Cooper et al. 2016 — Flame retardant screening of 1141 polyurethane-foam consumer products

This 8-page Environmental Science & Technology paper reports the results of a free public PUF-testing screening service operated by the Stapleton lab at Duke University between February 2014 and June 2016. The service received 1141 foam sub-samples from 567 U.S. households across 44 states and screened each for seven additive flame retardants (PentaBDE, Firemaster 550 / FM550, Firemaster 600 / FM600, TCIPP, TDCIPP, the isobutylated triaryl-phosphate mixture the authors call “TBPP,” and V6) by GC/MS in scan mode plus UPLC/MS/MS for V6. Across all samples, 52% (598/1141) contained a flame retardant at greater than 1% by weight; TDCIPP was the most frequently detected single FR (24% of all samples), followed by FM550 (10%), TCIPP (9%), and PentaBDE (6%). Within child car seats (n=98), FR detection frequency fell from 91% in seats purchased 2009–2013 to 50% in seats purchased 2014–2016 (p<0.0001), with TDCIPP detection in car-seat foam dropping from 77% to 48% and TCIPP detection rising from 29% to 52%, consistent with a TCIPP-for-TDCIPP substitution following TDCIPP’s 2011 addition to California’s Proposition 65 list. The paper measures no heavy metals and contributes no occurrence data to the HMI Pb/tAs/Cd/MeHg/tHg/iAs/Ni/Al/Cr-VI/Sn certification panel; it is ingested as an out-of-core-scope durable-infant-product contamination reference (per the precedent set by [[sources/stapleton2010-flame-retardants-baby-product-foam]] from the same lab, of which this paper is the larger 2014–2016 follow-on screening), routed to [[products/car-seats]] for discoverability as a child-car-seat additive-chemical exposure-pathway reference covering the post-TB-117-2013 regulatory transition window.

Key numbers

All detection-frequency figures below are pulled verbatim from Table 1 (p. 10655), Figure 1 (p. 10656), Figure 2 (p. 10656), Figure 3 (p. 10657), Figure 4 (p. 10658), and supporting narrative on pp. 10655–10658; n=1141 PUF sub-samples total, n=567 households.

Overall sample throughput (Table 1, p. 10655)

Notes from Table 1: “Values in parenthesis are percentages of products purchased in California under each column header.” The CA-purchase context matters because the paper interprets the temporal trends in light of California’s TB-117 (1972), TB-117-2013 amendment, and 2011 addition of TDCIPP to Proposition 65.

Per-product-category FR detection counts (Figure 1, p. 10656)

Figure 1 reports per-category detection counts for each of the seven analytes screened. Numbers below are read from the pie-chart annotations on Figure 1; sums per category exceed total-FR-positive counts in Table 1 because 90 of 1141 samples (8 %) contained more than one FR.

The paper does not report per-category counts for pillows, mattresses, pit cubes, other, or other furniture in Figure 1 in a form that survives reproduction in this table; the Supporting Information (Figure S1) is cited as containing the small-n categories. The “Other child products” row carries two cells the figure leaves unlabeled (TDCIPP and TBPP detection counts); the figure shows wedges of measurable size for both analytes but does not annotate them with integer counts.

Per-analyte detection frequency, all samples (narrative, p. 10657)

The remaining three screened analytes (FM600, TBPP, V6) are reported only at the per-category level (Figure 1) and at the per-purchase-year level (Figures 2–4); the paper does not report an overall-1141-sample detection frequency for them in the narrative.

Per-category detection frequencies and within-category headline figures (pp. 10655–10657)

