Trends and concentrations of selected polycyclic aromatic hydrocarbons in general US population: Data from NHANES 2003–2008

Polycyclic aromatic hydrocarbons (PAH) are potentially mutagenic and carcinogenic, and as such their exposure is of serious concern. I aimed to study the trends in the levels of selected PAHs in US for the period 2003–2008 and their distribution by gender, race/ethnicity, socioeconomic and smoking status, and by exposure to second-hand smoke (SHS) at home and work. Using data from the National Health and Nutrition Examination Survey, regression models were fitted for 10 individual urinary PAH metabolites. Smoking was statistically significantly associated with higher levels of naphthalene, fluorene, phenanthrene, and pyrene metabolites when compared to non-smokers. SHS exposure at home was also statistically significantly associated with higher concentrations of naphthalene, fluorene, phenanthrene, and pyrene metabolites. There was a statistically significant increase in the concentrations of total naphthalene and fluorene, 2-hydroxynaphthalene, 2and 9-hydroxyfluorene, 2and 4-hydroxyphenanthrene, and 1-hydroxypyrene during the study period of 2003–2008. Females were found to have statistically significantly higher concentrations of total naphthalene and phenanthrene metabolites as well as 1and 2-hydroxynaphthalene, 1-hydroxyphenanthrene, and 1-hydroxypyrene. For most of the PAH metabolites, non-Hispanic whites had the highest adjusted concentrations and Mexican Americans had the lowest adjusted concentrations. For the concentrations of 2-hydroxynaphthalene, however, the reverse was true. *Corresponding author: Ram B. Jain, Womack Army Medical Center, 2817 Reilly Road, Fort Bragg, NC 28310, USA; Empiristat, Inc., Mount Airy, MD, USA E-mail: Ram.b.jain.ctr@mail.mil


PUBLIC INTEREST STATEMENT
Human organ systems that are adversely affected by the exposure to polycyclic aromatic hydrocarbons (PAHs) include dermal, hepatic, and immunological. Humans can be exposed to PAHs during smoking, grilling, broiling, or other high temperature food processing. Tobacco smoke has also been shown to have PAHs as one of its constituents. This communication has shown how smoking, both mainstream and second-hand, can affect exposure to PAHs. In addition, the factors that affect PAH concentration levels have also been delineated. Finally, the change in the levels of PAHs during the period 2003-2008 has been investigated.

Introduction
Polycyclic aromatic hydrocarbons (PAHs) are a group of more than 100 chemicals usually produced during incomplete combustion of organic materials. Some of the many sources of PAHs are motor vehicle exhaust, residential and industrial heating sources, coal, crude oil and natural gas processing, waste incineration, and tobacco smoke. Organ systems affected by exposure to PAHs include dermal, hepatic, and immunologic (ATSDR, 2014). Smoking, grilling, broiling, or other high temperature processing can result in the formation of PAHs. The emitted PAHs can form or bind to particles in the air, and particle size depends in part on the source of the PAHs (CDC, 2013). Relatively higher concentrations of PAHs are associated with smaller particulates (Boström et al., 2002;Rehwagen, Müller, Massolo, Herbarth, & Ronco, 2005). Uncooked foods and vegetables generally contain low concentrations of PAHs but can be contaminated by airborne particle deposition (CDC, 2013). Some leafy vegetables, may, however, have relatively higher concentrations of PAHs .
Children aged 6-15 years old exposed to higher concentrations of fluorene metabolites had a twofold odds of having been enrolled in special education (Abid, Herbstman, & Ettinger, 2014). Total urinary PAH metabolites and naphthalene metabolites are associated with higher body mass index, waist circumference, and obesity in children 6-11 years of age (Scinicariello & Buser, 2014). Everett et al. (2010) analyzed association between serum C-reactive protein and nine urinary PAH metabolites. They suggested a role for monohydroxy PAHs in progression of atherosclerosis. However, after controlling for tobacco use, Clark et al. (2012) did not find an association between PAH exposure and serum biomarkers of cardiovascular disease. Huang, Caudill, Grainger, Needham, and Patterson (2006) found children aged 6-11 years to have higher concentrations of 1-hydroxypyrene than adolescents aged 12-19 years and adults. Smokers and persons exposed to second-hand smoke (SHS) were also found to have higher concentrations of 1-hydroxypyrene than non-smokers.
