Engel et al. 2016 — Campaign for Safe Cosmetics “Pretty Scary 2”: heavy metals in 48 Halloween face-paint items from 14 children’s kits

This Breast Cancer Fund / Campaign for Safe Cosmetics report (Engel, Nudelman, Rasanayagam, Witte, Palmer, October 2016) is the seven-year follow-up to the 2009 campaignsafecosmetics2009-pretty-scary-face-paints report. The heavy-metals component commissioned Weck Laboratories (City of Industry, California) to test 48 individual face-paint colors from 14 Halloween face-paint kits purchased from an unnamed online Halloween retailer for five metals: arsenic (tAs), cadmium (Cd), chromium (total Cr; no Cr-VI speciation), lead (Pb), and mercury (tHg). Cadmium was detected in 14 of 48 items (29%, range 0.58–14 mg/kg), chromium in 13 of 48 items (27%, range 1.4–12 mg/kg), lead in 9 of 48 items (19%, range 1.2–3.9 mg/kg), arsenic in 4 of 48 items (8%, range 1.1–1.9 mg/kg), and mercury in 0 of 48 items. Across the 48 colors, 21 items (about 44%) had detectable concentrations of at least one heavy metal, and several items contained two, three, or four detected metals. Concentrations and detection frequencies were qualitatively higher in black and other deeply pigmented colors than in lighter colors. A separate volatile-organic-compounds (VOC) component tested 65 items from 51 partner-purchased children’s lip balms, lip glosses, nail polishes, and stick-on nails; VOCs are organic chemicals outside HMI’s heavy-metals scope and the per-product VOC values are documented in the source’s Table 4 but not reproduced here. A label-reading exercise covered 187 unique ingredient lists from 158 children’s cosmetic products / kits and is documented at category-level frequencies only.

Note on evidence tier: This is a B-tier NGO-commissioned report rather than a peer-reviewed primary study. Weck Laboratories is an accredited independent laboratory and the analytical methods are EPA-validated (EPA 6010B for As/Cd/Cr/Pb; EPA 7471A for Hg), but the public document contains no replicate, matrix-spike, or method-blank QC data; no measurement uncertainty; and the n=48 sample (drawn from 14 kits = 14 manufacturer choices) cannot support population-level inference about the U.S. children’s-face-paint market. The 2009 predecessor’s self-limitation statement (“Testing 10 products does not provide a complete window on the entire face paint market”) applies to this 48-item follow-up as well, scaled up only modestly. Use the data as occurrence evidence demonstrating the persistence of heavy-metal contamination in commercially available U.S. children’s Halloween face paints in 2016, seven years after the 2009 report; do not treat the per-metal detection frequencies as representative of the broader market.

Key numbers

Sample frame — heavy-metals testing (What labels don’t tell us p. 21)

• Kits purchased: 14 Halloween face-paint kits ordered from an unnamed online Halloween retailer.
• Individual colors tested: 48 (kits contained multiple items; the laboratory tested each color separately).
• Purchase year: 2016 (report published October 2016).
• Laboratory: Weck Laboratories, City of Industry, California (accredited third-party laboratory; EPA-method-compliant).
• Analyte panel: arsenic (tAs), cadmium (Cd), chromium (total Cr), lead (Pb), and mercury (tHg) — five metals. Cobalt and nickel (which were on the 2009 panel) were not on the 2016 panel.
• Reporting limits stated by the report: “around .5 to 1 ppm depending on the test”; the per-sample detection-limit appendix shows Cd reporting limit of 0.5 mg/kg across all 48 items, Pb / As / Cr reporting limits of 1.0–2.2 mg/kg (typically 1.8–2.0 mg/kg for grease-makeup and palette colors and 1.0 mg/kg for crayon-format products), and Hg reporting limit of 0.01 mg/kg (one product 0.013 mg/kg).

