Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Passive sampling methods to determine household and personal care product use

Abstract

Traditionally, use of household and personal care products has been collected through questionnaires, which is very time consuming, a burden on participants, and prone to recall bias. As part of the SUPERB Project (Study of Use of Products and Exposure-Related Behaviors), a novel platform was developed using bar codes to quickly and reliably determine what household and personal care products people have in their homes and determine the amount used over a 1-week period. We evaluated the acceptability and feasibility of our methodology in a longitudinal field study that included 47 California households, 30 with young children and 17 with an older adult. Acceptability was defined by refusal rates; feasibility was evaluated in terms of readable bar codes, useful product information in our database for all readable barcodes, and ability to find containers at both the start and end of the week. We found 63% of personal care products and 87% of the household care products had readable barcodes with 47% and 41% having sufficient data for product identification, respectively and secondly, the amount used could be determined most of the time. We present distributions for amount used by product category and compare inter- and intra-person variability. In summary, our method appears to be appropriate, acceptable, and useful for gathering information related to potential exposures stemming from the use of personal and household care products. A very low drop-out rate suggests that this methodology can be useful in longitudinal studies of exposure to household and personal care products.

Introduction

Personal care and household care products, such as cleaning products and pesticides, are frequently used in most households although little is known about the quantity of these products used. Many personal care products contain toxic substances with potential adverse health effects (Daughton and Ternes, 1999; CDC, 2005), such as phthalates, fragrance, nanoparticles, petroleum byproducts (1,4-dioxane), and triclosan. Some chemicals in personal care products are known to cause endocrine disruption, birth defects, or cancer (Szczurko et al., 1994; Stickney et al., 2003; Golden et al., 2005; Schettler, 2006).

Health effects resulting from direct emissions of compounds found in cleaning products include eye irritation, asthma, allergy, and/or respiratory irritation (Wolkoff et al., 1998). Some of the compounds of concern are ethanolamine, laural dimethyl benzyl ammonium chloride, benzalkonium chloride, formaldehyde and other carbonyls, toluene, 4-nonylphenol, and glycol ethers such as 2-butoxy ethanol, 2-(2-ethoxyethoxy)ethanol, and 2-(2-butoxyethoxy)ethanol (reviewed in Nazaroff and Weschler, 2004). In addition, ambient ozone enters homes from the outdoors (Weschler et al., 1992; Reiss et al., 1995; Weschler, 2000) and reacts with unsaturated compounds in cleaning products forming secondary pollutants of concern (Weschler, 2006). There are a number of unsaturated terpenes found in cleaning products, such as α-pinine, used for its pine scent, and d-limonene, used for its lemon scent, as well as other terpene-related compounds, such as α-terpineol, linalool, and linalyl acetate. These compounds react with ozone to form OH radicals, which in turn form aldehydes and ketones (Weschler and Shields, 1997; Singer et al., 2006), hydrogen peroxide (Li et al., 2002), and secondary particulate matter (Weschler and Shields, 1999; Wainman et al., 2000; Singer et al., 2006), all of which are known to have adverse health effects. Many of these terpenes are also found in air freshener products.

To determine the extent of potential adverse health impacts in a population from compounds found in personal care and cleaning products using risk assessment methodology, data are needed about (1) the types of compounds found in various products, (2) the distribution of the frequency of use of products, and (3) the amount typically used (Van Engelen et al., 2007).

A number of models have been developed by the European Commission to assess exposure and health risks from chemicals in consumer products, for instance the EIS-ChemRisks Toolbox (http://web.jrc.ec.europa.eu/eis-chemrisks/toolbox.cfm). Exposure databases for European populations have also been established, for example, ExpoFacts Database (Vuori et al., 2006). The European Union has also initiated the REACH program (Registration, Evaluation, Authorization and Restriction of Chemicals) to ensure public safety of chemicals in consumer products. However, exposure assessment and policy enforcement for chemicals in personal care products are insufficient in the United States (GAO, 2005, http://www.gao.gov/products/GAO-05-458). The EPA Exposure Factors Handbook (EFH) covers only limited categories of consumer products (USEPA, 1997). The most recent national-scale study on usage patterns of personal care products was conducted >25 years ago.

This paper has several goals: first to evaluate the feasibility of Universal Product Codes (UPC or bar codes) readily found on personal care and household care products to quickly and reliably determine what products people used in their homes; second, to present data on amount of personal care and household care products used as collected by these methods; and third, to evaluate acceptability of the method. Feasibility was evaluated in terms of the availability of product data based on barcodes and the ability to determine the amount of a product used over a 1-week period. Based on the collected data, we determined the distribution of the amount used for participating households, compared with values in the literature, and determined the inter-individual and intra-individual variability of amount used across four 1-week periods. Acceptability was evaluated in terms of refusal rate and dropout rate. Additionally, we evaluated a motion sensor to collect data on cleaning product use.

