Abstract
Introduction:
Mercury intoxication is known to be associated with adverse symptoms of fatigue and sleep disturbances, but whether low-level mercury exposure could affect sleep remains unclear. In particular, children may be especially vulnerable to both mercury exposures and to poor sleep. We sought to examine associations between mercury levels and sleep disturbances in Mexican youth.
Methods:
The study sample comprised 372 youth from the Early Life Exposures to Environmental Toxicants (ELEMENT) cohort, a birth cohort from Mexico City. Sleep (via 7-day actigraphy) and concurrent urine mercury were assessed during a 2015 follow-up visit. Mercury was also assessed in mid-childhood hair, blood, and urine during an earlier study visit, and was considered a secondary analysis. We used linear regression and varying coefficient models to examine non-linear associations between Hg exposure biomarkers and sleep duration, timing, and fragmentation. Unstratified and sex-stratified analyses were adjusted for age and maternal education.
Results:
During the 2015 visit, participants were 13.3 ± 1.9 years, and 48% were male. There was not a cross-sectional association between urine Hg and sleep characteristics. In secondary analysis using earlier biomarkers of Hg, lower and higher blood Hg exposure was associated with longer sleep duration among girls only. In both boys and girls, Hg biomarker levels in 2008 were associated with later adolescent sleep midpoint (for Hg urine in girls, and for blood Hg in boys). For girls, each unit log Hg was associated with 0.2 hour later midpoint (95% CI 0 to 0.4), and for boys each unit log Hg was associated with a 0.4 hour later sleep midpoint (95% CI 0.1 to 0.8).
Conclusions:
There were mostly null associations between Hg exposure and sleep characteristics among Mexican children. Yet, in both boys and girls, higher Hg exposure in mid-childhood (measured in urine and blood, respectively) was related to later sleep timing in adolescence.
Keywords: Mercury, neurotoxicant, sleep, circadian, adolescence
1. INTRODUCTION
Sleep plays an integral role in health and well-being among adolescents. Mental and physical health, as well as academic performance and accident risk, have each been consistently related to sleep among adolescents(Owens et al., 2014). Yet, adolescents are consistently at risk for insufficient sleep and poor-quality sleep; for example, it has been estimated that 60% of US middle schoolers (approximate ages 11–13 years) and 70% of US high schoolers (14–18 years) do not obtain the recommended sleep given their age(Wheaton et al., 2018). Similarly, in a recent meta-analysis representing 17 different countries/regions, over 50% of adolescents did not obtain sufficient sleep(Gradisar et al., 2011). Thus, uncovering the role of potentially-modifiable environmental factors that could affect sleep is a critical public health issue(Owens et al., 2014). While there are a number of known factors that affect adolescent sleep (e.g. screen use, early school start times, caffeine intake), a few recent studies have highlighted the possible role that toxicants may play (Jansen et al., 2019; Shiue, 2017).
Mercury is a neurotoxicant that could plausibly affect sleep (Parmalee and Aschner, 2017). For example, autopsy reports of workers exposed to elemental mercury show that mercury can accumulate in the pineal gland (Falnoga et al., 2000), the brain organ responsible for regulating sleep and wake behaviors. In line with this pathway, adult workers at E-waste shops in Thailand who were exposed to mercury (average urinary Hg levels 19.0 ug/g creatinine) had higher reports of insomnia symptoms than office workers who were not occupationally exposed (Decharat, 2018). Case reports from both adult and child populations exposed to relatively high levels of mercury (with ranges of urinary Hg exposures from 1.6 to >200 ug/L) through accidental poisoning events or gold mining activities have documented sleep complaints including fatigue, excessive sleepiness, and sleep disturbances (Bose-O’Reilly et al., 2016; Do et al., 2017; Kasznia-Kocot et al., 2010). In addition, a study of 118 US adults showed that weekly consumption of fish with high mercury content was associated with more daytime fatigue, although neither mercury levels or sleep characteristics were measured directly (Kothari et al., 2015). It is important to note that these populations mostly had relatively high levels of mercury exposure; thus, very little is known about how lower-levels of exposure could affect sleep. Other limitations in these studies includes small sample sizes (all studies identified had <150 participants) and the fact that most are case reports. Finally, very few studies have assessed the association between Hg and sleep at younger life-stages. To our knowledge, there is only one study among children with lower-level exposures. In this cross-sectional study of 100 children aged 9 to 11 years, study authors showed that higher mercury levels were associated with shorter sleep duration (Gump et al., 2014).