Sample frame — VOC testing (What labels don’t tell us: Volatile organic compounds p. 26–27)

• Products / kits purchased: 51 children’s makeup products through 14 partner organizations and one individual purchaser; retail outlets named were Target, Toys R Us, Claire’s, Justice, and Dollar General. The report’s prose on p. 27 states “we received 51 products from 15 states including California.” Appendix 1 (p. 40) enumerates 14 partner-organization rows covering 13 distinct state codes (Alaska, Colorado, Connecticut, Georgia, Illinois, Kentucky, Maine, Massachusetts, Minnesota, North Carolina, Rhode Island, Tennessee, Washington); the California coordinator is acknowledged separately on the contributors page (p. 3 — Sara Schmidt, Breast Cancer Fund), bringing the verifiable state count to 14. The source-internal discrepancy between the prose’s “15 states” and the appendix’s 14 enumerable states is documented in Verification notes.
• Individual items tested: 65 (kits contained multiple items; the laboratory tested each item separately).
• Analyte panel: 50 volatile organic compounds (organic chemistry — out of HMI’s heavy-metals scope).
• Laboratory: an accredited third-party laboratory; the laboratory name is not given in the public report.
• Reporting sensitivity caveat: because personal-care products are complex matrices with many ingredients, samples were diluted to 1/1,000 of original concentration prior to analysis; this reduced sensitivity to parts-per-million-range detection rather than parts-per-trillion-range detection. VOCs present at very-trace levels would not be visible to this method.

Sample frame — label-reading exercise (What labels CAN tell us p. 14)

• Products / kits reviewed: 158 (93 multi-product kits + 22 Halloween products later tested for heavy metals + 39 VOC-tested products + 4 additional label-only items, per the report’s Table 1 reconciliation).
• Unique ingredient lists: 187 (kits often contained multiple items with different ingredient lists).
• Retail channels: Target, Toys R Us, Claire’s, Justice, Family Dollar, Dollar Tree, Dollar General.
• Frequencies reported: at category level (% of 187 unique ingredient lists containing each label-disclosed ingredient class) — see Table 2 in the source.

Heavy-metal occurrence — aggregate counts and concentration ranges (Table 3, p. 23)

The report’s Table 3 reports per-metal aggregate detection counts and concentration ranges across the 48 individual face-paint colors:

The report’s prose summary (p. 7 and p. 23) states that 21 of 48 items had at least one detected heavy metal — about 44% of the tested colors — and “some products contained as many as 4 metals” (p. 6). Mercury was not detected in any of the 48 items at the laboratory’s reporting limit of 0.01–0.013 mg/kg.

Heavy-metal occurrence — color/pigment pattern (Heat Map p. 24)

The report’s per-product / per-color heat map (p. 24) shows that heavy-metal detection frequencies and concentrations were qualitatively higher in black and other deeply pigmented colors than in lighter colors. Specifically (no brand attribution; the per-color findings are summarised by pigment colour and product format only — see Verification notes):

• Black colors (n=13 black-pigmented rows across the 48 items): highest cadmium concentrations in the dataset. The maximum cadmium detection (12–14 mg/kg, top heat-map band) was in a black face-paint color; additional black colors showed cadmium at 8–9.99 mg/kg, 6–7.99 mg/kg (two colors), and 4–5.99 mg/kg. The maximum chromium detection (10–11.99 mg/kg, top heat-map band) was also in a black grease-makeup face paint. Lead detections in black colors were generally in the 0.5–2.99 mg/kg range. Several black colors carried two or more co-detected metals (Cd + Cr; Cd + Cr + Pb).
• Blue colors (n=9 blue-pigmented rows): mostly light detections (0.5–2.99 mg/kg) of cadmium and chromium; one chromium detection at 2–3.99 mg/kg; one lead detection in the 2–3.99 mg/kg range.
• Red colors (n=11 red-pigmented rows): generally below detection across all five metals, except for one cadmium detection at 0.5–2.99 mg/kg and one chromium detection at 2–3.99 mg/kg in red face-paint colors.
• Yellow colors (n=5 yellow-pigmented rows): light cadmium detections (0.5–2.99 mg/kg) in three; otherwise below detection.
• Brown, green, orange, purple, white colors: generally light detections (predominantly cadmium 0.5–2.99 mg/kg), with one purple chromium detection at 4–5.99 mg/kg.