These methods minimize participant burden and make data collection easier by providing unique information on products as it is difficult for participants to recall not only the exact products they use but also to estimate the amount of a product used when asked to report this information in a typical questionnaire format.

Methods

This effort was part of the SUPERB project (Study of Use of Products and Exposure-Related Behaviors), an evaluation of alternative platforms for collecting longitudinal data related to non-occupational environmental exposures through food, daily activities, and household products. This field study nested within the larger SUPERB study included a total of 47 California households. Briefly, study staff visited the home and identified personal care and household care products using a bar code scanner. The products were weighed at the beginning of the week and again at the end of the week, allowing us to determine the amount of the product used over a 1-week period. The goal was to employ only low participant burden methods, thus questionnaires eliciting information on product use during the week were not administered. The process is outlined in the flowchart in Figure 1 and further described below.

Figure 1
figure1

Flow chart of the procedure of barcode scan and the determination of product use. DB, database.

Study Population

The SUPERB study includes two cohorts of participants. The first consisted of households with young children living in a 22 county area in Northern California, including much of the greater Sacramento and San Francisco Bay Area regions. Candidate households were randomly selected from birth certificate records of children born between 2000 and 2005 in this area, resulting in 499 participants. The second cohort included households with older adults, that is, above the age of 55, residing in three central California counties recruited after random selection of addresses from tax assessors’ records. All recruited participants completed a phone interview in what we refer to as the main study (Hertz-Picciotto et al., 2010).

For the research presented in this paper, we recruited a subset of households from the main study. Specifically, we recruited 30 households with young children and 17 households with older adults. Data were collected on a household basis, but in each household one individual has been identified as the participant, the one who completed the phone interview.

Data Collection

The goal of data collection was to reduce participant burden, yet obtain useful data on exposure-related behaviors through household visits.

Home Visit Protocol

For each household enrolled in this sub-study, data were collected over four 1-week periods across 16 months. During the first visit, study staff members visited the home, introduced themselves, and explained the study protocol. The participants were asked to show study staff where in the house personal and household care products were kept.

We selected the following categories of personal care products: shampoo/conditioner, liquid soap, antibacterial soap, hand sanitizer, hand and body lotion, facial moisturizer, body wash, hair styling products, nail polish, foundation make-up, aftershave, fragrance, sun block, bubble bath, baby shampoo, baby bath, and baby lotion. We included several classes of cleaning products: all-purpose cleaning products, glass cleaners (w/o ammonia), ammonia products, bathroom cleaning products, oven cleaners, and metal polish. We also evaluated two other categories of household care products, pesticides, and spray air fresheners. Finally, we assessed glues, whether used on household care or hobbies.

Bar Code Data

All products were scanned using a custom UPC tracking system application. A self-contained web application written in Hypertext Preprocessor (PHP) used a small local web server on a laptop to run in a secure and reliable manner. The laptop was connected to a scale (OHAUS Model Scout Pro SP4001) and barcode scanner (symbol LS2208) through USB connections to transfer data. The application tool was used to keep track of the household ID number, the products scanned, when they were scanned, and the weight of each product.

We used a publicly available database to match barcodes with product information. We scanned each container with a bar code and using our tool identified all available product information from the database (see flowchart in Figure 1). If a barcode was present but our database contained no information on the product, information from the container was manually entered into the database and was then available for all future encounters with the same product. If the product had either no barcode or a barcode that could not be read by the scanner, we tracked these products by writing the name of the product down in the participant binder and scanning a unique barcode that we generated and placed next to the written product name. We also entered information about the product into the electronic database, allowing us to track amount used as done for all other scanable products.

In addition, for unambiguous classification of all products, we generated our own additional barcodes, one for each product category (as listed above). After scanning the individual product barcode (when possible), we also scanned the corresponding category barcode. For example, a body wash product might be identified in the database as a “body shampoo” and but we preferred to classify the product as a body wash rather than a hair shampoo at the time of recording, rather than attempting to reclassify it at a later date without the product to help with these decisions.

This system of data tracking also allowed us to record whether or not a specific container had been scanned at a prior visit and track it over time. Every time a new product container was identified at a field visit, a mark was placed on the underside of the container with a felt tip pen. A container was considered a new product if it did not have a mark on it. At the end of the week, products were checked for the pen mark and reweighed. New products not previously recorded were also scanned and marked. During subsequent visits, we repeated the processes above, additionally noting if there was a pen mark on the product to indicate we had seen the bottle before at an earlier visit.

Quality Assurance/Quality Control

We determined the minimum weight change detectable by the scale. Two products were measured repeatedly (200 measurements) in our laboratory on multiple days. The minimum detectable weight change was calculated as three times the SD of the difference between each individual weight and the average weight of the product, 0.066 g.

Product Motion

The participant was asked to identify the two cleaning products they used most frequently. An Actical Accelerometer, a device designed to record human movement, which contains a biaxial piezoelectric accelerometer sensor to record physical motion in two planes, was strapped on each of the two products to record how frequently and for how long the product was moved.

