Study population and design
The NHANES is administered by the National Center for Health Statistics (NCHS) at the US Centers for Disease Control and Prevention (CDC). NHANES uses a multi-stage, stratified, probability sampling design to obtain unbiased national health information (detailed sample design: https://wwwn.cdc.gov/nchs/nhanes/tutorials/module2.aspx). Data from NHANES surveys for 2005–2016 includes exposures to multiple synthetic chemicals among the noninstitutionalized US population (NCHS) (Fangjian and Timothy, 2016). None of the authors of the current study were involved in the collection of data or the production of the database in use.
A total of 60,936 subjects were enrolled in the initial analysis. Participants missing data for phenols and parabens (n = 45,209), those pariticipants younger than 20 years and missing data regarding CVD outcomes (n = 5,299) were excluded from the present analysis. In total, 10,428 subjects with complete data on exposure to multiple phenols and parabens, primary covariates, and CVD outcomes were included for further analysis (Fig. S1).
Measurement of chemical exposures
Urinary samples were collected and stored at −20 °C until they were sent to the National Center for Environmental Health, where they were analyzed by the Organic Analytical Toxicology Branch. A total of 9 chemicals, BPA, BP3, 2,4-DCP, 2,5-DCP, TCS, BUP, EPB, MPB, and PPB were measured. The limits of detection (LOD) for BPA, BP3, 2,4-DCP, 2,5-DCP, TCS, MPB, and PPB were 0.2 ng/mL, 0.4 ng/mL, 1.0 ng/mL, 0.1 ng/mL, 1.7 ng/mL, 0.1 ng/mL, and 0.1 ng/mL, respectively (Supplement Table 1). The concentrations of BUP and EPB were not incorporated in further analyses due to low detectable concentrations.
The concentrations and distributions of the 7 remaining urinary chemical metabolites were analyzed. Urinary BPA, BP3, 2,4-DCP, and 2,5-DCP metabolites were extracted using online solid-phase extraction (SPE) and measured by high-performance liquid chromatography and tandem mass spectrometry (MS/MS). Urinary TCS, MPB, and PPB metabolites were extracted using online SPE and measured by high-performance liquid chromatography-isotope dilution MS/MS (34). All concentrations below the LOD were replaced by the LOD value divided by the square root of 2 (LOD/√2). The detailed protocol is described elsewhere (NHANES 2011–2012: Environmental Phenols & Parabens Data Documentation, Codebook, and Frequencies (cdc.gov)). All urinary measurements were adjusted for creatinine to account for differences in kidney function.
Assessment of CVD outcomes
Participants responded to questions related to medical conditions and CVD outcomes prior to undergoing a physical examination, which was documented by a Computer-Assisted Personal Interviewing (CAPI) system. Responses to follow-up questions related to CVD-specific medications and healthcare were also ascertained. Self-reported diagnosed and borderline diabetes/high blood pressure assessments were combined as ‘diabetes/hypertension’ for the purposes of this analysis.
Participants aged ≥20 years were asked, “Has a doctor or other health professional ever told you that you have…” congestive heart failure, coronary heart disease, angina, heart attack, or stroke. “Cardiovascular disease” (CVD) was defined as any reported diagnosis of congestive heart failure (CHF), coronary heart disease (CHD), angina, heart attack, or stroke. The data for each CVD outcome were collected for further analyses of associations with the 7 assessed urinary chemical metabolites.
Potential confounders of the CVD outcome measure, including demographic characteristics (age, sex, educational levels, race/ethnicity, and poverty income ratio [PIR]), lifestyle characteristics (BMI, smoking, and alcohol use), comorbidities (diabetes mellitus and hypertension), and urine creatinine (Ucrea) levels were collected. BMI was calculated as the ratio of weight in kilograms to height in meters squared (kg/m2). Ucrea data were collected by interview or laboratory detection performed by NHANES.
Covariates details were as follows: (1) age was modeled continuously; (2) sex was dichotomized as male or female; (3) education level was coded as blow/above high school and high school (or K-12 aged (6–19 years old), ≤ high school/GED, some college, ≥ Bachelor Degree); (4) the family PIR was assessed as the ratio of family income to the poverty threshold, poverty is characterized as a family PIR below 1 (Lesliam et al., 2018); (5) a smoker was identified as someone who smoked at least 100 cigarettes in his lifetime, alcohol use were categorized as never drinker, former drinker, current drinker; (6) diabetes mellitus (DM) and hypertension were assessed and determined by the questionnaire administered by NHANES personnel (Lesliam et al., 2018); Ucrea (mg/dL) concentrations were measured from the urine samples using an automated colorimetric method based on a modified Jaffe reaction on a Beckman Synchron AS/ASTRA clinical analyzer (Beckman Instruments, Inc., Brea, CA) and were used to correct for urine dilution in our statistical models.
A total of 7 urinary phenol and paraben metabolites (BPA, BP3, 2,4-DCP, 2,5-DCP, TCS, MPB, and PPB) were analyzed for their contributions to the risk of categorized CVD outcomes (CHD, CHF, heart attack, stroke, and angina).
The urinary chemical metabolite concentrations were log-transformed to normalize their distributions for further analysis. The Kolmogorov–Smirnov test was used to assess the normality of continuous variables. Continuous variables are expressed as the mean (standard deviation [SD]) and were compared using unpaired t-tests. Categorical or dichotomous variables are expressed as absolute values (percentages) and were compared using χ2 statistics. Pearson’s correlation coefficients were calculated among all urinary phthalate exposures. Multivariable logistic regression models were used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) to assess the prevalence of total and individual CVDs associated with urinary phenols, parabens, and TCS metabolites. The metabolite levels of urinary phenols, parabens, and TCS were divided into quartiles, and the lowest quartile was used as the reference category.
Regression models were adjusted for potential confounders, including age, sex, race, education levels, and poverty, which a previous study (Traci et al., 2020) reported being associated with urinary metabolite concentrations, and Ucrea was used to adjust for urine concentration (Ferguson et al., 2016).
Initial adjustment was for the following: age; sex; education; Advanced models were additionally adjusted for smoking, drinking, BMI, diabetes and hypertension history, and urinary creatinine concentration; Fully adjusted models were adjusted as advanced adjusted models plus other urinary 6 chemicals. Association between urinary chemicals metabolites concentration and CVDs were contacted. The calculate odds ratios (ORs) and 95% confidence intervals (CIs) used in multivariable logistic regression.
WQS regression was used to assess associations between mixtures of chemical exposures and CVD. Each exposure variable is assigned a weight within the index, which indicates the importance of the variable within the mixture (Bhatnagar, 2006; Han et al., 2016). WQS models were run using the gWQS v.2.0.0 in R v.3.6.1. Within the gWQS function, the reported results were estimated by specifying deciles for exposure weighting, with 40% of the data used as the test set and the remaining 60% used as the validation set, 1,000 bootstrap repetitions, the random seed set to 2016, and a binomial distribution used for the general linear model. In each case, the weights were estimated by pooling only the effects in the positive or negative direction as constrained. Individual chemical weights ≥0.1 were considered the most important. The value of 0.1 was chosen for ease of comparison across models, including different numbers of chemicals. Restricted cubic splines with knots were used at the 5th, 10th, 50th, and 90th percentiles for multiple phenols and parabens distributions with the highest weight proportions from the above results to further explore the dose-response curves between multiple phenols, parabens, and total CVD. Significance was set at P <0.05 (two-sided).