• Sofas and loveseats (n=451): 55.4 % FR detection overall. The paper reports that “45 % [of sofas and loveseats] contained no identifiable FR” (p. 10655) — i.e., GC/MS peak unmatched to the seven-analyte standards panel; this 45 % figure closely matches the Table 1 no-FR-detected fraction (201/451 = 44.6 %). The authors note this is “much higher than that reported by Stapleton et al. (2012) for couches purchased between 1985 and 2010, for which only 12 % contained no known FRs” (p. 10655). Temporal series (Figure 3): for samples purchased ≥ 2005 (n=150), PentaBDE detection 2 % (down from 44 % pre-2005, n=75), TDCIPP detection 44 % (up from 41 % pre-2005), TCIPP detection 13 % (up from 0 %), TBPP detection 4 % (up from 0 %), FM550 detection 28 % (up from 5 %).
• Child car seats (n=98): 73.5 % FR detection overall (72 of 98). All car-seat samples submitted were purchased after 2008 (the paper attributes the absence of older samples to “the short lifetime of child car seats”). Of 2009–2013 purchases (n=53), 91 % contained FRs. Of 2014–2016 purchases (n=42), 50 % contained FRs (p < 0.0001 by one-tailed Fisher’s exact test for the pre/post-2014 contrast). Within-product temporal substitution (Figure 4b, n=48 pre-2014, n=21 post-2014, with mixed-FR products driving the post-2014 sum above 100 %): TDCIPP detection dropped from 77 % to 48 %; TCIPP detection rose from 29 % to 52 % (consistent with TCIPP-for-TDCIPP substitution); TBPP detection was 0 % in both windows (“TBPP was not observed in any car seat,” p. 10658); FM550 detection 4 % pre-2014 and 5 % post-2014. The paper notes that child car seats are federally regulated under FMVSS 302, not TB-117, which “may reflect” why car-seat FR substitution trends differ from sofa/loveseat trends.
• Chairs, rocking chairs, and recliners (n=174 combined; per Figure 2, n=39 1950–1999, n=13 2000–2004, n=79 2005–2013, n=16 2014–2016): The narrative on p. 10658 gives chairs/rockers/recliners-specific FR-detection percentages for the first two purchase windows: 36 % (1950–1999, n=39) → 62 % (2000–2004, n=13), described as a statistically significant increase. The paper then states only that detection “continued to increase … through 2013” without giving a chairs-specific percentage for 2005–2013, before reporting a statistically significant decrease (p=0.018) in 2014–2016 vs 2005–2013. Figure 2’s chairs panel shows the 2005–2013 bar at approximately 80 % and the 2014–2016 bar at approximately 31 %, but these chart-reads are not numerically anchored in the narrative for chairs alone. The 35 % / 61 % / 68 % / 37 % series sometimes seen for these windows comes from the pooled five-category narrative on p. 10657 (“For these product categories” = sofas + chairs + rockers/recliners + mattress pads + child car seats, n=124/85/390/132), not from chairs/rockers/recliners alone.
• Mattress pads (n=106): 29.2 % FR detection overall. No statistically significant temporal trend in FR detection frequency. Of four pre-2000 samples, 0 % contained an FR. 2000–2004: 18 % (n=11). 2005–2013: the source is internally inconsistent on this cell — the narrative on p. 10658 states “32 % of the samples purchased between 2005 and 2013 (n=71) … did contain a flame retardant,” whereas the Figure 2 mattress-pads panel shows the 2005–2013 bar at 21.1 %. The two figures cannot both be right; the wiki preserves both rather than silently choosing. 2014–2016: 29 % (n=15).
• Mattresses (n=71): 31 % FR detection overall.
• Pillows (n=32): 28 % FR detection overall.
• Pit cubes / gymnastics equipment (n=39): 85 % FR detection overall (33 of 39).
• Other child products (n=49): 43 % FR detection overall (21 of 49).
• Child mattresses (n=36): 36 % FR detection overall (13 of 36).

TDCIPP-prevalence narrative figures (p. 10657)

• Across all samples, TDCIPP is the most frequently detected FR (24 % of n=1141).
• TDCIPP is also the most frequently detected FR within every individual major product category covered in the paper’s primary analysis.
• TDCIPP was added to California’s Proposition 65 list in 2011; the authors interpret the post-2014 declines in TDCIPP detection (most pronounced in sofas, loveseats, and child car seats) as a plausible consequence of the combined effect of the TDCIPP Prop 65 listing and the 2013 amendment to TB-117 (smolder test replacing the open-flame test).

Co-occurrence and replacement-pattern figures (pp. 10656–10658)

• 90 of 1141 samples (8 %) contained more than one FR. No clear trend related multiple-FR application to product type. Most-frequent FR co-occurrence pattern across all categories: TDCIPP + TCIPP.
• In sofas and loveseats purchased after 2005, the detection frequencies of TDCIPP, FM550, TCIPP, and TBPP all increased relative to pre-2005, “suggesting that all of these FRs may serve as PentaBDE replacements.”
• No TCIPP was observed in sofas or loveseats purchased prior to 2005; 13 % of sofas and loveseats purchased after 2005 contained TCIPP.
• Within sofas and loveseats purchased before vs after 2014 (Figure 4a, n=215 pre-2014, n=18 post-2014): TDCIPP detection 47 % (pre) vs 11 % (post); TCIPP detection 7 % (pre) vs 22 % (post); TBPP detection 12 % (pre) vs 17 % (post); FM550 detection 20 % (pre) vs 17 % (post). The post-2014 n is small (n=18) and the authors caveat that “purchase dates may not reflect manufacturing dates, which may be obfuscating some of our results.”