Participant Estimated Product Use

For personal care products, participants were also asked to provide a sample of the amount they typically used of shampoo and liquid soap by dispensing the quantity of shampoo or liquid hand soap that they typically used into the palm of their hand, which was weighed by staff. The methods and results for this effort are included in the Supplementary Information. Ultimately, we determined participants had difficulty estimating the quantity of product used. This finding further supports the need to weigh products to determine the amount used as this is not a quantity participants can estimate.

Data Analysis

Ability to Determine Amount Used

We evaluated whether or not the amount of a product used could be determined during the 1-week period. If the product was found at the beginning and the end of the week, and a dot indicated that the product had been seen and weighed at the last visit, the difference in the weight likely indicated the amount used during that week. However, if a product was found only at the beginning and not the end of the week the product might have been finished and discarded, thrown out without further use, or moved to a different location where study staff or the participant could not find it. In any case, the weight of the product could not be determined for products not found at the end of the week. For each week, each product encountered was placed into one of the following categories (see Figure 1):

  1. 1)

    Used — product found at both beginning and end of the week, and weight decreased.

  2. 2)

    Not used — product found at beginning and end of week, and weight did not change.

  3. 3)

    Increased weight — product found at beginning and end of week, and weight increased (for this change there is no clear interpretation; water could have been added to thin a product, for instance).

  4. 4)

    Removed — product found at beginning of week but not at end of the week.

  5. 5)

    New — product found only at end of week, indicating purchase of a new product or potentially a product that was not found/presented at the first visit.

  6. 6)

    Rediscovered — product found only at the end of the week, but not beginning of the week, and rather than being a new product, it was marked with a dot, indicating that the product had been seen during a previous visit in another season, but was missed on the initial visit of that season's week.

  7. 7)

    Replaced — product found at the beginning and end of the week, but there was no dot on it, indicating that the original container had been discarded and a new one purchased.

We integrated all individual products within a given category to determine a classification for the household:

  1. 1)

    All products found, used —all products found both times and some product was used during the week.

  2. 2)

    All products found, not used — all products found both times and no product was used during the week.

  3. 3)

    Not owned — no products were found either time.

For the households falling into in these first three groups, the amount used was quantifiable. If products were removed, purchased, replaced, or rediscovered in a product category, the total amount used could not be determined (see Supporting Information for additional details on categorization of these situations). The percentage of households falling into each product category was calculated for the study sample as a whole, as well as for the samples from the two cohorts separately.

Distribution of Amount Used

We calculated the amount of the products used in each category over a weeks’ time, including all products that had a reduced weight when encountered at the end of the week, even if some other products in the same category had been removed, newly purchased or replaced (evaluation of our inclusion criteria in the Supplementary Information).

We calculated the Spearman correlation coefficient between amount used and each of the following variables: the number of people living in the house, the number of products owned, and the frequency of use as reported in the phone survey.

Inter- and Intra-Household Variability

To assess between-individual variability and within-individual between-season variability, intra-class correlation coefficients (ICC) were calculated for categories of products more frequently used. ICC is defined as the ratio of between-subject variance to total variance, ranging from 0 to 1 (Xue et al., 2004). The higher the ICC is, the greater the proportion of total variability that is attributed to differences from one individual to the next. An ICC of 80% indicates a relatively strong correlation of measurements collected from an individual, which in this study translates to low seasonal variation.

A hierarchical mixed effects model (using PROC GLIMMIX in SAS) was fit to apportion the variance of household type (household with young children/household with older adults), season, order of visit (up to four visits), and number of people in household at each visit (including guests who stayed temporarily) with individual household as a random effect. As the distribution of amount used of all visits generally concentrated around zero, only weeks with a non-zero amount used were included in this calculation.

We were interested in determining whether a single week of sampling might be representative of the average usage of household or personal care products in households. We defined the average amount used across all available weeks to be the “true” usage in a household. We assigned each household to a tertile for both this “true” or average amount used, and also for the observed individual week amount used. The agreement between the “true” and “observed” tertiles provided an estimate for how well a single week sample predicts average usage level (Peck et al., 2003). Tertiles were based on all non-zero amount used data of the first three visits for each household (not all households had four visits) Note that this produced an unbalanced design (i.e., some tertiles contained zero households) for some products because of small sample size. Geometric means for each tertile were calculated to show the quantitative differences in usage amount defined by a single visit.

Product Motion

Actical accelerometer were attached to products to determine how frequently and for how long the product moved and we used this data to determine the number and duration of product use. Actical records contain the number of movements per 15 s, which were summed across each minute. Each minute in which there was movement was noted. Any series of minutes with movement with less than a 20-min period between uses was considered one product use. Actical counts during the use period were summed.

The duration of each product use was also determined. We determined the distribution of frequency of product use throughout the week, the duration of each use, and the time of use during the day from the Actical data. Finally, we calculated the correlation coefficient between frequency of use and amount used during the week, and calculated the amount used per application (divided total amount used for the week by the number of times the product was used).

