We selected 66 organic chemicals for inclusion in the study based on evidence of endocrine disruption or asthma exacerbation, expected presence in consumer products, and compatibility with analytical methods developed in our household exposure studies (Rudel et al. 2003, 2010). We tested 85 samples representing 213 products in two rounds of chemical analysis. The chemical groups, their typical uses, and the evidence of endocrine disruption or asthma exacerbation are listed in Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1104052).

Product selection. We first identified the types of products likely to contain the compounds of interest. Product types included personal care products (e.g., lotion, hair products, toothpaste), cleaners (e.g., laundry detergent, all-purpose cleaner), and other household goods. We then identified several “conventional” products and one “alternative” for each product type. Exclusion criteria for alternative products are listed in Table 1. A product was classified as alternative if the label did not include the terms listed in Table 1. Many of the products that met our criteria for alternative products were marked as “green.” We also identified as alternative products six items often used in recipes for homemade cleaners, such as bleach and vinegar. Products that did not meet the “alternative” criteria were classified as conventional. In selecting conventional products, we tried to choose products that are widely used in order to better represent typical exposures. Because we lacked comprehensive information from which to select products, we identified leading companies for the product sector (e.g., hair care) based on market share and selected candidate products from several leading companies. When possible, we also included a generic store-brand product. Final product selections were made informally on the basis of availability and shelf space.

We purchased most alternative products at a nationwide store specializing in natural products, so products met the store’s selection criteria, which favored non–petroleum-based—and especially plant-based—ingredients. Most of the conventional products were purchased at major grocery and pharmacy chain stores primarily in fall 2007. We added products for a second round of chemical analysis approximately 1 year later. Names of the products that were tested and their manufacturers are available from Silent Spring Institute (2012).

Sampling design and compositing. We analyzed 42 analytical samples composited from 170 conventional products and 43 samples of individual alternative products.

To cost-effectively evaluate typical exposures from conventional products, we composited 170 conventional products into a single sample for each product type (42 analytical samples). We combined equal masses of 1–7 products within a product type and analyzed the mixture as a single sample. The advantage of compositing is that samples may provide more generalizable exposure information. However, composited samples are more limited in that they a) will not reveal an unusually high concentration in a single product if that product is mixed with others having lower concentrations; b) will not reveal a concentration just above the limit of detection (LOD) in a single product if that product is mixed with others having concentrations < LOD; and c) may show a higher detection frequency for chemicals well within the detectable range.

We sought to identify specific products that were free of the chemicals of concern (alternative products), so the products could be used in an intervention study. Thus, we analyzed just 1 alternative product per product type (43 analytical samples, 1 for each of 43 individual products). Therefore, reported detection frequencies and concentrations for conventional and alternative product types are not directly comparable. To provide some information about variability in products within a category, we tested individual samples of 5 alternative sunscreens and calculated an average for the product type “alternative sunscreen.”

Chemical analysis. We analyzed samples in two rounds: 50 compounds in the first round and those same 50 compounds plus 16 other compounds in the second round. Products were composited as described above, and surrogate recovery standards were added. Samples were then extracted with dichloromethane:methanol, passed through a weak anion exchange cartridge, spiked with internal standard, and analyzed by gas chromatography/mass spectrometry in the full scan mode. A separate aliquot was derivitized and analyzed for phenolic compounds.

For each compound, the method reporting limit (MRL) was defined as the maximum analytical LOD and the 90th percentile of the blank concentrations within each analytical round. The reporting limit was 1 µg/g for chemicals in products, but it was reported as > 1 µg/g if there were detectable concentrations in the blank samples (1 chemical in analytical round 1 and 12 chemicals in analytical round 2).