A limited number of studies have evaluated the association between SHS exposure and PAH metabolite concentrations. Scherer, Frank, Riedel, Meger-Kossien, and Renner (2000), in a field study of 69 subjects, evaluated the influence of smoking, diet, SHS, and location of residence on urinary excretion of 1-hydroxypyrene and benzo[a]pyrene. They found that diet and smoking were the major sources of PAH exposure for persons not occupationally exposed to PAH but the influence of SHS exposure was negligible. Suwan-ampai, Navas-Acien, Strickland, and Agnew (2009) evaluated the impact of active and involuntary smoking on the concentrations of 23 PAHs using data from National Health and Nutrition Examination Survey (NHANES) for the years 1999-2002. Involuntary smoking was found to be associated with 1-hydroxypyrene, 2-hydroxyfluorene, 3-hydroxyfluorene, 9-hydroxyfluorene, 1-hydroxyphenanthrene, 2-hydroxyphenanthrene, and 3-hydroxyphenanthrene.
Relatively higher concentrations of PAHs have been reported in high-density traffic areas (Fischer et al., 2000;Tuntaviroon, Mahidol, Navasumrit, Autrup, & Ruchirawat, 2007). Tuntaviroon et al. (2007) reported total PAH concentrations to be 30-fold higher in high-density traffic areas of Bangkok as compared to roadsides in proximity to the low-density traffic areas near provincial schools. Outdoor concentration of total PAH was reported to be substantially higher (Fischer et al., 2000) in high intensity streets of Amsterdam as compared to low traffic intensity streets.
While there have been studies that have assessed the association of PAH concentrations with a limited number of factors, for example, body mass index (Scinicariello & Buser, 2014), a study that has evaluated association between PAH concentrations with factors like gender, race/ethnicity, and SHS exposure while controlling for the effects of other factors, using recent data are lacking. Consequently, this study was undertaken to assess the effect of various factors that may affect concentrations of PAH in US adults aged 20 years and over. Also, an important objective was to assess the trends in the concentrations of PAH over time. A study of trends over time is necessary to estimate the variability in adverse health risks associated with exposure to PAHs. This can help public health professionals estimate the resources that may be needed to provide general public information necessary to protect themselves from unnecessary exposure to PAHs. Data from National Health and Nutrition Examination Survey (NHANES) for the years 2003-2008 were selected for this purpose.

Materials and methods
Data from NHANES (www.cdc.gov/nchs/nhanes.htm) for the years 2003-2008 for demographics, PAH metabolites, body measures, serum cotinine, urine creatinine, and family smoking and occupational questionnaire data files were downloaded and match merged. The function of match merge is to assemble all data on each specific NHANES participant in one data file for the purpose of analysis. For example, gender of the participant XYZ which is provided in the data file on demographics need to be matched with his/her observed concentrations for PAH metabolites which are provided in a different data file. NHANES provides a unique participant ID, labeled as SEQN for each NHANES participant. SEQN was used to match and merge data from different data files. The sampling plan for NHANES is a complex, stratified, multistage, probability cluster designed to be representative of the civilian, non-institutionalized US population. NHANES provides sampling weights to account for the complex survey design, including oversampling, survey non-response, and post-stratification. All analyses incorporated information on sampling design variables.
In a household smoking questionnaire (http://www.cdc.gov/nchs/nhanes/nhanes2007-2008/ SMQFAM_E.htm) administered at the time of the home interview, questions were asked if somebody smoked at home, and if so, how many persons smoked cigarettes inside home and how many cigarettes were smoked inside home every day (CPD_Home). Using the data from this questionnaire, a categorical variable, presence and absence of SHS exposure at home (HmE) was created. Based on the results of preliminary analysis, only CPD_Home was selected for analysis for this research. For those who had no HmE, CPD_Home was set at zero. Using an occupational questionnaire (http://www.cdc. gov/nchs/nhanes/nhanes2007-2008/OCQ_E.htm), the respondent was asked if the NHANES participant was working at a job or business, and if so, how many hours per day, tobacco smoke was inhaled from other people's cigarettes, cigars, or pipes. Using the responses to the questions, if participants inhaled tobacco smoke at work, an indicator variable representing SHS exposure at work (WkE) was created. Number of hours smoke was inhaled per day from other people's cigarettes, cigars, or pipes (Hours_ Smk) was the continuous variable available for use. For this study, smokers were defined as those who had serum cotinine concentrations ≥ 10 ng/mL and non-smokers were defined as those who had serum cotinine concentrations < 10 ng/mL. A consideration was given to use WkE as a categorical variable but it would have increased the number of data cells from 32 to 64 possibly resulting in certain number of empty data cells and thus, negatively affecting the stability and reliability of statistical estimates and decreasing the number of degrees of freedom available for error terms. In the interest of having stable and reliable estimates, a decision was made to use WkE as an indicator variable.
