Naphthalene concentrations were monitored in four communities in southeast Michigan, USA (Figure 1). Ann Arbor (AA) is a largely suburban and affluent community. Ypsilanti (YP) is more commercial and industrial. Dearborn (DB) and Detroit (DT) are densely populated and industrialized cities, and household income tends to be considerably lower. Several techniques were used to recruit a total of 288 households in these four communities. Random sampling via telephone dialing and snowball recruitment techniques was used to recruit 65 households in AA, 35 in YP, and 61 in DB, as part of VOC exposure study. In DT, 127 households were recruited via questionnaires distributed to caregivers at community-based organizations, schools, community fairs and other venues, and households selected for this study were required to have at least one child 6–12 years old with symptoms or medication use consistent with persistent asthma. The DT component was conducted as part of a community-academic partnership asthma study (Parker et al. 2008). In each house, questionnaires and building walkthrough audits were used to obtain information on housing, smoking, work and family characteristics, hobbies and other factors potentially related to exposure. House characteristics in AA, YP and DB have been described by Jia et al. (2008a; 2008b), and DT homes have been described by Du et al. (Du et al. 2011). Study protocols followed informed consent and other procedures approved by the University of Michigan Institutional Review Board.

Each residence was visited in several seasons. AA and YP residences were visited in summer 2004 and winter 2005. DB residences were visited in fall 2004 and spring 2005. In the second season in these cities, 31 additional residences were recruited to replace 26 dropouts, and additional follow-up studies took place in an additional 8 AA residences in summer 2005. In DT, households entered the study on a rolling basis between March 2009 and February 2010. Follow-up studies in this city were completed by October 2010, and 11 homes had one visit, 17 had two visits, 87 had three visits, and 12 had four (or more) visits. In AA, YP and DB, indoor and outdoor samples were collected simultaneously. Indoor samplers were deployed in duplicate in the living room, and outdoor samplers were deployed in duplicate at the location close to the house, e.g., the backyard. Outdoor samples were not collected in a small portion (10%) of homes due to inclement weather or lack of appropriate location, and each location was sampled in duplicate or triplicate. In DT, indoor samples were collected (single or duplicate samples) in the living room, and duplicate samples in the child's bedroom. Outdoor concentrations in DT were monitored at a central site using active sampling (described later).

A technician completed a standardized walkthrough inspection to collect information on each home's characteristics and condition, e.g., type of heating and cooling system, presence of an attached garage, and emission sources such as candles, incense and room deodorizers. For brevity, we did not inquire about specific uses and application rates of naphthalene products in the inspections, but instead depended on the naphthalene measurements to indicate its use. Tracers of environmental tobacco smoke (ETS), 2,5-dimethyl furan and 3-ethenyl pyridine (3-EP), were also measured ((Charles et al. 2008). While low levels of ETS may not always be identified, ETS is nearly certainly present if these tracers are detected.

Naphthalene, the ETS tracers and other VOCs were measured using passive thermal desorption tube samplers over a 3 to 7 day period (Batterman et al. 2002). In each home, samplers were deployed ≥0.6 m above the floor and below the ceiling, away from windows, doors and obvious sources of potential contaminants, ≥0.5 m away from bookshelves and other potentially stagnant areas, and out of the reach of children. Except for the variance analyses (described later), concentrations in the two spaces were averaged and treated as a single observation, and concentrations in different seasons in each house were averaged. Thus, each observation represents the average of replicates, two rooms for the DT homes, and at least two seasons. Overall, the monitoring campaign collected 1,439 valid indoor samples (245 single samples, 487 duplicates, 48 triplicates, 19 quadruplicates) and 478 residential outdoor samples (1 single sample, 204 duplicates, 23 triplicates), excluding samples considered invalid due to sampling or analysis issues.