Methods (brief)

Sample collection and metadata (p. 10654)

Participant recruitment and sample submission were handled through the project’s public web site (http://foam.pratt.duke.edu), bolstered by news and media reports on flame retardant chemicals, blogs (≈31 % of site traffic), social media (≈5 % of site traffic), academic partners, and relevant interest groups; search engines and direct traffic accounted for the remaining > 50 % of site traffic. Participants were U.S. residents age 18 or older. Each participant completed a questionnaire capturing contact information and per-product details: product type, whether the product was intended for a child or adult, the flammability standard label present on the product (CA TB-117 or CFR 1632), and the year and state of product purchase. Product categories captured in the questionnaire included sofas and loveseats (two-seat furniture pieces); chairs; rocking chairs and recliners; other furniture; mattresses; mattress pads (cushions that sit on top of and are separate from mattresses); child mattresses; child car seats; pillows; other child products; pit cubes / gymnastics equipment; and other (miscellaneous). After completing the questionnaire, participants cut a small marble-sized piece of PUF from each product, wrapped it in aluminum foil, sealed it in a plastic bag, and mailed it to Duke University. Approximately 6–8 weeks after submission, participants were mailed a letter providing their results. All study protocols were approved by the Duke institutional review board.

Materials (p. 10654)

The standard for the PentaBDE technical mixture, V6, EHTBB (TBB), BEHTBP (TBPH), and 2-ethylhexyl 2,3,4,5-tetrabromobenzoate were purchased from Wellington Laboratories. The standard for Firemaster 550 (FM550) was a gift from Dr. Susan Klosterhaus (who had previously received the mixture from Great Lakes Chemical, West Lafayette, IN). The Firemaster 600 (FM600) reference was prepared from extracted PUF foam from a North Carolina manufacturer who had used FM600 in their foam (Stapleton’s lab extracted FM600 from foam from this manufacturer and used that extract as the FM600 standard). Triphenyl phosphate (TPHP) was purchased from Sigma-Aldrich (St. Louis, MO). Tris(2-chloroethyl) phosphate (TCEP), tris(1-chloro-2-propyl) phosphate (TCIPP), and tris(1,3-dichloroisopropyl) phosphate (TDCIPP) were purchased from Sigma-Aldrich (St. Louis, MO), Pfaltz & Bauer (Waterbury, CT), and Fluka (St. Louis, MO). Dichloromethane (DCM), acetonitrile (ACN), water, and methanol were purchased from Honeywell (Muskegon, MI).

Sample preparation (p. 10655)

Method adapted from Stapleton et al. 2012. Briefly: 5 mL of DCM was added to a 50 mg PUF sub-sample in a glass test tube. The mixture was sonicated for 10 minutes. A 100 µL aliquot of the extract was diluted to 1 mL for GC/MS analysis. A laboratory blank (test tube containing DCM but no foam) was included with each batch of samples processed. Extracts were first screened by GC/MS. For samples requiring further analysis by LC/MS/MS (i.e., samples in which V6 was a candidate identification), a 100 µL aliquot was exchanged for 200 µL of methanol and diluted to 1 mL with water. The paper notes that small sample size precluded thorough physical-property testing (density) or PUF-type classification.

GC/MS analysis (p. 10655)

A suite of mass-spectrometric analyses identified individual FR chemicals and evaluated commercial mixtures. FRs in extracts were analyzed by GC/MS in scan mode under both electron-ionization (Agilent 7890 GC, Agilent 5795C MS, Wilmington, DE) and electron-capture negative-ionization (Agilent 6890 GC, Agilent 5975N MS) modes. Chromatographic separation was conducted on a DB-5 column (Agilent) using a thermal gradient: 40 °C for 1 minute, 18 °C/min to 250 °C, 1.5 °C/min to 260 °C, 25 °C/min to 300 °C, hold at 300 °C for 20 minutes. Initial inlet temperature 80 °C for 0.3 minutes, then ramped to 600 °C/min to 275 °C. Transfer line held at 280 °C. Chromatograms were screened against a custom spectral library for the known target FRs. To assist in detecting FRs likely intentionally applied to the foam to meet flame-retardant standards, FRs were semiquantitatively analyzed using authentic standards listed above, with the comparison concentration set to 1 % FR by weight of foam. This concentration threshold was chosen “based on the amounts of FRs typically applied to foam to meet flame retardant standards to limit and avoid false positives.” All blanks were screened alongside samples; FRs were rarely detected in blanks, and whenever detected the blank levels were well below the 0.1 %-by-weight concentration that would have triggered a sample-positivity flag.