The report frames the pigment-color pattern as consistent with heavy-metal contamination originating in the colorant ingredients themselves (mineral / metal-oxide pigments used to achieve deep, opaque pigmentation), not from background contamination of all formula bases.

Comparison to the 2009 predecessor report (What labels don’t tell us p. 21, p. 23)

• 2009 report (campaignsafecosmetics2009-pretty-scary-face-paints): 10 face-paint kits, lead detected in 100% of products at 0.054 – 0.65 ppm (≈ 0.05 – 0.65 mg/kg). The 2009 study used a more sensitive Zeeman graphite-furnace AAS method (EPA Method 200.9) for lead, with detection capability into the low-ppb range.
• 2016 report (this source): 48 face-paint items, lead detected in 9 items (19%) at 1.2 – 3.9 mg/kg. The 2016 study used EPA Method 6010B (ICP family) with a reporting limit of roughly 1.0–2.2 mg/kg for lead.
• Report’s own interpretation (p. 23): “Our tests revealed that 19 percent (9 products) contained lead at levels above 1 ppm, the lower detection limit of our tests. In 2009, we only found lead at levels below 1 ppm (this was possible due to more specific and sensitive tests for lead). In other words, although we found lead in a higher percentage of products in 2009, the levels were actually lower than the levels we found in this year’s batch of products.” The drop in detection frequency between 2009 (100%) and 2016 (19%) is therefore primarily an analytical-method-sensitivity artifact, not a market improvement: the 2016 method could not detect lead below ~1 ppm and the 2009 detections were entirely below that floor, while the 2016 detections were entirely above it. The two datasets cover non-overlapping concentration ranges.

Heavy-metal regulatory context cited (Heavy metals and health p. 22; Heavy metals findings p. 23)

• U.S. EPA Safe Drinking Water Act lead MCL: 15 ppb (0.015 ppm). All 9 lead-positive face-paint items exceeded this drinking-water limit by approximately 80- to 260-fold (face-paint detections 1.2 – 3.9 ppm vs water limit 0.015 ppm). Comparison to a drinking-water limit is illustrative rather than directly applicable because face-paint exposure is dermal / hand-to-mouth, not via ingested water.
• U.S. FDA “no safe level of lead in food” framing: the report cites the FDA / CDC consensus that no safe blood-lead level has been identified, applied here as the public-health rationale for treating any lead detection in children’s face paint as a concern.
• U.S. CDC “no safe blood level of lead”: explicitly cited (p. 23) as the basis for the report’s framing of lead in children’s products as a serious concern.
• U.S. children’s-food lead benchmark: 0.5 ppm — referenced as a public-health comparator (p. 23). All 9 lead-positive face-paint detections (1.2–3.9 ppm) exceeded the children’s-food benchmark.
• NTP and IARC classifications cited (p. 22): arsenic and cadmium classified as known human carcinogens; lead classified as a reasonably anticipated human carcinogen; hexavalent chromium classified as a known human carcinogen (the analytical method used in this study cannot distinguish Cr-VI from total Cr); the report does not cite a numerical regulatory ceiling for any metal in cosmetics.

Detection limits — per-item reporting limits (Appendix 2 p. 41–42)

The report’s Appendix 2 documents per-item reporting limits across the 48 individual colors:

Per-item reporting limits varied by product matrix, with denser grease-makeup and palette colors requiring higher limits and crayon-format makeup achieving the 1.0 mg/kg floor. The report does not document method detection limits (MDLs), method blanks, matrix-spike recoveries, or replicate-measurement RPDs in the public document; the values in Appendix 2 are reporting limits / quantitation thresholds, not MDLs.