Results and discussion

Respondent Burden

Enrollment rates in the field study varied by age group, race, education, and employment status. Of 87 households with young children that were contacted, 30 participated. We contacted 33 households with an older adult, of which 17 participated. There were low enrollment rates among Asian households, with 14 households contacted and invited and only 2 enrolled (and in one of these two households, only one of the parents was Asian). The enrollment rate was reflective of the overall enrollment rate across all other races. Enrollment was highest among participants with 1–4 years of college, with 47% of recruited participants participating as compared with 26% and 38% for those with <1 year of college or with graduate degrees, respectively. Enrollment rates were also higher among retired participants and stay-at-home parents, with both groups enrolling at just over 50%, as compared with just over 30% for all other employment categories. For more details on the demographics of the sample, see Hertz-Picciotto et al. (2010).

Once enrolled in the field study, the vast majority of study participants completed the study. Of the 30 households with a young child, 27 completed all four scheduled seasons, 1 completed two seasons, and 2 completed only one season. Of the 17 households with older participants, 13 completed all four season visits, 2 completed three seasons, and 2 completed only one season. The fact that 90% and 76% of the participants in the two sub-cohorts completed all four seasons indicates that this protocol met the goal of being not too burdensome to those who had agreed to participate. The only indication of the protocol not being acceptable was that in some cases, the participant preferred to bring products to study staff, rather than have study staff go into other rooms in the house to scan the products.

Time spent with the participant at the home for the first visit was on average 25 min. This included not only the informed consent and scanning and weighing of personal care products, but numerous other features of this study: scanning and weighing of other household products and provision of other equipment to monitor food preparation and time-activity patterns (results will be reported elsewhere). Subsequent follow-up visits averaged a total of 8 min of the participant's time. Staff time spent in the home was on average 104 min for the first visit, and 90 min thereafter.

Availability of Barcode Product Data

In Table 1, we present for each product category the percentage of individual products found (a) with information in the public database, (b) with a readable barcode but without information in the public database, and (c) without a readable barcode. Products with a readable bar code and no information in the publicly available database were generally store-branded products.

Table 1 Frequency of identifiable barcodes, readable barcodes (no information), and missing/illegible (no) barcodes; prevalence of products in each category and number of products found per product category.

Products without barcodes generally included: (a) products that are sold inside a box (i.e., high-end cosmetic products that are often sold with the container in a cardboard box); (b) products that were sold with multiple units packaged inside a box or other outer packaging (i.e., two bottles of a cleaning product inside a plastic wrapper sold at warehouse stores); (c) products that were sold through distribution channels outside of stores (e.g., Avon and Amway); (d) store brands for a limited number of stores; and (e) products with a bar code too small to be read by our scanner.

Of the personal care products, the majority (61–90%) of facial moisturizers, aftershaves, nail polishes, foundations, and fragrances had no barcodes; these products are frequently sold inside boxes. The majority of products from other categories (liquid soaps, antibacterial soaps, shampoos and conditioners, hand and body lotions, hand sanitizers, body washes, sun blocks, and hair styling products), had a readable barcode (55–88%). Almost all of the cleaning products, pesticides, and air fresheners included a readable bar code. Of the products with readable bar codes, the majority of oven cleaners and metal polishes had product information available, approximately half of all-purpose cleaners, ammonia products, bathroom cleaner products, disinfectant sprays, and pesticides had product information available, while glass cleaners, air fresheners, and hobby products were less likely to have product information.

This indicates that UPC scanners can be used efficiently to identify many household and personal care products in a home. The high prevalence in some categories of readable bar codes without product information indicates a preference to buying store brands that lack the product information in available commercial databases, indicating that relying on publicly available databases alone will not be as effective for these categories.

Number of Products Found

Summary statistics on products found in homes are shown in Table 1. Of personal care products, shampoo, liquid soap, hand and body lotions, and hair styling products were present in over 90% of the households. Men's aftershave and fragrance were present infrequently (6–17% of all households). A higher portion of households with young children owned antibacterial soap, body wash, facial moisturizer, hair styling products, sun block, and baby products. Many of these products are marketed to parents of children, specifically baby products, sun block, and antibacterial soap. The number of different products present in each of the categories shampoo, body wash, and sun block was significantly greater in households with young children than in households with older adults.

Of household care products, all-purpose, glass and bathroom cleaners were found in almost all households, while other product categories were found less frequently. The percentage of households with the product and the number of products found per household was similar for most product types between both cohorts. One exception was pesticide products, with a higher fraction of older adults having a pesticide product in their home (82% of older adults vs 73% of families with young children), but the families with young children who did have products owned a larger number of different types of products. Hobby glues were found almost exclusively among families with young children. Homes with an older adult typically had more types of bathroom cleaners.