To address the current gaps in the literature, we utilized the ELEMENT study(Perng et al., 2019), a prospective study of Mexican children who have been followed since prenatal life and who have assessments of exposures to mercury through biomarker measurements. The primary aim was to evaluate whether concurrent adolescent mercury exposure (measured in urine) was related to actigraphy-assessed sleep duration, timing, and fragmentation measured at the same time. The secondary aim was to assess whether mercury exposure measured in mid-childhood (through hair, blood, and urine) were related to adolescent sleep duration, timing, and fragmentation.
2. MATERIALS AND METHODS
2.1. Study Population
The study sample included adolescent participants from 2 of 3 sequentially-enrolled cohorts of the Early Life Exposure in Mexico to ENvironmental Toxicants (ELEMENT) study(Perng et al., 2019). Between 1997 and 2004, 1012 mother/child dyads were recruited from prenatal clinics of the Mexican Social Security Institute in Mexico City, which serves low- to middle-income populations formally employed in the private sector. Beginning in 2015, a subset of 554 participants from the original birth cohorts 2 and 3, who were undergoing the pubertal transition (ages 9 to 18 years), were selected to participate in a follow-up study. This study included 7 consecutive days of actigraphy for sleep/wake behavior along with a urine sample for mercury analysis. The primary analysis includes 372 children with both sleep measurements and urine mercury analysis. As part of a secondary analysis, among a subsample of 259 children, we also assessed mercury levels in urine, blood and hair in in 2008, when they were between 8–12 years of age (called ELEMENT 2008). See Figure 1 for a schematic diagram of the measures. The Institutional Review Boards at the Mexico National Institute of Public Health and the University of Michigan approved the research protocols. Informed consent was obtained from parents for all participants, and assent was also received from the participants.
Fig. 1. Sample sizes of different Hg measures.
2.2. Mercury analysis
2.2.1. ELEMENT 2015
Total mercury was measured in urine (a biomarker that typically reflects elemental and inorganic mercury exposures) using a Direct Mercury Analyzer 80 (DMA-80, Milestone Inc., CT) as previously described (Paruchuri et al., 2010).
Mercury in urine was adjusted for concurrent specific gravity, as recommended in the literature (Sauvé et al., 2015). The adjusted urine mercury measurement for subject i is calculated as the original urine mercury measurement for subject i multiplied by 1000*(median-1)/[1000*(specific gravity-1)], where median is the median specific gravity for all samples in the 2008 or 2015 study visit, and specific gravity is the specific gravity measurement for subject i.
2.2.2. ELEMENT 2008
Total mercury was also measured in blood and hair samples taken at two visits in the ELEMENT 2008 follow-up visit, 18 months apart(Basu et al., 2014). Hair and blood are typically reflective of exposure to organic/methylmercury derived from seafood consumption with hair reflecting longer-term exposure than blood. Urine mercury was only measured once at the ELEMENT 2008 visit, and was adjusted for specific gravity as described above.
Due to the proximity of the two measurements in blood and hair from ELEMENT 2008, we constructed one blood and one hair mercury measurement for childhood for each subject by averaging both visit measurements if they were both observed, and retaining the observed value if one was missing. Among adolescents with data from both timepoints, the correlations were 0.56 and 0.49 for blood and hair, respectively. Further, sensitivity analysis that used each timepoint separately did not show substantial differences.
To reduce the influence of outliers and to evaluate the potential nonlinearity of the associations between mercury and sleep, we discretized each of the mercury measurements and created six indicator variables. For each subject i’s mercury measurement Hgi, the jth indicator variable corresponds to 1(quantilej-1 ≤ Hgi < quantilej) for the jth percentile in (0.15, 0.3, 0.45, 0.6, 0.75, 1), j=1, …, 6.
