Salles et al. 2023 — Potentially toxic elements in costume cosmetics (face paints + pancakes), Brazil

This study quantifies twelve potentially toxic elements (PTEs: Al, As, Ba, Cd, Co, Cr, Cu, Ni, Pb, Sb, Sn, Sr) in 95 costume cosmetics (90 face paints + 5 pancakes) purchased in São Paulo, Brazil, and runs USEPA risk-assessment models for dermal-absorption and incidental-ingestion (hand-to-mouth) exposure pathways for both children (age 2 to <11 y) and adults (≥21 y, occupational scenario). Cancer risk for children ranged from 10⁻⁸ to 10⁻⁵; for adults in occupational scenarios, 10⁻³ to 10⁻⁵. Non-cancer hazard quotients (HQ) stayed below 1 except for dermal exposure in adults for cardiovascular/dermal targets (driven by As, ≈100% contribution) and “other” targets in pancakes (driven by Cr(VI), ≈99% contribution). Arsenic contributed approximately 90% of total cancer risk across all scenarios. Pancake exposure scenarios were associated with higher total risk values than face paint scenarios.

The paper cites the 2009 USA Campaign for Safe Cosmetics finding (its reference [10]) that all commercial face paints tested contained lead and 60% contained known skin allergens (Ni, Co, Cr) at higher-than-recommended levels.

Key numbers

Sample design:

• 95 samples total: 90 face paints + 5 professional pancakes
• 12 PTEs quantified: Al, As, Ba, Cd, Co, Cr, Cu, Ni, Pb, Sb, Sn, Sr
• Below-LOD substitution: LOD/√2 for values <LOD (18.7% for Cu; 9.4% for Cd; 6.3% for Sn)
• Excluded from statistical means (RSD >30%): Cu 28%; Cd 24%; Sn 21.9%; Sb 18.7%; Al 10.4%; As, Sr, Pb 3%; Ba 2%; Co, Ni 1%

Table 3 — Concentrations (mg/kg) in face paints and pancakes (n = number of samples retained after RSD filtering):

Two table cells transcribed exactly as printed in the paper but look anomalous: Ba pancakes P95 = 0.21 (below the reported min 3.42) and Pb pancakes P95 = 72.97 (well above the reported max 11.69). Both are likely typesetting transpositions in the published Table 3 — but the wiki preserves the source as printed rather than silently “correcting” it.

Compliance counts against published cosmetic-impurity limits (face paint samples only):

• German BVL technically-avoidable limits: Pb (2 ppm) exceeded by 2.2% of samples; Sb (0.5 ppm) by 2.7%; As (0.5 ppm) by 10.9%; Cd (0.1 ppm) by 12.7%
• Health Canada limits (Pb 10 ppm, As 3 ppm, Cd 3 ppm, Sb 5 ppm): Pb exceeded by 1.1% of samples
• US FDA color-additive limits (As 3 ppm, Pb 20 ppm, Hg 1 ppm): the paper does not tabulate Salles study compliance counts against these
• Brazil MERCOSUR technical regulation: lists permitted color additives (20 ppm Pb, 100 ppm other heavy metals) but does not set general cosmetic impurity limits

Color-stratified statistical signals (Kruskal–Wallis, p-values from the paper):

• White, blue, and purple paints had highest mean Pb and Cd (p = 0.01)
• Lilac, brown, and white paints had highest mean As (p < 0.01)
• Lilac, blue, and green had highest Cu (p < 0.01)
• Red had highest Ba (p < 0.01); brown and yellow had highest Co (p = 0.03)
• For most elements (Al, As, Ba, Cd, Co, Cr, Ni, Pb, Sb, Sn), means were higher in pancakes and liquid face paints than in cream/fluorescent face paints (p < 0.05)
• Cream and professional pancakes had higher means for Cd, Cr, Pb (p < 0.0001)
• Sr was higher in fluorescent and liquid paints
• Cu did not differ across product types

Table 4 — Cancer risk (CR) and non-cancer HQ summed across all PTEs (selected target systems with HQ approaching or exceeding 1 in italics in the source; values reproduced from the published table for child and adult exposure scenarios):

Children (2 to <11 y), face paints:

• Cancer risk: 8.97 × 10⁻⁸ (female dermal) to 7.70 × 10⁻⁶ (male total); range 10⁻⁸ to 10⁻⁵
• Non-cancer HQ all targets <1 (highest: cardiovascular/dermal in males, 1.71 × 10⁻³)

Children (2 to <11 y), pancakes:

• Cancer risk: 3.26 × 10⁻⁷ to 9.13 × 10⁻⁵; range 10⁻⁷ to 10⁻⁵
• Non-cancer HQ all targets <1

Adults (≥21 y, occupational), face paints:

• Cancer risk: 6.57 × 10⁻⁵ (female ingestion) to 1.28 × 10⁻³ (female total); range 10⁻⁵ to 10⁻³
• Non-cancer HQ: cardiovascular/dermal (female total = 7.13 × 10⁻¹; male = 6.16 × 10⁻¹); “other” (female total 1.28 × 10⁻¹; male 1.11 × 10⁻¹)

Adults (≥21 y, occupational), pancakes:

• Cancer risk: female 3.09 × 10⁻⁴ (ingestion) to 6.82 × 10⁻³ (total), with female dermal 6.51 × 10⁻³; male 2.67 × 10⁻⁴ (ingestion) to 3.65 × 10⁻³ (total), with male dermal 3.39 × 10⁻³; overall range 10⁻⁴ to 10⁻³
• Non-cancer HQ: cardiovascular/dermal (female total 1.57 × 10⁰; male 1.36 × 10⁰) — exceeds 1; “other” (female total 2.44 × 10⁰; male 2.11 × 10⁰) — exceeds 1

Carcinogenic risk contributions: arsenic ≈ 90% of total cancer risk; in pancake “other”-target HQ, Cr(VI) contributed ≈ 99%; in cardiovascular/dermal HQ, As contributed ≈ 100%.

USEPA model parameter values used by the paper (Table 1):

• Dermal absorption fractions (ABS): As 0.03; Cr(VI) 0.04; all other PTEs 0.001
• Conversion factor (CF): 1 × 10⁻⁶ mg/kg
• Adherence factor (AF, mg cm⁻² event⁻¹): children 196.08 (age 2-3), 166.67 (3-6), 151.52 (6-11); adult females 260.42, adult males 207.68
• Mass per application: 1000 mg children, 20,000 mg adults
• Skin surface area (SA): children head only 510 / 600 / 660 cm²; adult head + trunk 9630 (F) / 7680 (M) cm²
• Exposure frequency (EF, days yr⁻¹): children 2 (4 h/day, 12 d/yr); adults 83 (8 h/day, 250 d/yr — occupational)
• Exposure duration (ED, yr): children 1 / 3 / 5; adults 35
• Hand-to-mouth frequency (λ_D, contacts/hour): children 13; adults 8
• Hand-to-mouth transfer efficiency (f_D): 0.24
• Time of oral exposure (t, h/day): children 4; adults 8
• Body weight (BW, kg): female children 14.45 / 18.70 / 30.05; male children 14.95 / 19.02 / 29.46; adults 63.35 (F) / 73.25 (M)
• Averaging time: carcinogenic AT = 70 × 365 d; non-carcinogenic AT = ED × 365 d

Toxicity-value sources (Table 2): oral slope factor (SF_o) from California OEHHA for As (9.5) and Pb (0.0085) and from USEPA for Cr(VI) (0.5); oral RfD from USEPA/IRIS for As (0.0003), Cd (0.001), Cr(III) (1), Cr(VI) (0.003), Ni (0.02), Sb (0.0004), Sr (0.6); ATSDR MRL used for Al, Co (0.01), Cu (0.02), Sn (0.3); USEPA absorbed RfD (RfD_abs = RfD_o × ABS_gi) computed for Ba (ABS_gi 7), Cd (5), Cr(III) (1.3), Cr(VI) (2.5), Ni (4), Sb (15).

Methods

Sample collection: 95 samples (90 face paints across 4 Brazilian brands × 3 textures (liquid, cream, fluorescent) × multiple colors; 5 pancakes from 1 brand × 5 colors) purchased at the largest high-street commercial center in São Paulo. Two batches per type+color collected when available.

Sample preparation: 150–200 mg sample in triplicate, + 2 mL conc. HNO₃ (14 mol/L, sub-distilled, ~65% m/m; Synth, Brazil; sub-distilled in DST-1000, Savillex, USA). Overnight room-temperature pre-digestion. Heating in graphite-covered EasyDigest digester block (Analab, France) at 120 °C × 4 h (per Paniz et al. [16]). Dilute to 40 mL with 18.2 MΩ·cm deionized water (Master System All, Gehaka, Brazil). Filter 0.2 µm cellulose acetate prior to ICP-MS.

