Naphthalene

IAP intake. To calculate pollutant inhalation in U.S. residences, we used a data compilation described by Logue et al. (2011) (link) that includes summary statistics from 77 studies reporting residential air pollutant concentration measurements in the United States and other countries with similar lifestyles. The aggregate data were used to calculate concentrations relevant to assessing chronic residential exposures to 267 chemical air pollutants. Seventy of the pollutants had sufficient toxicological and epidemiological data to calculate chronic health impact using the methodology described below and were included in this study (Table 1). Our analysis did not extend to contaminants from biological sources such as molds and allergens. We thus refer to the suite of pollutants considered as “nonbiological.”
Determining annual population health impact. The annual health impact of residential IAPs was calculated by considering the total intake in residences as an increment adding to intake in other environments. The increment was calculated by considering in-home inhalation of air containing the population-mean exposure concentrations relative to the theoretical case of the population inhaling residential air containing no pollutants.
The DALY metric allows quantification and comparison of the health costs from varied disease end points that can result from various pollutants. As a measure of equivalent years of life lost (YLL) due to illness or disease, DALY loss quantifies overall disease costs (impacts) due to both mortality and morbidity. DALY losses include YLL due to premature mortality and equivalent YLL due to reduced health or disability (YLD). For each disease, the DALYs lost per incidence are calculated as follows:
DALY_disease = YLL_disease + YLD_disease. [1]
The equivalent life-years lost to reduced health are weighted from 0 to 1 based on the severity of the disease. For example, a 5-year illness that reduces quality of life to 4/5 that of a healthy year is valued at 1 DALY lost.
Several authors have determined the DALYs lost per incidence of specific diseases using the preeminent work of Murray and Lopez (1996a , 1996b ) [Huijbregts et al. 2005 (link); Lvovsky et al. 2000 ; World Health Organization (WHO) 2009]. Multiplying disease incidence by a “DALY factor” yields total DALYs lost per disease incidence:
DALYs = (∂DALYs/∂disease incidence) × disease incidence. [2]
Equation 2 uses a partial derivative in recognition that DALY losses are incrementally affected by causes other than disease. The total burden of disease in a community can be calculated as the aggregate, across all diseases, of DALY factors multiplied by disease incidence rates.
Our analysis used two approaches to calculate DALY losses from estimated exposure concentrations. For criteria pollutants [ozone, nitrogen dioxide (NO₂), particulate matter ≤ 2.5 μm in aerodynamic diameter (PM_2.5), sulfur dioxide (SO₂), and carbon monoxide (CO)] we used an intake–incidence–DALY (IND) method that uses epidemiology-based C-R functions to quantify disease incidence rates; these are combined with estimates of DALY losses per disease incidence reported in the literature. For noncriteria pollutants we used an intake–DALY (ID) approach using the work of Huijbregts et al. (2005) (link) to calculate the health impact associated with intake of noncriteria pollutants based on animal toxicity literature. The IND approach is preferred because it does not require interspecies extrapolations, which generally involve larger uncertainties than the epidemiologically based C-R functions. However, the IND approach can be used only for pollutants with information on C-R functions in humans. Ozone was the only pollutant for which both the IND and ID approaches could be applied.
Although the disease incidence relationships in the IND and ID approaches are accepted health impact models, they are nevertheless simplifications of populationwide responses to chronic inhalation exposure. Our approaches use linear (i.e., IND) and nearly linear (i.e., ID) disease incidence models without effect thresholds. For these types of disease incidence models, only the mean of the concentration distribution is needed to estimate population impact. Existence of a threshold concentration for disease incidence, or a strongly nonlinear disease-to-intake response, would necessitate accurate determination of the shape of the population intake distribution. A discussion of the impact of threshold effects on our DALY loss estimates is included in the “Discussion.” The potential impacts of nonlinear response functions are beyond the scope of the present study.
The IND approach. The first step of the IND method comprises the application of C-R functions to determine disease incidence. For almost all of the disease outcomes, the C-R function follows the formula:
ΔIncidence = – {y₀ × [exp(–βΔC_exposure) – 1]} × population, [3]
where y₀ is the baseline prevalence of illness per year, β is the coefficient of the concentration change, C_exposure is the exposure-related concentration, and population is the number of persons exposed. For each pollutant and outcome, y₀ and β vary. Respiratory illness due to long-term NO₂ intake requires a slightly different C-R functional form but still relies on a β with specified uncertainty.
When aggregating C-R functions and DALY factors, we tried to include all of the diseases with available established relationships between concentrations and disease incidence. We did not include diseases/outcomes that were negligible compared with the other diseases included. The health end points selected and DALY loss per incidence of disease are summarized in Table 2.