Label-reading exercise — frequencies of label-disclosed chemicals of concern (Table 2 p. 15)

For HMI’s heavy-metals focus, the label-reading data is documented as ancillary context rather than primary occurrence data. Frequencies are expressed as percentage of 187 unique ingredient lists (no brand attribution):

• Fragrance (undisclosed compound mixture): 50% of products
• Mineral oil (potentially PAH-contaminated): 40%
• Propylparaben (parabens family): 33%
• Ethoxylated ingredients (potentially 1,4-dioxane / ethylene-oxide-contaminated): 28%
• Talc (potentially asbestos-contaminated): 18%
• BHT, petrolatum, silica: 13% each
• Styrene acrylates copolymer (potentially residual-styrene-contaminated): 8%
• Ethylparaben: 4%
• Butylparaben: 3%
• DMDM hydantoin (formaldehyde releaser), epoxy resin (potentially residual-BPA-contaminated): 2% each
• Benzophenones (BP-1, BP-4), polymethyl methacrylate, diazolidinyl urea, imidazolidinyl urea, styrene copolymer, styrene resin: ≤ 1% each

None of these label-disclosed ingredients are heavy metals; the report’s framing is that they are common cosmetic ingredients whose manufacturing processes can leave behind residual heavy-metal or organic-chemical contaminants, not that they are themselves heavy metals.

VOC findings — aggregate counts and ranges only (Table 4 p. 29; Table 5 p. 30)

Volatile organic compounds are organic chemicals (not heavy metals) and the per-product VOC table is documented in the source but not reproduced here. For sample-frame completeness only: VOCs were detected in 13 of 65 items (20%). Seven distinct VOCs were observed across the 13 positive items — toluene (most common; 7 items, range 1.3 – 143 mg/kg), styrene (1 item, 22.4 mg/kg), ethylbenzene (2 items, 1.05 – 1.10 mg/kg), vinyl acetate (2 items, 78.9 – 229 mg/kg), acetone (2 items, 2.92 – 148 mg/kg), m,p-xylene (2 items, 3.62 – 3.80 mg/kg), and o-xylene (3 items, 0.58 – 1.39 mg/kg). Toluene and styrene are EPA-regulated in drinking water at 1.0 mg/L and 0.1 mg/L respectively; the report draws the regulatory comparison to drinking water rather than to a cosmetics ceiling because no U.S. cosmetics ceiling exists for these compounds in children’s products. VOCs are out of scope for HMI and are not propagated to the heavy-metals occurrence dataset.

Methods (brief)

Heavy-metals testing: 14 Halloween face-paint kits were ordered from an unnamed online Halloween retailer in 2016. Each kit’s individual colors were tested separately, yielding 48 individual face-paint colors. Samples were sent to Weck Laboratories (City of Industry, California), an accredited independent third-party laboratory. The laboratory used U.S. EPA Method 6010B (inductively coupled plasma optical emission spectrometry, ICP-OES, with EPA-prescribed acid-digestion sample prep) for arsenic, cadmium, chromium, and lead, and U.S. EPA Method 7471A (cold-vapour atomic absorption spectroscopy with manual digestion of solids using aqua regia + permanganate-sulfate-persulfate oxidation) for mercury. Per-item reporting limits were 0.5 mg/kg for cadmium across all 48 items, 1.0 – 2.2 mg/kg for As / Cr / Pb (varying by sample matrix; lower for crayon-format products, higher for dense grease-makeup and palette colors), and 0.01 – 0.013 mg/kg for mercury. Method detection limits (MDLs), method blanks, matrix-spike recoveries, replicate-measurement RPDs, and other quality-control parameters are not documented in the public report; the values stated in Appendix 2 are reporting / quantitation limits.

Speciation discipline: arsenic is reported as total arsenic (tAs); mercury is reported as total mercury (tHg); chromium is reported as total Cr (the analytical method cannot distinguish hexavalent chromium from trivalent chromium); cadmium and lead are reported as element-total. The 2009 predecessor report’s nickel and cobalt analytes were not included in the 2016 panel.

Volatile-organic-compounds testing (organic chemistry — outside HMI’s heavy-metals scope, documented for sample-frame completeness): 51 children’s makeup products (lip balm, lip gloss, nail polish, stick-on nails) were purchased by 14 partner organizations and one individual purchaser across 15 U.S. states. The 51 products yielded 65 individual items (some products contained multiple items). All 65 items were sent to an unnamed accredited third-party laboratory and tested for 50 volatile organic compounds. Because cosmetic matrices are complex and dilution to 1/1,000 of original concentration was required for analysis, reporting sensitivity was limited to the parts-per-million range; sub-ppm-trace VOC concentrations would not be detectable.

