1.1 Study Population
The National Health and Nutrition Examination Survey (NHANES) is a national survey assessing the health and nutritional status of children and adults in the U.S. (National Center for Health Statistics, 2017). Urinary arsenic samples were collected in participants aged 18 years and older who met the one-third subsample selection criteria. To oversample adult smokers, participants 18 years and older who were not in the one-third subsample but who smoked at least 100 cigarettes in their entire lifetime were included (CDC, 2018b). For sleeping disorders, participants aged 16 years and older were sampled by NHANES (CDC, 2018a). In our study, only participants aged 20 years and older were included in the data analysis. Demographic data for the NHANES study is included in the "2015-2016 Demographics File" (CDC, 2017a). We restricted the sample to only individuals who had no missing data on the independent (speciated arsenic) and dependent (trouble sleeping) variables, and depression, a key confounder in this study (n=1,611).
1.2 Urinary Arsenic Assessment
The NHANES 2015-2016 "Speciated Arsenics - Urine - Special Sample (UASS_I)" dataset was used in this analysis. Samples were collected at Mobile Examination Centers (MECs) and sent to the Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA, for analysis. Detailed information on the laboratory procedures conducted by NHANES can be found in the "Arsenics - Speciated - Urine Laboratory Procedure Manual." Samples were analyzed for the following types of speciated arsenic and corresponding lower level of detection (LLOD): urinary arsenous acid 0.12 µg/L, urinary arsenic acid 0.79 µg/L, urinary arsenobetaine 1.16 µg/L, urinary arsenocholine 0.11 µg/L, urinary dimethylarsinic acid (DMA) 1.91 µg/L, urinary and monomethylarsonic acid (MMA) 0.20 µg/L (CDC, 2018b).
1.3 Sleep Disorder Assessment
The NHANES 2015-2016 "Sleep Disorders (SLQ_I)" dataset was used in this analysis. Questions were asked in the home by interviewers using a Computer-Assisted Personal Interview (CAPI) system. Several questions related to sleep patterns were asked, including usual sleep time on weekdays or workdays, usual wake time on weekdays or workdays, sleep hours, how often participants snore, snort, or stop breathing, if participants have ever told a doctor, they had trouble sleeping, and how often they feel overly sleepy during the day (CDC, 2018a). In our study, to determine if the participant had a sleep disorder, we used the NHANES variable SLQ050 (ever told a doctor or other health professional that you had trouble sleeping).
We included covariates from social and demographic factors such as gender (male, female), race/ethnicity (Mexican American, other Hispanic, Non-Hispanic White, Non-Hispanic Black, Other Race- Including Multi-Racial), highest level of education (less than high school, high school graduate/GED or equivalent, some college or A.A. degree, college graduate or above), marital status (married, widowed/divorced/separated, never married, living with partner), family poverty-to-income ratio (<130%, 130% - 350%, >350%), moderate physical activity (yes, no), cotinine, BMI (underweight, normal weight, overweight, obese), and depression (yes, no). For demographic variables used in Table 1, the following data files were used: 2015-2016 Body Measures (BMX_I), Physical Activity (PAQ_I), Cotinine and Hydroxycotinine - Serum (COT_I), Mental Health - Depression Screener (DPQ_I), and Demographic Variables (DEMO_I). Other data files utilized in this study are Albumin and Creatine (ALB_CR_I), Arsenic Total Urine Special Sample (UTASS_I), and Speciated Arsenics-Urine Special Sample (UASS_I).
Creatinine (mg/dL), antimony (ug/L) and total urinary arsenic (ug/L) were all defined as continuous variables. Serum Cotinine (COT_I), a valid marker, was used to adjust for smoking. BMI was categorized as underweight <18.5, normal weight 18.5-24.9, overweight 25-29.9, and obese ≥30 (CDC, 2003). Moderate physical activity was assessed by questionnaire, which asked participants, “In a typical week do you do any moderate-intensity sports, fitness, or recreational activities that cause a small increase in breathing or heart rate such as brisk walking, bicycling, swimming, or volleyball for at least 10 minutes continuously?”. The responses were yes and no (CDC, 2017b). The poverty-to-income ratio (PIR) is determined annually using the poverty threshold determined by U.S. Department of Health and Human Services. PIR is the annual household income divided by the poverty threshold, categorized by < 130%, 130–350%, and >350% (Bailey et al. 2017). The Patient Health Questionnaire (PHQ-9), a nine-question screening tool, was used to assess for depression. The questionnaire determines the frequency of depression symptoms with a score placed on the frequency of each symptom over the previous 2 weeks. Scores range from 0 to 3 as follows: 0: “not at all,” 1: “several days,” 2: “more than half the days,” or 3: “nearly every day.” The maximum score is 27; depression was defined as a score of ≥ 10.
1.5 Statistical Analyses
Survey design procedures were used to create weighted analysis, which incorporated sampling design stratification and clustering procedures. To assess the distribution of participants' demographic and socioeconomic status by trouble sleeping, descriptive statistics were used. Percentages were reported for the categorical predictors, and medians with interquartile ranges were written for the continuous variables. A bar-chart of the median levels of the six speciated urinary arsenic examined was plotted. The Wilcoxon-Mann-Whitney test was used to determine the differences between the underlying distributions of age, serum cotinine, creatinine, antimony and total urinary arsenic by trouble sleeping. In contrast, the chi-square test was utilized to assess the relationship between the categorical covariates and trouble sleeping.
Unadjusted logistic regression was used to compute odds ratios (ORs) and 95% confidence intervals (95% CI) for the association between urinary arsenic and sleep disorders. Multivariate logistic regression was used to calculate the 95% CI for the exposure-disease relationship of interest and ORs; potential confounders were controlled. Each independent arsenic variable had individual models run. In the bivariate analysis, confounders controlled for in the analyses that were significantly associated with trouble sleeping included race/ethnicity, age, serum cotinine, and depression. All analyses were computed using SAS® (version 9.4, Cary, NC) with p-value < 0.05 considered statistically significant.