A total of 5,028 NHANES participants aged 20 years and older were available for analysis. There were 72 participants for whom sampling weight was recorded as zero in the NHANES database. These 72 participants were removed from the database. There were 209 females who were pregnant at the time of NHANES participation. Since pregnancy may affect PAH concentrations, these 209 participants were removed from the database for this study leaving a sample size of 4,747 available for analysis ( Figure 1). Data were available for 2,374 males, 2,373 females; 2,370 non-Hispanic whites (NHW), 1,014 non-Hispanic blacks (NHB), 890 Mexican Americans (MA), 473 participants for whom race/ethnicity was not classified (OTH); 3,262 non-smokers, 1,246 smokers; 932 participants who were exposed to SHS at home and 3,775 participants who were not exposed to SHS at home (Figure 1). Data were missing for 239 participants for smoking status and for 40 participants for SHS exposure status at home (Figure 1). However, actual sample size used in the analyses was much smaller because of missing values for dependent and independent variables. General characteristics of the study population are given in Table 1.
The independent variables considered for regression modeling were age as a continuous variable, gender (males, females), race/ethnicity (NHW, NHB, MA, OTH), smoking status (nonsmoker, smoker), body mass index (BMI), HmE, WkE, CPD_Home, and Hours_Smk. It is important to consider the association of BMI with the concentration concentrations of PAHs because PAHs are transported to all tissues of the human body containing fat and are strongly lipophilic (Scinicariello & Buser, 2014). In addition, since a positive association between obesity (BMI ≥ 30 kg/m 2 ) and PAHs has been previously reported among children (Scinicariello & Buser, 2014), it is possible that an association may be found among adults too. Other reasons to include BMI as a covariate may be increased caloric intake and increased volume of distribution possibly resulting in lower concentrations of target chemicals. Family poverty income ratio (PIR) was used as a surrogate variable for socioeconomic status. PIR No should be expected to have association with the concentrations of PAHs because of the possible association of PIR with higher smoking levels and/or higher levels of coal burning for heating. Data for PAH metabolites were log10 transformed before being used as dependent variables in the regression models. A total of 13 regression models were fitted.
A consideration was given to use consumption of broiled/smoked meat/fish as an independent variable as has been used by Suwan-ampai et al. (2009). I identified those participants from the 24-h recall dietary databases for NHANES years 2003-2008 (http://www.cdc.gov/nchs/nhanes/nhanes2005-2006/ DR1IFF_D.htm) who specifically were reported to have eaten broiled and/or smoked food items. However, in certain cases, it was not possible to determine if the food that was eaten was broiled or baked. For example, for the United States Department of Agricultural (USDA) food code database, the description provided in the NHANES 24-h recall dietary database for USDA food code 21101120 was "BEEF STEAK, BROILED OR BAKED, LEAN & FAT". All such food items were not considered to be broiled or smoked. I identified a total of 37 participants who specifically reported having consumed broiled/smoked food items. Since non-availability of complete data did not allow creation of a variable representing consumption of broiled/smoked food, these 37 participants were removed from the analysis.
All data were analyzed using SAS 9.2 (www.sas.com) and SUDAAN 11.0 (www.rti.org/sudaan). SUDAAN Proc DESCRIPT was used to compute unadjusted geometric means and SUDAAN Proc REGRESS was used to fit regression models and to compute adjusted geometric means. It should be noted that unadjusted geometric means are simply the means of the log-transformed PAH variables, for example, for males irrespective of how these unadjusted geometric means may be affected by other factors like race/ethnicity. On the other hand, adjusted geometric means are means of logtransformed variables after mathematical adjustments have been made for the contribution of other variables in the regression models. Two-way interactions between gender, race/ethnicity, HmE, and smoking status were also considered but were retained in the final model only if one or more of them were statistically significant at α = 0.05. An interaction term between HmE and smoking status was considered because of the possibility that HmE may affect PAH metabolite concentrations differently among smokers and non-smokers, and retained in the final model if the interaction was found to be statistically significant at α = 0.05.