To examine trends in ambient air, 24-hr VOC samples were collected daily at the Detroit Department of Health and Wellness Promotion Herman Kiefer complex located in central Detroit, at a site 60 m ESE of the Lodge Freeway (M-10). At this near-road location, the freeway is slightly below grade, the annual average daily traffic (AADT) was 143,300 vehicles/day, and the daily commercial annual average daily traffic (CADT) was 2,600 vehicles/day (Michigan Department of Transportation 2009). The area surrounding the Kiefer complex is primarily residential. Using an automated sequential sampler, an active sample flow regulated by a mass flow controller, and the same thermal desorption tubes described, calendar day 24-hr samples were collected from August 15, 2009 to August 16, 2010 at 5 ml min⁻¹, and from August 17, 2009 to April 19, 2011 at 2.5 ml min⁻¹. Possible impacts of rush hour traffic were investigated by collecting 3 hour samples daily from 6:00 to 9:00 am at the same site using a second sequential sampler at a flow rate of 10 ml min⁻¹ from January 26 to March 29, 2011. These flow rates represent compromises between obtaining sufficient sampling volume and minimizing problems associated with excess water. The active sampling at the Kiefer site represented another 548 daily/valid samples. About 10% of the Kiefer samples failed due to rain events and instrument failures.

All samples were analyzed using an automated thermal desorption–gas chromatography–mass spectrometry (ATD-GC-MS) system (Jia et al. 2006). Given the high naphthalene concentrations sometimes encountered indoors, our standard 7-point calibration, which used concentrations between 0.2 to 200 µg m⁻³, was extended to 500 µg m⁻³ where it continued to show excellent linearity. Quality assurance measures to ensure reproducibility and data quality included the use of standard operating protocols, routine collection and analysis of blanks (weekly at each site), regular flow checks, quarterly calibrations, and duplicate and sometime triplicate samples. The method detection limit (MDL), established using low concentration spiked samples, was <0.08 µg m⁻³, and replicate precision was generally ≤20%. Non-detects were set to 1/2 MDL.

Random effects models were applied to differentiate the variance of the naphthalene concentrations into five component parts (Jia et al. 2011): (1)Var(Yijkl)=σ2C+σ2R+σ2H+σ2S+σ2E where σ²_C and σ²_R = spatial variability calculated as the variances between cities and between residences, respectively; σ²_H = variability between bedroom and living measurements in the same home; σ²_S = seasonal variability calculated as the variance between seasons; and σ²_E = measurement uncertainty calculated as the variance between replicates. This analysis was performed for both indoor (AA, YP, DB and DT) and outdoor (AA, YP and DB) concentrations. Naphthalene concentrations were right-skewed, thus statistical analyses used log-transformed data since the random effects models assume normality. Variance components were computed using MIXED procedure in SAS (v9.1.3, SAS Institute, Cary, NC, USA).

For the DT outdoor data, effects of season, day-of-week and weekday vs. weekend were evaluated using Kruskal-Wallis tests, and concentrations during the 3 hour morning rush hour period and the corresponding 24-hour sample were compared using Wilcoxon signed ranks test. Trends were fit using an exponential smoother (α = 0.05 per day) and linear regression. Differences among indoor measurements also used Kruskal-Wallis tests. These analyses used SPSS 17 (IBM Corp., Somers, NY, USA). Data and results were organized in Microsoft Excel 2003.

Excess lifetime cancer risk was estimated as the product of the naphthalene concentration (µg m⁻³) and the draft US EPA (US EPA 2004) cancer unit risk estimate (URE). This simple "screening" level estimate makes several important assumptions: it does not account for exposures in environments other than the home; measurements are assumed to be representative of long-term or lifetime exposure; and the outcome represents an upper bound (95th percentile) risk estimate.