Label-reading exercise: the report documents 158 cosmetic products / kits marketed to children, yielding 187 unique ingredient lists. Products were obtained from Target, Toys R Us, Claire’s, Justice, Family Dollar, Dollar Tree, and Dollar General. Frequencies of label-disclosed ingredients of concern (parabens, fragrance, mineral oil, ethoxylated ingredients, formaldehyde releasers, etc.) are reported in Table 2 as percentages of the 187 unique ingredient lists.

Limitations the report itself or this ingest notes:

• Sample size and selection: 14 kits and 48 individual face-paint colors are too few and too purposively selected to support inference about the U.S. children’s-face-paint market. The 2009 predecessor’s caveat (“Testing 10 products does not provide a complete window on the entire face paint market”) applies, scaled up only modestly.
• Lead-detection floor: the 2016 method’s reporting limit for lead (1.0 – 2.2 mg/kg) is approximately 20-fold higher than the 2009 predecessor’s Zeeman graphite-furnace AAS method (which detected lead at 0.054 mg/kg). The 19%-detection figure (9/48) refers only to lead at concentrations above 1 ppm; the report’s prose explicitly notes that the apparent drop from 100% (2009) to 19% (2016) detection is a sensitivity artifact, not a market improvement.
• Chromium speciation: EPA 6010B cannot distinguish Cr-VI from Cr-III. The report flags chromium-oxide-pigment as one likely source of detected chromium without measuring speciation directly.
• Reporting discipline: no documented method blanks, matrix-spike recoveries, replicate measurements, or measurement-uncertainty estimates in the public document. The per-item Appendix 2 values are reporting limits, not MDLs.
• Heat-map interpretation: the report’s heat-map (p. 24) bins concentrations into seven colour bands (0.5–2.99, 2–3.99, 4–5.99, 6–7.99, 8–9.99, 10–11.99, 12–14 mg/kg). Exact per-item concentrations are not given in the public document outside Table 3’s aggregate ranges; only the binned heat-map and the aggregate per-metal range are usable.
• Internal Table 3 discrepancy on lead percentage: the printed Table 3 shows “Lead 9 (number) 4.6% (percentage)”; the prose on p. 7 and p. 23 both state 19% (9 / 48 = 18.75%). The Table 3 “4.6%” entry is a transcription error in the source. This wiki source page reports the prose-consistent value (19%) in the table above and documents the source-internal discrepancy in Verification notes.

Implications

Contributes occurrence data for product category face-paint at a larger sample size (n=48) than the 2009 predecessor (n=10) and with an expanded analyte panel for cadmium (which the 2009 predecessor’s panel did not measure as a primary analyte but the 2016 follow-up identifies as the most-frequently-detected metal: 14 / 48 = 29%, range 0.58 – 14 mg/kg). Per-metal aggregate counts and ranges plus the per-color pattern (dark / black pigments carry higher loads than light pigments) are usable as documented presence-evidence for tAs, Cd, total Cr, Pb, and tHg in U.S.-marketed Halloween face paints in 2016.

Contributes occurrence data for product category children’s-makeup through the VOC sub-study’s 51-product / 65-item children’s-lip-balm / lip-gloss / nail-polish / stick-on-nail sample frame, which documents that children’s cosmetic kits are sold through Target, Toys R Us, Claire’s, Justice, and Dollar General in U.S. states from coast to coast. The VOC measurements themselves are organic-chemistry and out of HMI’s heavy-metals scope.

Contributes time-series context to the 2009 → 2016 trajectory. The 2009 predecessor and this 2016 follow-up together demonstrate that heavy-metal contamination in U.S. Halloween children’s face paints persisted across the seven-year window despite intervening regulatory attention (Consumer Product Safety Improvement Act 2008–2009 children’s-products lead provisions; California’s Safe Cosmetics Act 2005 disclosure requirements; the EU Cosmetics Regulation 1223 / 2009 … superseding the prior Cosmetics Directive). The 2009 study’s Zeeman graphite-furnace AAS lead-detection floor (≈ 0.05 mg/kg) and the 2016 study’s EPA 6010B ICP lead-reporting floor (≈ 1 – 2 mg/kg) cover non-overlapping concentration ranges; a directly comparable temporal trend cannot be inferred from the count percentages alone. The 2016 detections (1.2 – 3.9 mg/kg) are above the 2009 detection range (0.054 – 0.65 mg/kg), but this is partly a function of the higher 2016 reporting limit excluding low-level positives from the 2016 count.

