Hair Product Use and Urinary Biomarker Concentrations of Non-Persistent Endocrine Disrupting Chemicals among Reproductive-aged Black Women

Abstract

Background:

Studies have shown an association between hair product use and adverse health outcomes. Scientists have hypothesized that exposure to endocrine-disrupting chemicals (EDCs) drives these associations, but few studies have directly evaluated associations between hair product use and biomarkers of EDCs. Even more limited are studies of Black women, who frequently use EDC-containing products (e.g., hair relaxers).

Objective:

We estimated associations between hair product use and EDC biomarker concentrations.

Methods:

We leveraged cross-sectional data from the Study of Environment, Lifestyle, and Fibroids, a cohort of females aged 23–34 years who self-identified as Black/African American from the Detroit-metropolitan area (USA; n=425). On structured questionnaires, participants reported their past 24-hour and past 12-month use of hair products, including relaxers/straighteners/perms, styling products, moisturizers, oils, and hair food. We quantified urinary concentrations of 19 phthalate/phthalate alternative metabolites, 7 phenols, and 4 parabens using high performance liquid chromatography isotope dilution tandem mass spectrometry. EDC biomarker concentrations were creatinine-adjusted and natural log-transformed. We used multivariable linear regression to estimate mean percent differences in EDC biomarker concentrations and 95% confidence intervals (CIs) associated with hair product use, adjusting for sociodemographic confounders.

Results:

Hair product use was associated with greater concentrations of multiple EDC biomarkers. Notably, use of hair products in the previous 24 hours (compared with non-use) was associated with 16.2% (95% CI=0.7%, 35.9%), 35.0% (95% CI=2.6%, 77.6%), and 32.3% (95% CI=8.8%, 92.0%) higher concentrations of mono-isobutyl phthalate, methyl paraben, and ethyl paraben, respectively. Use of hair relaxers/straighteners/perms, styling products, moisturizers, oils, and hair food in the past 12 months was also associated with higher concentrations of multiple phthalate, phenol, and paraben biomarkers.

Conclusion:

Hair product use was associated with higher biomarker concentrations of multiple phthalates, phenols, and parabens. These findings suggest that hair products are potentially important exposure sources for hormonally-active chemicals among Black women.

Graphical Abstract

1. INTRODUCTION

There is accumulating evidence that hair product use is associated with adverse health outcomes in women. Numerous studies, for example, report associations between hair product use (e.g., hair relaxers/straighteners, dyes) and cancers (breast, uterine, and ovarian) (Chang et al., 2022; Eberle et al., 2020; Gera et al., 2018; Heikkinen et al., 2015; Llanos et al., 2017; Rao et al., 2022; Stiel et al., 2016; Tzonou et al., 1993; White et al., 2021a, 2021b; Zhang et al., 2020), fecundability (Wise et al., 2023), adverse birth outcomes (Chan et al., 2023; Kazemi Shishavan et al., 2021), and uterine fibroids (Wise et al., 2012) (Table 1). It is hypothesized that these associations are driven by exposure to endocrine disrupting chemicals (EDCs) in hair products, such as phthalates, phenols, parabens, metals, and formaldehyde, among others (Chang et al., 2022; Ghosh et al., 2022; Wise et al., 2012). Non-persistent EDCs, such as phthalates, phenols, and parabens, have relatively short biological half-lives in the body and can interfere with hormonal systems (Centers for Disease Control and Prevention, n.d.; Stahlhut et al., 2009; Wei et al., 2021). Because of their extensive use in personal care products, including hair products (Helm et al., 2018), exposure to non-persistent EDCs among product users is ubiquitous (Centers for Disease Control and Prevention, n.d.; Stahlhut et al., 2009; Wei et al., 2021).

Table 1.

A summary of hair products, EDC exposures related to product usage, and associations of product use with risk of health outcomes in women.

Hair product	Use(s) and examples of active ingredient(s)	Known EDC exposure(s)1	Associations of hair product use with health outcomes2
Hair relaxers/straighteners	Straightens curly hair Sodium hydroxide (lye); potassium hydroxide, lithium hydroxide, guanidine carbonate (non-lye); ammonium thioglycolate (thio); keratin (temporary) (Brinton et al., 2018; de Sá Dias et al., 2007; Weathersby and McMichael, 2013)	Formaldehyde/formaldehyde releasing chemicals (Aglan and Mansour, 2020; Chang et al., 2022; Galli et al., 2015; Weathersby and McMichael, 2013), phthalates (Helm et al., 2018; Wise et al., 2012), metals (e.g., cadmium, lead) (Iwegbue et al., 2016), parabens (Helm et al., 2018; Zota and Shamasunder, 2017), ethanolamines (Helm et al., 2018), phenols (Helm et al., 2018), fragrances (e.g., linalool) (Helm et al., 2018), cyclosiloxanes (Helm et al., 2018), UV filters (e.g., octinoxate) (Helm et al., 2018)	Uterine cancer (Bertrand et al., 2023; Chang et al., 2022), breast cancer (Eberle et al., 2020; Rao et al., 2022; Stiel et al., 2016; White et al., 2021a), uterine fibroids (Wise et al., 2012), scalp inflammation and hair loss (Hatsbach de Paula et al., 2022), ovarian cancer (White et al., 2021b), fecundability (Wise et al., 2023)
Hair oils, lotions	Provides lubrication for hair shaft, prevents breakage, hydrates hair Natural oils (e.g., castor)(Mysore and Arghya, 2022)	Parabens, phthalates, phenols, cyclosiloxanes, fragrances (e.g., linalool), UV filters (e.g., octyl dimethyl PABA3) (Helm et al., 2018)	Breast cancer (Stiel et al., 2016), dermatitis (Rucker Wright et al., 2011), lower birthweight (Chan et al., 2023)
Conditioners, moisturizers, petroleum jelly, shea butter, hair food	Hydrates hair Silicone, cationic surfactants (e.g., cetyltrimethylammonium chloride), natural oils, emulsifiers (e.g., ethoxylated fatty oxides), petroleum (Dias, 2015; D’Souza and Rathi, 2015)	Formaldehyde/formaldehyde releasing chemicals (Malinauskiene et al., 2015), phthalates (Braun et al., 2014; Parlett et al., 2013), metals (e.g., cadmium, cobalt) (Iwegbue et al., 2016), parabens (Helm et al., 2018), cyclosiloxanes (Helm et al., 2018), phenols (Helm et al., 2018), fragrances (e.g., linalool) (Helm et al., 2018), ethanolamines (Helm et al., 2018)	Breast cancer (Llanos et al., 2017)
Styling products (e.g., gels, mousses, sprays, pomade, anti-frizz, polishes)	Creates texture, prevents frizz, enhance hair texture Surfactants, natural oils, vinyl acetate copolymers (Dias, 2015; He et al., 2022)	Phthalates (Helm et al., 2018; Hsieh et al., 2019; Parlett et al., 2013; Sakhi et al., 2017), parabens (Braun et al., 2014), cyclosiloxanes (Helm et al., 2018), fragrances (e.g., limonene) (Helm et al., 2018), UV filters (e.g., octinoxate) (Helm et al., 2018)	Lower birthweight (Chan et al., 2023)
Root stimulators	Promotes root growth Biotin	Parabens, phthalates, phenols fragrances (e.g., linalool), glycol ethers, cyclosiloxanes, UV filters (e.g., octinoxate) (Helm et al., 2018)	Breast cancer, mammographic density (Stiel et al., 2016)
Dyes	Colors hair (permanently or temporarily) Aromatic amines, lawsone (He et al., 2022; Rubio et al., 2022)	Metals (e.g., cobalt, lead) (Bocca et al., 2014; Kaličanin and Velimirović, 2016), aromatic amines (Heikkinen et al., 2015; Turesky et al., 2003), phthalates (He et al., 2022), phenols (He et al., 2022)	Breast cancer (Eberle et al., 2020; Gera et al., 2018; Heikkinen et al., 2015; Llanos et al., 2017; Rao et al., 2022; White et al., 2021a; Zhang et al., 2020), bladder cancer (Koutros et al., 2011), lymphoma (De Sanjosé et al., 2006; Zhang et al., 2020), ovarian cancer (Tzonou et al., 1993; White et al., 2021b; Zhang et al., 2020), skin cancer (Zhang et al., 2020), lower birthweight (Kazemi Shishavan et al., 2021)

¹Based on studies testing concentrations of EDCs in hair products or studies assessing associations of hair product use with biomarker EDC concentrations in human populations.