The mean and median naphthalene concentrations outside homes in AA, YP and DB were 0.28 and 0.16 µg m⁻³, respectively (n=145; Table 1). The variance proportions analysis showed that effects of season were the strongest (49% of the total variance), followed by city (28%), measurement uncertainty (17%), and house-to-house variation (7%; Table 2). The relatively high city-to-city variation likely reflects differences in terms of urbanization and industrialization, while the rather negligible between-house variation indicates that concentrations are homogeneous within a neighborhood, a result driven largely by the AA results. Variance analyses computed for each city showed several differences (Table 2), e.g., in AA, effects of season were the strongest (78%) followed by measurement uncertainty, while in YP and DB, house-to-house variations were the strongest (50–58%), followed by season (28–31%) and measurement uncertainty (12–22%). The house-to-house variation in YP and DB reflected multiple and unevenly distributed emission sources. Because the outdoor samplers were deployed close to houses, the measurements may reflect both the local neighborhood as well as very local activities, e.g., use of barbeques, lawnmowers, gardening, etc. While the measurement reproducibility deteriorated at low concentrations, it remained within the 25% criterion for the TO-15 method (US EPA 1999).

At the near-road monitoring site, the median naphthalene concentration was 0.15 µg m⁻³ and nearly identical to the neighborhood samples just discussed, and the 24-hour samples reached a maximum of 1.2 µg m⁻³ (Table 3). Seasonal variation was significant (p<0.05), and concentrations were highest in winter 2010 (0.29 ± 0.15 µg m⁻³) and lowest in spring 2011 (0.09 ± 0.04 µg m⁻³). Day-of-week and weekday-weekend differences were not significant (p=0.63 and p=0.33, respectively). Figure 2 shows the trend of daily concentrations monitored for nearly two years. Seasonal patterns were not consistent, e.g., the smoothed trend showed the highest levels in winter, summer and fall of 2010, and a relatively small number of peaks was influential. A linear trend fit to the data shows a gradual decline in concentrations over the study period with a decrease of about 30% per year, however, the linear model may not have much explanatory or predictive value. Declining trends of toluene, ethylbenzene, p-m-xylene and other VOCs were seen at this site, while benzene and carbon tetrachloride showed slightly ascending trends. We could not identify the specific cause of these trends, which might include decreases in traffic and reductions in other emission sources. Analytical issues were not implicated based on our calibrations and quality assurance measures.

Naphthalene concentrations were moderately to highly correlated with concentrations of most aromatic and alkane VOCs measured at the same site, e.g., benzene (r=0.46), toluene (r=0.77), ethylbenzene (r=0.76), p-m-xylene (r=0.60), o-xylene (r=0.60), n-C_7–15 alkanes (r=0.43– 0.64; all significant at p=0.01), suggesting that vehicle emissions was the main source for these VOCs. Conversely, correlation coefficients were not statistically significant between naphthalene and trichloroethylene (r=−0.03), 1,4-dichlorobenzene (r=0.07), cyclohexane (r=0.04), and carbon tetrachloride (r=0.00), indicating that these VOCs had different emission sources. Although use is banned under the Montreal Protocol, carbon tetrachloride is long lived in the atmosphere and globally distributed, and local sources are not suspected.

During the 3-hr morning rush hour period, naphthalene concentrations were elevated by an average of 33% compared to the daily mean (Table 2, Figure 3). These increments occurred on most days and were statistically significant (p<0.001). Day-of-week and weekday-weekend differences were not significant for the morning rush hour period (p=0.26, p=0.35, respectively). This rush hour differential suggests vehicle emissions, although poor ventilation and a low ceiling for atmospheric mixing can also raise levels of pollutants emitted by other local sources. Such meteorological conditions are common during the early morning period in Detroit. Concentrations of other VOCs were also elevated during this period, e.g., benzene concentrations rose by 17% compared to the daily mean, and did not show day-of-week and weekday-weekend differences, while toluene concentrations rose by about 23%, and showed higher concentrations on weekdays than weekends (Kruskal-Wallis test, p<0.05, both rush hour and daily averages).