2.3. Sleep Measures
The sleep outcome measures were sleep duration, sleep midpoint, and sleep fragmentation. At the ELEMENT 2015 visit, adolescents were given an actigraph (ActiGraph GT3X+; ActiGraph LLC, Pensacola, FL) to wear on their non-dominant wrist continuously for 7 days. Nightly sleep duration was estimated from the actigraphic and sleep diary data with the use of a fused lasso (least absolute shrinkage and selection operator)-based calculator package developed in R (R Foundation for Statistical Computing, Vienna, Austria). The obtained estimates were highly correlated with manual sleep duration detection in a subset of 50 randomly selected participants (r=0.95). We examined sleep duration (averaged over the wear period), as well as one measure of sleep quality: sleep fragmentation index. In accordance with the Actilife software (ActiGraph LLC, Pensacola, FL) and used in previous research(Chung et al., 2016), sleep fragmentation index was calculated as the percentage of one-minute (or shorter) periods of sleep out of the total number of sleep bouts of any length, with higher values representing more fragmented sleep. The sleep versus wake episodes (during the previously determined night-time sleep duration window) were identified with the Sadeh algorithm.(Sadeh et al., 1994)
2.4. Potential confounders
Potential confounders examined included sex, age, BMI for age Z scores, maternal education, socioeconomic status, pubertal status, and docosahexaenoic acid levels (marker of fatty fish intake and related in our prior work to sleep timing and duration (Jansen et al., 2020)). All confounders were assessed during the ELEMENT 2015 visit, except for maternal education (from the original recruitment visit). Age was categorized into: 9.5 to <12 years, 12 to <14 years, 14 to <16 years, and 16 to 18 years. Trained research assistants measured height (in cm, Tonelli E120 A) and weight (in kg, Inbody 370 Biospace Co, Ltd, South Korea) using standard protocols. BMI-for-age Z scores were calculated based on the World Health Organization reference,(de Onis et al., 2007) and divided into: <0, 0 to <1, 1 to <2, and ≥2. Maternal education was abstracted from questionnaires mothers completed during the original enrollment visit and categorized as <9 years, 9 to <12 years, 12 years, or >12 years. At ELEMENT 2015, adolescents (with input from mothers when available) answered questions regarding family socioeconomic status. This questionnaire was created by the Asociación Mexicana de Agencias de Investigación de Mercados y Opinión Pública (Mexican Association of Marketing Research and Public Opinion Agencies) to evaluate household resources, including education of the household head, number of rooms in the house, number of vehicles owned, ownership of particular appliances (e.g. microwave and washing machine), and access to internet. The index consists of 7 categories ranging from A/B (highest SES) to E (lowest SES). Finally, pubertal status, assessed by Tanner staging and testicular volume assessment (for boys), was completed by trained physicians during the visit using standard methods(Chavarro et al., 2017). Girls were also asked whether they had started menstruating. To assess pubertal status, we classified participants into those who had reached the latter stages of puberty by the visit (testicular volume ≥15 mm for boys and onset of menarche for girls) and those who had not. Plasma docosahexaenoic acid was measured as described previously and was analyzed as a percentage of total fatty acids (Jansen et al., 2020).
2.5. Statistical Analysis
Analyses of the primary aim (2015 mercury exposure) and secondary aim (2008 mercury exposures) proceeded in the same manner. We first examined pairwise associations of mercury exposure (each separately) with sociodemographic and anthropometric characteristics to identify potential confounders. We then regressed each weekday sleep characteristic – weekday sleep duration, weekday sleep midpoint, and averaged estimated fragmentation index – on each discretized mercury exposure with five indicator variables, in which the first category (the lowest Hg) was treated as the reference level, adjusting for confounders. The final set of confounders included in the models were sex, age (as a continuous variable), and socioeconomic status, as addition of the remaining potential confounders did not alter the findings and thus were not retained. These models were fit on the full dataset (not stratified by sex), as well as on data stratified by sex to investigate sex-specific associations between mercury and sleep. Second, we ran the same set of models, but with continuous log-transformed Hg values. Finally, we explored the associations between sleep characteristics and continuous concurrent urine mercury and mid-childhood hair, blood, and urine mercury exposures through varying coefficient models(Chambers, 2017). Varying-coefficient models, in contrast to regression, do not give fixed coefficient values to interpret constant direction of association between variables, but enable us to detect non-linear associations and interactions. As a conservative approach to interpretation, we only consider the Hg and sleep relationships with statistically significant findings from the varying coefficient models to be meaningful (since they capture overall associations, both linear and non-linear), and we then use the linear regression results to interpret the direction of the associations. All statistical analyses were run using R software (version 3.6), and the varying coefficient models were fit using the gam function in the mgcv R package.
3. RESULTS
During the ELEMENT 2015 visit, participants were 13.3 ± 1.9 years, and 48% were male. Median (IQR) urine Hg measured at the ELEMENT 2015 visit was 0.44 (0.47) ug/L. Higher urine mercury was associated with higher socioeconomic status.