²Does not include studies examining associations of EDC biomarkers with health outcomes.

³PABA= 4-aminobenzoic acid.

Prior epidemiologic evidence supports the hypothesis that hair product use is associated with increased exposure to non-persistent EDCs. In four studies, hair product use was associated with increased phthalate biomarker concentrations in Taiwanese, Norwegian, and U.S. women (Braun et al., 2014; Hsieh et al., 2019; Parlett et al., 2013; Sakhi et al., 2017). However, these studies had few Black participants (<9.0%) (Braun et al., 2014; Hsieh et al., 2019; Parlett et al., 2013; Sakhi et al., 2017). Black women tend to have higher exposure to several non-persistent EDCs and also experience higher rates of multiple hormone-dependent diseases (Eltoukhi et al., 2014; Schildroth et al., 2022). Although recent regulatory policies aimed to limit exposure to certain chemicals (i.e., formaldehyde) in personal care products (i.e., hair relaxers) (Franklin, 2023), product testing data indicate that hair products marketed to Black women contain detectable levels of multiple EDCs, and that EDCs in products may not be accurately labeled (Helm et al., 2018). Therefore, understanding the effects of hair product use on EDC exposure among Black women is an important public health goal.

Hairstyling among Black Americans has a long cultural and political history (see Byrd and Tharps (Byrd and Tharps, 2014)). Hairstyling practices, and thus use of hair products, among Black Americans often begin in childhood (Byrd and Tharps, 2014; Gaston et al., 2020; James-Todd et al., 2011; White et al., 2021a). For example, in a previous publication from the Study of Environment, Lifestyle, and Fibroids (SELF), we found that the prevalence of hair relaxer/straightener use ≥2 times/year was 9% at 5 years of age, and this prevalence increased to 34% and 73% by 10 and 15 years of age, respectively (Gaston et al., 2020). Specific hairstyles and hair products (like relaxers) that are commonly used in Black communities are partly influenced by discriminatory hair texture preferences that historically idealized European beauty norms, such as straight, long hair (Byrd and Tharps, 2014; Zota and Shamasunder, 2017). These discriminatory hair standards persist today, particularly in the workplace (Lee and Nambudiri, 2021), where Black women frequently report experiencing discrimination for wearing natural hairstyles (Hair et al., 2014). Pressure to conform to such hair texture policies and standards may influence patterns of hair product use: Black women, compared with other racial groups, are more likely to report using hair relaxers and root stimulators (Collins et al., 2021; Dodson et al., 2021; James-Todd et al., 2012; JOY Collective, 2019; Zota and Shamasunder, 2017). Further, hair products targeted towards Black women have detectable concentrations of multiple EDCs, including phthalates, phenols, and parabens (Helm et al., 2018). Because use of multiple hair products (e.g., relaxers, conditioners, dyes) begins early in life and persists throughout the life course, often with high frequency (i.e., every few weeks) (Byrd and Tharps, 2014; Gaston et al., 2020; James-Todd et al., 2011; White et al., 2021a), Black women may have high cumulative exposure to EDCs from these products.

Few studies have directly evaluated the effects of hair product use on EDC biomarker concentrations in populations with Black participants (Braun et al., 2014; Hsieh et al., 2019; Parlett et al., 2013; Sakhi et al., 2017). In the current study, we examined associations between use of hair products—including styling products, relaxers, and oils—with biomarker concentrations of non-persistent EDCs (i.e., phthalates, phenols, parabens) in a cohort of reproductive-aged Black women.

2. METHODS

2.1. Study population

We analyzed cross-sectional data from SELF, a prospective cohort of reproductive-aged Black individuals from the Detroit, Michigan (USA) area. Details of study recruitment have been described previously (Baird et al., 2015). Briefly, 1,693 participants were enrolled between 2010–2012. Participants met the following enrollment requirements 1) self-identified as Black or African American, 2) were 23–34 years of age, 3) had an intact uterus, 4) had no prior diagnosis of uterine leiomyomata (UL), and 5) had no prior diagnosis of cancer or an autoimmune disease requiring medication use. Pregnant participants were eligible, but their enrollment was delayed until three months postpartum. All participants provided written informed consent following orientation to all study protocols. Institutional Review Boards at the National Institute of Environmental Health Sciences (NIEHS), Henry Ford Health (HFH), and Boston University approved all SELF protocols. The involvement of the Centers for Disease Control and Prevention (CDC) did not constitute human subjects research.

SELF participants completed a series of self-administered questionnaires, phone interviews, a transvaginal ultrasound, and urine collection at baseline and at the 20-month, 40-month, and 60-month follow-ups (Figure S1). In 2015, SELF received supplemental funding for a case-cohort substudy to examine associations between EDC exposure and UL incidence. The case-cohort substudy included a randomly-selected subcohort of 592 UL-free participants at baseline (139 of whom developed incident UL during 60 months of follow-up), as well as 162 participants outside of the subcohort who developed incident UL during 60 months of follow-up. Hair product use (described below) in the past 24 hours was assessed at each study visit, but hair product use in the past 12 months was only queried at the 40-month follow-up. As such, the current analysis utilized cross-sectional data from a subset of participants from the case-cohort substudy that had both urinary EDC and hair product use data at the 40-month follow-up.

2.2. Measurement of hair product use

Information on past 24-hour and past 12-month hair product use was ascertained using 24-hour recall questionnaires and computer-assisted web interviews (CAWI) concurrently with urine sample collection (Figure S1). Participants reported their use (yes/no) of hair products (sprays, gels, oils, or mousses) and chemical treatments (hair color or relaxers) in the past 24 hours on the 24-hour recall questionnaire.

The CAWI queried participants about their past 12-month use of the following hair products: 100% petroleum jelly; 100% shea butter, natural plant-based oils; hair food; moisturizing creams or lotions; conditioners, crème rinses or detanglers; and styling products (gel, mousse, pomade, grease, oil sheens, sprays, spritzes, setting lotion) (see Supplemental Information). Response options for each of these products were ≥2 times/day, 1 time/day, 2–6 times/week, 1 time/week, 1–3 times/month, <1 time/month, or did not use. We queried participants about their frequency of use of hair coloring products, including hair gloss; henna; rinses that fade; semi-permanent hair dye; permanent hair dye; and products that bleach, frost or highlight hair. Response options for each of these products were >1 time/month, 9–12 times/year, 4–8 times/year, 1–3 times/year, or did not use. Participants further reported their past 12-month use of hair relaxers, straighteners, or perms (≥12 times/year, 6–11 times/year, 2–5 times/year, 1 time/year, or did not use), and the type of hair relaxer ever used (yes/no): permanent with lye, permanent with low lye, permanent with thio, permanent but unsure of content, temporary with keratin, temporary without keratin, and temporary but unsure of content.

We categorized response options for each hair product on the 12-month questionnaire based on distributions of reported use to optimize power in statistical analyses (Table 2). We dichotomized petroleum jelly, hair gloss, henna, rinses that fade, semi-permanent dye, and products that bleach, frost, or highlight hair (any use in the past 12-months vs. non-use). We categorized permanent hair dye as non-use, moderate use (1–3 times/year), and high use (≥4 times/year); hair oils, 100% shea butter, and hair food as non-use, moderate use (<1 time/month or 1–3 times/month), and high use (≥1 time/month); and moisturizing creams or lotions, styling products, and conditioners, crème rinses or detanglers as non-use, low use (<1 time/month), moderate use (1–3 times/month), or high use (≥1 time/week). Finally, we categorized use of hair relaxers, straighteners, or perms as non-use, low use (1 time/year), moderate use (2–5 times/year), or high use (≥6 times/year).

