Study population
National Health and Nutrition Examination Survey (NHANES) is a representative, nationwide, cross-sectional survey, with a complex multistage probability sampling design of the noninstitutionalized U.S. population, conducted by the National Centers for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC)(CDC, 2022). The survey combines in-person interviews, physical examinations, and laboratory tests to assess the health and nutritional status of the general U.S. population. The NHANES was approved by the NCHS Ethics Review Board, and written informed consent was obtained from each participant. More detailed information regarding the NHANES datasets and related documentation can be obtained elsewhere(CDC, 2022).
In this study, we used data from four consecutive cycles of the NHANES (2007–2008, 2009–2010, 2011–2012, 2013–2014) as cross-sectional data. Pyrethroid exposure measurements were collected from a one-third subsample of participants aged ≥6 years(CDC, 2013). There were 10789 participants in these four cycles, but 3325 participants under 20 years of age were excluded. In addition, we excluded 1060 participants who lacked information on depressive symptoms and urinary 3-PBA concentrations. Furthermore, we excluded two participants with missing urinary creatinine concentrations. Given that the percentage of missing values of alcohol intake (12.4%) and family income-to-poverty ratio (8.3%) were >5%(Chen et al., 2019), we performed multiple imputations for all missing values of covariates using the fully conditional specification (FCS) method in PROC MI(Yuan, 2011) (Table A.1). Finally, 6402 participants were included in the analysis. The screening process for study participants is shown in Fig.1.
Urinary pyrethroids exposure assessment
Spot urine samples were collected each 2-year cycle and stored at -20°C until they were shipped to the CDC’s National Center for Environmental Health laboratory for analysis. Pyrethroid metabolites, including 3-PBA, 4-fluoro-3-phenoxybenzoic acid (4-F-3PBA), trans-3-(2,2-dichlorovinyl)-2,2-dimethylcylopropane carboxylic acid (trans-DCCA), and cis-(2,2-dibromovinyl)-2,2-dimethylcyclopropane-1-carboxylic acid (cis-DBCA), were measured in spot urine samples using high-performance liquid chromatography/electrospray chemical ionization/tandem mass spectrometry(Baker et al., 2004; Barr et al., 2010). The value of each metabolite concentration below the limit of detection (LOD) was estimated for each LOD and divided by the square root of 2(CDC, 2013). In our study, 3-PBA was detectable in approximately 80% of the adult participants, whereas other metabolites were detectable in only a small percentage (Table 1). According to previous studies, to avoid undue influence on the estimates due to large amounts of value imputations, we only carried out statistical analysis on metabolites with a sufficient detection rate (>60%)(Barr et al., 2010). Therefore, we used urinary 3-PBA as the primary marker of pyrethroid pesticide exposure in the current study. 3-PBA is a common metabolite that is exposed to permethrin, cypermethrin, deltamethrin, allethrin, resmethrin, fenvalerate, cyhalothrin, cyhalothrin, fenpropathrin, and tralomethrin(Barr et al., 2010). To account for urine dilution, urinary creatinine concentrations were used to correct urinary 3-PBA concentrations(Barr et al., 2005). Urine creatinine concentrations were measured by an automated colorimetric method using a modified Jaffe reaction Beckman Synchron AS/ASTRA clinical analyzers(Barr et al., 2010).
Table 1 Weighted detection rates of pyrethroid metabolites among adults in NHANES 2007–2014.
Metabolites
|
Sample size(n)
|
Limit of detection(μg/L)
|
Weighted detection rate(%)
|
3-PBA
|
7095
|
0.1
|
81.3
|
4-F-3PBA
|
7232
|
0.5
|
8.9
|
trans-DCCA
|
7086
|
0.6
|
16.0
|
cis-DBCA*
|
3769
|
0.1
|
1.1
|
Abbreviation: 3-PBA, 3-phenoxybenzoic acid; 4-F-3PBA, 4-fluoro-3-phenoxybenzoic acid; trans-DCCA, trans-3-(2,2-dichlorovinyl)-2,2-dimethylcyclopropane-1-carboxylic acid; cis-DBCA, cis-3-(2,2-dibromovinyl)-2,2-dimethyl-cyclopropane-1-carboxylic acid.
*Cis-DBCA data was from two consecutive cycles of NHANES (2007–2008 and 2009–2010).
Assessment of depressive symptoms
Depressive symptoms were assessed using the Patient’s Health Questionnaire (PHQ-9), which is considered a validated and reliable measurement tool(Spitzer et al., 1999). The PHQ-9 has nine items to evaluate symptoms of depression in the past 2 weeks. Each item was given a score of 0 (not at all), 1 (several days), 2 (more than half the days), or 3 (nearly every day), with a total score ranging from 0 to 27. Participants with a total score of ≥10 were defined as having depressive symptoms; otherwise, they were classified as having no depressive symptoms(Kroenke et al., 2001; Levis et al., 2019).