Ability to Determine Amount Used

Figure 2 includes the percentage of 1-week periods in which a given container for a product was used (weight decreased), was not used, or we were unable to determine the change in mass (product increased in mass, was removed, new, rediscovered, or replaced, rates of each sub-category in Supplementary Table 2) for personal care and household care products, respectively. In some cases, a product may be misclassified as new or removed because the container may have been moved within the house and thus it was not found both times. In some cases, participants did not allow field staff into the bathroom, and participants may not have always brought out all containers.

Figure 2
figure2

Percentage of products that were used, not used, increased in weight, or classified as other (removed, new, rediscovered, or replaced) during a week of observation.

Households stored a significant number of both personal care and household care products in their homes that they did not use on a weekly basis (10–70% of all products found per product category). For some product categories, such as nail polish and fragrance, which are items used infrequently or, by some individuals on special occasions only, we expected products to be stored and rarely used. We found this to be correct, with only 4–22% of such products used in any given week. One would anticipate metal polish and oven cleaner to only be used occasionally and therefore expect these products not to be used in a given week. However, we also found that a number of products generally thought to be used frequently were not used, perhaps because the individual had changed brands or “stocked-up” when products were on sale.

For household care products, the same data are also presented for the time periods between seasons, in other words, we assessed which products were found between the end visit during one season and the first visit during the next season, approximately 4 months later in the Supplementary Information.

Figures 3a and b present the percentage of visits for each product category for which we could determine (a) the amount of product used, (b) that the product was not used, (c) that the product was not owned, or (d) determine that the product fell into one of many categories for which we could not determine the amount used (rates in Supplementary Tables 3 and 4). The data for the two cohorts are presented separately.

Figure 3
figure3

Fraction of households for which we could and could not determine the amount used over a 1-week period for (a) personal care products and (b) household care products. OA, older adults; PYC, parents of young children.

Many personal care products were not present for a majority of household-weeks, especially fragrances, nail polish, make-up foundations, and many baby products. For several product categories, amount used could not be measured in >40% of the weeks in households with young children: specifically shampoo, liquid soap, hand and body lotion, and hair styling products. These product categories also had the highest proportion of weeks with difficulties in interpreting the data for households with older adults, but the fraction falling into this category was generally lower. We speculate that the smaller proportion of items with a quantifiable amount used for households with young children is likely due to more people living in the home and using these products; thus, these households may be buying more products and also products may be more likely to be moved around and end up in various locations within the home where staff did not find them.

In contrast, for many household care product categories, we were able to measure use for 80–90% of the households. In some cases, very few households owned the products, including oven cleaners, disinfectant sprays, ammonia products, metal polish, and glues. For some product categories, we had some difficulty interpreting the data because the same products were not found or other problems occurred, specifically this occurred 55% of the time for all-purpose cleaner and 25% of the time for glass cleaners in households with young children, with a lower percentage of difficulties in homes with an older adult. Bathroom cleaners were the final product category for which we were able to interpret data for <80% of the households.

During the course of the study, the scanner failed while four household visits had already been underway. While waiting for the new scanner to arrive, we recorded the end of the week data by hand. Data from these four visits were incomplete as we were unable to match many of the products from the beginning to the end of the week manually as households owned products with similar names. This problem did not occur with the barcode technology. Although it was unfortunate that we lost data, it also pointed out the benefits of unique product identification made possible by the bar code technology we employed.

Distribution of the Amount Used

Next, we calculated summary statistics for the amount used for a given product category (Table 2) in both cohorts for households that used the product. Of personal care products, a greater amount of shampoo, soaps, and body lotions were used than other products. The categories of household care products with the greatest amount used were all-purpose cleaners in both populations, and ammonia products in the older adult population. However, there is considerable variability between households, with the SD generally exceeding, or being similar to, the mean value.

Table 2 Distribution of amount (g) used in 1 week per product category in the populations with young children and with older adults.

To determine whether number of products owned was a good surrogate for the amount used, we calculated Spearman correlation coefficients and found moderate but statistically significant correlations for a number of product categories, specifically for facial moisturizers (0.28, P=0.03), hair styling products (0.27, P=0.008), liquid soap (0.30, P=0.003), hand and body lotion (0.24, P=0.012), and shampoo (0.30, P=0.0004). There were no significant correlations for household care products. As these correlations were only moderate, and only significant for limited product categories, this is not an ideal surrogate.

We also calculated the correlation coefficient between reported frequency of use in the annual interview and amount used for a type of product. Moderate correlations were found for a limited number of product categories, specifically antibacterial soap (0.28, P=0.046), hair styling products (0.33, P=0.0014), hand and body lotion (0.36, P=0.0002), and shampoo (0.29, P=0.0014). The lack of stronger correlations is likely due to differing numbers of household members and variability of the amount used each time by household members. Among the household care products, we could only make comparisons for all-purpose cleaner and glass cleaner and there was no correlation for these products. Supporting the accuracy of interviews for determining if a product category is used, we generally found that people did actually use the types products that they reported using frequently in the phone interview, more details on this comparison can be found in the Supporting Information.