Table 2.

Summary statistics for sociodemographic, anthropometric, hair product, and personal care product variables in SELF (n=425).

Characteristic	Mean (SD) or N (%)
Sociodemographic variables
Age (years)	32.2 (3.4)
Annual household income
<$20,000	155 (36.5%)
$20,000–50,000	168 (39.5%)
>$50,000	102 (24.0%)
Educational attainment
High school degree/GED	63 (14.8%)
Some college/Associate’s/Technical school	224 (52.7%)
≥Bachelor’s degree	138 (32.5%)
Employment status
Not employed	77 (18.1%)
Employed <30 hours/week	54 (12.7%)
Employed ≥30 hours/week	294 (69.2%)
Marital status
Never married	166 (39.1%)
Currently married or living as married	82 (19.3%)
Previously married	177 (41.6%)
Hair product use in the past 12 months
100% petroleum jelly1
Non-use	357 (84.0%)
Ever use	68 (16.0%)
Hair gloss1
Non-use	377 (88.7%)
Ever use	48 (11.3%)
Henna1,2
Non-use	405 (95.3%)
Ever use	20 (4.7%)
Rinses that fade1
Non-use	352 (82.8%)
Ever use	73 (17.2%)
Semi-permanent dye1
Non-use	346 (81.4%)
Ever use	79 (18.6%)
Bleach, frost, or highlight1
Non-use	357 (84.0%)
Ever use	68 (16.0%)
Permanent dye3
Non-use	313 (73.6%)
Moderate use	92 (21.7%)
High use	20 (4.7%)
100% shea butter4
Non-use	300 (70.6%)
Moderate use	58 (13.6%)
High use	67 (15.8%)
Natural plant-based oils4
Non-use	254 (59.8%)
Moderate use	76 (17.9%)
High use	95 (22.3%)
Hair food4
Non-use	280 (65.9%)
Moderate use	72 (16.9%)
High use	73 (17.2%)
Moisturizing creams or lotions5
Non-use	194 (45.7%)
Low use	38 (8.9%)
Moderate use	60 (14.1%)
High use	133 (31.3%)
Conditioners, crème rinses, or detanglers5
Non-use	96 (22.6%)
Low use	45 (10.6%)
Moderate use	142 (33.4%)
High use	142 (33.4%)
Styling products5
Non-use	53 (12.5%)
Low use	57 (13.4%)
Moderate use	102 (24.0%)
High use	213 (50.1%)
Relaxers, straighteners, or perms6
Non-use	245 (57.6%)
Low use	43 (10.1%)
Moderate use	90 (21.2%)
High use	47 (11.1%)
Hair product use in the past 24 hours
Hair products (spray, gel, oil, or mousse)
No	233 (54.8%)
Yes	192 (45.2%)
Chemical treatments (hair color, relaxers)2
No	420 (98.8%)
Yes	5 (1.2%)
Type of hair relaxer, straightener, or perm ever used 7
Permanent relaxer with lye
No	264 (62.1%)
Yes	161 (37.9%)
Permanent relaxer with low lye
No	295 (69.5%)
Yes	130 (30.5%)
Permanent relaxer with thio
No	413 (97.2%)
Yes	12 (2.8%)
Permanent relaxer, unsure of content
No	298 (70.1%)
Yes	127 (29.9%)
Temporary relaxer with keratin
No	348 (81.9%)
Yes	77 (18.1%)
Temporary relaxer without keratin
No	390 (91.8%)
Yes	35 (8.2%)
Temporary relaxer, unsure of content
No	317 (74.6%)
Yes	108 (25.4%)
Frequency of use by relaxer type8
Never use, temporary relaxer use only, or non-use in past 12 months	305 (71.8%)
Ever use of both temporary and permanent relaxers	29 (6.8%)
Use of permanent relaxers only with low use in past 12 months	21 (4.9%)
Use of permanent relaxers only with moderate use in past 12 months	41 (9.7%)
Use of permanent relaxers only with high use in past 12 months	29 (6.8%)
Personal care product use in the past 24 hours7
Perfume
No	154 (36.2%)
Yes	271 (63.8%)
Lotions or cream before bed
No	270 (63.5%)
Yes	155 (36.5%)
Make-up
No	91 (21.4%)
Yes	334 (78.6%)
Nail polish
No	393 (92.5%)
Yes	32 (7.5%)

¹Ever use vs. never use in the previous 12 months.

²Excluded from regression analyses due to low reported use.

³Moderate use= 1–3 times/year, high use= ≥4 times/year.

⁴Moderate use= <1 time/month or 1–3 times/month, high use= ≥1 times/week.

⁵Low use= <1 time/month, moderate use= 1–3 times/month, high use= ≥1 times/week.

⁶Low use= 1 time/year, moderate use= 2–5 times/year, high use= ≥6 times/year.

⁷Variables used in secondary analyses only.

⁸Variable created by combing ever use of relaxer type variables with frequency of relaxer use in the past 12-month variable.

Our primary analyses focused on hair products that were used in the past 24 hours or in the past 12 months. We excluded past 12-month use of henna and past 24-hour use of chemical treatments from analyses because of a low prevalence of use (<5.0%) (Table 2).

2.3. Quantification of urinary EDC biomarkers

Urine samples were provided by participants at baseline and each follow-up study visit (20-month, 40-month, and 60-month) at HFH. Following collection, samples were stored at −80°C at HFH until they were shipped overnight on dry ice to the NIEHS biorepository for long-term storage (at −80°C). During 2016–2019, we shipped the 40-month urine samples in two batches (n=376 and n=49) overnight from the NIEHS biorepository to the CDC on dry ice for measurement of EDCs. Nineteen phthalate or phthalate alternative metabolites, 7 phenols, 4 parabens, and triclocarban were quantified in urine using high performance liquid chromatography isotope dilution tandem mass spectrometry (Silva et al., 2007; Ye et al., 2006, 2005).

Phthalate metabolites included monoethyl phthalate (MEP), mono-n-butyl phthalate (MBP), mono-hydroxybutyl phthalate (MHBP), mono-isobutyl phthalate (MiBP), mono-hydroxyisobutyl phthalate (MHiBP), monobenzyl phthalate (MBzP), mono-3-carboxypropyl phthalate (MCPP), mono (2-ethylhexyl) phthalate (MEHP), mono (2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono (2-ethyl-5-oxohexyl) phthalate (MEOHP), mono (2-ethyl-5-carboxypentyl) phthalate (MECPP), mono-isononyl phthalate (MNP), mono-carboxyisooctyl phthalate (MCOP), mono-carboxyisononyl phthalate (MCNP), mono-2-ethyl-5-carboxypentyl terephthalate (MECPTP), and mono-2-ethyl-5-hydroxyhexyl terephthalate (MEHHTP). We also measured 1,2-cyclohexane dicarboxylic acid-monohydroxy isononyl ester (MHiNCH) and 1,2-cyclohexane dicarboxylic acid-monocarboxy isooctyl ester (MCOCH), two metabolites of di(isononyl)cyclohexane-1,2-dicarboxylate (DINCH), a non-phthalate alternative.

Phthalate biomarkers were classified as metabolites of either low (alkyl side chain length ≤4 carbons) or high (alkyl side chain length >4 carbons) molecular weight parent compounds (National Research Council Committee on the Health Risks of Phthalates, 2008; Wang et al., 2019). Low molecular weight phthalates included diethyl phthalate (DEP; metabolite: MEP), di-n-butyl phthalate (DnBP, metabolites: MBP and MHBP), and di-isobutyl phthalate (DiBP, metabolites: MiBP and MHiBP). High molecular weight phthalates included benzylbutyl phthalate (BzBP; metabolite: MBzP), dioctyl phthalate (DOP; metabolite: MCPP), di(2-ethylhexyl) phthalate (DEHP; metabolites: MECPP, MEHHP, MEHP, and MEOHP), diisononyl phthalate (DiNP; metabolites: MCOP and MNP), and diisodecyl phthalate (DiDP; metabolite: MCNP). Phthalate alternatives included DINCH metabolites (MHiNCH and MCOCH) and di-2-ethylhexyl terephthalate (DEHTP; metabolites: MECPTP and MEHHTP). Limits of detection (LODs) for all metabolites were 0.2–1.2 ng/mL (Table 3).

