Flame Retardant Exposure in Vehicles Is Influenced by Use in Seat Foam and Temperature

Overview

This ten-page peer-reviewed Environmental Science & Technology article reports the first study to use silicone passive samplers to characterise flame-retardant (FR) chemicals in the cabin air of US personal vehicles, paired with sub-sampled vehicle-seat polyurethane (PUR) foam from a subset of the same vehicles. Authored by the Stapleton lab (Duke Nicholas School) in collaboration with the Green Science Policy Institute (Blum, Jahl, Soehl) and the Diamond group at University of Toronto, the study deployed silicone samplers (suspended from the rearview mirror for 7 days) in n = 101 vehicles in winter 2022 and a paired summer sampler in n = 54 of those vehicles, with n = 52 paired vehicle-seat-foam sub-samples. Forty-eight analytes were measured by GC/MS and one additional analyte (2,4,6-tribromophenol) was measured by LC/MS (silicone samplers); the foam sub-cohort was separately analysed for 15 BFR/OPE compounds by a dual GC/EI-MS + GC/ECNI-MS instrument pair as described in Methods (brief) below. Of the 49 targeted FRs, 17 were detected in at least one silicone sampler, dominated by organophosphate esters (OPEs, 12 detected) over brominated flame retardants (BFRs, 6 detected). Tris(1-chloro-isopropyl) phosphate (TCIPP, a sum of three isomers) had a 99% silicone-sampler detection frequency and was the dominant FR detected in seat foam (23 of 52 foam samples positive). Silicone-sampler TCIPP concentrations were 2–5× higher in summer than winter, with linear-mixed-effects modelling showing a ~12% TCIPP-sampler-concentration increase per 1 °C rise in average ambient temperature, and TCIPP-positive seat foam associated with cabin-sampler concentrations ~4× higher in winter and ~9× higher in summer than vehicles whose seat foam was TCIPP-negative. Estimated cabin-air TCIPP concentrations (derived from the Okeme et al. 2018 generic silicone-wristband sampling rate of 1.5 m³/day/dm²) had a median of 56 ng/m³ in winter and 180 ng/m³ in summer. The study identifies vehicle seat foam as one of multiple in-cabin TCIPP sources (other FR-containing components named by the authors include headrests, ceiling headliners, interior padding, and expanded-polystyrene foam components), interprets the post-2011 TCIPP-for-TDCIPP substitution pattern as a ‘regrettable substitution’ given the 2023 NTP report of TCIPP carcinogenicity, and concludes that FMVSS 302 (the US Federal Motor Vehicle Safety Standard for cabin-interior burn resistance, unchanged in substance since the 1970s) should be re-evaluated against the cabin-air exposure burden it imposes via additive FR use.

The paper measures no heavy metals and contributes no occurrence data to the HMI / HMTc 10-analyte certification panel (Pb, total arsenic (tAs), cadmium (Cd), methylmercury (MeHg), total mercury (tHg), inorganic arsenic (iAs), nickel (Ni), aluminum (Al), hexavalent chromium (Cr-VI), Sn). It is ingested as an out-of-core-scope exposure-pathway reference per the precedent set by wiki/sources/stapleton2010-flame-retardants-baby-product-foam (same first-author lab, baby-product foam FRs) and wiki/sources/anderson2002-uk-us-diaper-voc-emissions-mice (diaper VOC emissions, also metals: []). It is routed to [[products/car-seats]] for discoverability as the cabin-environment context in which infant car-seats are deployed; this is an indirect-relevance routing (the paper measures the vehicle’s own seat foam and cabin air, not the infant car-seat itself) and is documented as such in Verification notes below.

Key numbers

All concentration values are in ng FR per g silicone sampler (ng/g) unless otherwise noted; cabin-air estimates are in ng/m³. Source citations refer to the published paper Hoehn et al. 2024 (Environ. Sci. Technol. 58:8825–8834), DOI 10.1021/acs.est.3c10440. Page numbers refer to the published-article page numbering (8825–8834).

Sample population (Table 1, p. 8827)

Vehicle model years 2013–2022 (representation skewed toward 2015–2022 per the study inclusion criterion ‘model year 2015 or newer’ stated in the Abstract; the n = 1 (2013) and n = 1 (2014) winter rows in Table 1 represent two vehicles that fell slightly outside the headline inclusion criterion, and the n = 1 (2013) and n = 0 (2014) summer rows reflect the same).

Silicone-sampler detection frequencies, ranges, and medians (Table 2, p. 8828)

All units ng FR per g silicone sampler (ng/g). <x.yz indicates the method detection limit (MDL); concentrations below the MDL were imputed as MDL ÷ √2 per the Statistical Analysis section (p. 8827). n/a (not applicable) is the source’s notation for median not calculated due to detection frequency below the 60% statistical-analysis threshold.