We included extensive quality assurance/quality control (QA/QC) samples in our analyses. Chemical detection in blanks was infrequent, and elevated MRLs were ≤ 5 µg/g except for cyclosiloxane decamethylcyclopentasiloxane (D₅; the only compound detected in > 75% of blanks). Results were blank corrected by subtracting the median blank value from the reported value. Precision was assessed with 13 duplicate samples (relative percent difference was generally < 50%); accuracy was assessed by determining spike recovery for all target compounds in six different matrices (median recoveries across products were generally within 50–150%) and by calculating recoveries of surrogates in all samples (median percent recoveries were within the 50–150% acceptance range for all surrogates in both analytical rounds). For additional details regarding chemical analysis and QA/QC measures, see Supplemental Material, pp. S-9–S-10 (http://dx.doi.org/10.1289/ehp.1104052).

Data analysis. Our analysis of this large data set is visual and exploratory. We graphed product type against compounds detected using a “heat map” approach for conventional and alternative products (Figures 1 and 2, respectively). Only values > MRL or > 1 µg/g are presented. We graphed results for sunscreens in a similar format [see Supplemental Material, Figure S1 (http://dx.doi.org/10.1289/ehp.1104052); results are presented for a composited sample of conventional sunscreens, the calculated composite obtained by averaging results for five alternative sunscreens, and individual results for the five alternative sunscreens).

To identify chemicals that tend to co-occur because they are used together in a product, we estimated correlations for chemicals simultaneously detected within a product type (e.g., laundry detergent, lipstick). We calculated Kendall’s tau adjusted for censored data and with p-values obtained from 10,000 bootstrap replications (Newton and Rudel 2007). The magnitude of Kendall’s tau coefficients tends to be smaller than those of the more familiar Spearman’s correlation coefficients. We limited this analysis to chemicals detected in more than three analytical samples, and we conducted analyses separately for conventional and alternative products.

We tested 213 conventional and alternative consumer products in 85 analytical samples for up to 66 compounds of interest. We detected 55 compounds: 50 chemicals in 42 conventional samples representing 170 products, and 41 compounds in 43 alternative samples representing 39 product types.

The most frequently detected compounds in conventional products were two natural fragrance compounds that may be derived from plant materials, two synthetic fragrance compounds, and diethyl phthalate (DEP) (Figure 1). The most frequently detected compounds in alternative products, including the calculated composite sunscreen, were two natural fragrance compounds, an alkylphenol, methyl paraben and DEP (Figure 2).

We detected 11 compounds at concentrations > 1% (10,000 µg/g) and 26 compounds at > 0.1%. DEHP was detected at 28% and 14% by weight in the vinyl shower curtain composite and the vinyl pillow protector, respectively. The glycol ether 2-butoxyethanol was detected at nearly 5% in the carpet cleaner. The sunscreen samples contained 2.5–6.2% of the UV filters octinoxate and benzophenone-3 (BP-3). The fragrance/​perfume composite contained almost 3% hexahydrohexamethyl cyclopentabenzopyran (HHCB), a synthetic fragrance chemical, and 1.4% DEP. The car air freshener contained the fragrance chemicals isobornyl acetate and limonene (a natural fragrance compound) at approximately 2% each. The alternative shampoo sample had 2.4% diethanolamine. Additional findings are described by chemical class.

Parabens. Parabens are added to many consumer products, pharmaceuticals, and foods as preservatives and antimicrobial agents (Soni et al. 2001). Previous studies found parabens, particularly methyl paraben, in most cosmetic samples (Rastogi et al. 1995; Shen et al. 2007). In a study of 100 demographically diverse adults, Ye et al. (2006) detected methyl and propyl paraben in > 96% of urine samples. Parabens are weakly estrogenic in vitro, and butyl paraben (100 mg/kg) has been reported to affect reproductive tract development in rats (Kang et al. 2002).

We detected parabens in personal care products but not in cleaners. Methyl paraben was detected most frequently and at the highest concentrations; ethyl and butyl paraben were found only if methyl paraben was also detected. The highest concentration was in an alternative sunscreen (methyl paraben; 1,600 µg/g). Of the 11 conventional samples with detectable parabens, 10 included products with “paraben” on the label. With the exception of shaving cream, products were not considered alternative if parabens were listed as an ingredient. Nevertheless, in alternative products, we detected parabens in 7 products, including 3 sunscreens, that did not list parabens on the label.