Average weekday sleep duration (SD) measured in the ELEMENT 2015 visit was 507.27 (71.3) minutes, while weekend sleep duration was 539.67 (69.04) minutes. The average sleep midpoint was 3.77 (1.32) hours, and during the weekend it was 4.71 (1.28) hours. Average sleep fragmentation did not differ from weekdays to weekends, and was 11.96 (3.71) %.
3.1. Concurrent Hg and sleep in adolescence
There were no overall associations between concurrent urine Hg and sleep characteristics, neither in unstratified nor sex-stratified analyses (Tables 2A–2C).
Table 2 a).
Cross-sectional associations between mercury levels in urine at P01 and actigraphy-assessed sleep characteristics in a sample of 371 Mexican youth
| N | Weekday Duration (minutes)1, Median (IQR) | Adjusted Difference2 | Weekday Midpoint (decimal hour)3, Median (IQR) | Adjusted Difference2 | Fragmentation Index4, Median (IQR) | Adjusted Difference2 | |
|---|---|---|---|---|---|---|---|
| Hg, urine in P01 | 371 | 508.4 (92.5) | 3.4 (2.1) | 11.5 (5) | |||
| Q1 (median=0.18) | 56 | 487.5 (79.2) | Reference | 3.4 (2.1) | Reference | 11.5 (4.9) | Reference |
| Q2 (median=0.25) | 55 | 492 (102.7) | 13.8 (−12.4,40) | 3.8 (2.4) | 0.1 (−0.3,0.6) | 12.2 (5) | 0.1 (−1.3,1.5) |
| Q3 (median=0.34) | 56 | 497.2 (86.2) | 3.7 (−22.3,29.8) | 3.4 (1.9) | 0 (−0.5,0.4) | 12 (5.4) | 0 (−1.3,1.4) |
| Q4 (median=0.45) | 55 | 510.4 (109.2) | 12.2 (−14,38.3) | 3.5 (1.8) | 0.2 (−0.2,0.7) | 10.6 (5) | −0.6 (−1.9,0.8) |
| Q5 (median=0.57) | 56 | 530.3 (81.1) | 36.4 (10.1,62.6)* | 3.5 (2.1) | 0.3 (−0.2,0.7) | 11.5 (4) | 0.1 (−1.3,1.5) |
| Q6 (median=1.09) | 93 | 510.2 (101.8) | 21.3 (−2,44.6) | 3.5 (2.1) | 0.2 (−0.2,0.6) | 11.7 (4.7) | 0 (−1.2,1.2) |
| β, log transformed5 | 10.6 (−7,28.3) | −0.1 (−0.4,0.3) | 0 (−1,0.9) | ||||
| P value for VCM6 | 0.18 | 0.27 | 0.32 | ||||
Sleep duration averaged over the weekday nights (Sunday through Thursday)
From linear regression models with sleep characteristic (separately for duration, midpoint, and fragmentation index) as the dependent variable, indicator variables for quartiles of mercury as the independent predictor; and sex, continuous age, and socioeconomic status.
Sleep midpoint (median of bedtime and wake time) averaged over the weekdays
Calculated as the percentage of one-minute periods of sleep out of total number of sleep bouts of any length, averaged over the 7-day wear time
From linear regression models with log transformed Hg exposure.
VCM= varying coefficient model; VCM= varying coefficient model; tests the significance of an overall association and allows for non-linear associations
Table 2 c).
Cross-sectional associations between mercury levels in urine at P01 and actigraphy-assessed sleep characteristics in a sample of 180 Mexican boys
| N | Weekday Duration (minutes)1, Median (IQR) | Adjusted Difference2 | Weekday Midpoint (decimal hour)3, Median (IQR) | Adjusted Difference2 | Fragmentation Index4, Median (IQR) | Adjusted Difference2 | |
|---|---|---|---|---|---|---|---|
| Hg, urine in P01 | 180 | 500.4 (82.3) | 3.6 (2.2) | 12 (5.6) | |||
| Q1 (median=0.19) | 22 | 481.2 (86.4) | Reference | 3.2 (2.1) | Reference | 11.6 (4.8) | Reference |
| Q2 (median=0.25) | 32 | 482.4 (92.8) | 14.2 (−21.4,49.8) | 3.1 (2.3) | 0.2 (−0.5,0.9) | 12.9 (5.6) | 1.2 (−0.9,3.4) |
| Q3 (median=0.34) | 30 | 499.6 (78.3) | 14.5 (−21.7,50.7) | 3.3 (2.3) | 0.2 (−0.6,0.9) | 11.9 (5.9) | 1.2 (−1,3.4) |
| Q4 (median=0.45) | 24 | 513.1 (90.8) | 29.6 (−8.4,67.7) | 3.8 (2.1) | 0.5 (−0.3,1.3) | 10.8 (6.7) | 1.4 (−0.9,3.7) |
| Q5 (median=0.57) | 29 | 523.6 (82.2) | 35.2 (−1.3,71.7) | 3.8 (2.5) | 0.6 (−0.2,1.3) | 12.1 (4.7) | 1.8 (−0.4,4.1) |
| Q6 (median=1.18) | 43 | 496.6 (83.4) | 19.9 (−14.2,53.9) | 3.7 (1.8) | 0.2 (−0.5,0.9) | 12.6 (5.9) | 1.1 (−1,3.2) |
| β, log transformed5 | 4.1 (−18.8,27) | −0.2 (−0.6,0.3) | 0.1 (−1.3,1.5) | ||||
| P value for VCM6 | 0.24 | 0.19 | 0.18 | ||||
Sleep duration averaged over the weekday nights (Sunday through Thursday)
From linear regression models with sleep characteristic (separately for duration, midpoint, and fragmentation index) as the dependent variable, indicator variables for quartiles of mercury as the independent predictor; and continuous age, and socioeconomic status.