Contributes evidence on the dark-pigment / metal-load relationship for children’s face paint. The report’s heat-map indicates that black and other deeply pigmented colours systematically carry higher cadmium, chromium, and lead concentrations than red, yellow, white, and light colors, supporting the interpretation that the dominant contamination source is the mineral / metal-oxide pigment itself rather than background contamination of the formulation base. This pattern is consistent with the chromium-oxide-pigment / cobalt-blue-pigment / lead-chromate-yellow-pigment framework discussed in the 2009 predecessor and in the Washington Department of Ecology children’s-cosmetics testing programs.

Contributes regulatory context for the U.S. cosmetics-regulation gap. The report documents that the U.S. Federal Food, Drug, and Cosmetic Act dedicates only two pages to cosmetic safety (vs. 112 pages for food and drugs), that the FDA cannot require cosmetic recalls without going to court, and that the FDA cannot require manufacturers to register their cosmetic establishments. These framing claims are documented at the regulatory-context level and do not introduce numerical thresholds.

Methodological precedent: EPA Method 6010B (ICP-OES with acid digestion) for arsenic, cadmium, chromium, and lead in children’s cosmetics; EPA Method 7471A (CVAAS with aqua-regia / permanganate / persulfate digestion) for mercury in children’s cosmetics. These methods overlap the Washington Department of Ecology children’s-cosmetics testing program (stone2012-wa-qapp-parabens-metals-childrens-cosmetics / trumbull2017-wa-childrens-seasonal-products-2014-2015) and the broader EPA cosmetics-testing precedent ecosystem. The lack of documented MDL / MQL / QC parameters in the public document limits independent replicability and is the principal driver of this report’s B-tier rating.

Wiki pages this source may touch

• lead
• cadmium
• chromium
• arsenic
• mercury
• face-paint
• childrens-makeup
• children-personal-care
• makeup-body-paints-bases-fixatives
• childrens-lip-balm-plain
• childrens-nail-polish