Results
In the interest of journal space, results for individual naphthalene, fluorene, and phenanthrene metabolites are presented as supplemental material. Specifically, regression coefficients with associated p-values are presented in Table S1 and adjusted geometric means (AGM) in Table S2. Results for ∑NAP, ∑FLU, ∑PHE, and 1-hydroxypyrene are presented here.
Actual sample sizes used in regression models for ∑NAP, ∑FLU, ∑PHE, and 1-hydroxypyrene were 3,804, 3,821, 2,133, and 3,858, respectively. R 2 for the models for ∑NAP, ∑FLU, ∑PHE, and 1-hydroxypyrene was 39.5, 63.3, 45.5, and 50.2%, respectively. No statistically significant interactions were found for the models for ∑NAP and ∑PHE. For the model for ∑FLU, statistically significant interactions were found between gender and smoking status (p < 0.01), and between smoking status and HmE (p < 0.01). For the model for 1-hydroxypyrene, statistically significant interactions were found between gender and race/ethnicity (p < 0.01), and between smoking status and HmE (p = 0.03).
Unadjusted geometric means (UGM) with 95% confidence intervals based on unweighted data for ∑NAP, ∑FLU, ∑PHE, and 1-hydroxypyrene by survey period, gender, and race/ethnicity are given in Table 2. While no formal statistical tests were conducted, it is quite obvious that the levels of ∑NAP, .9 ng/L) and 1-hydroxypyrene (111.2 vs. 79.9 ng/L) than females and females had higher UGM than males for ∑NAP (8,623.5 vs. 7,468.9 ng/L). Among the three major racial/ethnic groups, the order in which UGMs for ∑NAP, ∑FLU, and 1-hydroxypyrene were observed was: NHB > NHW > MA.
Concentrations of ∑NAP, ∑FLU, ∑PHE, and 1-hydroxypyrene were positively associated with urine creatinine concentrations (Table 3). Statistically significant positive associations were also observed between urine creatinine and every metabolite of naphthalene, fluorene, and phenanthrene (Table S1).
Smokers had statistically significantly higher AGM for ∑NAP, ∑FLU, ∑PHE as well as 1-hydroxypyrene than non-smokers (p < 0.01, Table 4). In fact, for both ∑NAP and ∑FLU, AGMs for smokers were more than three times of what they were for non-smokers (17,556.7 vs. 5,205.2 ng/L for ∑NAP, and 1,964.9 vs. 574.9 ng/L for ∑PHE). Non-smokers had statistically significantly lower AGMs for every metabolite of naphthalene, fluorene, and phenanthrene than smokers (p < 0.01, Table S2). It should be noted that for 3-hydroxyfluorene, AGM for smokers was more than six times of what it was for non-smokers (450.9 vs. 73.1 ng/L, Table S2).
Exposure to SHS at home was associated with higher concentrations of ∑NAP, ∑FLU, ∑PHE as well as 1-hydroxypyrene as compared to no exposure to SHS at home. However, statistical significance (Table 4, p ≤ 0.01) was reached for ∑FLU, ∑PHE, and 1-hydroxypyrene only. For almost every metabolite of naphthalene, fluorene, and phenanthrene, exposure to SHS at home was associated with statistically significantly higher AGM concentrations than when there was no exposure to SHS at home (p < 0.01, Table S2). The only exception was 2-hydroxynaphthalene for which statistical significance was not reached.