Chronic PM_2.5 exposure affects both the respiratory and cardiovascular systems. The three outcomes that we included were all-cause mortality, chronic bronchitis, and stroke. Pope et al. (2002) (link) predicted incidence rates of all-cause mortality and the average YLL per unit increase in PM_2.5 (Pope et al. 2009 (link)); we divided the former by the latter to get DALYs lost per incidence. The 95th percentile range of the DALYs lost per death was set to represent the span of values seen in the literature (Lvovsky et al. 2000 ; Pope et al. 2009 (link)). Recent studies have shown that chronic PM_2.5 exposure can lead to heart disease and thickening of arterial walls (Künzli et al. 2004 (link)). The total impact of PM_2.5 on cardiovascular health is not known. However, recent work by Miller et al. (2007) (link) has shown associations between chronic PM_2.5 and stroke, an outcome of heart disease, in women. The end point of nonfatal stroke was included in the analysis using the hazard ratios derived by Miller et al. (2007) (link) for both men and women. This is likely an underestimation of the total impact of PM_2.5 on heart disease. The DALYs lost per nonfatal stroke incidence were taken from Brook et al. (2010) (link). The incidence of stroke predicted was split among 0, 1, and > 1 complications, and the percentage of stroke that resulted in death was determined based on the findings of Brook et al. (2010) (link). Burnett et al. (1999) (link) developed a C-R function for hospital admissions associated with long-term PM_2.5 exposure. However, because the impact was negligible compared with the impact of mortality, chronic bronchitis, and stroke, we did not include this outcome. There is evidence that PM_2.5 exposure is associated with other health outcomes, including diabetes and reduced lung function; however, these findings are relatively new and have not been included in this work.
For CO and SO₂, the only outcomes relevant to chronic exposure appear to be hospital admissions. Chronic ozone and NO₂ exposure have been associated with early death and respiratory illness, respectively. The input parameters into the C-R functions for these outcomes are the same as those used in the U.S. EPA cost–benefit analysis of the Clean Air Act (U.S. EPA 1999). For hospital admission and respiratory illness, we used the DALY loss/incidence values available in the literature. For ozone mortality, as with PM_2.5, it is unclear how much life is lost because of early death. Values in the literature range from a few weeks to 10 years; we chose a large range of values to represent this uncertainty (Levy et al. 2001 (link); Lvovsky et al. 2000 ).
The C-R functions are formulated to calculate the increment of disease incidence per increment of exposure concentration, not total disease incidence for a given exposure concentration. According to population-weighted demographics (Klepeis et al. 2001 (link); U.S. Census Bureau 2010), summarized in Table 3, the “average” American spends 70% of the time in residences. The chronic exposure-relevant concentration contributed from indoor exposure was therefore set to 70% of the indoor concentration:
ΔC_exposure = 0.7C_indoors. [4]
Incidence rates were combined with DALY factors to calculate total health impacts by pollutant (Equation 2). A Monte Carlo approach was used to calculate impacts by pollutant by sampling with replacement from the available distributions of DALY factors and β. We assumed that all DALY factor distributions are log-normal.
The ID approach. The ID approach extrapolated directly from indoor concentrations to total DALYs lost due to intake of specific pollutants. From this standpoint, it is convenient to rewrite Equation 2 as follows:
DALYs = (∂DALY/∂disease incidence) × (∂disease incidence/∂intake) × intake, [5]
where intake is the mass of pollutant that an individual inhales over a given time frame. Huijbregts et al. (2005) (link) computed expected ranges of human impact for cancer and noncancer chronic effects of 1,192 substances, applying equal weightings for a year lost, independent of age (i.e., zero discounting). Using the values determined by Huijbregts et al. (2005) (link), the DALYs lost for 1 year of breathing pollutant i is calculated using the following equations:
DALYs_i = (∂DALY/∂intake) × intake, [6]
DALYs_i = C_i × V × [(∂DALY_cancer/∂intake)_i × ADAF + (∂DALY_noncancer/∂intake)_i], [7]
where ∂DALY/∂intake_i are the cancer and noncancer mass intake-based DALY factors, C_i is the indoor concentration, V is volume of air breathed in the residence each year, and ADAF is the age-dependent adjustment factor for cancer exposure as described below.
The age at which carcinogens are inhaled has an appreciable effect on total toxicity, and the U.S. EPA has developed ADAFs to calculate cancer health impact as a function of exposure age (U.S. EPA 2005). To align with U.S. EPA-recommended ADAFs, we considered three age groups: < 2, 2–16, and > 16 years of age (U.S. EPA 2005). A population-weighted average annual air intake volume and ADAF were calculated by combining age distribution of the U.S. population, age-specific inhalation rates, and time spent at home (Table 3).