Inter- and Intra-household variability

To determine the effectiveness of collecting data during a single 1-week period, we assessed temporal variability using two methods, calculation of ICC values and analysis of tertiles, presented in Table 3 and Supplementary Table 5.

Table 3 Inter-class correlation coefficients and agreement of tertile analysis for each product category.

Moderate intra-individual variability was observed for the majority of personal care products, with ICCs of 0.22–0.52; very high intra-individual correlation was only found for hand sanitizer use (ICC=0.91). Significant temporal variability was observed for the majority of household care products, with ICCs of 0.20. A relatively high ICC was found for air freshener products (ICC=0.70). Other factors we examined to determine their predictiveness, for example, season, visit number and number of people living in a household, did not influence usage amount for the majority of products.

There was good agreement (70–90%) for personal care products and moderate agreement (63–82%) for household care products between the tertile based on the average use across all weeks (considered the “true” usage) and the tertile based on the usage from a single week (considered the “observed” usage), as seen in Table 3. Full results including agreement in each tertile, along with the average amount used in each tertile, are presented in Supplementary Table 5. Of the households in the “true” high usage tertile, the percentage of visits that had all “observed” visits in the highest tertile ranged from 62% to 100% across the product categories. Similarly, for the households in the “true” lowest tertile, the percentage of visits with all “observed” visits in the lowest tertile ranged from 72% to 100% across the product categories.

Motion Sensors

A total of 120 household weeks of monitoring data were extracted from 45 households. The use frequencies of the monitored products were summed for each sampling week (Figure 4a). Twenty-one percent of the resulting product use records had a sum of Actical counts, with each count representing one movement, below 10, with a definite peak occurring between 4 and 9 counts, and we suspect that these were not a result of actual product use as it does not seem feasible to move the bottle so few times while using the product (the average number of movements per use was nearly 1000). Rather this may have resulted from a different mechanism, perhaps the product was accidentally bumped in the cabinet when other products were retrieved from the cabinet; thus these instances were excluded from the analysis.

Figure 4
figure4

(a) Use frequency of two most commonly used cleaning products during the sampling week. (b) Distribution of the number of minutes for each use.

Data showed that major household cleaning products were used seven or more times in 47% of the sampling weeks; in 12% of the weeks, cleaning products were used 14 or more times. The frequency of use during different sampling weeks for each household was moderately consistent, with an ICC of 0.43.

The majority of cleaning products (92%) were used between 0700 and 2100 hours, with peaks at 0800, 1200, 1500, and 1900 hours. The distribution of the length of cleaning periods is presented in Figure 4b. More than half of the periods over which the product moved lasted 1 or 2 min (58%), and 80% lasted 10 min or less. People may spray the cleaning products in places where they are about to clean and then put the product away while continuing to clean. Unfortunately, Actical only records the movement of product bottles. We were not able to measure the time that people were in contact with the cleaning products. Thus, the duration of exposure to cleaning products may be an underestimate.

The correlation between frequency of use and amount used during the sampling week was statistically significantly (P<0.0001). Given that the change in weight of the wipe products include both the weight of the cleaning product and the wipe, while the change in weight of the spray products is just the weight of the cleaning product, we calculated correlations separately, with a correlation of 0.57 for wipe products and 0.76 for non-wipe products. However, no significant difference was observed in the mass used each time between wipe and non-wipe products (median: 7.6 g vs 4.3 g, P=0.25).

Comparison with Existing Literature

The EPA EFH reports the single use mass used of many personal care products, based on the averages of the values reported by 20 companies interviewed. We do not have the actual number of times a product was used in the household over a specific week. Such information was only reported by our participants in a telephone interview at another time when they were asked to report general frequencies of use for one or two members of the household. In Table 2, we list the amount used per use and the mean frequency of use per week as specified in the EFH. Using these two numbers, we then calculate the amount of product that would be used per person (in parentheses in the last column of the table), in order to provide a very rough estimate of potential per week use. For the most part, comparing the measured amount used (which includes only households that used the product) with the estimated amount used using the EFH, many of the values estimated by the EFH seem reasonable, or perhaps underestimate the amount likely used in a week. For example, average amount used for baby lotion was 6.4 g compared with an estimated change of 10 g. In some instances, the frequencies of use provided by the handbook are potentially low resulting in low estimates of amount used. For example, the reported per use mean value of bubble bath is 11.8 g according to the EPA handbook, and the mean use frequency is once per week, which would result in an estimated 11.8 g used during the week. In comparison, the median amount used per week was 30.1 g in our population; it is possible that our population uses bubble bath at a higher frequency than the national average.

There were, however, three categories of products where the estimated amount used in the EFH may very well be too low. First, the handbook reports use of foundation as 0.3 g, with an expected use frequency of 3.3 times per week, resulting in an estimated 0.9 g used during the week, whereas in the two SUPERB populations, the median amount used were 2.7 and 2.9 g, and the means were even higher. Even if we assume that the use frequency reported in the handbook was too low, it would have to be greater than once per day to reach the actual amount used we observed. Additionally, since foundation is most likely used by only one member of the household, use by other family members would not explain this discrepancy.