Instrumentation: Agilent 7900 ICP-MS (Hachioji, Japan). External calibration with multi-element standard (1000 µg L⁻¹ PerkinElmer) at 1, 5, 10, 20, 50, 100, 200 µg L⁻¹. Linearity ≈1.00 for most elements; 0.9999 for Pb. LOD = 3σ procedural blank / calibration slope × dilution factor (10 independent blank measurements).

Quality assurance: CRMs analyzed alongside samples — NIST 2709 (San Joaquim soil), ERM CC 141 (loam soil), NIST 1573 (Tomato Leaves), Agro 1003a (Tomato Leaves). Recoveries reported in supplementary Table S2.

Risk assessment: USEPA exposure modelling for dermal absorption (Eqs 1–6) and incidental hand-to-mouth ingestion (Eqs 7–8), with cumulative cancer risk and HQ across PTEs (Eqs 9–10). Dermal absorption fractions from USEPA Risk Assessment Guidance for Superfund. Slope factors for As and Pb from California OEHHA; for Cr(VI) from USEPA. Cancer risk computed for As, Cr(VI), and Pb. For Cr(VI), the entire measured total Cr concentration was assumed to be Cr(VI) (worst-case assumption — no speciation was performed).

Statistical analysis: Descriptive statistics (arithmetic mean, min, max, 95th percentile). Kruskal–Wallis test across colors and product types; Dunn’s test for multiple comparisons after significant Kruskal–Wallis (p < 0.05). Software: R [31].

Speciation: Total Cr by ICP-MS only — no Cr(VI) speciation; the paper applied a worst-case “all Cr is Cr(VI)” assumption to risk math but did not measure Cr(VI) directly. Arsenic reported as total (no inorganic/organic speciation). Total Hg not measured (Hg not on the analyte list).

Implications

Occurrence data this paper contributes:

• Brazilian high-street face-paint and pancake concentrations for 12 PTEs (Al, As, Ba, Cd, Co, Cr, Cu, Ni, Pb, Sb, Sn, Sr), n = 90 face paints + 5 pancakes. The Table 3 means, ranges, and P95 values are the principal occurrence data the wiki should carry into product-page pooling for face paint and professional pancakes.
• Color-stratified concentrations for face paints across 10–11 colors, with statistical-significance flags. White / blue / purple drive Pb and Cd; lilac / brown / white drive As; red drives Ba; brown / yellow drive Co.
• Texture-stratified differences: pancake and liquid forms tend to carry higher PTE concentrations than cream and fluorescent forms; cream and professional pancakes have the highest Cd, Cr, Pb.

Exposure-pathway data this paper contributes:

• A worked USEPA dermal + incidental-ingestion exposure model for a face-paint and pancake exposure scenario, with parameter values (Tables 1 and 2) sourced from USEPA, ATSDR, OEHHA, and the published exposure-factors literature. Dermal-absorbed-dose dominates cancer risk for adults under the occupational scenario (8 h/day, 250 d/yr). Incidental ingestion (hand-to-mouth) dominates cancer risk for children under the recreational scenario (4 h/day, 12 d/yr).
• Per-element contribution analysis: arsenic ≈ 90% of total cancer risk; Cr(VI) ≈ 99% of “other”-target HQ in pancakes (driven by the worst-case all-Cr-is-Cr(VI) assumption); As ≈ 100% of cardiovascular/dermal HQ.
• A 2.2%–12.7% non-compliance rate against the German BVL technically-avoidable limits across Pb (2.2%), Sb (2.7%), As (10.9%), and Cd (12.7%) for face-paint samples, and 1.1% non-compliance with Health Canada Pb (10 ppm) for face paints.

Cr(VI) caveat for downstream use: this paper’s Cr(VI) risk numbers should be treated as upper-bound, not measured — total Cr was substituted into the Cr(VI) risk equations rather than speciated. Any downstream synthesis on Cr(VI) in face paints needs a speciated source.

Courses: Worked example of USEPA dermal + incidental-ingestion exposure modelling for cosmetics. Equations 1–8 and the full parameter table are usable as-is for teaching how exposure assessment translates concentration data into risk.

App: Face paints and theatrical pancakes warrant a per-color contamination-likelihood signal rather than a single product-form signal, given the strong color-stratified signals for Pb, Cd, As, Ba, Co.

Microbiome: Not addressed.

