samples were purchased from U.S. retail and online channels in 2011 by Frontier Global Sciences on FDA contract. The regulatory frame discussed is U.S. federal (21 CFR §73, 74, 82, 700.13 for color-additive elemental limits; FDA non-regulation of cosmetics-themselves elemental content except for Hg).- Empty
ingredients:is intentional: the paper measures total-content metals in finished cosmetic products, not in specific cosmetic ingredients. The Discussion attributes the 22,000 mg/kg Cr eye-shadow outlier to “chromium oxide greens” or “chromium hydroxide greens” pigments and notes that high solid-filler content (clay, talc) and pigment content in eye shadows / blushes / compact powders is the likely source of the higher mean values in these product types — but these are qualitative attributions, not direct ingredient-level measurements. doi: null/no_doi_assigned: true: J. Cosmet. Sci. (Society of Cosmetic Chemists journal) did not consistently register Crossref DOIs for its 2014 articles; no DOI string appears anywhere in this paper’s text (front matter, footers, or reference list).access_url: https://library.scconline.org/points to the SCC Online publication library (the source of the “Purchased for exclusive use” stamp visible on every page of the PDF), which is the canonical online access channel.- Brand firewall (Part 12, strict reading locked 2026-05-17): the source identifies samples by anonymised manufacturer/brand letter codes (Manufacturer A, B, C, …, AA, BB, CC, DD, etc.; Brand a, b, c) and never names specific brands or manufacturers in any of Tables V-XVI. No brand-attribution stripping was necessary. Vendor names cited per Exception 2 (scientific-method vendors): Perkin Elmer Nexion 300X (ICP-MS), Tekran 2600 (CVAFS), Savillex (digestion vessels), Tyco Healthcare Group (centrifuge tubes), VWR International (centrifuge tubes), Frontier Global Sciences (contract analytical laboratory under FDA contract), CPI International (ICP-MS standards), Inorganic Ventures (check standards), High Purity Standards (primary Hg stock), Absolute Standards (secondary Hg stock), SCP Science (Conostan oil standard), SPEX CertiPrep (mercury custom oil standard), NIST (reference materials 695, 1641d, 1641e, 1573a, 2702, 1643e).
- Wiki/HMTc firewall (Part 2): the source’s overall framing is descriptive — “the information in this study will help FDA and others make appropriate decisions regarding elemental contaminants in cosmetics” (p. 144) — without proposing thresholds or harmonising with any specific regulatory ceiling beyond noting that no product exceeded the FDA’s 1 ppm Hg limit. This wiki page reports those findings as the paper presents them without adopting any threshold proposal as the wiki’s voice and without translating the per-product-type medians into HMTc threshold proposals. The Implications section frames the source as a direct percentile-pool input for Cat 13 and Cat 14 cosmetic rows (and Cat 2 face paint) without specifying what percentile or what threshold; that is the Standards Workbench’s job, not the wiki’s.
- Outlier handling. The source applies Grubbs test (p = 0.05) to identify and reject statistical outliers from category-mean t-tests: 22,000 mg/kg Cr (eye shadow) and 15,000 mg/kg Cr (face paint), 1,600 mg/kg Ni (eye shadow), 64 mg/kg Co (eye shadow), 1.2 mg/kg Cd (blush). The outliers are real measured values on real U.S.-market products (verified by replicate analysis and, for the 1,600 mg/kg Ni / 64 mg/kg Co eye shadow, by qualitative XRF scan of the eye-shadow holder confirming the metals were not in the container itself). The source reports both with-outlier and without-outlier summary statistics throughout Tables XVII-XXIX so downstream pooling can choose either basis without re-reading the PDF; this wiki page preserves both columns. Whether pigment-based product-design choices belong inside or outside a given row’s pooling envelope is a Standards Workbench question, not a wiki-page question, and is not adjudicated here.