The central tendency of outdoor concentrations in Detroit are typical of levels found in 24 previous urban and suburban studies, which show medians from 0.02 to 0.31 µg m⁻³ and averages from 0.01 to 0.82 µg m⁻³ (Jia and Batterman 2010). The highest outdoor levels, 6.0 µg m⁻³ in DB, 2.5 µg m⁻³ in AA, and 1.2 µg m⁻³ at the DT roadway site, also are similar to the previously reported upper bound concentration of 5 µg m⁻³ (Jia and Batterman 2010). Like other pollutants, naphthalene concentrations reflect the strength of local emission sources, the sampling location, averaging time, season, meteorology, and occasional outliers. For example, naphthalene concentrations were very low, only 0.06 µg m⁻³, in a southern Californian community with light traffic, and 0.58 µg m⁻³ in areas near high traffic roads (average of 200,000 vehicles per day(Eiguren-Fernandez et al. 2004). Several tunnel studies also have shown elevated concentrations, e.g., 0.07–0.57 µg m⁻³ in the Lundby tunnel in Gothenburg, Sweden (Wingfors et al. 2001), and 0.4–0.8 µg m⁻³ in the Shing Mun tunnel in Hong Kong (Ho et al. 2009). As another indicator of vehicle emissions, we calculated ratios of naphthalene to BTEX (total) concentrations of 0.03 (daily average) and 0.04 (rush hour average). Using data in Fujita et al. (2003) intended for receptor modeling, naphthalene:BTEX ratios were 0.005 along a truck-free highway in California, and 0.07 for diesel exhaust as measured at truck stops. The Detroit results fall between these ratios, quite reasonably suggesting that naphthalene sources included a mix of diesel and gasoline vehicles. Additional sources may include industry and, in rural areas, biomass burning (Miller et al. 2009). In the present study, the higher naphthalene levels seen in DB likely reflect both traffic and industrial sources, as compared to levels in YP and especially AA, which are largely suburban in nature (Jia et al. 2008).

Like many other pollutants, naphthalene concentrations undergo diurnal variation with peaks at night, due to low mixing heights that build up levels from local sources, and morning peaks associated with vehicle emissions at rush-hour and diminished dispersion, as discussed earlier (Park et al. 2002; Lu et al. 2005). Higher naphthalene concentrations have been reported in winter (Zielinska et al. 1998; Park et al. 2002; Lu et al. 2005; Reisen and Arey 2005), which has been attributed to decreased photochemical reaction rates (Mohamed et al. 2002), increased emissions from heating sources (Ravindra et al. 2008), and decreased dispersion due to more stable air and lower mixing heights (Cheng et al. 1997). While several studies have used a large number of measurements, long-term trends have not been investigated in any of the studies located. While the near-road measurements in DT suggest a downward trend, the variation is large and this inference is based upon measurements at a single site.

Outdoor concentrations of naphthalene were far below indoor levels in most homes (discussed in the next section) and thus will have only minor contributions to the total exposure for most persons. However, for those homes that do not contain naphthalene-emitting products, the outdoor level forms a “floor” for indoor concentrations, thus indoor and outdoor levels will be very similar in these homes.

Indoor concentrations in the four cities are summarized in Table 1. Residence-average naphthalene concentrations ranged from near detection limits (<0.1 µg m⁻³) to 201 µg m⁻³. Naphthalene levels (medians) in DT and DB tended to exceed levels in AA and YP, and median indoor levels varied among the four cities (Kruskal-Wallis p=0.02). However, the variance components analysis, used to differentiate among the variance contributions from home, city and season, shows that the variability of indoor naphthalene concentrations was predominantly due to home-to-home variation (62%) and seasonal variation (30%), while between-city effects (0%) and measurement error (7.9) were small or negligible (Table 2). This pattern was consistent among the cities. The large between-residence variation reflected differences in naphthalene uses, source strengths, and ventilation conditions among homes. The low measurement uncertainty indicates the high precision of measurements (Jia et al. 2006). The seasonal variation reflects temporal variation in source strengths and ventilation conditions.

In Detroit, where samplers were deployed in both living rooms and bedrooms, the within-home variation in naphthalene concentration was modest, e.g., concentrations in these rooms had an average absolute relative difference of 42% (n=279). As observed elsewhere (Batterman et al. 2007; Dodson et al. 2008), air in U.S. homes is generally well-mixed (although not completely mixed) and thus concentration gradients within a household are often modest. This was expected in the Detroit homes given the predominance (87%) of forced air heating/cooling systems. Nearly the same number of high naphthalene levels (e.g., >30 µg m⁻³) were found in bedrooms and living rooms, suggesting that naphthalene's use in the child's bedroom and in other spaces occurs with about the same frequency. Because only two locations in each home were monitored, however, we could not definitively identify the locations where naphthalene was used.