Verification notes

• Brand firewall (CLAUDE.md Part 12 strict reading 2026-05-17): The source’s Table 3 aggregate (p. 23) is brand-neutral and is reported as-is. The source’s heat map (p. 24) lists the 48 individual face-paint colors with per-product brand identifiers (BATMAN, PAW PATROL, FACE PAINT, HALLOWEEN PALETTE, HORROR VALUE, HOWLEEN WOLF, LIPS ‘N LASHES, ABBEY BOMINABLE, SCARY CLOWN, SPIRIT MAKEUP CRAYON, ZOMBIE, ZOMBIE BOYS, ZOMBIE KIDS, and one “FACE PAINT” generic kit) including character-licensed merchandise (DC Comics, Nickelodeon, Monster High character properties). Per Part 12 strict reading, none of these brand identifiers is reproduced in this wiki source page; the per-color findings are summarised by pigment colour (black / blue / red / yellow / etc.) and product format (grease-makeup / palette / crayon) only. The aggregated heat-map binned concentrations (0.5–2.99, 2–3.99, … 12–14 mg/kg) are documented as range bands by pigment colour without brand attribution. The two strict-reading exceptions reviewed: Exception 1 (regulatory-event subject) does not apply (this is an NGO advocacy report, not an FDA recall or congressional investigation targeting specific brands). Exception 2 (scientific-method vendor / material names) applies to the analytical-laboratory name (Weck Laboratories, City of Industry, CA) and is preserved.
• Brand firewall, VOC component: The source’s Table 4 (p. 29) lists VOC concentrations against named children’s-makeup brands (Blow Pop, Disney Frog, Disney Frozen, Disney Princess, Hershey’s, Lisa Frank, Minions). Per Part 12 strict reading, no brand identifiers from Table 4 are reproduced in this wiki source page. The VOC component is also out of HMI’s heavy-metals scope; the aggregate per-VOC counts and ranges are documented at the sample-frame level only, without brand attribution.
• Source-type field: nonprofit selected, matching the precedent set by the 2009 predecessor campaignsafecosmetics2009-pretty-scary-face-paints and by houlihan2019-hbbf-whats-in-baby-food (HBBF 2019). The publisher (Breast Cancer Fund) and the coalition it coordinates (Campaign for Safe Cosmetics) are a U.S. 501(c)(3)-coordinated NGO advocacy structure, not a regulatory agency, peer-reviewed journal, or government body.
• Evidence tier B: small (n=48 individual colors from n=14 kit-level manufacturer choices) convenience sample, no documented QC data in the public report, advocacy framing. B-tier matches the 2009 predecessor and the HBBF 2019 precedent.
• Source-internal discrepancy — Table 3 lead percentage: Table 3 (p. 23) prints “Lead 9 (number) 4.6% (percentage)“. Both the report’s prose summary (p. 7: “we found lead in nearly 20% of the Halloween face paints tested”) and the report’s prose body (p. 23: “Our tests revealed that 19 percent (9 products) contained lead at levels above 1 ppm, the lower detection limit of our tests”) give 19% (9 / 48 = 18.75% ≈ 19%). The Table 3 “4.6%” entry is a transcription error in the source — likely a typesetting mistake in which “19%” was rendered as “4.6%” (perhaps a digit-swap or copy-paste error from a different denominator). This wiki source page reports the prose-consistent value (19%) in the Key-numbers Table 3 row and flags the source-internal discrepancy here so future readers and any downstream synthesis pass have the documentation trail. No correspondence with the report’s authors was attempted; the published document stands as the cited record.
• Speciation: tAs and tHg retained — no inorganic / organic speciation was performed (tAs was detected in 4 of 48 items at 1.1 – 1.9 mg/kg; tHg was below the 0.01 mg/kg reporting limit in all 48 items). Cr retained as total chromium — EPA Method 6010B cannot distinguish Cr-VI from Cr-III. Cd and Pb are element-totals. The metals frontmatter array [Pb, Cd, Cr, tAs, tHg] matches the report’s five-metal panel; cobalt and nickel (which were on the 2009 predecessor’s six-metal panel) are not on the 2016 panel and are not included.
• Units: all concentrations preserved in source units (mg/kg = ppm on a whole-product basis). The report uses ppm and mg/kg interchangeably and converts to ppb / mg/L only when drawing the EPA drinking-water-MCL comparison. The report does not specify wet-vs-dry-weight basis for the digested face-paint samples; the analytical methods (acid digestion of weighed wet-product subsamples, EPA 6010B / 7471A) report on a sample-as-digested basis without separate wet/dry differentiation, consistent with cosmetics matrix convention.