Covariates
Sociodemographic factors (age, sex, race/ethnicity, education, marital status, family income-to-poverty ratio, and health insurance), lifestyle factors (physical activity, smoking status, alcohol intake, body mass index [BMI]), trouble sleeping, and disease status were selected as covariates based on previous literature(Lehmler et al., 2020; Yu et al., 2020) and biological considerations. Age was categorized into three groups (20–39, 40–59, and ≥60 years), and was modeled as a continuous variable. Sex was dichotomized as male or female. Race/ethnicity was classified as Hispanic (including Mexican and other Hispanics), non-Hispanic white, non-Hispanic black, and others. Education was categorized into three groups (less than high school, high school, and college or higher). Marital status was categorized into two groups (living together and single)(Yu et al., 2020). The family income-to-poverty ratio was defined as total family income divided by the federal poverty threshold (<1.3, 1.3–3.5, and >3.5), with a higher family income-to-poverty ratio indicating a higher family income status(Bao et al., 2020). Health insurance was assessed by the (yes/no) question “Are you covered by health insurance or some other kind of health care plan?”. Physical activity was classified as inactive (no regular activity), low (regular activity, but <150 min/week), moderate (150–300 min/week), or high (>300 min/week) according to the 2018 Physical Activity Guidelines for Americans(HHS, 2018). Smoking status was defined as never (participants who smoked <100 cigarettes in their lifetime), former (participants who smoked ≥100 cigarettes in their lifetime but did not smoke during the interview), or current smoker (participants who smoked ≥100 cigarettes in their lifetime and smoked during the interview)(Wang et al., 2018). Alcohol intake was defined as none (consumed <12 alcoholic drinks in the past year), moderate (consumed <1 drink per day for women or consumed <2 drinks per day for men), and heavy (consumed ≥1 drink per day for women or consumed ≥2 drinks per day for men)(Xue et al., 2021). The BMI was calculated as weight (kg) divided by height squared (m2) and categorized as normal (<25 kg/m2), overweight (25–29.9 kg/m2), and obese (≥30 kg/m2). Trouble sleeping was assessed through the (yes/no) question, “Ever told by a doctor that you have trouble sleeping?”(Li et al., 2021). Disease status was defined as the presence of hypertension, diabetes, or CVD. Hypertension was defined by any of the following conditions: ever been told to have high blood pressure; ever been told to take a prescription for hypertension; currently taking prescribed medicine for hypertension; and having an average systolic blood pressure (SBP) ≥140 mmHg or diastolic blood pressure (DBP) ≥90 mmHg(Yu et al., 2020). Diabetes was defined by any of the following conditions: ever to have diabetes, now taking insulin, ever took diabetic pills to lower blood sugar, has a fasting glucose concentration ≥126 mg/dl, and HbA1c ≥6.5%(Yu et al., 2020). CVD was defined by self-reported doctor diagnosis as any of the following: coronary heart disease, myocardial infarction, angina pectoris, congestive heart failure, or stroke(Zhang et al., 2022).
Statistical analyses
To ensure that our results could be generalized to the national population, all analyses included sample weights to adjust for the complex sampling survey design. Continuous variables were presented as means and standard deviations and were compared using the t-test between groups with and without depression. Categorical variables are presented as cases (n) and percentages (%) and were compared using the chi-square test. Due to the highly right-skewed distribution of urinary 3-PBA concentrations according to the histogram, urinary 3-PBA and creatinine-corrected urinary 3-PBA concentrations were reported as geometric means (GMs) with 95% confidence intervals (CIs) and were compared between groups using t-test or Wald F-tests after ln-transformation. Consistent with the previous recommendation, we used urinary creatinine concentrations as a covariate(Xue et al., 2021) and further transformed it using the natural logarithm for normalization in the model analysis. Urinary 3-PBA was modeled as a continuous variable with ln-transformation and as a categorical variable with the lowest tertile as the reference in multivariable logistic regression models. We used the Taylor series method to estimate the odds ratios (ORs) and 95% CIs for the association between urinary 3-PBA and depressive symptoms. Three models were established: Model 1 was adjusted for urinary creatinine concentration; Model 2 was additionally adjusted for age, sex, race/ethnicity, education, marital status, family income-to-poverty ratio, and health insurance; and Model 3 was further adjusted for physical activity, smoking status, alcohol intake, BMI, trouble sleeping, and disease status. We further explored the nonlinear association using restricted cubic splines with three knots (10th, 50th, and 90th percentiles of ln-transformed urinary 3-PBA concentrations). In addition, we conducted a series of subgroup analyses by adding the 3-PBA×covariate term to the model to estimate the modification of certain variables on the association between 3-PBA exposure and depressive symptoms.
We conducted three sensitivity analyses to test the robustness of our results. First, we reanalyzed the association using creatinine-corrected urinary 3-PBA concentrations as an exposure variable rather than adjusting for urinary creatinine as a covariate in the logistic regression model. Second, based on previous findings that urinary 3-PBA concentrations were more than an order of magnitude higher in the 95th percentile(Lehmler et al., 2020), we conducted another sensitivity analysis by excluding participants with 3-PBA >150μg/L (n=5), according to the boxplot distribution as extreme values (Fig A.1). Finally, we repeated the main analysis by further adjusting for dietary factors, including energy, fat, protein(Li et al., 2020), fiber(Swann et al., 2020), and caffeine(Hall et al., 2015) intake.
All data analyses were conducted using the SAS software (version 9.4; SAS Institute, Cary, NC, US). Statistical significance was defined as a two-sided P-value of <0.05.