The NHANES is a nation-wide representative survey assessing the health and nutritional status of residents in the United States. The detailed study design has been described previously . Briefly, for each biennial NHANES survey a complex sampling process was used to randomly select U.S. residents who are representative of the civilian non-institutionalized population. Participants completed interviews at their homes with trained health professionals and underwent physical examinations in a mobile examination center. During the physical examination, a single spot urine sample was collected from NHANES participants aged six years and older. Written informed consents were obtained from all participants. The study protocol was approved by the institutional review board at the Centers for Disease Control and Prevention (Atlanta, Georgia) .
The current study combined data from six NHANES cycles (2001–2002, 2003–2004, 2005–2006, 2011–2012, 2013–2014, 2015–2016) among participants ≥20 years old. We excluded two cycles (2007–2008, 2009–2010) because the measurements on body fat distribution were unavailable. Since cigarette smoking is a predominant exposure source of PAHs among smokers, and smoking is known to modulate body weight , the association between PAHs and body fat distribution can be largely driven by smoking status. To avoid the strong confounding of cigarette smoking on the associations between PAHs and body fat contents, we restricted the current analysis to non-smokers. In addition, participants below 20 years old (n = 27,842), without medical examination data (n = 2284) or complete DXA scan (n = 10,214), with medical conditions that may influence body fat distribution (kidney disease, physical limitations, diabetes, cardiovascular disease, asthma, pulmonary disease, and cancer) (n = 2504), and with missing values in smoking status (n = 3) were excluded. Of the 61,049 surveyed residents, 7531 individuals had measurements of both PAH levels and body fat distribution. Of them, 2691 non-smokers were included in the main analysis after exclusions. The flowchart of the study design is shown in Supplemental Fig. S1.
Measurements of urinary PAH metabolites
In NHANES study, urinary monohydroxylated metabolites of PAHs were measured in approximately one-third subsample of all participants 6 years and older in the six NHANES cycles. Ten urinary PAHs were measured, including two naphthalene metabolites (1-napthalene, 2-napthalene), three fluorene metabolites (2-fluorene, 3-fluorene, 9-fluorene), four phenanthrene metabolites (1-phenanthrene, 2-phenanthrene, 3-phenanthrene, 4-phenanthrene), and one pyrene metabolite (1-pyrene). In addition to the ten individual PAHs, we created four other variables: the sum of the molar mass of all PAH metabolites, total naphthalene metabolites, total phenanthrene metabolites, and total phenanthrene metabolites, respectively. Of note, the number of PAHs measured in each NHANES cycle varied: it was nine in 2001–2002, ten in 2003–2004/2005–2006/2011–2012, and six in 2013–2014/2015–2016 cycles. We included PAHs measured in at least two survey cycles to ensure a reasonable sample size.
The detailed procedure of laboratory measurement has been described previously . In brief, PAHs were measured in the single spot urine sample by capillary gas chromatography and high-resolution mass spectrometry. To minimize the variability of urinary PAH levels due to differential dilutions, creatinine-adjusted PAHs (dividing PAH concentrations [nanograms per liter urine] by creatinine concentrations [grams per liter urine]) were used in the current analysis as recommended previously . Most PAHs had a detection rate of > 99%, and the detection rate for 4-phenanthrene and 9-fluorene was 89 and 93%, respectively. The coefficient of variation (CV) of PAH measurements ranged from 5 to 13%.
Dual-energy X-ray absorptiometry (DXA) measurements
DXA scans were administered to eligible survey participants 8 years of age and older in the NHANES mobile examination centers. Females with a positive pregnancy test or those who reported being pregnant at the time of the exam were excluded from the DXA examination. Individuals who reported taking tests with radiographic contrast material in the past 72 hours, participants in nuclear medicine studies in the past 3 days, or those who had a self-reported weight (> 300 pounds) or height (> 6′5″) over the DXA table limit were also excluded from the DXA examination . The whole body DXA was taken with a Hologic QDR-4500A fan-beam densitometer (Hologic, Inc., Bedford, Massachusetts) and provided fat mass and lean mass measurements for the total body, both arms, both legs, the trunk, and the head. Hologic Discovery software was used to derive fat mass and lean mass including total fat, trunk fat, leg fat, bone mineral density, and total lean mass. Multiple imputations were conducted to impute missing data for attenuating any potential biases due to missing DXA data. Specifically, missing readings for DXA data were imputed five times by the sequential regression multivariate imputation . Fat mass percentage (FM%) for total body, trunk and legs were calculated as: FM% = fat mass (kg)/total mass (kg) × 100%.
Measurements of body mass index and waist circumference
Body measurements such as weight (kg), height (cm), and WC (cm) were measured by trained health professionals following standardized protocols . BMI was calculated as weight divided by height squared (kg/m2).
Information on demographics, lifestyle, medical history, and dietary intake was collected using survey questionnaires during the face-to-face interview . Demographic information included age (years), gender (men, women), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or others), education (high school or below, any college, and college graduate or above), and poverty income ratio (< and ≥ 1). The poverty income ratio was used as an indicator of socioeconomic status and was calculated by dividing family income by the poverty threshold adjusted for family size and inflation. Lifestyle factors consisted of alcohol use (nondrinkers, 1–3 drinks/day, and ≥ 4 drinks/day) and moderate-to-vigorous physical activity (yes, no). Dietary intake included total calorie intake (kcal/day) and protein intake (gram/day). Serum cotinine levels (ng/mL), an indicator for environmental tobacco smoke among non-smokers , were measured by the isotope-dilution high-performance liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometric method with a lower detection limit of 0.015 ng/ml. C-reactive protein (CRP) levels were quantified by latex-enhanced nephelometry, with a lower detection limit of 0.1 mg/L.
Since the number of missing values varies among PAHs, sample size may vary for individual PAHs. Log-transformation (e base) was applied for creatinine-adjusted PAHs and FM% (total fat, trunk fat, leg fat, trunk/leg ratio), bone mineral density, total lean mass, BMI, and WC to improve the normality. In this cross-sectional analysis, we calculated partial Pearson correlation coefficients (r) weighted by the NHANES sampling weight with adjustment of potential confounding factors to examine correlations between log-transformed PAHs and FM%, bone mineral density, total lean mass, BMI, and WC among non-smokers. Covariates adjusted included age, gender, race/ethnicity, education, poverty income ratio, moderate-to-vigorous physical activity, alcohol use, total calorie intake, protein intake, and serum cotinine and CRP levels. Missing values of continuous covariates were replaced with median values. For categorical variables, a missing indicator variable was created for missing values. To examine the potential differences by race/ethnicity, we performed stratified analyses by main racial/ethnic groups (non-Hispanic White, non-Hispanic Black, and Hispanic). Correlation coefficients of creatinine-adjusted PAHs with FM% were transformed to Fisher z scores and compared using the Student’s t-test between two groups. We also compared correlations of PAHs with trunk FM% versus leg FM% using the Wolfe’s method [31, 32]. Due to the multiple-imputation procedure, each participant has five sets of measured and imputed values of body fat; correlations were calculated within each DXA dataset, and combined into a single composite estimate using the method of Rubin and Schenker .
To account for multiple comparisons, False Discovery Rate (FDR) Benjamini-Hochberg Procedure was used for the correction of P values. Statistical significance was determined by a two-sided FDR P value smaller than 0.05. All data were analyzed in RStudio version 1.4.1106 (Rstudio, PBC) and SAS version 9.4 (SAS Institute, Cary, NC).
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.