Phthalates. Phthalates are used as plastic additives, as solvents in cosmetics and perfumes, and as an inert ingredient in pesticides. Higher molecular weight phthalates (e.g., DEHP) are typically used in plastics (10–60% by weight) and readily migrate out of products (Rakkestad et al. 2007). Lower molecular weight phthalates [e.g., di-n-butyl phthalate (DBP), DEP] are typically used as solvents in personal care products and in lacquers, varnishings, and coatings (Meeker et al. 2009b). Several different phthalates have been reported in cosmetics and other personal care products, sometimes at concentrations > 1% (Hubinger and Havery 2006; Koniecki et al. 2011; Shen et al. 2007). Near universal detection of phthalates in urine samples shows widespread exposure (CDC 2009; Heudorf et al. 2007). In humans, phthalates have been associated with adverse reproductive system outcomes, including reduced semen quality and altered male genital development, as well as respiratory symptoms (Bornehag et al. 2004; Engel et al. 2010; Hauser and Calafat 2005; Hauser et al. 2006; Kimber and Dearmna 2010; Meeker et al. 2009a, 2009b; Mendiola et al. 2011; Swan 2008; Swan et al. 2005). Many phthalates are identified as antiandrogenic EDCs in mammalian models, whereas DEP is not generally characterized as an endocrine-active compound (Hannas et al. 2011; Heindel et al. 1989; Howdeshell et al. 2008). Among the EDCs in the present study, phthalates are the only chemical group for which there is supporting evidence of health effects from human studies.

We analyzed samples for 12 phthalates. DEP, a common solvent for fragrance (Hubinger and Havery 2006), was detected most frequently; the highest DEP concentrations were in fragrance/perfume (14,000 µg/g) and car air freshener (8,000 µg/g). Vinyl products had the highest concentrations of any phthalate, with DEHP at 28% in the shower curtains and 14% in the pillow protector. DBP and benzylbutyl phthalate (BBP) were detected in the conventional pillow protector, dryer sheet, polish/wax, car interior cleaner, and in alternative tub/tile cleaner, bar soap, shaving cream, and lipstick. Three phthalates were found only in alternative products: dicyclohexyl phthalate (DCP), diisononyl phthalate (DINP), and di-n-propyl phthalate (DPP). These compounds may have been introduced as substitutes for the better-known antiandrogenic phthalates (DBP, BBP, DEHP), even though they are also EDCs and have similar (DCP) or lesser (DINP, DPP) potency (Boberg et al. 2011; Hannas et al. 2011; Heindel et al. 1989; Saillenfait et al. 2009). The alternative shaving cream contained 5 different phthalates, illustrating the potential for simultaneous exposures to multiple phthalates, which act cumulatively on endocrine targets (National Research Council 2008). None of the products we tested had “phthalate” on the label, including personal care products requiring that intentional ingredients must be labeled. However, the conventional nail polish sample with measurable DEHP contained a product labeled “phthalic anhydride copolymer.”

BPA. BPA is used in a variety of consumer products containing epoxy resins, polyester-styrene, and polycarbonate plastics. It can be an ingredient in vinyl and in dental sealants, protective coatings, flame retardants, and adhesives (Meeker et al. 2009b). Biomonitoring studies indicate that exposure is widespread; BPA was detected in > 93% of urine samples in the National Health and Nutrition Examination Survey (NHANES) (Calafat et al. 2008b). A wide body of laboratory evidence shows BPA-induced endocrine disruption in a number of organ systems (Food and Agriculture Organization of the United Nations/World Health Organization 2010).

We detected BPA in 15 conventional samples, including the vinyl shower curtain and pillow protector, dish and laundry detergent, tub and tile cleaner, soaps, lotions, shampoo, conditioner, shaving cream, nail polish, and sunscreen. Concentrations were < 100 µg/g, with most < 10 µg/g. BPA was not detected in alternative samples except sunscreen, so selecting alternative products according to our criteria appears to avoid BPA. No labels listed BPA.