The mean and median indoor naphthalene concentrations in southeast Michigan, 5.4 and 0.89 µg m⁻³, respectively, fell within ranges reported in the recent review paper (0.8 to 9.5 µg m⁻³ for means, 0.17 to 4.1 µg m⁻³ for medians, based on 21 residential studies; (Jia and Batterman 2010). Two recent papers also reported fairly similar concentrations: the mean level measured in 28 homes in Detroit in 2006 was 0.50 ± 0.53 µg m⁻³ (Johnson et al. 2010); and the mean estimated among 9 North American and European studies was 1.2 µg m⁻³ (Logue et al. 2011). Previously, we calculated median indoor/outdoor (I/O) ratios of 2.4 to 12 for naphthalene, which depended on the city (AA, YP or DB were considered) and season (Jia et al. 2008).

Distributions of indoor naphthalene concentrations were highly skewed as shown by Figure 4, the large difference between means and medians, and the skewness coefficient (4.2 to 7.5, depending on the city). The single highest (7-day) measurement was 556 µg m⁻³ at a Detroit house. When averaged across rooms and seasons, this house also had the highest concentration (201 µg m⁻³), which exceeds the highest previously reported for residences, 144 µg m⁻³ (Jia and Batterman 2010). Table 4 shows that 14% of homes exceeded the chronic RfC concentration (3 µg m⁻³), 8% exceeded the WHO guideline of 10 µg m⁻³, and 5% exceeded 30 µg m⁻³, ten times the chronic RfC. Again, some differences among the cities are shown, e.g., the Detroit homes had the highest concentrations, and differences were largest at the highest concentrations. Houses that showed high naphthalene concentrations in one season tended to show high levels in other seasons, e.g., half of the homes with levels above 30 µg m⁻³ in one season also showed levels above this value in a second season.

Residences contain several sources of naphthalene. As noted, use of naphthalene as a repellent and/or deodorizer is a key source. The most common use these repellents is against moths, but naphthalene-containing products also are marketed to repel mosquitoes and other insects, as well as rabbits, dogs, pets and strays (NIH 2007). The very highest indoor levels, e.g., greater than perhaps 10 to 100 µg m⁻³, may suggest off-label uses of naphthalene. Griego et al. (2008) notes that Recochem, Canada is the sole U.S. registrant for pesticide applications of naphthalene, and that off-label uses of mothballs as area fumigants (e.g., placing many mothballs on open trays in attics or other portions of homes) can elevate indoor levels by 10 to 300 µg m⁻³. Such uses have been poorly documented. Other naphthalene sources, discussed below, include the presence of an attached garage, cigarette smoking/environmental tobacco smoke, wood combustion, incense burning, and outdoor air. The study cities differed with respect to the fraction of homes containing smokers and attached garages (Table 5). The significance of the indoor sources is shown by the large variation between-homes, the significant between-city effects in outdoor air but not in indoor air and, most directly, the elevated indoor levels.

Using the draft estimate of the cancer URE, and assuming that an individual's long term exposure is equal to the median concentration (0.89 µg m⁻³) in the four cities, the lifetime excess cancer risk is 9 × 10⁻⁵. The 90^th percentile concentration (6.57 µg m⁻³) gives a risk estimate of 7 × 10⁻⁴. Homes with measurements in the 100 µg m⁻³ range represent cancer risks in the 10⁻² range, which places naphthalene among the top environmental risks. These values far exceed risks from other VOCs, e.g., benzene (Jia et al. 2008), and they appear to eclipse risks attributable to other indoor air pollutants that have been identified to pose chronic health risks with the exception of particulate matter and perhaps radon. The risk estimates for naphthalene are biased upwards in that they do not account for the amount of time that individuals spend in homes, the temporal variation of concentrations, other exposure sources, compartments and pathways, and they are based on a draft URE. As noted earlier (Table 4), 8 and 14% of homes exceeded US EPA and WHO guideline concentrations. respectively.