For baby shampoo, per-use amount reported in the handbook was 0.5 g, with a use frequency of once per week. We measured a mean change of 11 g, which would translate into 22 uses during the course of a week, which is unlikely and high for typical use.

Finally, there were few users of nail polish, but with reported use of 0.3 g in the handbook and a use frequency of less than once per week, this amount seems low compared with our observed amount used, two of our four values being 1.3 and 3.9 g, or it would reflect multiple nail polish uses per week. Although our data may not be definitive with regard to per use values, these discrepancies may justify further investigation into the assumptions of per use amount for these three product categories.

There is very little published research on the amount of cleaning products used. One study enrolled 30 households, 10 families with children, 10 couples, 9 people living alone, and a commune, in which 9 of the 30 households weighed their all-purpose cleaner before and after every use (Weegels and van Veen, 2001). The mean use was 27 g, with a SD 30 g. Our mean amount used over the week was 126 g for the families with young children and 59 g for the households with an older adult. When considering typical use frequencies for this product, these numbers are comparable.

Conclusions and lessons learned

The bar code scanner proved to be a useful method to obtain detailed information relatively quickly and with high specificity. A majority of products could be identified by the scanner and thus this data collection approach allowed for comprehensive information to be collected quickly and with little burden to the participant. The barcode scanner was critical to successfully match products at the beginning and end of the week, as evidenced by the difficulties encountered during the 1-week period when the scanner did not work.

Based on our experiences, we suggest several modifications to the protocol. Given the high proportion of products removed from a household within a week's time, we suggest asking the participants to only show the staff products that they personally (or the index person in a household) use in order to prevent recording products that may be thrown away without being used before the next visit. This would also further reduce the time spent by staff in the home. Also, a 1-week sampling time is likely too short to reliably assess household care product use. Either a 2- or 4-week sampling period is likely the best interval to maximize products being used during the sampling time while minimizing the number of products being removed or newly purchased. Recording the product type was critical as product names are not always intuitive, preventing the need to obtain full product details from an on-line search.

Alternatively, in studies that allow for only one home visit, one could still identify and record all personal care products in a household, but also record whether the participants still use the products regularly or whether they used them in the past but no longer use them or are storing them for future use.

We found a significant portion of consumers using store brand cleaning products, a trend which may continue. If this methodology were to be used in a larger scale study, it would make sense to obtain a proprietary database of product information linked to store brand bar code data from popular chains; however, this was beyond the scope of our project. Of note is that for pesticides the EPA registration numbers are not uniquely linked to the bar code, as the formulation may change, yet the product maintains the same name and thus barcode. For this reason, either a database linking the product barcode to the pesticide ID code would need to be maintained or the pesticide ID code should additionally be recorded.

The motion sensor proved to be useful in collecting unbiased data on product use frequency. An additional approach in need of pilot testing would be to skip a home visit and send the motion sensors to the home by mail. The participant could attach the sensor to the product and return the sensor at the end of a fixed time period.

Overall, the use of bar code data is a promising method for evaluating use of household and personal care products with minimal burden to the participants.

References

  1. Centers for Disease Control and Prevention.. Third National Report on Human Exposure to Environmental Chemicals. Centers for Disease Control and Prevention, Atlanta, GA, USA, 2005.

  2. Daughton C.G., and Ternes T.A. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ Health Perspect 1999: 107: 907–938.

    CAS  Article  Google Scholar 

  3. Golden R., Gandy J., and Vollmer G. A review of the endocrine activity of parabens and implications for potential risks to human health. Crit Rev Toxicol 2005: 35: 435–458.

    CAS  Article  Google Scholar 

  4. Hertz-Picciotto I., Cassady D.L., Lee K., Bennett D.H., Ritz B., and Vogt R. Study of use of products and exposure-related behaviors (SUPERB): study design, methods, and preliminary results. Environ Health 2010: 9: 54.

    Article  Google Scholar 

  5. Li T.H., Turpin B.J., Shields H.C., and Weschler C.J. Indoor hydrogen peroxide derived from ozone/d-limonene reactions. Environ Sci Technol 2002: 36: 3295–3302.

    CAS  Article  Google Scholar 

  6. Nazaroff W.W., and Weschler C.J. Cleaning products and air fresheners: exposure to primary and secondary air pollutants. Atmos Environ 2004: 38: 2841–2865.

    CAS  Article  Google Scholar 

  7. Peck J.D., Hulka B.S., Savitz D.A., Baird D., Poole C., and Richardson B.E. Accuracy of fetal growth indicators as surrogate measures of steroid hormone levels during pregnancy. Am J Epidemiol 2003: 157: 258–266.

    Article  Google Scholar 

  8. Reiss R., Ryan P.B., Tibbetts S.J., and Koutrakis P. Measurement of organic-acids, aldehydes, and ketones in residential environments and their relation to ozone. J Air Waste Manag Assoc 1995: 45: 811–822.