Antimicrobials. We analyzed four antimicrobials: ortho-phenylphenol, triclocarban, triclosan, and 1,4-dichlorobenzene. Both triclocarban and triclosan are broad-spectrum agents commonly added to personal care products, such as toothpastes and soaps, detergents, toys, plastics, and textiles (Adolfsson-Erici et al. 2002; Calafat et al. 2008c; Perencevich et al. 2001). A national survey detected triclosan and triclocarban in 76% of liquid soaps and 29% of bar soaps (Perencevich et al. 2001), and triclosan was detected in 76% of NHANES urine samples (Calafat et al. 2008c). Triclosan has been shown to disrupt thyroid homeostasis in mammalian models (Paul et al. 2010; Rodríguez and Sanchez 2010), and current human exposure levels are in the range of those predicted to have this activity based on laboratory tests (Rotroff et al. 2010). Triclocarban has been shown to amplify endogenous androgen response in mammalian models (Chen et al. 2008). Personal care product labels must list antimicrobial concentrations (FDA 2009).

We detected triclocarban and triclosan but not the other two antimicrobials. When they were detected at higher concentrations, they were listed as active ingredients on the product labels, consistent with FDA labeling requirements. For example, the conventional bar soap sample contained triclocarban at 1,520 µg/g, and one of the four soaps in the composite was labeled “antibacterial” with triclocarban (0.6%). Concentrations of triclosan in conventional hand soap and toothpaste were slightly lower than predicted from labeling of active ingredient concentrations. Some products (conventional facial cleanser and lipstick) listed triclosan in the ingredient list but not as an active ingredient; however, we did not detect it in those composite samples. Also, we found relatively low levels (6 µg/g) of triclosan in conventional dish liquid composite, although it was not listed on the product labels. We did not detect these compounds in any of the alternative products.

Ethanolamines. Monoethanolamine (MEA) is used in cleaners and degreasers, detergents, soaps, cosmetics, hair dyes, and as an emulsifier in lotions and creams; diethanolamine (DEA) is used as an emulsifier in shampoos, cleaners, detergents, polishes, and auto products (National Library of Medicine 2010b). Exposure studies are limited. MEA and DEA have been associated with occupational asthma (Association of Occupational and Environmental Clinics 2010; Kamijo et al. 2009; Mäkelä et al. 2011; Piipari et al. 1998; Savonius et al. 1994). The European Commission prohibits DEA in cosmetics and restricts products with MEA to < 0.5% amine content because of concerns about formation of carcinogenic nitrosamines (European Commission 2011).

We detected MEA in conventional glass cleaner and laundry detergent (> 1,000 µg/g) and in alternative surface cleaner, glass cleaner, and shampoo (< 400 µg/g). The highest DEA concentration was in the alternative shampoo (24,000 µg/g; 2.4% by weight). DEA was detected in the composite sample of conventional dryer sheets (840 µg/g) and in four other conventional and alternative cleaning and personal care product samples (< 150 µg/g): conventional face lotion and alternative tub and tile cleaner, mascara, and shaving cream.

No product labels listed MEA or DEA as ingredients. Several product labels included the derivatives “cocamide MEA,” “lauramide DEA,” or “cocamide DEA,” but MEA and DEA were not detected in samples of these products. We detected DEA in mascara labeled with triethanolamine (TEA), but MEA and DEA were not detected in other TEA-labeled products. Commercial mixtures of TEA may contain small amounts of DEA and MEA [International Agency for Research on Cancer (IARC) 2000].

Alkylphenols. Alkylphenol polyethoxylates (APEOs), including nonylphenol and octylphenol ethoxylates, are used as surfactants in consumer products such as detergents, disinfectants, and surface cleaners, and as “inert” ingredients in pesticides. Mixtures containing ethoxylate chains of varying lengths are typical and can degrade to nonylphenol and octylphenol, both identified as weakly estrogenic (Jie et al. 2010). Nonylphenol also originates from vinyl and other plastics containing tris(nonylphenol) phosphite and may have other uses, including in epoxy resins. The branched chain para-substituted nonylphenol and octylphenol are the commercially prevalent compounds (European Commission 2002; Rudel and Perovich 2009; Rudel et al. 2010).