When interaction between smoking status and HmE was considered for ∑NAP, while non-smokers did have statistically significantly lower AGM than smokers (4,531.2 vs. 26,788.5 ng/L with SHS exposure at home, 5,382.2 vs. 15,856 ng/L with no SHS exposure at home) irrespective of SHS exposure at home, it was only for smokers that SHS exposure at home was associated with higher AGM than when there was no SHS exposure at home (26,788.5 vs. 15,856 ng/L, Figure 2, Panel A, p < 0.01). However, for ∑FLU (Figure 2, Panel B), SHS exposure at home was associated with higher AGM than when there was no SHS exposure at home for both smokers (2,952.8 vs. 1,779.6 ng/L) and nonsmokers (675.5 vs. 552.7 ng/L). There were no statistically significant gender differences in AGM for ∑FLU for non-smokers (593.2 for males vs. 557.9 ng/L for females, Figure 3, Panel A) but for smokers, males had statistically significantly lower AGM than females (1,791.7 vs. 2,145.8 ng/L, p < 0.01, Figure 3, Panel A). The racial/ethnic differences in the concentrations of 1-hydroxypyrene were limited to males only (96.9 ng/L for NHW, 65.4 ng/L for NHB, 92.5 ng/L for MA, and 93.4 ng/L for OTH,    Figure 3, Panel B). In addition, while non-smokers had lower AGM for 1-hydroxypyrene than smokers irrespective of SHS exposure at home (82.5 vs. 217.3 ng/L with SHS exposure at home, 72.3 vs. 149.3 ng/L with no SHS exposure at home, Figure 4), it was for only smokers that SHS exposure at home was associated with statistically significantly higher AGM than when there was no exposure to SHS at home (217.3 vs. 149.3 ng/L, Figure 4). Ding, Trommel, Yan, Ashley, and Watson (2005) presented data on 14 PAHs present in the mainstream smoke of various US cigarette brands. Naphthalene was found to be the most abundant PAH irrespective of cigarette brand followed by fluorene and phenanthrene. As provided in Supporting Information to Ding et al. (2005), the mean concentrations of naphthalene, fluorene, phenanthrene, and pyrene in full flavor Marlboro were found to be 386 ng/cigarette, 182 ng/cigarette, 145 ng/cigarette, and 59.1 ng/cigarette respectively. AGMs for these four PAHs observed in this study followed the same pattern not only for smokers but for non-smokers also ( Table 4). The same was true for those who were and were not exposed to SHS at home (Table 4). St.Helen et al. (2012) found hydroxyfluorene metabolites to be most discriminative of smokers from non-smokers followed by metabolite of 2-naphthalene and 1-pyrene. Similar results were observed in this study also. The ratios of AGMs for smokers divided by non-smokers were found to be 1.45, 3.42, 1.64, and 2.17 for naphthalene, fluorene, phenanthrene, and 1-pyrene metabolites respectively. The ratios of these AGMs for 1-naphthalene and 2-naphthalene metabolites were 4.2 and 3.43 respectively.

Discussion
In this study, for both aggregated (Table 4) as well as individual metabolite levels (Table S2) for naphthalene, fluorene, phenanthrene, and pyrene, smokers were found to have statistically significantly higher concentrations than non-smokers (p < 0.01) indicating the excess risk of exposure to PAHs carried by smokers. Suwan-ampai et al. (2009) also found active smokers (defined as those who self-reported themselves to be current smokers or had serum cotinine levels ≥ 10 ng/mL) to have statistically significantly higher concentrations of all PAH metabolites than those who were not exposed to tobacco smoke (defined as those who did not self-report themselves to be current smokers, were not living with a smoker, and had serum cotinine levels below LOD). Also, for both aggregated (Table 4) as well as individual metabolite levels (Table S2) for fluorene, phenanthrene, and pyrene, SHS exposure at home was associated with statistically significantly higher concentrations than when there was no exposure to SHS at home. The same was true for 2-hydroxynaphthalene. Thus, even the exposure to SHS at home is associated with excess risk of exposure to PAHs. Suwan-ampai et al. (2009) also found involuntary smokers (defined as those who did not self-report themselves to be current smokers but were living with a smoker in the home and had detectable but < 10 ng/mL level of serum cotinine) to have statistically significantly higher levels of 1-hydroxypyrene, 1-, 2-, and 3-hydroxyphenanthrene, and 2-and 3-hydroxyfluorene than those who were not exposed to tobacco smoke. I, in addition to evaluating the effect of SHS exposure at home, also evaluated the impact of the extent of SHS exposure at home. The variable that indicated the extent of SHS exposure at home was the number of cigarette smokers inside home every day. Suwan-ampai et al. (2009) did not consider the role of the extent of SHS exposure at home in their study. In addition, instead of generating adjusted geometric means as I generated, they generated ratios of geometric means which may not have a practical value if someone is looking for the numeric concentrations of PAH exposure. I considered the interactions between various categorical variables in models in our analyses, they did not. I evaluated the co-exposure of SHS at both home and work in our analyses for adults, they did not.