Table 3.

Limits of detection (LODs), detection frequency, and distributions for non-persistent EDC biomarkers in SELF (n=425).^1,2

Biomarker	LOD (ng/mL)	Detection frequency (%)	Median (25th, 75th percentiles (μg/g creatinine)
Phenols
Benzophenone-3	0.4	98.1%	9.4 (4.7, 27.7)
BPA	0.2	98.1%	1.1 (0.7, 1.7)
BPF	0.2	62.6%	0.3 (<LOD, 0.9)
BPS	0.1	96.7%	0.5 (0.2, 1.0)
2,4-dichlorophenol	0.1	96.5%	0.4 (0.3, 0.8)
2,5-dichlorophenol	0.1	96.5%	1.2 (0.6, 2.6)
Triclosan	1.7	79.3%	3.9 (1.8, 14.9)
Triclocarban	0.1	58.1%	NC
Phthalates
MEP	1.2	100%	53.2 (25.9, 108.8)
MBP	0.4	100%	0.9 (0.7, 1.0)
MHBP	0.4	84.5%	0.7 (0.5, 1.2)
MiBP	0.8	99.5%	9.4 (6.2, 16.0)
MHiBP	0.4	97.2%	2.6 (1.6, 4.4)
MBzP	0.3	99.1%	4.0 (2.4, 8.7)
MCPP	0.4	92.0%	1.2 (0.7, 2.5)
MEHP	0.8	85.2%	1.7 (1.0, 3.1)
MEHHP	0.4	99.5%	7.1 (4.9, 11.3)
MEOHP	0.2	99.5%	4.1 (2.7, 6.6)
MECPP	0.4	100%	9.4 (6.4, 15.2)
MNP	0.9	54.8%	NC
MONP	0.4	98.6%	2.6 (1.5, 6.7)
MCOP	0.3	100%	12.5 (5.7, 33.0)
MCNP	0.2	99.8%	2.2 (1.3, 4.0)
MECPTP	0.2	100%	13.6 (6.3, 37.8)
MEHHTP	0.4	97.9%	2.8 (1.4, 7.4)
Phthalate alternatives
MHiNCH	0.4	54.4%	NC
MCOCH	0.5	41.9%	NC
Parabens
Methyl paraben	1.0	100%	82.8 (30.0, 199.0)
Ethyl paraben	1.0	78.6%	3.1 (0.9, 14.0)
Propyl paraben	0.1	100%	12.3 (4.1, 30.3)
Butyl paraben	0.1	20.0%	NC

¹NC= Not calculated due to low detection frequency.

²BPA= bisphenol A; BPF= bisphenol F; BPS= bisphenol S; MEP= monoethyl phthalate; mono-hydroxybutyl phthalate; MBP= mono-n-butyl phthalate; MHBP = mono-hydroxybutyl phthalate; MiBP= mono-isobutyl phthalate; MHiBP= mono-hydroxyisobutyl phthalate; MBzP= monobenzyl phthalate; MCPP= mono-3-carboxypropyl phthalate; MEHP= mono (2-ethylhexyl) phthalate; MEHHP= mono (2-ethyl-5-hydroxyhexyl) phthalate; MEOHP= mono (2-ethyl-5-oxohexyl) phthalate; MECPP= mono (2-ethyl-5-carboxypentyl) phthalate; MNP= mono-isononyl phthalate; MONP = monooxononyl phthalate; MCOP= mono-carboxyisooctyl phthalate; MCNP= mono-carboxyisononyl phthalate; MECPTP= mono-2-ethyl-5-carboxypentyl terephthalate; MEHHTP= mono-2-ethyl-5-hydroxyhexyl terephthalate; MHiNCH= 1,2-cyclohexane dicarboxylic acid-monohydroxy isononyl ester; MCOCH= 1,2-cyclohexane dicarboxylic acid-monocarboxy isooctyl ester.

Phenols included benzophenone-3, bisphenol A (BPA), bisphenol F (BPF), bisphenol S (BPS), 2,4-dichlorophenol, 2,5-dichlorophenol, and triclosan. Parabens included methyl paraben, ethyl paraben, butyl paraben, and propyl paraben. LODs for phenols and parabens were 0.1–1.7 ng/mL and 0.1–1.0 ng/mL, respectively. We also measured triclocarban; the LOD for triclocarban was 0.1 ng/mL (Table 3).

Triclocarban, three phthalate or phthalate alternative metabolites (MNP, MHiNCH, and MCOCH), and butyl paraben were detected in <60% of samples and were therefore excluded from analyses (Table 3). EDC biomarker concentrations <LOD for EDCs included in the analyses were imputed as the $LOD/\sqrt{2}$ (Hornung and Reed, 1990). We accounted for urinary dilution by dividing EDC biomarker concentrations by creatinine concentrations, which were measured at the NIEHS with the Beckman Coulter clinical analyzer AU400e. EDC concentrations are presented as μg/g creatinine. For statistical analyses (described below), we further summed creatinine-adjusted molar concentrations of metabolites from the same phthalate for phthalates that had at least two metabolites with detection frequencies >60% (Table 3); these phthalates included DnBP (MBP, MHBP), DiBP (MiBP, MHiBP), DEHP (MECPP, MEHHP, MEHP, MEOHP), and DEHTP (MEHHTP, MECPTP).

2.4. Measurement of covariates

We collected sociodemographic information at each study visit using structured questionnaires and interviews. All covariate information used in this analysis was collected at the 40-month follow-up clinic visit. These variables included age (years), annual household income (<$20,000, $20,000–$50,000, or >$50,000), educational attainment (High school degree or GED, some college/Associate’s degree/technical school, or ≥Bachelor’s degree), employment status (not employed, employed <30 hours/week, or employed ≥30 hours/week), and marital status (never married, previously married, or currently married/living together as married).

We also queried use (yes/no) of personal care products in the past 24 hours: perfume or cologne, lotion or creams before bed, cosmetics (i.e., foundation, blush, eye make-up, lipstick, lip gloss, ChapStick, lip balm, Vaseline), and nail polish.

2.5. Statistical analysis

2.5.1. Analytic sample and missing data.

The current analysis leverages cross-sectional data from the 40-month follow-up, which was the only follow-up with data available on both past 12-month hair product use and urinary biomarkers (Figure S1). We restricted our analytic sample to 425 case-cohort study participants with complete urinary EDC biomarker data at the 40-month follow-up. There were little missing data in the analytic sample (<0.2% for all variables; Table S1). We imputed missing data using Monte Carlo Markov Chain methods, assuming data were missing at random (Zhou et al., 2001). We generated five imputed datasets using all available EDC biomarker, hair product, and confounder data.

2.5.2. Descriptive statistics and variable distributions.

We calculated summary statistics (means, standard deviations [SDs], percentages [%], medians, and 25^th–75^th percentiles) for all variables (see Table 2 and Table 3). Spearman correlation coefficients were calculated for the EDC biomarkers (Figure S2). We examined distributions for all variables; EDC biomarker concentrations were right-skewed and we naturally log (ln)-transformed the biomarkers prior to statistical modeling to meet assumptions of the normality of residuals.

2.5.3. Regression analysis.

We used multivariable linear regression models to estimate associations of hair product use in the past 24 hours and past 12 months with biomarker concentrations of EDCs. Our primary models adjusted for sociodemographic variables (age, household income, educational attainment, employment status, marital status), which were selected using a directed acyclic graph (DAG; Figure S3). Models were also co-adjusted for all hair product use variables, an approach we have previously used in SELF (Schildroth et al., 2022; Wesselink et al., 2021a). For all hair products, the “non-use” category was the reference group. We used generalized additive models to assess the linearity of associations between age and EDC biomarker concentrations, adjusting for all other sociodemographic confounders and hair product use variables; age was subsequently modeled continuously as a linear term.