We measured 4-t-octylphenol and 4-t-nonylphenol, as well as their mono and diethoxylates, and detected them in about half of the samples, including plastics, cleaners, and personal care products. Concentrations were low (< 20 µg/g) except in the conventional car cleaner and vinyl shower curtain.

Product labels did not list alkylphenols. Of the 37 alkylphenol-containing samples, 7 included products labeled as containing “surfactants” of any type (e.g., ionic, nonionic). Ten samples contained at least one product labeled as containing “surfactants,” but alkylphenols were not detected. Because many products containing APEOs did not list surfactants on labels, a significant fraction of the products may contain 4-t-nonylphenol from other uses, such as plasticizers, or are simply unlabeled. Thus, it appears that exposure to alkylphenols cannot be avoided by reading product labels.

Fragrances. Fragrances are added to products to achieve a desired scent or mask other scents in the product. More than 3,000 fragrance ingredients have been reported, and a fragranced product may contain 50–300 different chemicals (Bickers et al. 2003). Exact formulations are typically protected from disclosure (Bridges 2002; International Fragrance Association 2010). Fragrances can be either synthetic or natural/plant-based; however, some natural fragrance chemicals can be artificially synthesized, and these may or may not reflect the natural stereoisomer composition (Ravid et al. 2010) and may have different health effects (Smith 2009). In the present study we classified fragrance chemicals as natural if they are readily available from plant materials and synthetic if they are most commonly synthesized, based on information in the Hazardous Substances Data Bank (National Library of Medicine 2010a); however, we did not independently verify that the natural fragrance chemicals were not synthesized. Synthetic fragrance compounds, which include polycyclic and nitro musks, have been found in many personal care and household products and at concentrations up to several thousand micrograms per gram (Reiner and Kannan 2006; Zhang et al. 2008). Synthetic and natural fragrance compounds have been reported in cleaning products (Rastogi et al. 2001), cosmetics, and perfumes. In a 1996 study, a high proportion of perfumes and cosmetics labeled as containing natural ingredients in fact contained synthetic fragrances (Rastogi et al. 1996). Fragrances, particularly terpenes such as limonene, are associated with secondary chemical reactions in indoor air and can contribute to the production of formaldehyde, glycol ethers, ultrafine particles, and secondary organic aerosols (Nazaroff and Weschler 2004; Singer et al. 2006). Exposure to fragrances has been associated with a range of health effects, including allergic contact dermatitis, asthma and asthmatic exacerbations, headaches, and mucosal symptoms (Heydorn et al. 2003; Kumar et al. 1995; Steinemann 2009). Synthetic musks have been shown to have estrogenic effects (Bitsch et al. 2002; Schreurs et al. 2005; Seinen et al. 1999; van der Burg et al. 2008).

We frequently detected synthetic and natural fragrance chemicals in conventional samples. In our alternative products selected to allow only plant-based fragrance, synthetic fragrance chemicals were detected only in the facial cleanser, floor cleaner, and one sunscreen (< 100 µg/g). Bucinal, HHCB, and methyl ionone were the most frequently detected synthetic fragrance chemicals in conventional product samples. Concentrations of these and AHTN (acetyl hexamethyl tetralin), isobornyl acetate, and phenethyl alcohol included detects > 1,000 µg/g in fragrance/perfume, car and home air fresheners, and dryer sheets. Natural fragrance chemicals were common in conventional and alternative products. Most common natural fragrance chemicals included the terpenes limonene, hexyl cinnamal, and linalool. Concentrations of fragrance compounds were generally higher in conventional (21 instances > 1,000 µg/g) than in alternative products (2 instances > 1,000 µg/g), reflecting that product types used specifically to create scent (e.g., air freshener, perfume) were categorized as conventional and can be avoided altogether. We identified 26 alternative samples with no detectable fragrance chemicals.