Huijbregts et al. (2005) (link) presented, for each chemical, both a central estimate (50th percentile value) and the estimated uncertainty of the DALY losses per mass intake of pollutant; uncertainty was assumed to be log-normal, characterized by a factor, k_i, calculated as follows:
k_i = (97.5th percentile/2.5th percentile)^0.5, [8]
which includes the aggregated uncertainty of the rate of disease incidence as well as the uncertainty in the DALY losses per incidence of disease. We used a Monte Carlo approach to sample with replacement from uncertainty distributions of DALY factors derived from the central estimate of the DALY factor and the k_i value, to determine the central estimate and 95% confidence interval (CI) for combined cancer and noncancer DALY losses for each pollutant. A Monte Carlo approach was also used to determine the total DALYs lost from all of the pollutants analyzed using the IND and ID methods.
Despite the availability of a DALY factor for bromomethane, DALY-based impacts are not presented for this compound because the limited available concentration data (New York State Department of Health 2006 ) appear more indicative of a local outdoor source than of general conditions in U.S. homes (Logue et al. 2011 (link)).
Radon, SHS, and acute CO poisoning deaths. The population-average DALYs lost to radon, SHS, and acute CO poisoning deaths were determined based on estimates of disease incidence from the literature. We included DALY loss estimates for these pollutants for two reasons: a) to compare health impacts calculated for a subset of SHS pollutants using the IND and ID methodologies to independent estimates of overall DALY losses associated with SHS exposure (as described below), and b) to compare estimated IAP-associated DALY losses calculated in the present study with estimates for these three established indoor health hazards.
To estimate the health impact from radon, SHS, and acute CO, we used Equation 2 with disease incidence estimates from the literature, summarized in Table 4. For radon and acute CO poisoning, only the end point of premature death was used to estimate DALY losses. The DALYs lost per incidence of various SHS outcomes and per early mortality due to acute CO poisoning and radon were taken from the literature and are also summarized in Table 4.
Comparison with DALY losses estimated by other methods. Results from this study were compared with three other estimates of populationwide DALY losses for the United States. Although our study used an impact assessment approach, the studies used for comparison are cumulative risk assessment (CRA) and burden of disease studies (Ezzati and Lopez 2004 ; McKenna et al. 2005 (link); WHO 2009). The burden of disease studies used available statistics to determine the disease incidence rate as a function of age, sex, and geographical location. A DALY value was then assigned based on YLL and disability incurred. The CRA studies determined the fraction of disease or death attributable to a specific risk factor based on epidemiological studies of specific populations. This is similar to, but far more complex than, our method of estimating health impacts due to SHS and radon. If the disease rate and DALY factors were accurate, and if we used the same discount ratings and time weightings for the age at which years of life are lost, both methods should estimate the same number of DALYs lost associated with a specific risk factor. Indoor air, independent of the impact of household use of solid fuels, had not been studied in a CRA analysis thus far. We compared results from our methodology with CRA results with the caveat that the methods are far from equivalent and the comparison should be seen only as a point of reference. The comparison also provides a useful tool for bounding uncertainties for our impact assessment method.
The WHO compiled disease incidence data for all communicable and noncommunicable diseases and injuries to determine the total number of DALYs lost per year for 192 countries (WHO 2009). McKenna et al. (2005) (link) aggregated U.S. mortality and morbidity data to determine the top 20 causes of DALY losses for men and women in 1996. Ezzati and Lopez (2004) estimated the total DALYs lost due to smoking and tobacco use in industrialized nations by determining the impact of disease beyond what would be expected in nonsmoking homes. The total DALY losses that we estimated for all IAPs analyzed with the IND and ID methods were compared with estimates from these studies to discern whether the full CI of the aggregate IAP impact of indoor residential air is plausible. Additionally, we used our IND and ID methodology to calculate health impact for a suite of measured SHS components, and we compared the aggregate CI of the DALYs lost for these components with CRA-derived DALY estimates.
SHS is a complex mixture of chemicals. Nazaroff and Singer (2004) (link) estimated increases in specific volatile organic compound concentrations (1,3-butadiene, 2-butanone, acetaldehyde, acetonitrile, acrolein, acrylonitrile, benzene, ethyl benzene, formaldehyde, naphthalene, phenol, styrene, toluene, and xylenes) expected for average smoking activity. Simons et al. (2007) (link) found that homes with smokers had PM_2.5 concentrations that averaged 16 µg/m³ higher than those in the homes of nonsmokers. We applied the IND and ID modeling frameworks established here to determine the additional DALYs lost due to living in a household that had indoor concentrations elevated by the specified levels. We used the Monte Carlo sampling to determine an aggregate CI for the DALYs lost due to exposure to this chemical mixture.