Since 1999, the CDC has conducted the National Health and Nutrition Examination Survey (NHANES) annually. NHANES provides data, released in 2-year intervals, to evaluate the health and nutritional status of the civilian, noninstitutionalized U.S. population of all ages. NHANES includes household interviews, standardized physical examinations, and collection of medical histories and biological specimens. Some of these specimens are used to assess exposure to environmental chemicals.
For this study, we analyzed 2,548 spot urine specimens collected during one of three daily examination sessions from a one-third subset of 2005–2006 NHANES participants ≥ 6 years of age. The representative design of the survey was maintained, because the subset was a random sample of the total NHANES population. The National Center for Health Statistics (NCHS) Institutional Review Board reviewed and approved the study protocol. All participants gave informed, written consent; parents or guardians provided consent for participants < 18 years of age.
The urine samples were shipped on dry ice to the National Center for Environmental Health at the CDC and stored at −20°C or below until analyzed. The analytical method for measuring 15 phthalate monoesters, including MNP and oxidized metabolites of DINPs (MCOP) and DIDPs (MCNP), in 100 μL urine has been described in detail elsewhere (Silva et al. 2007b
). The analytical approach involved enzymatic hydrolysis of the conjugated species of phthalate metabolites, followed by online solid-phase extraction, separation with high-performance liquid chromatography, and detection by isotope-dilution negative ion electrospray ionization tandem mass spectrometry. We used the calibration curves constructed with mono(2,6-dimethyl-6-carboxyhexyl) phthalate (for MCOP), mono(2,7-dimethyl-7-carboxyheptyl) phthalate (for MCNP), and mono(3,5,5-trimethyl-1-hexyl) phthalate (for MNP) and their isotopically labeled analogs as the internal standards for quantification, as described previously (Silva et al. 2007b
). Calibration standards, quality control, and reagent blank samples were included in each analytical batch along with the study samples (Silva et al. 2007b
Under our experimental conditions, MCOP and MCNP, the metabolites of DINP and DIDP, respectively, were not chromatographically resolved, and both MCOP and MCNP eluted separately as broad peaks. For quantification, we integrated the whole area under the cluster of peaks encompassing the various isomers of MCOP and MCNP. The hydroxy- and oxo-oxidative metabolites of DINP (Koch and Angerer 2007
; Koch et al. 2007
) could not be separated adequately; as a result, we could not estimate their concentrations. The limits of detection (LODs)—calculated as 3S0
, where S0
is the standard deviation as the concentration approaches zero (Taylor 1987
)—were 0.8 μg/L (MNP), 0.7 μg/L (MCOP), and 0.6 μg/L (MCNP). We prepared low-concentration (4–9 μg/L) and high-concentration (27–58 μg/L) quality control materials with pooled human urine that was analyzed with standards, reagent blanks, and urine samples. The precision of measurements, expressed as the relative standard deviation of multiple measures, depending on the phthalate metabolite, was 8–10% for low-concentration and 6–10% for high-concentration quality control samples.
We used SAS (version 9.2; SAS Institute Inc., Cary, NC) and SUDAAN (version 10; Research Triangle Institute; Research Triangle Park, NC) to perform statistical analyses. SUDAAN calculates variance estimates that account for the complex, clustered design of NHANES. As recommended by NCHS, we used sample population weights to produce estimates that are representative of the U.S. population. We used the log10
-transformed urinary metabolite concentrations for the statistical analyses and assigned a value equal to the LOD divided by the square root of 2 (Hornung and Reed 1990
) to the concentrations below the LOD.
We stratified age, reported in years at the last birthday, in four groups (6–11 years, 12–19 years, 20–59 years, and ≥ 60 years). On the basis of self-reported data, we categorized race/ethnicity as non-Hispanic black, non-Hispanic white, and Mexican American. Participants not defined by these racial/ethnic categories (n = 195) were included only in the total population estimate. For each age, sex, and race/ethnic group, we calculated geometric means (GMs) (if the overall weighted frequency of detection was > 60%) and distribution percentiles for both volume-based (micrograms per liter) and creatinine-corrected concentrations (micrograms per gram creatinine). We also determined weighted Pearson correlations among the creatinine-corrected concentrations (log10 transformed) of MCOP, MCNP, and MNP in the 334 samples with detectable concentrations of all three compounds. Statistical significance was set at p < 0.05.
We used multiple regression to examine whether several variables [i.e., age group, sex, race/ethnicity, creatinine concentration, household income, and examination session (i.e., morning, afternoon, evening)] were associated with the log10-transformed urine concentrations of MCOP and MCNP. On the basis of questionnaire responses, annual household income was available in increments of $5,000 (ranging from < $5,000 to > $75,000). We categorized income as < $20,000, $20,000–$45,000, and > $45,000 to obtain a comparable number of participants per group. For the multiple regression models, we used the variables described previously and all their possible two-way interactions to calculate the adjusted GM concentrations (in micrograms per liter) of MCOP and MCNP. These variables were log10 transformed, because the distributions of concentrations of these phthalate metabolites and creatinine were skewed.
To arrive at the final model for each analyte, we used backward elimination with SUDAAN to remove the nonsignificant interactions one at a time. Covariates with nonsignificant main effects were then removed one at a time, and the model was rerun to determine whether the beta coefficients for covariates with significant main effects or interactions changed by > 10%. If any did, we retained the relevant nonsignificant covariate in the model. Once the backward procedure was completed, covariates and interactions between covariates were added back into the model one at a time to determine whether any were significant, in which case they were retained in the final model.
We also constructed a 2 × 2 table to examine the suitability of the urinary concentrations of MNP and MCOP as DINP exposure biomarkers.