Of the 34 conventional samples with detectable fragrance chemicals, 22 contained a product labeled with “fragrance” or other similar descriptors (e.g., “parfum”). Products that contained fragrance chemicals with no label indication were generally cleaners. Of the 17 alternative samples with detectable fragrance chemicals, 14 did not include “fragrance” or similar descriptors as ingredients. The other 3 were labeled “essential oil fragrance” or “plant based fragrance” and contained only natural fragrance compounds. Only 1 sample (sunscreen) of the 26 alternative samples with no detectable fragrance compounds had a reference to “fragrance” on the label, specifically listing “fragrance oil blend.”

Glycol ethers. Glycol ethers, a chemical class with > 80 compounds, are used in a broad array of cleaning applications because of their combined hydrophilic and lipophilic nature. They are often used in paints, varnishes, and cosmetics and have been detected in a variety of household products (Kwon et al. 2008; Plaisance et al. 2008). Biomonitoring methods are currently being developed, so large-scale studies are limited. In human studies, exposure to glycol ethers has been associated with low sperm mobility (Cherry et al. 2008), hematological effects (Starek et al. 2008), and asthma and allergies (Choi et al. 2010).

In the present study, we analyzed all samples for 2-butoxyethanol and 2,2-methoxyethoxyethanol, and in a later second sampling round, we analyzed 14 additional samples for six other glycol ethers. We detected glycol ethers in 3 conventional cleaners, face lotion, polish/wax, sunscreen, and in alternative shaving cream, pillow protector, and sunscreen samples. Of the 5 conventional samples with detectable 2-butoxyethanol, only the carpet cleaner was labeled as containing 2-butoxyethanol. When analyzed and detected, other glycol ethers were not listed on labels. Although we detected phenoxyethanol in conventional and alternative sunscreen samples, we did not detect this chemical in some conventional and alternative samples comprising products labeled as containing this compound; levels may have been < LOD.

Cyclosiloxanes. Cyclosiloxanes (cyclic volatile methylsiloxanes) are added to consumer products to enhance conditioning and spreading (Silicones Environmental, Health and Safety Council of North America 2011). Cyclosiloxanes are widely used, with the most common types being hexamethylcyclotrisiloxane, octamethylcyclotetrasiloxane (D₄), D₅, and dodecamethylcyclohexylsiloxane (D₆). They have been found in cleaning products, personal care products, and baby products at concentrations as high as 1,010 µg/g (Horii and Kannan 2008; Wang et al. 2009). Cyclosiloxanes appear to be persistent and have relatively long half-lives in humans. D₄ has weak estrogenic potential (Quinn et al. 2007) and D₅ is potentially carcinogenic in rats (Wang et al. 2009).

Cyclosiloxanes were analyzed in 10 product types that were added during the second sampling round. All three cyclosiloxanes (D₄, D₅, and D₆) were detected in the alternative composite sunscreen (D₅ and D₆ at > 4,000 µg/g) and in the conventional car interior cleaner (< 100 µg/g). One cyclosiloxane was detected in the conventional sunscreen (D₅; 50 µg/g) and in the alternative shaving cream (D₆; 10 µg/g). No product analyzed for cyclosiloxanes indicated “siloxane” on the label; however, two alternative sunscreens were labeled cyclomethicone, a generic name for polydimethylsiloxane, which includes D₄, D₅, and D₆.

UV filters. Organic compounds that act as UV filters are added to many personal care products for skin protection and product stability. Three UV filters included in this study—benzophenone-3 (BP-3; oxybenzone), octyl dimethyl PABA (p-aminobenzoic acid), and octinoxate (octyl methoxycinnamate)—were detected in a previous study of 75 sunscreen products from European and U.S. manufacturers (Rastogi 2002). Biomonitoring data have indicated widespread exposure to some UV filters; BP-3 was detected in 96% of urine samples in NHANES 2003–2004 (Calafat et al. 2008a). Benzophenone-1, BP-3, and octinoxate are estrogenic in vitro and in vivo (Schlumpf et al. 2004) and act additively as mixtures (Kunz and Fent 2006).

We analyzed UV filters in sunscreens and eight other samples added during the second analytical round. We detected them at > 1% concentration in conventional and alternative sunscreen samples for which they were labeled as active ingredients. We detected lower concentrations of three UV filters in conventional sunscreen and shaving cream and in four of five alternative sunscreens, and none of these were labeled as containing these chemicals.