We fit linear regression models for all five imputed datasets and pooled results using Rubin’s rule (Rubin, 2004). We estimated beta coefficients (β), reflecting the mean difference in ln-transformed EDC biomarker concentrations, and 95% confidence intervals (CIs). For ease of interpretation, we back-transformed β coefficients (and corresponding 95% CIs) as the percent (%) difference in EDC biomarker concentrations using the following equations:

$$\mathrm{\%}differenceinEDCconcentrations=(e^{\beta }-1)\mathrm{*}100$$ [1]

$$\mathrm{\%}difference95\mathrm{\%}CIs=(e^{\beta \pm 1.96\mathrm{*}SE}-1)\mathrm{*}100,whereSE=standarderror.$$ [2]

In secondary analyses, we investigated associations of hair relaxer type with EDC biomarker concentrations by cross-classifying the relaxer type variables with the past 12-month frequency of relaxer use variable (Table 2). We defined the frequency of use by relaxer type variable as: 1) non-use of any relaxer type/ever use of temporary relaxers only/non-use of relaxers in the past 12 months (reference group); 2) ever use of both temporary and permanent relaxers; 3) low use of relaxers in the past 12 months among ever users of permanent relaxers only; 4) moderate use of relaxers in the past 12 months among ever users of permanent relaxers only; and 5) high use of relaxers in the past 12 months among ever users of permanent relaxers only. This secondary analysis assumed that the frequency of use of relaxers in the past 12 months reflected use of permanent relaxers only. We also assumed that temporary relaxers contained different and less harmful chemicals than permanent relaxers (Wise et al., 2012).

2.5.4. Sensitivity analyses.

We performed several sensitivity analyses by fitting the following multivariable linear regression models: 1) reduced models that were not co-adjusted for all other hair product use variables (i.e., models included only an individual hair product use variable plus sociodemographic variables); and 2) models additionally adjusted for personal care product use (perfume, lotion or cream, make-up, and nail polish) in the past 24-hours.

5. RESULTS

5.1. Participant characteristics

Among 425 participants, the mean age at the 40-month follow-up visit was 32.2 years (SD=3.4 years). Most participants had annual household incomes ≤$50,000 (76.0%), had fewer than 16 years of education (67.5%), were employed ≥30 hours per week (69.2%), and were either currently (19.3%) or previously (41.6%) married. The majority of participants reported using perfume (63.8%) and make-up (78.6%) in the previous 24 hours, whereas use of lotion or cream before bed (36.5%) and nail polish (7.5%) was less frequent (Table 2).

Nearly half (45.2%) of participants reported using hair products (sprays, gels, oils, or mousses) in the previous 24 hours, but few (1.2%) reported using chemical hair treatments (Table 2). In the previous 12 months, the most commonly used hair products were moisturizing creams or lotions (low use: 8.9%, moderate use: 14.1%, high use 31.3%), conditioners, crème rinses, or detanglers (low use: 10.6%, moderate use: 33.4%, high use: 33.4%), and styling products (low use: 13.4%, moderate use: 24.0%, high use: 50.1%). Forty-two percent of participants reported any use of hair relaxers, straighteners, or perms in the past 12 months.

Ever use of type of hair relaxer, straightener, or perm was highest for permanent relaxers with lye (37.9%), permanent relaxers with low lye (30.5%), and permanent relaxers for which participants were unsure of the content (29.9%). Frequency of use for permanent relaxers, straighteners, or perms in the past 12 months varied, where 4.9%, 9.7%, and 6.8% of participants reported low, moderate, and high use, respectively (Table 2).

Summary statistics for the EDC biomarker concentrations are provided in Table 3. Among the phenols, the highest median concentrations were observed for benzophenone-3 (9.4 μg/g creatinine; 25^th, 75^th percentiles= 4.7, 27.7 μg/g creatinine) and triclosan (3.9 μg/g creatinine; 25^th, 75^th percentiles= 1.8, 14.9 μg/g creatinine). Median concentrations for the phthalate metabolites were highest for MEP (53.2 μg/g creatinine; 25^th, 75^th percentiles= 25.9, 108.8 μg/g creatinine), MCOP (12.5 μg/g creatinine; 25^th, 75^th percentiles= 5.7, 33.0 μg/g creatinine), and MECPTP (13.6 μg/g creatinine; 25^th, 75^th percentiles= 6.3, 37.8 μg/g creatinine). Methyl paraben had the highest median concentration among the parabens (82.8 μg/g creatinine; 25^th, 75^th percentiles= 30.0, 199.0 μg/g creatinine). SELF participants had appreciably higher biomarker concentrations of several non-persistent EDCs, particularly phthalate metabolites and parabens, compared to 20–39-year-old non-Hispanic White women from the 2009–2010 National Health and Nutrition Examination Survey (NHANES) cycle (Schildroth et al., 2022). Spearman correlation coefficients between the EDC biomarker concentrations ranged from −0.11 (2,4-dichlorophenol-BPF) to 0.92 (MEOHP-MEHHP) and tended to be strongest for biomarkers within the same EDC class and for metabolites from the same parent compound (Figure S2).

3.2. Associations of Hair Product Use with Urinary EDC Biomarker Concentrations

Use of hair products (sprays, gels, oils, and mousses) in the 24 hours before urinary sample collection was associated with several biomarker concentrations (Figure 1). Notably, past 24-hour use of hair products was associated with higher concentrations (% difference) of metabolites of DiBP (MiBP: β= 16.2%, 95% CI= −0.7%, 35.9%; MHiBP: β= 19.7%, 95% CI= 2.3%, 40.0%) and two parabens: methyl paraben (β= 35.0%, 95% CI= 2.6%, 77.6%) and ethyl paraben (β= 32.3%, 95% CI= −8.8%, 92.0%), but lower concentrations of MCNP (β= −17.3%, 95% CI= −32.0%, 0.6%). Associations tended to be imprecise, particularly for the parabens (Figure 1). Past 24-hour hair product use was not materially associated with biomarker concentrations for any of the phenols (Figure 1).

Figure 1.

Mean percent differences (95% CIs) in urinary EDC biomarker concentrations associated with past 24-hour hair product use in SELF (n=425). Non-use was set as the reference group. Models are adjusted for age, household income, educational attainment, employment status, and marital status, and are co-adjusted for all past 12-month and past 24-hour hair product variables.

¹Hair product use in the past 24-hours included sprays, gels, oils, or mousses.

²BP3 = benzophenone-3; BPA= bisphenol A; BPF= bisphenol F; BPS= bisphenol S; MEP= monoethyl phthalate; mono-hydroxybutyl phthalate; MBP= mono-n-butyl phthalate; MHBP = mono-hydroxybutyl phthalate; MiBP= mono-isobutyl phthalate; MHiBP= mono-hydroxyisobutyl phthalate; MBzP= monobenzyl phthalate; MCPP= mono-3-carboxypropyl phthalate; MEHP= mono (2-ethylhexyl) phthalate; MEHHP= mono (2-ethyl-5-hydroxyhexyl) phthalate; MEOHP= mono (2-ethyl-5-oxohexyl) phthalate; MECPP= mono (2-ethyl-5-carboxypentyl) phthalate; MONP = monooxononyl phthalate; MCOP= mono-carboxyisooctyl phthalate; MCNP= mono-carboxyisononyl phthalate; MECPTP= mono-2-ethyl-5-carboxypentyl terephthalate; MEHHTP= mono-2-ethyl-5-hydroxyhexyl terephthalate.

Associations between past 12-month hair product use and EDC biomarker concentrations were less consistent; use of several products was associated with higher concentrations for some EDC biomarkers, but lower concentrations for others. There was also limited evidence of clear dose-response relationships for products with multiple categories of use (e.g., low, moderate, and high; Figures 2–4; Figures S4–S7).

Figure 2.