• Detection-floor mismatch with 2009 predecessor: documented explicitly above (Heavy-metal occurrence — comparison section and Methods Limitations bullet). The 2009 lead range (0.054 – 0.65 mg/kg) and the 2016 lead range (1.2 – 3.9 mg/kg) cover non-overlapping concentration intervals because the 2016 reporting limit (1 – 2 mg/kg) sits above the entire 2009 detection range. The headline “100% → 19%” trend reported informally in the source is therefore not interpretable as a market improvement; it is primarily an analytical-sensitivity artifact and the source itself notes this on p. 23.
• VOC component: the report’s VOC sub-study (Tables 4 and 5, p. 29–30) is documented for sample-frame and methods purposes only. VOCs are organic chemistry and are out of HMI’s heavy-metals scope; the per-product VOC values and the regulatory-comparison table are not reproduced. If a future HMI scope expansion includes organic contaminants in cosmetics, this source’s 65-item VOC dataset would re-enter scope and the per-product table can be reconstructed from the source PDF.
• Label-reading exercise: Table 2 frequencies (parabens, fragrance, mineral oil, ethoxylated ingredients, formaldehyde releasers, etc.) are documented at category-level only. These ingredients are not themselves heavy metals; the report frames them as ingredient classes whose manufacturing processes can leave behind residual contaminants. The label-reading dataset is not propagated to HMI’s heavy-metals contamination_profile generators.
• Heat-map interpretation: the source’s heat map (p. 24) bins per-item concentrations into seven colour bands. Exact per-item concentrations outside the binned bands are not given in the public document; the aggregate per-metal ranges in Table 3 are the only exact concentration data. The colour-binned per-color summary above (Black / Blue / Red / Yellow / etc.) is the maximum-fidelity brand-firewall-compliant representation of the heat-map’s pigment-pattern signal.
• PDF page references: all in-text page numbers in this source page refer to the printed page numbers in the report (page 1 = cover; page 4 = Introduction; page 21 = What labels don’t tell us: Heavy metals; page 23 = Table 3; page 24 = Heat map; page 26 = VOC introduction; page 29 = Table 4; page 30 = Table 5; page 40 = Appendix 1; page 41–42 = Appendix 2). These match the PDF reader’s page numbering 1:1 in the file at raw/Manual Fetch Kimi /June 1 Infant cosmetics metals papers/02_Regulatory_Reports/REG-025_CampaignSafeCosmetics_2016_Pretty_Scary_2_Kids_Makeup.pdf (raw_sha256 3f779dc3f658a56b7ef24dfdae46b4a155cdf1a9f008b8394802455fe27f3645).
• Audit subagent 2026-06-02 — findings applied:
  • Audit flagged a source-internal discrepancy in the VOC sample-frame state count: p. 27 prose says “15 states including California” while Appendix 1 (p. 40) enumerates 14 partner-organization rows covering 13 distinct partner-state codes. Verified against source — p. 27 does state 15 states; Appendix 1 lists 14 rows with state codes AK, CO, CT, MA, RI, GA, TN, ME, IL, MN, NC, KY, IL, WA (CT appears with one Clean Water Action row, MA and RI each one row; IL appears twice — Learning Disabilities Association and US PIRG; so 13 distinct partner states), plus the California coordinator Sara Schmidt named on p. 3 = 14 verifiable states. The source’s “15 states” claim is one off versus its own Appendix 1 + p. 3 enumeration. Wiki was tightened to document both the source’s prose count (15) and the verifiable enumeration count (14), so a hostile reviewer can reconcile both values against the source. Finding correct; revisions applied.
  • Audit flagged heat-map per-color row counts: wiki said n=14 BLACK and n=11 BLUE; auditor recounted to n=13 BLACK and n=9 BLUE. Verified against heat-map p. 24 — auditor counts confirmed (BLACK rows: ABBEY BOMINABLE, BATMAN, FACE PAINT, HALLOWEEN PALETTE, HORROR VALUE, HOWLEEN WOLF, LIPS ‘N LASHES, PAW PATROL, SCARY CLOWN, SPIRIT MAKEUP CRAYON, ZOMBIE, ZOMBIE BOYS, ZOMBIE KIDS = 13; BLUE rows: ABBEY BOMINABLE, BATMAN, FACE PAINT, FACE PAINT CRAYON, HALLOWEEN PALETTE, HORROR VALUE CRAYON x2, SPIRIT CLOWN, SPIRIT MAKEUP CRAYON = 9). Wiki Key-numbers Heat-map section updated from 14 → 13 BLACK and 11 → 9 BLUE.
  • Audit flagged “scary clown kit color” and “character-licensed kit” descriptors as borderline brand-firewall narrowing: “SCARY CLOWN” is the verbatim kit-identifier on the heat map (p. 24) and “character-licensed” narrows the field to ~5 BATMAN / PAW PATROL / Monster High rows. Verified — strict Part 12 reading 2026-05-17 prohibits even narrowing identifiers in body text when they map 1:1 to heat-map rows. Wiki tightened: “scary clown” descriptor removed, “character-licensed kit” qualifier dropped; per-color findings now refer to color/format only without kit-theme identifiers.
• Audit subagent 2026-06-02 — findings rejected: none. All three ⚠️ concerns were verified correct and applied.

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.