Mixtures: chemicals that co-occur within and across products. Our results show that one product can be a source of many chemicals of interest and that use of multiple products can result in exposure to an even larger number of chemicals.

We detected 0–22 analytes in a single product type (Figure 1). For composited samples, we do not know how many chemicals were in any one of the products; for alternative products, the number of detects ranged up to 11 analytes in shaving cream (Figure 2) and 17 in an individual sunscreen [see Supplemental Material, Figure S1 (http://dx.doi.org/10.1289/ehp.1104052)], illustrating the exposure to multiple compounds from a single product.

We identified chemicals that co-occur within a product type by estimating Kendall’s tau correlation coefficients [see Supplemental Material, Figure S2 (http://dx.doi.org/10.1289/​ehp.1104052)]. Many fragrance compounds were significantly correlated with each other and with DEP. For example, the natural fragrance limonene was correlated with natural fragrances linalool (τ_conventional = 0.43; τ_alternative = 0.59) and pinene (τ_conventional = 0.52; τ_alternative = 0.52) in both conventional and alternative samples. Limonene and linalool also were positively correlated with DEP in both conventional and alternative samples (τ = 0.31–0.52). In the conventional samples, DEP was positively correlated with several fragrance compounds (AHTN, benzyl acetate, bucinal, hexyl cinnemal, HHCB, linalool, limonene, and methyl ionone; τ = 0.34–0.56), which supports the idea that DEP is a common carrier for fragrances. The finding of positive correlations among the fragrance compounds may be influenced by compositing. For example, if each individual product within a product type uses a different fragrance, these compounds will be correlated in the composites, even though an individual product may contain only one of the compounds. We also found that 4-t-nonylphenol and DEHP were correlated in conventional samples (τ = 0.4), consistent with use of both compounds as plastics additives. Nonylphenol monoethoxylate and nonylphenol diethoxylate were positively correlated (τ_conventional = 0.35; p = 0.1), consistent with their presence in commercial APEO mixtures. In alternative samples, methyl paraben was positively correlated with all three cyclosiloxanes (τ = 0.69–0.87), and the cyclosiloxanes were positively correlated with each other (τ = 0.62–0.73).

Our results also indicate that use of multiple products can lead to exposure to an even larger mixture of compounds, even if a consumer selected products considered alternative according to our criteria. For example, a consumer who used the alternative surface cleaner, tub and tile cleaner, laundry detergent, bar soap, shampoo and conditioner, facial cleanser and lotion, and toothpaste (a plausible array of product types for an individual) would potentially be exposed to at least 19 compounds: two parabens, three phthalates, MEA, DEA, five alkylphenols, and seven fragrances.

The impact of exposures via one product or multiple products is of concern because of the potential combined effects of EDCs or compounds associated with asthma. Our analysis demonstrates that chemical combinations are common in consumer products, and results highlight combinations for toxicity testing, risk assessment, and epidemiological study. Toxicity testing can identify common modes of action for co-occurring chemicals, and risk assessment can then investigate cumulative exposures to multiple chemicals. Considering effects of co-occurring compounds in risk assessment would advance the recommendations of the National Research Council (2008). Similarly, in epidemiological studies, co-occurring exposures need to be understood together, because they may have additive or interacting effects or result in confounding. As an example of possible confounding, several studies have shown an association of endocrine-related health effects with DEP, which does not show activity in animal studies (Duty et al. 2003; Hauser et al. 2007; López-Carrillo et al. 2010; Swan et al. 2005); instead, DEP could be a marker for a large number of synthetic and natural fragrances, which do have activity. This suggests an important area for future research is to characterize the endocrine activity of fragrances and to measure these compounds in epidemiological studies. Epidemiological studies should include collaborations with toxicologists to help design and interpret findings.

Variability within product types. Although our study was not designed to focus on variability in the composition of different individual products within a type, we examined this question for sunscreens. Our study provides some information about how exposure may differ depending on brand selection and allows us to investigate the effect of compositing, which is a limitation.