As would be expected, the concentrations of ∑NAP increased with increase in the number of hours per day tobacco smoke was inhaled at work (p = 0.005, Table 3) but contrary to what should be expected, those who were exposed to SHS at work were found to have lower concentrations of ∑NAP than those who were not exposed to SHS at work (p = 0.012, Table 3). The reason for this is unknown but there may be one possibility. All those who were not at a job and/or were not working were coded as being not exposed to SHS at work while, in fact, they may have been exposed. This may have resulted in a false statistically significant finding. Thus, caution should be exercised in interpreting this finding. It should be noted only when 1-hydroxynaphthalene was statistically significantly (p = 0.002, Table S1) negatively associated with exposure to SHS at work.
When the primary source of exposure to PAHs is through inhalation, the concentrations of exposure can be enhanced with increase in physical activity which may be associated with increased inhalation rate and of course, exposure levels do depend upon the concentration of these chemicals in the air. While, exposure to other PAHs can occur through multiple sources, exposure to naphthalene occurs mainly through inhalation (Li, Sandau, et al., 2008). When route of exposure is through diet, enhanced physical activity may lead to relatively higher energy intake which depending upon the concentrations of PAHs in the diet may lead to higher levels of exposure and which may affect the rate of metabolism.
Out of 10 PAH metabolites for which data were available for this study, concentrations of 6 metabolites increased statistically significantly over the study period of [2003][2004][2005][2006][2007][2008] (Tables 3 and S1). The increasing trends seen in this study should be of concern because of established mutagenicity and carcinogenicity (Castano-Vinyals et al., 2008;Perera, Tang, Whyatt, Lederman, & Jedrychowski, 2005;Petruzzelli et al., 1998) associated with exposure to certain PAHs. Since the source of exposure to PAHs was not available, it is not possible to speculate the reasons for these increases. However, diet is one source of exposure that could be a possible reason. Unfortunately, in spite of the availability of data on diet from 24-h recall questionnaires, it was not possible to assess this as a source of exposure to PAHs because of the unavailability of complete data on the consumption of the smoked and broiled food. Suwan-ampai et al. (2009) used data from NHANES for the period 1999-2002 and reported females to have higher model adjusted geometric means than males for 1-hydroxyphenanthrene and 2-hydroxynaphthalene. The geometric means for other metabolites analyzed by them were similar for males and females. Some of the results reported by Suwan-ampai et al. (2009) could not be confirmed by me because of the differences in design between this study and study by Suwan-ampai et al. (2009). Suwan-ampai et al. (2009 used participants aged 6 years and older in their study. This study used participants aged 20 years and older. There may be several reasons why for this study, females were found to have statistically significantly higher concentrations of 1-and 2-hydroxynaphthalene, 1-hydroxyphenanthrene, 1-hydroxypyrene, and total naphthalene, fluorene, and phenanthrene metabolites. First, though unlikely, females may be consuming diets cooked at high temperatures more often and/or in larger quantities than males. Adjustments for diets that expose consumers to relatively higher levels of exposure to PAHs could not be made for reasons previously mentioned. Secondly, which seems to be a more likely explanation, there may be differences in how certain PAHs are metabolized by males and females. It appears that females metabolize certain PAHs like naphthalenes more slowly than males or in other words, females have longer half-life for these PAHs than males. Gender differences in PAH-induced CYP1A1 expression have been reported by Mollerup, Ryberg, Hewer, Phillips, and Haughen (1999) among other authors.
It should be noted that UGMs for males were higher for ∑FLU and 1-hydroxypyrene (Table 2) than females but when adjustments were made for differential contributions of race/ethnicity, poverty income ratio and other factors, these differences disappeared for ∑FLU and the differences were reversed for 1-hydroxypyrene (Table 4).