Verification notes

• 2026-05-18 merge-enhance (Claude Opus 4.7) against the original PDF (17 pp.):
  • Key numbers section rewritten — prior page said “Specific per-element concentrations are tabulated in the paper but not in the abstract-level excerpts read here.” Replaced with the full Table 3 (per-PTE n, mean, min-max, P95 for face paints and pancakes), Table 4 risk summary (cancer + HQ by sex, pathway, age group), and the color-stratification statistical signals from §3.1. USEPA model parameters from Table 1 and toxicity-value sources from Table 2 added.
  • metals: array expanded from [Al, tAs, Cd, Cr, Cr-VI, Ni, Pb, Sb, Sn] to [Al, tAs, Ba, Cd, Co, Cr, Cu, Ni, Pb, Sb, Sn, Sr] to cover all 12 PTEs actually measured. Removed Cr-VI from metals: because the paper explicitly did not speciate (Section 2.5: “the entire concentration of total chromium determined was considered hexavalent chromium”). Per Part 14 speciation discipline, total-Cr-only papers list Cr, not Cr-VI. The Cr(VI) assumption is preserved in Methods and Implications as a worst-case modelling choice.
  • products/costume-cosmetics removed from frontmatter — not in current taxonomy (taxonomy snapshot 2026-05-17). The face-paint product page covers the form; children-personal-care and childrens-makeup cover the audience. Added childrens-makeup as a closer-fit slug than the deprecated costume-cosmetics.
  • Jurisdictions narrowed from [BR, US] to [BR] — the paper samples are 100% Brazilian-manufactured and Brazil-purchased; the US references are background-literature citations only, not study samples.
  • Synthesis claim deleted from opening paragraph: prior page said “This complements the Arshad 2020 Pakistani cosmetics finding… both papers converge on the conclusion that chronic dermal exposure…is the dominant carcinogenic risk pathway” — Part 2 firewall (no cross-source synthesis on source pages; synthesis is a Part 9 workflow).
  • HMTc threshold-policy proposals reframed in Implications. Prior numbered items 1–6 contained recommendations like “HMTc Cat 2 thresholds for face paints should anchor on dermal-absorbed-dose × lifetime-application-frequency” and “HMTc Cat 2 face-paint certification needs to default to ‘regular use’ assumptions.” Per Part 2 boundary (allowed: “contributes occurrence data”; not allowed: “suggests HMTc should certify at Y”), reframed as a description of the occurrence and exposure-pathway data the paper contributes, without recommending threshold logic or scenario defaults.
  • Legacy heading ”## Wiki pages updated on ingest” removed along with the bulleted-wikilink list — current schema uses the routing audit to compute fan-out automatically from frontmatter, not a hand-curated list at the bottom of the source page.
  • Two anomalous Table 3 values preserved as printed (Ba pancakes P95 = 0.21 vs min 3.42; Pb pancakes P95 = 72.97 vs max 11.69) with a one-line note that they look like typesetting transpositions but the wiki defaults to faithful transcription. Flag for any future synthesis pass that wants to use the pancake P95s.
• Preserved from prior page: cite_key, raw_handle (kimi-children-personal-care-salles2023 — pre-MFK_ format), raw_path, license, evidence_tier, sample_n.
• 2026-05-18 audit-application (Claude Opus 4.7, fresh-context subagent verdict REVISE, 3 ⚠️):
  • Applied (Table 4 attribution): prior “Cancer risk: 6.51 × 10⁻³ to 6.82 × 10⁻³ (female total) and 2.67 × 10⁻⁴ to 3.65 × 10⁻³ (male)” was a loose attribution — 6.51 × 10⁻³ is female dermal, not female total. Restated as: female 3.09 × 10⁻⁴ (ingestion) to 6.82 × 10⁻³ (total), with female dermal 6.51 × 10⁻³; male 2.67 × 10⁻⁴ (ingestion) to 3.65 × 10⁻³ (total), with male dermal 3.39 × 10⁻³. Verified against PDF Table 4 p. 9.
  • Applied (BVL non-compliance range): prior prose said “1.1%–12.7% non-compliance rate” but the BVL exceedances are Pb 2.2%, Sb 2.7%, As 10.9%, Cd 12.7% (range floor is 2.2%, not 1.1%). The 1.1% is Health Canada Pb. Restated as “2.2%–12.7%” with per-element percentages spelled out. Verified against PDF p. 11.
  • Rejected (Ni face paints min — false positive): subagent flagged Ni face paints min as “<0.01” (claimed wiki’s “0.45” was a misread). Independent re-verification against PDF Table 3 p. 8: PDF prints “0.45–1.24” for Ni face paints. Wiki value is correct; subagent finding was a transcription error on the subagent’s part, not the wiki’s. No change.

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.