    CAS  Article  Google Scholar 

  9. Schettler T. Human exposure to phthalates via consumer products. Int J Androl 2006: 29: 134–139.

    CAS  Article  Google Scholar 

  10. Singer B.C., Coleman B.K., Destaillats H., Hodgson A.T., Lunden M.M., Weschler C.J., and Nazaroff W.W. Indoor secondary pollutants from cleaning product and air freshener use in the presence of ozone. Atmos Environ 2006: 40: 6696–6710.

    CAS  Article  Google Scholar 

  11. Stickney J.A., Sager S.L., Clarkson J.R., Smith L.A., Locey B.J., Bock M.J., Hartung R., and Olp S.F. An updated evaluation of the carcinogenic potential of 1,4-dioxane. Regul Toxicol Pharmacol 2003: 38: 183–195.

    CAS  Article  Google Scholar 

  12. Szczurko C., Dompmartin A., Michel M., Moreau A., and Leroy D. Photocontact allergy to oxybenzone — 10 years of experience. Photodermatol Photoimmunol Photomedicine 1994: 10: 144–147.

    CAS  Google Scholar 

  13. US Environmental Protection Agency. Exposure Factors Handbook. National Center for Environmental Assessment, Office of Research and Development, 1997.

  14. US Government Accountability Office.. Chemical Regulation: Options Exist to Improve EPA's Ability to Assess Health Risks and Manage Its Chemcial Review program (Report to Congressional Requesters). Wahsington, DC, USA, 2005.

  15. Van Engelen J.G.M., Heinemeyer G., and Rodriguez C. Consumer exposure scenarios: development, challenges and possible solutions. J Expo Sci Environ Epidemiol 2007: 17: S26–S33.

    CAS  Article  Google Scholar 

  16. Vuori V., Zaleski R.T., and Jantunen M.J. ExpoFacts - an overview of European exposure factors data. Risk Anal 2006: 26: 831–843.

    CAS  Article  Google Scholar 

  17. Wainman T., Zhang J.F., Weschler C.J., and Lioy P.J. Ozone and limonene in indoor air: a source of submicron particle exposure. Environ Health Perspect 2000: 108: 1139–1145.

    CAS  Article  Google Scholar 

  18. Weegels M.F., and van Veen M.P. Variation of consumer contact with household products: a preliminary investigation. Risk Anal 2001: 21: 499–511.

    CAS  Article  Google Scholar 

  19. Weschler C.J. Ozone in indoor environments: concentration and chemistry. Indoor Air 2000: 10: 269–288.

    CAS  Article  Google Scholar 

  20. Weschler C.J. Ozone's impact on public health: contributions from indoor exposures to ozone and products of ozone-initiated chemistry. Environ Health Perspect 2006: 114: 1489–1496.

    CAS  Article  Google Scholar 

  21. Weschler C.J., Brauer M., and Koutrakis P. Indoor ozone and nitrogen-dioxide — a potential pathway to the generation of nitrate radicals, dinitrogen pentaoxide, and nitric-acid indoors. Environ Sci Technol 1992: 26: 179–184.

    CAS  Article  Google Scholar 

  22. Weschler C.J., and Shields H.C. Measurements of the hydroxyl radical in a manipulated but realistic indoor environment. Environ Sci Technol 1997: 31: 3719–3722.

    CAS  Article  Google Scholar 

  23. Weschler C.J., and Shields H.C. Indoor ozone/terpene reactions as a source of indoor particles. Atmos Environ 1999: 33: 2301–2312.

    CAS  Article  Google Scholar 

  24. Wolkoff P., Schneider T., Kildeso J., Degerth R., Jaroszewski M., and Schunk H. Risk in cleaning: chemical and physical exposure. Sci Total Environ 1998: 215: 135–156.

    CAS  Article  Google Scholar 

  25. Xue J.P., McCurdy T., Spengler J., and Ozkaynak H. Understanding variability in time spent in selected locations for 7–12-year old children. J Expo Anal Env Epid 2004: 14: 222–233.

    Article  Google Scholar 

Download references

Acknowledgements

The project was supported by the United States Environmental Protection Agency (Grant # RD-83154001-0). We thank all of our study participants and our field staff (Jessica Riley, Valerie Moore, Laura Gonzalez, Brianna Diaz, and Karen Wagner).

Author information

Affiliations

Authors

Corresponding author

Correspondence to Deborah H Bennett.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies the paper on the Journal of Exposure Science and Environmental Epidemiology website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Bennett, D., Wu, X., Teague, C. et al. Passive sampling methods to determine household and personal care product use. J Expo Sci Environ Epidemiol 22, 148–160 (2012). https://doi.org/10.1038/jes.2011.40

Download citation

Keywords

  • personal care products
  • cleaning products
  • passive sampling
  • SUPERB
  • longitudinal

Further reading

Search

Quick links