Using sunscreens as an example, we observed substantial variability in composition of products within this product type [see Supplemental Material, Figure S1 (http://dx.doi.org/10.1289/ehp.1104052)]. Among the 5 alternative sunscreens, we detected 4–17 compounds per sample, with a total of 24 chemicals detected in the sunscreens. The product with the highest number of detects was marketed for children and favorably rated by a popular environmental health site. The variable composition of individual products within a class is important to consider in exposure modeling and in epidemiological studies that rely on self-reported product use as a proxy for exposure.

Limitations. To our knowledge, this study is one of the first to look for a large and varied suite of compounds in a broad range of product types; however, the product types and chemicals we included are still a small fraction of those in use, so this report is not comprehensive. In addition, the alternative and conventional products in this study may not be representative. In particular, alternative products, selected according to criteria in Table 1, were mostly purchased at one store with its own criteria, and we do not know how these criteria influenced our product selection. All products were purchased in 2007 and 2008; because formulations may have been changed, products purchased today could be different.

We chose to composite conventional products to increase representativeness and limit analytical costs; however, this strategy limits interpretation in several ways. First, compositing does not allow observation of extreme high and low concentrations because it is meant to optimize the estimate of the average concentration. Second, compositing may increase or decrease the number of compounds detected. Twelve chemical concentrations in individual sunscreen samples were diluted to lower concentration categories in the calculated composite, including to values < LOD [see Supplemental Material, Figure S1 (http://dx.doi.org/10.1289/ehp.1104052)]. Conversely, the number of detects could increase if manufacturers use different chemicals to achieve a particular function in the product (e.g., scent), thereby increasing the number of different chemicals in a composite. To evaluate the effects of compositing on number of detects, we calculated composites from varying numbers of individual sunscreens. The number of detected chemicals in possible sunscreen composites ranged from 5 to 21 [see Supplemental Material, Figure S3 (http://dx.doi.org/10.1289/ehp.1104052)] and was positively correlated with the number of products in the composite (see Supplemental Material, Figure S4). These results indicate the varying chemical formulations within a product type. Finally, because we composited conventional and not alternative products and the composites comprise varying numbers of products per sample, direct comparisons between conventional and alternative products and some comparisons between conventional product types could be misleading.

We tested an exceptionally wide range of products, including 50 types of personal care and cleaning products as well as selected household goods, for 66 compounds identified as EDCs or asthma related. We detected 55 compounds, suggesting a wide range of exposures from common products. Results suggest that vinyl products are an important source of DEHP in homes. In other products, the highest concentrations and numbers of detects were in fragranced products (e.g., perfume, air fresheners, and dryer sheets) and sunscreen. To our knowledge, this is the first test of sunscreens for a range of EDCs. In addition to the labeled ingredients, sunscreens contained up to seven target chemicals that were not included on the product label. The highest number of detects in sunscreen was in a product favorably rated by a popular environmental health website and marketed for babies, children, and sensitive adults; this illustrates the limitations of rating products based on ingredients disclosed on product labels. In addition to a broad assessment of chemicals in widely used personal care and cleaning products, one of our goals was to identify a strategy for reducing exposure by removing or substituting products. Our shopping criteria did identify a set of alternative products containing no BPA or antimicrobials and limited synthetic fragrance. We detected DCP, DINP, and DPP only in alternative products, suggesting the possibility that manufacturers have substituted these antiandrogenic phthalates for the better known and also antiandrogenic phthalates DEHP, DBP, and BBP, which are common in conventional products. Our observations of multiple chemicals of concern in composites of high-market-share products coupled with consumers’ use of multiple product types (e.g., laundry detergent plus dish soap plus shampoo) highlight the importance of considering the cumulative toxicological effects of combined exposures. Our correlation analysis identifies mixtures for evaluation and also raises caution that associations in epidemiological studies may be due to unmeasured chemicals that co-occur with the study target. Disclosure of product ingredients would enable researchers to identify exposures for study and risk evaluation and allow consumers to make decisions consistent with their personal values.