- Speciation note. ICP-MS without chromatographic separation provides total elemental Cr — not Cr(III) vs Cr(VI). The chromium pigments used in cosmetic eye shadows are typically Cr(III)-based “chromium oxide greens” (Cr₂O₃) or “chromium hydroxide greens” (Cr₂O(OH)₄), per 21 CFR §73 color-additive listings; nothing in this paper measures the Cr(VI) fraction. Downstream HMTc work that needs Cr(VI)-specific data on cosmetics must use a different source (e.g., Kang et al. 2006, ion-chromatography post-column derivatisation, cited as reference 20 in this paper). Similarly the total-As values include both inorganic As (As(III), As(V)) and organic As species; the paper does not speciate. Hg values are total Hg by BrCl-oxidised CVAFS; methylmercury content is not measured.
- Audit subagent (2026-05-18, fresh-context general-purpose, autonomous v2.0 manual-fetch skill): REVISE verdict; 2 ❌ findings applied, 0 rejected; 1 ⚠️ wording concern applied; 6 product-slug ❌/⚠️ findings rejected as snapshot-staleness false-positive.
- CHECK 1 ❌ Eye-shadow Ni outlier-removed mean transcription error: prior wiki text said “16 (9.8); 5.5 (9.6) after 1600 mg/kg outlier removed” but Table XVII (p. 136) shows Ni mean = “16 (5.2)” not “16 (5.5)“. The “5.5” appears to have been mis-copied from the Co mean column (“5.5 (3.5)”). Verified independently against Table XVII. Corrected to “16 (9.8); 5.2 (9.6) after 1600 mg/kg outlier removed”.
- CHECK 1 ❌ Blush Cd outlier-removed mean transposition: prior wiki text said “0.056 (NF); 0.012 (NF) after 1.2 mg/kg Cd outlier removed” but Table XVIII (p. 137) shows Cd mean = “0.056;NF” — the outlier-removed parenthetical is NF, not 0.012. The “0.012” is the Table XXIX all-categories outlier-removed Cd mean (
0.020(0.012)) and was wrongly transposed into the blush row. Verified independently against Table XVIII. Corrected to “0.056 (NF); NF (NF) after 1.2 mg/kg Cd outlier removed”. - CHECK 5 ⚠️ Implications-section “should not be silently dropped during HMTc pooling” softened from prescriptive to descriptive register. Rewrote to “The source reports both with-outlier and without-outlier summary statistics throughout Tables XVII-XXIX so downstream pooling can choose either basis without re-reading the PDF; this wiki page preserves both columns. Whether pigment-based product-design choices belong inside or outside a given row’s pooling envelope is a Standards Workbench question, not a wiki-page question, and is not adjudicated here.” This preserves the substantive point (both bases are available; the choice is upstream of the wiki) while removing the wiki’s prescriptive voice toward HMTc pooling discipline.
- CHECK 2 ❌ FALSE POSITIVE on six product slugs (eye-makeup, makeup-foundation-powders-blush, body-hand-leave-on-skin-care, sun-suntan-products, dusting-talc-powders-adult, shaving-cream-gel-foam) flagged as “NOT in snapshot”: all six pages confirmed to exist at
wiki/products/<slug>.mdwith HMTc Cat 13 / Cat 14 row bindings populated. The taxonomy snapshot atdocs/gpt-collaboration/taxonomy-snapshot.mdwas last generated 2026-05-17 from a commit predating the Cat 13/14 product-page buildout and is therefore stale on the cosmetic-leave-on/rinse-off rows. This is the same snapshot-staleness false-positive pattern logged in 2026-05-18 audits of bae2018-absorbent-hygiene-pads-safety-review, lai2025, and barabash2025. Snapshot regeneration is a separate maintenance task and does not require source-page changes. Findings rejected; no slug changes applied. - CHECK 2 ⚠️
matrices: [cosmetic-personal-care]flagged as bare-string-not-in-snapshot: same bare-string matrix vocabulary precedent across 10+ cosmetic-occurrence sources documented in Verification notes; finding rejected for the same precedent reason; no change. - CHECK 3, CHECK 4: ✅ clean per audit.
Page history
The five most recent substantive edits to this page. The full version history lives in git; when DOI minting comes online (see schema docs), each entry below will also link to a version-pinned DataCite DOI.