UPLC/MS/MS analysis for V6 (p. 10655)

V6 confirmation/quantification used an Accela UPLC system (ThermoFisher Scientific, Bremen, Germany) coupled to a ThermoFisher TSQ Vantage triple-quadrupole mass spectrometer. UPLC separation used a Hypersil Gold C18 column (ThermoFisher, 100 × 2.1 mm, 1.9 µm particle size) with an ACN:water gradient (20 % ACN held 2 minutes, ramped to 99 % ACN at 13 minutes, held at 99 % ACN for 2 minutes). Injection volume 20 µL, flow rate 300 µL/min. Ion transition 582.9 m/z → 360.8 m/z used for quantification; transitions 582.9 m/z → 296.8 m/z and 582.9 m/z → 98.9 m/z used for confirmation. Seven FR commercial mixtures total were evaluated in this screen using authentic standards: FM550 (TPHP, isopropylated TPHP isomers, TBB, TBPH); FM600 (TDCIPP); V6; PentaBDE (containing TPHP, isopropylated TPHP isomers, and BDE congeners 47, 85, 99, 100, 153, and 184); “TBPP” (containing TPHP and monoiso- and triisobutylated TPHP). Structures of all FR compounds evaluated in the screen are listed in Table S1 of the Supporting Information.

Quantitation discipline (p. 10655)

This is a screening study, not a quantitative occurrence study. The paper reports detection / non-detection of each FR at a 1 %-by-weight threshold; it does not report continuous concentration values (ppm or mg/kg) per sample, does not report per-sample concentration distributions or summary statistics, and does not report per-FR limits of detection in units below the 1 %-by-weight screening threshold. The “FR detected” call is a binary above/below the 1 %-of-foam threshold. Per the authors (p. 10655): “this concentration was selected based on the amounts of FRs typically applied to foam to meet flame retardant standards to limit and avoid false positives.” The 1 %-by-weight threshold was chosen because below that level, GC/MS screen-mode peaks become difficult to distinguish from background; sub-1 % FR-treated foam is therefore counted as “FR not detected” in this study even when some FR may genuinely be present at lower concentrations.

Statistical reporting (pp. 10656–10658)

Pre/post-2005 and pre/post-2014 within-category contrasts were tested with one-tailed Fisher’s exact tests. The paper reports two-sample comparisons with reported p-values: chairs, rocking chairs, and recliners post-2013 decrease (p = 0.018); 2014–2016 vs 2005–2013 FR-detection decrease across the primary categories combined (p < 0.001); child car seats pre/post-2014 FR-detection decrease (p < 0.0001).

Limitations explicitly acknowledged by the authors (p. 10658)

• Public-service-derived survey: participants may not report product information (purchase state, year, label information) accurately.
• Demographic non-representativeness: not all demographic groups participate equally; participants are self-selected via web/blog/social-media outreach, with California residents disproportionately represented (24 % of households vs ~12 % U.S. population share).
• Purchase date ≠ manufacture date: post-2014 purchases may include products manufactured earlier; this caveat is most consequential for the post-2014 FR-detection-trend analyses given the relatively short post-TB-117-2013 enforcement window (TB-117-2013 enforcement began 2014).
• 1 %-by-weight detection threshold: products treated with FRs below 1 % by weight are counted as “FR not detected.” The screening method is not designed to catch sub-1 % additive levels or reactive/polymeric FRs covalently bound to the foam matrix.
• Small n for post-2014 product subsets: the post-2014 mattress-pad n is 15; the post-2014 sofa/loveseat n is 18 (or n=132 across all four sofa/loveseat purchase windows combined); the post-2014 chair/rocker n is 16. Trend estimates with small post-2014 n carry wide uncertainty.