A total of 17 of the 49 targeted FRs were detected in at least one silicone sampler (Abstract; Results, p. 8827), comprising 12 OPEs and 6 BFRs. Among the BFRs, 2,4,6-tribromophenol (2,4,6-TBP, the one analyte measured by LC/MS rather than GC/MS) was the most frequently detected, at 22% of winter and 43% of summer silicone samplers (Results, p. 8827).

Winter–summer paired silicone-sampler comparison (n = 54 paired sub-cohort, Figure 2, p. 8828)

Median sampler-concentration increases from winter to summer in the n = 54 vehicles sampled in both seasons:

• TCIPP: 56 ng/g (winter) → 231 ng/g (summer); 4.0–4.6-fold (the paper text states ‘increased ∼4–6-fold’ in the published wording on p. 8828, paired with the table-level 56 → 231 median values; the in-text fold range is the source’s wording).
• TEP: 1.6-fold higher in summer.
• TIBP: 4.9-fold higher in summer.
• TNBP: 5.3-fold higher in summer.

Wilcoxon Signed Rank (paired) test: p < 0.001 for all four (TCIPP, TEP, TIBP, TNBP). Average exterior temperature in this paired sub-cohort rose from 7.4 °C (winter) to 21.9 °C (summer) (Results, p. 8828).

Vehicle-seat-foam detection (Vehicle Seat Foam Analyses, p. 8830)

• Foam samples analysed: n = 52 (Recruitment text: ‘Fifty-one of 101 participants collected a foam sample from a vehicle seat,’ Abstract; Results-narrative: ‘Of the foam samples analyzed (n = 52),’ p. 8830; Figure 4 caption: ‘n = 51 winter and n = 28 summer’). The source-internal count discrepancy (51 vs 52) is documented in Verification notes.
• FRs detected in 33 of 52 foam samples.
• TCIPP: 23 of 52 foam samples positive (most commonly detected FR in seat foam).
• TDCIPP: 11 of 52 foam samples positive (second most commonly detected).
• Other FRs detected in foam at lower frequency (counts not reported in body text): V6 (BCMP-BCEP), Thermolin 101, 2,4,6-tribromophenol (2,4,6-TBP), tris(2-chloroethyl) phosphate (TCEP), decabromodiphenyl ether (BDE-209).

Foam-as-source effect on cabin-sampler TCIPP (Figure 4, p. 8830)

Median silicone-sampler TCIPP, stratified by whether TCIPP was detected in the paired seat foam:

Wilcoxon Rank Sum test: p < 0.05 in both seasons (Figure 4 caption).

Temperature association — linear mixed effects model (Figure 3, Table S5)

Exponentiated β coefficient = fractional change in silicone-sampler concentration per 1 °C increase in average ambient temperature. Linear mixed-effects model run on log₁₀-transformed concentrations, restricted to the four >60%-detection-in-both-seasons OPEs:

Estimated cabin-air TCIPP concentrations (Predicting Cabin Air Concentrations of TCIPP, p. 8830)

Air concentrations were estimated from the silicone-sampler concentrations using the generic indoor silicone-wristband sampling rate of 1.5 m³/day/dm² derived by Okeme et al. 2018 (Environ. Pollut. 239:109–117), accounting for sampler deployment time:

For comparison, the authors cite the following prior literature values reported on p. 8831:

• Schreder et al. 2016 (Chemosphere 150:499–504): TCIPP active personal-air sampler air concentrations from 10 adults, range 16–1,180 ng/m³, median 262 ng/m³.
• Vykoukalova et al. 2017 (Environ. Int. 106:97–104) using polyurethane-foam disc passive samplers: median TCIPP home-air concentrations 26.3 ng/m³ (USA), 73.6 ng/m³ (Canada), 16.4 ng/m³ (Czech Republic).
• Prior indoor-air bedroom and home maxima cited (Hoehn et al. cite Ortiz & Harrad 2023 and the unnumbered home reference for reference 46): 3,300 ng/m³ (bedrooms) and 4,190 ng/m³ (homes).

Trends by vehicle engine type (Figure 1, p. 8828)

TCIPP silicone-sampler concentrations were significantly lower in all-electric vehicles than in either gasoline or hybrid vehicles (Kruskal–Wallis p < 0.05, with significant post-hoc Wilcoxon Each Pair Method differences indicated by asterisks in Figure 1):

• TCIPP electric vs gasoline: ~6× lower in electric (Trends by Vehicle Engine Type, p. 8828).
• TCIPP electric vs hybrid: ~8× lower in winter, ~14× lower in summer (Trends by Vehicle Engine Type, p. 8828).
• TNBP electric vs gasoline: ~3× lower in electric (Trends by Vehicle Engine Type, p. 8828).
• TCIPP between electric and hybrid vehicles also differed significantly (Trends by Vehicle Engine Type, p. 8828).