Some of the racial/ethnic differences and associations with socioeconomic status noted by Suwaniampai et al. (2009) also could not be confirmed in this study. In this study, family poverty income ratio (PIR) as a continuous variable was used as an indicator of socioeconomic status. Low income was associated with statistically significantly higher concentrations for ∑FLU and 1-hydroxypyrene (Table  3) as well as 3-hydroxyphenanthrene (Table S1). Suwani-ampai et al. (2009) used both respondents' education as well as PIR as indicators of socioeconomic status but instead of using PIR as a continuous variable, they used it as a categorical variable. Suwani-ampai et al. (2009) found low PIR as compared with medium and/or high PIR to be associated with higher concentrations of 1-hydroxypyrene, 1-, 2-, and 3-hydroxyphenanthrene, 1-and 2-hydroxynaphthalene, 2-and 3-hydroxyfluorene. Once again, the differences in results in this and their study may be due to differences in study design. They found NHW to have higher concentrations of all PAH metabolites except 2-hydroxynaphthalene than MA with or without statistical significance having been reached. Also, they found NHW to have higher concentrations of all PAH metabolites except 2-hydroxynaphthalene and 9-hydroxyfluorene than NHB with or without statistical significance having been reached. Similar results were observed in this study also except for distribution of the concentrations of 1-hydroxynaphthalene and 2-hydroxynaphthalene (Table S2). While MA had the lowest concentrations of 1-hydroxynaphthalene at 2,457.8 ng/L and NHW had the highest concentrations at 3,165.2 ng/L, for 2-hydroxynaphthalene, MA had the highest concentrations at 4,713.1 ng/L, and NHW had the lowest concentrations at 3,562.6 ng/L. Racial/ethnic differences in concentrations for various PAHs seen in this study are also probably due to differences in how various PAHs are metabolized by different race/ethnicities. More work will be needed to explain why MA had the highest concentrations of 2-hydroxynaphthalene and lowest concentrations of 1-hydroxynaphthalene or, in other words, metabolize them differently. However, racial differences in CYP1A1 gene inducibility have been reported by Cosma et al. (1993).  (Tables S1 and 3) as well as total naphthalene, fluorene, and phenanthrene metabolites (Table 3) varied over time from 2003-2004 to 2007-2008. In addition, I also computed unadjusted concentrations of three aggregated metabolites for each of the three NHANES cycles (Table 2), namely, 2003-2004, and 2007. Second, while Suwan-ampai (2009 did use data on exposure to SHS at home to define involuntary smokers, they did not evaluate if the extent of exposure to SHS at home may also affect PAH concentrations. I used model-based approach to determine if the extent of exposure to SHS at home defined by the number of smokers smoking inside home may affect PAH concentrations (Table  S1 and 3). The results indicate that both the exposure as well the extent of exposure to SHS at home affect PAH concentrations. Third, while Suwan-ampai (2009) did not determine if the SHS exposure at work may affect PAH concentrations, I used model-based approach to determine if the exposure as well the extent of exposure to SHS at work may affect PAH concentrations (Tables S1 and 3).
In summary, (i) statistically significantly higher concentrations of selected metabolites of naphthalene, fluorene, phenanthrene, and pyrene are associated with exposure to SHS at home, (ii) smoking was associated with statistically significantly higher concentrations of every metabolite of naphthalene, fluorene, phenanthrene, and pyrene analyzed in this study, (iii) there was a statistically significant increase in the concentrations of total naphthalene and fluorene, 2-hydroxynaphthalene, 2-and 9-hydroxyfluorene, 2-and 4-hydroxyphenanthrene, and 1-hydroxypyrene during the study period of 2003-2008, and (iv) poverty income ratio was negatively associated with the concentrations of 2-, 3-, and 9-hydroxyfluorene, 3-hydroxyphenanthrene, and 1-hydroxypyrene.
The limitations of this study should be carefully considered. Consumption of food cooked at high temperatures like broiled and smoked food has been found to be associated with elevated concentrations of PAHs. An attempt was made to include the consumption of such food as an independent variable in this study but since it was not fully possible to identify all those who consumed smoked or broiled food, a few data points where it was possible to identify smoked or broiled food were deleted from the analysis. To what degree, the results of this study could have been affected because of this limitation is unknown. Neither the magnitude nor the source of exposure to PAHs was available. The timing of the exposure was also not available for either SHS or PAHs. Data on exposure to PAHs from traffic related sources were also not available and as such, exposure to PAHs from traffic sources could not be included as one of the independent variables in the regression models. A relatively long-term follow-up study (or may be a short term well designed study) that can keep track of the ongoing PAH and SHS exposure and observed urinary PAH metabolite concentrations may better be able to evaluate association between PAH exposure and SHS exposure.