Implications

• Certification (HMTc): No direct relevance. The paper measures only chlorinated organophosphate (TDCIPP, TCIPP), brominated (PentaBDE, FM550 brominated components TBB and TBPH), and triaryl-phosphate (TBPP, FM550 phosphate component TPHP, V6) flame retardants, and contributes no data to the HMTc 10-analyte panel (Pb, tAs, Cd, MeHg, tHg, iAs, Ni, Al, Cr-VI, Sn) for the HMTc rows that would cover the studied product categories (child car seats, plus the broader furniture categories which do not currently carry HMTc rows). The exposure pathway documented (additive flame retardants leaching from PUF into indoor dust and onto product surfaces, with sustained dermal and inhalation exposure to occupants) is also out of scope for the HMTc heavy-metals certification model. The paper is preserved as a documentation reference for the post-TB-117-2013 (and post-2011-Prop-65) chemical-substitution pattern in U.S. child-car-seat foam (TCIPP replacing TDCIPP) and for the broader context that durable-infant-product foams have been a route of recurring chemical-additive exposure shifts over the 2005–2016 window. The comparable HMI-scope question on car-seat foam (acid-digestion + ICP-MS determination of total-metal content in PUF, plus the upholstery textile and dye/pigment layers that sit on top of the foam) is not addressed in the current corpus and is flagged here as an evidence-base gap for the [[products/car-seats]] page.
• Courses: Useful as a methodology reference for GC/MS scan-mode screening of polyurethane foam at the 1 %-by-weight FR threshold, for the design of public-participation screening services as a route to large-n consumer-product chemistry datasets, and as a documented case study of how a regulatory shift (TB-117 → TB-117-2013, 2011 Prop 65 TDCIPP listing) reshapes additive-chemical use patterns in a consumer-product foam class (with a clearly detectable substitution signal from TDCIPP to TCIPP in child car seats specifically). The 91 % → 50 % FR-detection drop in child car seats across 2009–2013 vs 2014–2016 purchase windows is a useful illustration of how rapidly market formulations can shift in response to regulatory pressure on a single chemical.
• App: No app-relevant content. The contamination_profile data structures consume metals concentrations from ingredient and product pages, not flame-retardant-mixture detection-frequency data from PUF-substrate consumer products.

Verification notes

Audit subagent (2026-06-09) returned REVISE verdict with one ❌ and three ⚠️ findings; all four were verified correct against the PDF and applied: (a) the chairs/rocking-chairs/recliners 1950–1999 detection percentage was corrected from “35 %” to “36 %” (the wiki had earlier taken the pooled five-category 35 % figure from p. 10657 and misattributed it to chairs alone; the chairs-specific narrative on p. 10658 says 36 %), and the unsourced “81 % (2005–2013)” was reframed as a Figure 2 chart-read rather than presented as a narrative figure; (b) the mattress-pads 2005–2013 cell was rewritten to surface the source-internal contradiction between the p. 10658 narrative (32 %) and the Figure 2 panel (21.1 %) rather than silently picking one; (c) the sofas/loveseats “45 %” sentence was rewritten to match the source claim (“45 % of sofas and loveseats contained no identifiable FR,” closely matching the Table 1 no-FR-detected fraction 201/451 = 44.6 %) rather than the earlier misleading gloss that read the 45 % as the unidentifiable-FR fraction within the FR-positive subset; (d) the frontmatter sample_population now notes that the source is internally inconsistent on whether the collection window began January or February 2014 (abstract and Methods say February; Results header says January). No false positives in the audit subagent’s findings; no findings rejected.

Source page newly created 2026-06-09 from MFK_05-unknown (raw file raw/Manual Fetch Kimi /June 8/Kimi_Agent_Download Corruption Issue/_extracted_infantdurable_03_Carriers_HighChairs_CarSeats/03_Carriers_HighChairs_CarSeats/05_Unknown.pdf, raw_sha256 03bf0334a43cb4a8db96b3dfc4b13801ae3c144e0ade53df2ac3300d00ca6950). Identity checks against wiki/sources/ (DOI 10.1021/acs.est.6b01602, raw_handle MFK_05-unknown, cite-key cooper2016) returned no hits. The Stapleton 2010 Dioxin-proceedings precursor paper from the same lab ([[sources/stapleton2010-flame-retardants-baby-product-foam]]) is recorded as a near_duplicate; the present paper is the 2014–2016 follow-on screening at much larger scale (n=1141 vs n=101). Both papers are out of HMI-core-scope (organic flame retardants, no heavy metals) and are routed through the wiki for their child-durable-product additive-chemical-exposure context per the precedent the Stapleton 2010 page established. Routed to [[products/car-seats]] only (the paper’s child-car-seat sample is n=98); the paper has no stroller, nursing pillow, or other-HMI-product-page sample populations. License recorded as ACS AuthorChoice per the open-access banner on p. 10653 (“This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.”). Evidence tier B (peer-reviewed ES&T paper from an established lab, public-participation survey-based design with explicit selection-bias caveats acknowledged by the authors and a 1 %-by-weight screening threshold that censors sub-1 % FR-treated samples — both of which prevent A-tier).

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.