Mean percent differences (95% CIs) in urinary EDC biomarker concentrations associated with hair relaxer, straightener, or perm use in the past 12-months in SELF (n=425). Non-use in the past 12-months was set as the reference group. Models are adjusted for age, household income, educational attainment, employment status, and marital status, and are co-adjusted for all past 12-month and past 24-hour hair product variables.

¹Low use= 1 time/year, moderate use= 2–5 times/year, high use= ≥6 times/year.

²BP3 = benzophenone-3; BPA= bisphenol A; BPF= bisphenol F; BPS= bisphenol S; MEP= monoethyl phthalate; mono-hydroxybutyl phthalate; MBP= mono-n-butyl phthalate; MHBP = mono-hydroxybutyl phthalate; MiBP= mono-isobutyl phthalate; MHiBP= mono-hydroxyisobutyl phthalate; MBzP= monobenzyl phthalate; MCPP= mono-3-carboxypropyl phthalate; MEHP= mono (2-ethylhexyl) phthalate; MEHHP= mono (2-ethyl-5-hydroxyhexyl) phthalate; MEOHP= mono (2-ethyl-5-oxohexyl) phthalate; MECPP= mono (2-ethyl-5-carboxypentyl) phthalate; MONP = monooxononyl phthalate; MCOP= mono-carboxyisooctyl phthalate; MCNP= mono-carboxyisononyl phthalate; MECPTP= mono-2-ethyl-5-carboxypentyl terephthalate; MEHHTP= mono-2-ethyl-5-hydroxyhexyl terephthalate.

Figure 4.

Mean percent differences (95% CIs) in urinary EDC concentrations associated with (a) hair food and (b) hair oil use in the past 12-months in SELF (n=425). Non-use in the past 12-months was set as the reference group. Models are adjusted for age, household income, educational attainment, employment status, and marital status, and are co-adjusted for all past 12-month and past 24-hour hair product variables.

¹<1 time/month or 1–3 times/month, high use= ≥1 times/week.

²BP3 = benzophenone-3; BPA= bisphenol A; BPF= bisphenol F; BPS= bisphenol S; MEP= monoethyl phthalate; mono-hydroxybutyl phthalate; MBP= mono-n-butyl phthalate; MHBP = mono-hydroxybutyl phthalate; MiBP= mono-isobutyl phthalate; MHiBP= mono-hydroxyisobutyl phthalate; MBzP= monobenzyl phthalate; MCPP= mono-3-carboxypropyl phthalate; MEHP= mono (2-ethylhexyl) phthalate; MEHHP= mono (2-ethyl-5-hydroxyhexyl) phthalate; MEOHP= mono (2-ethyl-5-oxohexyl) phthalate; MECPP= mono (2-ethyl-5-carboxypentyl) phthalate; MONP = monooxononyl phthalate; MCOP= mono-carboxyisooctyl phthalate; MCNP= mono-carboxyisononyl phthalate; MECPTP= mono-2-ethyl-5-carboxypentyl terephthalate; MEHHTP= mono-2-ethyl-5-hydroxyhexyl terephthalate.

Past 12-month use of hair relaxers, straighteners, or perms (compared with non-use) was associated with greater concentrations of 2,5-dichlorophenol (low use: β= 44.8%, 95% CI= −14.7%, 145.8%), MBzP (low use: β= 41.9%, 95% CI= 1.7%, 98.0%; high use: β= 28.4%, 95% CI= −6.2%, 75.7%), and ethyl paraben (moderate use: β= 32.3%, 95% CI= −17.3%, 111.8%; high use: 29.7%, 95% CI= −29.4%, 138.1%), though there was little evidence of a dose-response relationship (Figure 2). Alternatively, high use of relaxers, straighteners, or perms was associated with 52.3% lower (95% CI= −73.0%, −15.8%) BPF concentrations.

In secondary analyses, we examined frequency of use (low, moderate, or high) by type of hair relaxer, straightener, or perm. High use of permanent relaxers in the past 12 months (compared with never users, ever users of temporary relaxers only, or non-users of any relaxers in the past 12 months) was associated with higher concentrations of MBzP (β= 37.7%, 95% CI= −6.9%, 103.8%) and both metabolites of DEHTP (MECPTP: β= 52.2%, 95% CI= −15.5%, 174.0%; MEHHTP: β= 50.7%, 95% CI= −11.2%, 155.8%). However, moderate use of permanent relaxers in the past 12 months and ever use of both temporary and permanent relaxers was associated with lower concentrations for three metabolites of DEHP (MEHP, MEHHP, and MEOHP; Figure S4). These associations tended to be imprecise given the low number of users of permanent relaxers in the past 12 months.

Consistent with findings from past 24-hour use of hair products (sprays, gels, oils, and mousses), past 12-month use of hair styling products and moisturizing creams or lotions was consistently associated with increased concentrations for several EDC biomarkers (Figure 3). Compared with no reported use in the past 12 months, styling product use was associated with higher concentrations for several phenols (benzophenone-3, BPA, and BPS). Low use of styling products was further associated with both metabolites of DEHTP (MECPTP: β= 75.1%, 95% CI= −0.8%, 209.1%; MEHHTP: β= 80.4%, 95% CI= 8.4%, 200.3%), while low, moderate, and high use were each associated with the parabens (Figure 3a), where associations were strongest for ethyl paraben concentrations (low use: β= 78.6%, 95% CI= −11.8, 261.7%; moderate use: β= 89.6%, 95% CI= −0.7%, 262.1%; high use: β= 40.5%, 95% CI= −23.5%, 158.0%).

Figure 3.

Mean percent differences (95% CIs) in urinary EDC biomarker concentrations associated with (a) styling product and (b) moisturizing cream use in the past 12 months in SELF (n=425). Non-use in the past 12 months was set as the reference group. Models are adjusted for age, household income, educational attainment, employment status, and marital status, and are co-adjusted for all past 12-month and past 24-hour hair product variables.

¹Low use= <1 time/month, moderate use= 1–3 times/month, high= ≥1 times/week.

²BP3 = benzophenone-3; BPA= bisphenol A; BPF= bisphenol F; BPS= bisphenol S; MEP= monoethyl phthalate; mono-hydroxybutyl phthalate; MBP= mono-n-butyl phthalate; MHBP = mono-hydroxybutyl phthalate; MiBP= mono-isobutyl phthalate; MHiBP= mono-hydroxyisobutyl phthalate; MBzP= monobenzyl phthalate; MCPP= mono-3-carboxypropyl phthalate; MEHP= mono (2-ethylhexyl) phthalate; MEHHP= mono (2-ethyl-5-hydroxyhexyl) phthalate; MEOHP= mono (2-ethyl-5-oxohexyl) phthalate; MECPP= mono (2-ethyl-5-carboxypentyl) phthalate; MONP = monooxononyl phthalate; MCOP= mono-carboxyisooctyl phthalate; MCNP= mono-carboxyisononyl phthalate; MECPTP= mono-2-ethyl-5-carboxypentyl terephthalate; MEHHTP= mono-2-ethyl-5-hydroxyhexyl terephthalate.

Past 12-month use of moisturizing creams or lotions was associated with increased concentrations of several phenols, including a 64.9% increase (95% CI= 7.1%, 153.8%) in benzophenone-3 concentrations among high users, compared with non-use (Figure 3b). Low or moderate use of moisturizing creams or lotions was also associated with increased concentrations of BPA, BPS, 2,4-dichlorophenol, and 2,5-dichlorophenol. As with product use in the past 24 hours, moisturizing cream or lotion use was associated with increased concentrations of MHiBP (low use: β= 30.0%, 95% CI= −5.2%, 77.5%), methyl paraben (moderate use: β=36.3%, 95% CI= −9.7%, 105.8%), and ethyl paraben (moderate use: β= 129.3%, 95% CI= 29.9%, 304.9%), as well as MECPTP (high use: β= 35.0%, 95% CI= −10.6%, 103.7%) and MEHHTP (high use: β=58.4%, 95% CI= 9.2%, 129.9%; Figure 3b). We observed similar associations for past 12-month use of hair conditioners, crème rinses, or detanglers, though these products were also associated with increased BPF concentrations (e.g., high use: β= 71.6%, 95% CI= 1.1%, 191.3%; Figure S5). There was also evidence that use of moisturizing creams or lotions was associated with lower concentrations of MCPP, MONP, and MCNP (Figure 3b).