The authors explicitly caveat (Trends by Vehicle Engine Type, p. 8828) that the engine-type trends may be confounded by vehicle brand: ‘8 brands are included in the electric vehicles category (n = 26), and among these, 53% are exclusively one brand. This is also true of the hybrid category (n = 26), where 9 brands are included, and 50% of these vehicles are exclusively one brand. It is difficult to determine if this trend is driven by engine type or another factor associated with manufacturers/brands.’

The specific brand identities are not reported in the paper (Part 12 brand-firewall compliance — the source itself anonymises brand attribution).

Methods (brief)

Study design (Materials and Methods, p. 8826):

• Recruitment via Green Science Policy Institute newsletter advertisement; inclusion criteria: US residence and personal-vehicle ownership of model year 2015 or newer. Selection prioritised geographic distribution across the United States and engine-type representation (~50% ICE / 25% electric / 25% hybrid target).
• IRB determination: Duke University IRB classified the study as ‘exempt human subjects research’ on the basis that the unit of analysis was the vehicle, not the human participant.
• Winter sampling round: February–May 2022, n = 101 vehicles. Summer sampling round: July–August 2022, n = 54 of the same vehicles (paired subset). Foam-collection sub-sample: n = 52 vehicles in winter, of which n = 51 paired with winter-sampler analyses and n = 28 paired with summer-sampler analyses (per Figure 4 caption).
• Vehicle VIN was collected for production-year and engine-type validation against the manufacturer database. Participant zip-code was used to extract NOAA Integrated Surface Dataset (NCEI DSI 3505_03) hourly temperature for the deployment period; the 7-day average ambient temperature for each vehicle’s deployment was the temperature covariate.

Silicone passive sampler deployment and collection (p. 8826):

• Pre-cleaned silicone passive samplers (Soxhlet-extracted) were mailed to participants in a kit with nitrile/vinyl gloves, pre-combusted aluminium foil, zip-ties, a Ziplock bag, and a preaddressed stamped envelope.
• Participants suspended the sampler from the rearview mirror for 7 days using zip-ties, then (wearing gloves) wrapped the sampler in aluminium foil, sealed it in the Ziplock bag, and returned it to the laboratory.
• Lab handling: gloves at all times; samplers transferred to airtight trace-clean glass vials and stored at −20 °C.

Vehicle-seat-foam collection (p. 8826):

• Participants were instructed to cut a ~1 cm³ piece of unupholstered foam from the underside of the front seat per the Supporting Information sampling kit instructions, wrap in pre-combusted aluminium foil, and return it.
• Lab handling: foam samples transferred to airtight trace-clean glass vials, stored at room temperature.

Silicone-sampler processing (p. 8826):

• Silicone-sampler segments (~0.7 g) were spiked with a suite of isotopically labelled internal standards (Table S1).
• Extracted with 50:50 hexane:acetone; concentrated under purified nitrogen prior to GC/MS analysis.
• 48 of 49 targeted analytes analysed by GC/MS; one analyte (2,4,6-tribromophenol, 2,4,6-TBP) analysed by LC/MS following solvent exchange to methanol.
• Full method, Table S2 (analyte list with CAS numbers), and GC/MS and LC/MS parameters in Supporting Information.

Foam processing (p. 8826):

• ~50 mg of foam extracted by sonication for 10 min in 2 mL of dichloromethane; extraction performed twice; extracts combined and concentrated by nitrogen evaporator to ~1 mL.
• Foam extracts analysed for 15 compounds (BFRs and OPEs) in full-scan mode using two GC/MS systems:
  • Agilent GC Model 7890A coupled to Agilent MS Model 5975C (Agilent Technologies, Santa Clara, CA, USA) in electron-ionization mode (GC/EI-MS).
  • Agilent GC Model 6890N coupled to Agilent MS Model 5975 (Agilent Technologies, Santa Clara, CA, USA) in negative-chemical-ionization mode (GC/ECNI-MS).
• Chromatograms inspected manually; significant peaks compared to in-house and NIST mass-spectral libraries; positive detections were ≥80% probability match.
• 2,4,6-TBP detections further confirmed by LC tandem MS: Agilent LC Model 1260 with Agilent MS Model 6460 (Agilent Technologies, Santa Clara, CA, USA) in negative-electrospray-ionization mode (LC/ESI-MS).
• Full target-analyte list (foam): Table S3.