While long-term concentrations of naphthalene were below the current chronic non-cancer RfC of 3 µg m⁻³ in most study homes (89% of AA and DB homes, 83% in YP and DT homes), the median concentration of 0.89 µg m⁻³ confers an individual excess lifetime cancer risk near 10⁻⁴. Contemporary notions of acceptable risk fall in the 10⁻⁶ to 10⁻⁴ range. Importantly, the very skewed distribution of naphthalene concentrations produces upper bound individual cancer risks for a subset of residences that exceeds 10⁻³ and even 10⁻². Moreover, potentially millions of individuals are exposed to high levels. Such risks are exceptionally high, and they reflect homes in which this product is inappropriately used. The cancer risk estimates are based on a draft and controversial assessment of naphthalene's carcinogenic potential, which depends heavily on a study in male rats. However, naphthalene exposure remains of concern using the old URE, and the frequency and extent of exceedences over the non-cancer chronic RfC and WHO guideline also remain problematic.

The current strategy for managing exposures and risks associated with naphthalene and other chemicals in consumer and industrial products is through right-to-know requirements. Naphthalene was listed as a carcinogen in 2002 under California's Proposition 65, the Safe Drinking Water and Toxic Enforcement Act of 1986, and as a hazardous chemical under the European Union's Regulation on Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) with a registration deadline of 2010. Encouraged by the California rule, most of the California and much of the US market for pest repellents has shifted away from naphthalene to paradichlorobenzene. This substitute product has its own risks, and it also is listed under both California and European regulations. Despite these changes, naphthalene continues to find extensive use as a pest repellent, which undoubtedly accounts for the high concentrations found in Michigan homes and elsewhere. Unlike most of the literature, the present study emphasized high-end concentrations, which not infrequently reached high levels that are commensurate with risks that exceed health-based guidelines and other benchmarks. These results demonstrate that further actions to manage naphthalene exposures and risks are warranted. Appropriate actions could include sales bans or restrictions, improved labeling, consumer education, and promotion of non-toxic alternatives. WHO (2010) supports the former policy, stating that the "most efficient way to prevent high exposures" is to ban the use of naphthalene-containing mothballs. The present study did not undertake a full risk-benefit analysis, which might consider the magnitude of other indoor risks, address cumulative exposures, examine in detail policy and management options, and justify the rationale needed to implement these actions in residential environments. Still, the concentrations and predicted risks for a subset of homes are strikingly high, and thus naphthalene exposure represents a widespread public health concern. In our community-based study, we intend to provide individualized feedback to the participants in the form of a "fact sheet" and other communications regarding these findings.

There are several limitations to this study. With respect to the experimental measurements in Michigan, while each home was measured at least twice, we did not characterize temporal (e.g., seasonal) variability. In AA, YP and DB homes, indoor samples included only the living room. We did not account for other indoor locations (vehicles, bathrooms, workplaces, etc.) where people might be exposed to naphthalene. We were unable to document the specific uses and application rates of naphthalene in each home. It would be helpful to query occupants regarding their usage practices given its significance and the lack of information on this topic. Households in southeast Michigan may not be representative of Michigan or US residences. Unfortunately, naphthalene was not measured in NHANES, RIOPA or other large exposure studies, nor has it been measured in studies using biomarkers or personal samples that can better account for multiple exposure compartments. A draft estimate for naphthalene's carcinogenic potential (URE) was used, although the earlier URE would not dramatically change conclusions. While several factors associated with high naphthalene concentrations are identified, and preliminary apportionments of indoor sources are made, the assessment is semiquantitative due to large differences in the building characteristics and unmeasured covariates, e.g., naphthalene application rates. Lastly, we did not measure particulate phase naphthalene, although nearly all naphthalene is expected to be in the vapor phase.