Similar to findings for use of conditioners, crème rinses, or detanglers, high reported use of hair food and hair oils in the past 12 months was associated with 89.6% (95% CI= 9.5%, 228.3%) and 49.2% (95% CI= −10.4%, 148.3%) higher BPF concentrations, respectively, compared with non-use (Figure 4). In addition, moderate use of hair food was associated with 24.6% (95% CI= 0.4%, 54.6%) and 39.1% (95% CI= 3.7%, 86.6%) higher MHBP and MBzP concentrations, but lower concentrations of other phthalate metabolites (MECPTP and MEHHTP) and several phenols (2,4-dichlorophenol, 2,5-dichlorophenol, and triclosan; Figure 4a).

Conversely, high use of hair oils in the past 12 months was associated with 43.3% (95% CI= −6.9, 120.6%) greater MECPTP concentrations, compared with non-use, but lower concentrations of MHBP, MCOP, and MCNP (Figure 4b). Consistent with other hair products, past 12-month use of 100% shea butter, compared with non-use, was associated with higher concentrations of phenols (e.g., high use, BPS: β= 71.6%, 95% CI= 18.2%, 149.0%), and DEHP metabolites (moderate use, ∑DEHP metabolites: β= 23.4%, 95% CI= −4.4%, 59.2%; Figure S6).

Past 12-month use of hair coloring products was consistently associated with higher EDC biomarker concentrations (Figure S7 and Figure S8a–d). High use (vs. non-use) of permanent hair dye, for example, was associated with 89.6% (95% CI= 1.3%, 255.1%), 85.9% (95% CI= −10.0%, 283.9%), and 80.4% (95% CI= −10.9%, 265.3%) increases in MCPP, MONP, and MCOP concentrations, respectively (Figure S7). Concentrations of multiple phthalate biomarkers were similarly higher for participants reporting any use, compared with non-use, of semi-permanent dye (∑DnBP: β= 25.9%, 95% CI= −0.6%, 59.2%; ∑DiBP: β= 31.0%, 95% CI= 1.5%, 69.0%) and bleach, frosts, or highlights (∑DEHTP: β= 80.4%, 95% CI= 10.5%, 194.5%), but lower for use of hair gloss (∑DiBP: β= −20.5%, 95% CI= −38.4%, 2.5%; ∑DEHP: β= −18.9%, 95% CI= −35.9%, 2.6%); associations were generally null for use of dyes that fade (Figure S8a–d). Finally, any past 12-month use of 100% petroleum jelly was associated with higher paraben concentrations, but lower concentrations of benzophenone-3 and several phthalate metabolites (MBP, MHBP, MEHP, MEHHP, and MEOHP; Figure S8e).

3.3. Sensitivity analyses

Compared with our primary analyses, we observed similar associations in sensitivity analyses in which we: 1) adjusted for past 24-hour personal care product use (data not shown), and 2) did not co-adjust for all hair products (used in the past 24 hours and past 12 months) in regression models, with some exceptions. Notably, we observed stronger associations between several hair products and BPF concentrations in reduced models that were not co-adjusted for all hair products, whereas these associations were null or attenuated in the main findings: 1) product use in the past 24-hours (any use vs. non-use: β= 33.6%, 95% CI= −4.2%, 86.5%); moisturizing creams or lotions (high use vs. non-use: β= 47.7%, 95% CI= −0.2%, 118.6%); and 100% shea butter (high use vs. non-use: β= 71.6%, 95% CI= 7.2%, 174.7%) (Figures S9–S11).

4. DISCUSSION

In this study of reproductive-aged Black women, recent use of several hair products was associated with concentrations of multiple non-persistent EDC biomarkers. Specifically, use of hair sprays, gels, oils, or mousses in the 24 hours before urine collection was associated with higher biomarker concentrations of DiBP metabolites and parabens. Similarly, past 12-month use of styling products, moisturizing creams or lotions, conditioning products, hair food, hair oil, and 100% shea butter was associated with greater concentrations of several EDC biomarkers. High use of permanent hair relaxers, straighteners, or perms was also associated with increased biomarker concentrations of some phthalates. However, the direction and magnitude of associations for each product varied by EDC biomarker and, for most hair products, there was no evidence of monotonic dose responses with greater frequency of use (i.e., low, moderate, and high).

Several prior epidemiologic studies have reported similar associations between use of hair products and increased EDC biomarker concentrations (Braun et al., 2014; Hsieh et al., 2019; Parlett et al., 2013; Sakhi et al., 2017). Hair sprays, conditioners, shampoos, and gels were associated with higher concentrations of urinary phthalates and parabens in reproductive-aged women (Braun et al., 2014; Hsieh et al., 2019; Parlett et al., 2013; Sakhi et al., 2017). Two of these studies were conducted in Taiwanese and Norwegian women (Hsieh et al., 2019; Sakhi et al., 2017). and the two studies in U.S. populations were comprised primarily of white women (>91%) (Braun et al., 2014; Parlett et al., 2013). To our knowledge, this is the first study to examine associations between hair product use and EDC biomarker concentrations in a large cohort of Black women.

Hairstyling has a long socio-cultural and political history in Black communities (Byrd and Tharps, 2014; Hair et al., 2014; Teteh et al., 2017). Discriminatory standards of beauty that historically idealized Eurocentric hairstyles (long, straight hair) have created both implicit and explicit biases against natural hair (Johnson et al., 2017; Zota and Shamasunder, 2017), such that an estimated one-third of U.S. Black women have experienced some form of discrimination when wearing natural hairstyles (Hair et al., 2014). This discrimination has sustained pressure at various levels (i.e., workplace, school, social settings) on Black individuals to conform to Eurocentric hair texture preferences in the U.S. for hundreds of years (Teteh et al., 2017; Zota and Shamasunder, 2017), with long-lasting effects on patterns of hair product use: Black women, for example, report higher rates of use for multiple hair products (e.g., relaxers, oils, root stimulators) than other racial groups (Collins et al., 2021; Dodson et al., 2021; James-Todd et al., 2012; Zota and Shamasunder, 2017). Further, Black women are twice as likely to report feeling pressure to straighten their hair to conform with workplace hair texture preferences compared with white women (Johnson et al., 2017).

Findings from the present study suggest that Black women may experience EDC exposure from hair product use. Product testing data support our results: hair products used by Black women were found to contain an array of non-persistent EDCs, and several hair products were hormonally active (Helm et al., 2018; James-Todd et al., 2021). Hair oils, conditioners, straighteners, and root stimulators, for example, contained chemicals from multiple chemical and/or functional classes (e.g., phthalates, parabens, UV filters) (Helm et al., 2018). Further, ingredients in hair products, including lotions, conditioners, oils, and root stimulators, acted as agonists (estrogen) and antagonists (androgen, progesterone, glucocorticoid) in assays (James-Todd et al., 2021), demonstrating the ability of hair products targeted to Black women to disrupt hormonal systems, which is consistent with evidence in human populations (Geczik et al., 2023; Rivera-Núñez et al., 2022).