Quality assurance / quality control (p. 8826):

• Field blanks: n = 5 winter and n = 5 summer kits mailed to study collaborators in California but not opened or deployed.
• Method detection limit (MDL) calculation, per analyte:
  • Detected in ≥3 field blanks: MDL = 3 × SD of blanks.
  • Detected in exactly 2 field blanks: MDL = 2 × mean of blanks.
  • Detected in 0–1 field blanks: MDL = lowest concentration visible in the calibration curve.
• Per-compound MDLs (normalised to the average ~0.7 g silicone-sampler mass) in Table S2.
• All sampler concentrations are blank-corrected by subtracting the average blank concentration from the raw measurement.
• Recovery of isotopically labelled internal standards: average 79–119% across all compounds (Table S1).

Statistical analysis (p. 8826–8827):

• Statistical analyses performed in JMP Pro 17 and SAS 9.4 (SAS Institute, Cary, NC, USA).
• Statistical analyses restricted to compounds with >60% detection in silicone samplers in both winter and summer deployment periods.
• Sub-MDL values imputed as MDL ÷ √2, adjusted for the mass of sampler extracted.
• Shapiro–Wilk normality tests indicated FR concentrations in silicone samplers were not normally distributed; non-parametric tests used throughout:
  • Wilcoxon Rank Sum (Mann–Whitney) for unpaired comparisons (e.g., winter vs summer across all samplers).
  • Wilcoxon Signed Rank for paired comparisons (winter vs summer in the n = 54 paired sub-cohort; sampler concentration in vehicles with vs without TCIPP in paired foam).
  • Kruskal–Wallis with post-hoc Wilcoxon Each Pair Method for >2-group comparisons (engine type).
  • Linear mixed-effects models (repeated measures) for temperature–concentration relationships, on log₁₀-transformed concentrations.
• Statistical significance threshold p < 0.05.

Cabin-air concentration estimation (Predicting Cabin Air Concentrations of TCIPP, p. 8830):

• Air concentrations contextualised from silicone-sampler concentrations using indoor-air sampling rates derived for silicone wristbands by Okeme et al. 2018 (Environ. Pollut. 239:109–117).
• Generic sampling rate used: 1.5 m³/day/dm², compiled from several indoor calibration studies in Okeme et al.
• Per-sampler air concentration calculated accounting for deployment period.
• Lower- and upper-bound sampling rates (also from Okeme et al.) reported in Table S7.
• Authors explicitly caveat: ‘the actual sampling rate will vary based on temperature, humidity, and air flow … those behaviors would influence temperature and ventilation and could thus influence air concentrations and sampler uptake’ (p. 8831).

Implications

• Certification (HMTc): No direct relevance to the HMTc 10-analyte panel (Pb, tAs, Cd, MeHg, tHg, iAs, Ni, Al, Cr-VI, Sn) for any HMTc product row. The paper measures only organic flame retardants (OPEs and BFRs); zero heavy-metal occurrence data are reported. The closest HMTc product row in scope is Category 10 (Infant and Child Durable Goods and Textiles (Ages 0–5)) Row 8 ‘Car seats (if retained in HMTc scope),’ which covers the infant-car-seat product that is deployed inside the vehicle cabin environment characterised here — but the paper measures the vehicle’s own seat foam and cabin air, not the infant car-seat product, so the relevance is environmental-context rather than direct product measurement. The vehicle-cabin-air exposure burden documented here applies to any occupant of a US personal vehicle including infants and small children riding in car seats, and the authors explicitly note (Discussion, p. 8831) that ‘children, who breathe a greater amount of air per kg body weight compared to adults, would also be at risk of greater exposures for equivalent commuting times.’ The analogous HMI-scope question on the same product surface (heavy-metal content of vehicle-seat polyurethane foam or vehicle-cabin dust) is not addressed in the current corpus and is flagged as an evidence-base gap for vehicle-cabin built-environment exposure.
• Courses: Useful as a methodology reference for silicone passive sampling of OPEs/BFRs in microenvironments (rearview-mirror sampler deployment for 7-day cabin-air integration), and for the post-2011 chlorinated-OPE substitution-cascade pattern (TDCIPP added to California Prop 65 in 2011 → industry substitution with TCIPP → 2023 NTP carcinogenicity finding for TCIPP, framed by the authors as a ‘regrettable substitution’). Also useful for illustrating non-parametric statistics on right-skewed environmental exposure data (Shapiro–Wilk non-normality → Wilcoxon Rank Sum / Signed Rank / Kruskal–Wallis suite) and for engine-type / brand-confounding caveats in vehicle-environment studies.
• App: Not applicable to the heavy-metals consumer app.

Update history

No substantive edit history is available in this build. The full commit record is available in git.