Hair product use, and associated exposure to EDCs, has important public health implications. A large body of evidence suggests that hair products, in particular relaxers/straighteners and dyes, are associated with earlier age at menarche (James-Todd et al., 2011), reduced fecundability (Wise et al., 2023), and greater incidence of uterine fibroids and several cancers (e.g., breast, uterine, and ovarian) (Table 1) (Chang et al., 2022; Coogan et al., 2021; White et al., 2021b; Wise et al., 2012). Black women have higher incidence (relative to white women) of several of these health outcomes (e.g., uterine fibroids) (Eltoukhi et al., 2014; Nnorom and Wilson, 2022). Addressing EDC exposure from hair product use may provide an opportunity to reduce racial disparities in rates of adverse health outcomes. However, micro-/individual-level public health interventions focused on reducing EDC exposure from hair products through consumer education may not be impactful because studies have found that Black women are aware of and try to avoid harmful chemicals in consumer products where possible (Collins et al., 2021; Teteh et al., 2017). Moreover, chemicals are often not accurately reported on hair product labels or may be concealed as proprietary ingredients, suggesting that exposure is still possible even if consumers are actively trying to avoid certain harmful chemicals (Helm et al., 2018). Companies, supported by regulatory oversight, should test, accurately label, and, where possible, remove chemicals used in their products, especially those targeting Black women. Additional structural interventions are needed: beginning in 2019, several states passed laws (known as CROWN Acts) that ban school and workplace discrimination against natural hairstyles (Gonzales, 2023). Whether such policies will influence patterns of hair product use, and therefore EDC exposure, among Black women remains to be seen; but it is clear that the use of some products (i.e., relaxers) declines as women embrace natural hairstyles (Council, 2022). Therefore, future research should investigate the efficacy of such policies at reducing EDC exposure among Black women.

Although there was consistent evidence that hair product use was associated with higher concentrations of some EDC biomarkers in the SELF cohort, few of these associations were monotonic in nature. In addition, we observed some unexpected inverse associations. We hypothesize that the inverse associations may reflect residual confounding from changing product formulations (e.g., replacements) over time or other hair products for which we lacked data. For example, we found consistent evidence of inverse associations between petroleum jelly use and ∑DEHP metabolite concentrations; these inverse associations may instead reflect lower rates of use for hair products with similar uses among petroleum jelly users that we were not able to adjust for (e.g., anti-frizz products), which also contain EDCs (Helm et al., 2018). We further hypothesize that the non-monotonic associations we observed (i.e., stronger adverse associations among low or moderate users comparatively to high users) may reflect non-differential misclassification. Notably, the EDC biomarkers included in this study have relatively short biological half-lives (e.g., hours, days) and product use in the few days preceding urine collection would be most relevant for biomarker measurement (Centers for Disease Control and Prevention, n.d.; Stahlhut et al., 2009; Wei et al., 2021). Although we had data on frequency of use for hair products in the previous 12 months, we did not collect data on time since most recent product use in relation to urine collection. Therefore, some participants reporting low use of products may have used the product temporally closer to urine collection than participants reporting high use, potentially explaining the non-monotonic dose responses. We had insufficient data to validate this hypothesis, though only 58% of participants reporting high use of styling products in the past 12 months also reported using hair products in the past 24 hours (Table S2). We also had no data on the formulations or the quantities of product used, which could have introduced error in our estimation of EDC exposure patterns. We also did not quantify biomarker concentrations for several EDCs that have been detected in hair products targeted to Black women (Table 1).

Our study had additional limitations. As discussed above, there was likely misclassification of hair product use in relation to the timing of EDC exposure assessment. Misclassification of hair product use likely arose given product use was self-reported; notably, we were able to compare reported use of styling products in the past 12 months and reported use of hair products in the past 24 hours, given the overlap in products included for these variables (e.g., mousses, gels). Twenty-three percent of participants who reported no styling product use in the past 12 months also reported product use in the past 24 hours (Table S2), highlighting the potential for misclassification of self-reported hair product use. Removing these participants from regression analyses yielded slightly stronger results for styling product use in the past 12 months (data not shown), suggesting that misclassification of hair product variables was likely non-differential with respect to EDC biomarker measurements. We were not, however, able to assess the degree of potential misclassification of other hair product variables. Given the relatively short half-lives of the EDCs and episodic nature of exposure from hair product use, it is also likely there was some degree of misclassification in the single measurement of EDC biomarker concentrations, where the measurement did not fully reflect cumulative exposures. However, we would also expect this misclassification to be non-differential. Our sample size was also limited (n=425), and there were low frequencies of use for some hair products. As such, several of our estimates were imprecise. The case-cohort design of our study further included participants who were randomly selected into the case-cohort substudy at baseline (some of whom developed incident UL during follow-up), as well as participants who were not randomly selected but had an incident UL over follow-up (Figure S1). In addition, as part of the prospective design, participants who developed incident UL between baseline and 40 months did not have their 40-month urine samples analyzed for EDCs. Therefore, incident UL cases were underrepresented in the current analysis of 40-month EDC data. This could have potentially induced a selection bias if EDC exposure was related to UL incidence, although we have not observed strong associations between non-persistent EDC biomarker concentrations and UL incidence in SELF (Fruh et al., 2021; Wesselink et al., 2021b), suggesting any potential bias would be small. Finally, there may be residual confounding due to other sources of exposure to EDCs, including other personal care products that we not able to account for and that may be correlated with hair product use.

Study strengths include our investigation of hair product use with EDC biomarker concentrations among Black women, who have a high prevalence of hair product use but remain an understudied population for EDC exposure. We assessed this research question in an established cohort of U.S. Black women who provided extensive information on hair product use, potential confounders, and urine samples. As such, we were able to examine associations between use of multiple hair products and biomarker concentrations for a suite of EDCs, while controlling for relevant socioeconomic and demographic factors. We used state-of-the-art methods to assay non-persistent EDCs in urine, thereby reducing potential for misclassification of EDCs.

In conclusion, we found that recent use of some hair products was positively associated with biomarker concentrations of several non-persistent EDCs in a cohort of reproductive-aged Black women. Additional studies with larger sample sizes and better temporal resolution for hair product use and EDC measurement are needed to support the findings from this study. However, our results suggest that hair product use may be an important source of exposure to several non-persistent EDCs among Black women, which has important public health and policy implications.

HIGHLIGHTS.

• We assessed associations of hair product use and biomarkers of endocrine disrupting chemicals (EDCs).

• Hair product use was associated with higher biomarker concentrations of EDCs.

• Hair product use may be an important route of exposure to EDCs among Black women.

Abbreviations:

BPA

bisphenol A

BPF

bisphenol F

BPS

bisphenol S

BzBP

benzylbutyl phthalate

CAWI

computer assisted web interview

CDC

Centers for Disease Control and Prevention

CI

confidence interval

DAG

directed acyclic graph

DCP24

2,4-dichlorophenol

DCP25

2,5-dichlorophenol

DEHP

di(2-ethylhexyl) phthalate

DEHTP

di-2-ethylhexyl terephthalate

DEP

diethyl phthalate

DiBP

di-isobutyl phthalate

DiDP

diisodecyl phthalate

DINCH

di(isononyl)cyclohexane-1,2-dicarboxylate

DiNP

diisononyl phthalate

DnBP

di-n-butyl phthalate

DOP

dioctyl phthalate

EDC

endocrine disrupting chemical

EPB

ethyl paraben

HFH

Henry Ford Health

LOD

limit of detection

MBP

mono-n-butyl phthalate

MBzP

monobenzyl phthalate

MCNP

mono-carboxyisononyl phthalate

MCOCH

1,2-cyclohexane dicarboxylic acid-monocarboxy isooctyl ester

MCOP

mono-carboxyisooctyl phthalate

MCPP

mono-3-carboxypropyl phthalate

MECPP

mono (2-ethyl-5-carboxypentyl) phthalate

MECPTP

mono-2-ethyl-5-carboxypentyl terephthalate

MEHHP

mono (2-ethyl-5-hydroxyhexyl) phthalate

MEHHTP

mono-2-ethyl-5-hydroxyhexyl terephthalate

MEHP

mono (2-ethylhexyl) phthalate

MEOHP

mono (2-ethyl-5-oxohexyl) phthalate

MEP

monoethyl phthalate

MHBP

mono-hydroxybutyl phthalate

MHiBP

mono-hydroxyisobutyl phthalate

MHiNCH

1,2-cyclohexane dicarboxylic acid-monohydroxy isononyl ester

MiBP

mono-isobutyl phthalate

MNP

mono-isononyl phthalate

MPB

methyl paraben

NIEHS

National Institute of Environmental Health Sciences

PABA

4-aminobenzoic acid

PPB

propyl paraben

SD

standard deviation

SE

standard error

SELF

Study of Environment, Lifestyle, and Fibroids