Sleep midpoint (median of bedtime and wake time) averaged over the weekdays
Calculated as the percentage of one-minute periods of sleep out of total number of sleep bouts of any length, averaged over the 7-day wear time
From linear regression models with log transformed Hg exposure.
VCM= varying coefficient model; tests the significance of an overall association and allows for non-linear associations
3.2. Mid-childhood Hg and sleep in adolescence
In analysis that included both males and females, there was little evidence of an association between 2008 Hg measures and 2015 sleep characteristics. A non-linear association between urine Hg and sleep fragmentation was noted from the varying coefficient model, but neither of the linear models confirmed any statistically significant differences.
In sex-stratified analysis, there was evidence for a non-linear positive association between blood mercury and sleep duration among girls (Table 3B; varying coefficient P value=0.02). Compared to girls with the lowest and highest levels of blood Hg, those with moderate Hg levels (especially around 1 ug/L) had longer sleep duration. In addition, higher urine mercury was associated with later sleep midpoint, such that each unit log Hg was associated with 0.2 hour later midpoint (95% CI 0,0.4; varying coefficient P value=0.02; Table 3B).
Table 3 b).
Associations between mercury levels and actigraphy-assessed sleep characteristics in a sample of Mexican girls
| N | Weekday Duration (minutes)1, Median (IQR) | Adjusted Difference2 | Weekday Midpoint (decimal hour)3, Median (IQR) | Adjusted Difference2 | Fragmentation Index4, Median (IQR) | Adjusted Difference2 | |
|---|---|---|---|---|---|---|---|
| Hg, blood | 113 | 508.2 (123.2) | 3.4 (1.9) | 11.4 (4) | |||
| Q1 (median=0.7) | 19 | 499.4 (104.6) | Reference | 3.3 (1.7) | Reference | 11.4 (3.7) | Reference |
| Q2 (median=1) | 16 | 581 (69.9) | 82.3 (28.4,136.2)* | 4.7 (1.9) | 0.7 (−0.1,1.5) | 12 (4.1) | 0.6 (−1.8,3) |
| Q3 (median=1.4) | 19 | 501 (94.7) | 16.2 (−34.6,67.1) | 3.8 (1.6) | 0.2 (−0.5,1) | 11.2 (4.5) | −0.2 (−2.5,2) |
| Q4 (median=1.6) | 14 | 494.3 (95.1) | 5.1 (−48.8,58.9) | 3.2 (1.2) | −0.3 (−1.1,0.5) | 11.6 (3.9) | −0.1 (−2.4,2.3) |
| Q5 (median=2.3) | 16 | 496.9 (103.4) | 19.7 (−32.3,71.7) | 3.2 (1.3) | 0 (−0.8,0.8) | 11.5 (5.2) | −0.9 (−3.2,1.4) |
| Q6 (median=3.6) | 29 | 501.4 (124.7) | −1.7 (−46.8,43.4) | 3.2 (1.7) | 0.2 (−0.5,0.8) | 11.2 (3.2) | −0.2 (−2.2,1.8) |
| β, log transformed5 | −8.4 (−28.7,12) | −0.1 (−0.4,0.2) | −0.2 (−1,0.7) | ||||
| P value for VCM6 | 0.027 | 0.63 | 0.79 | ||||
| Hg, hair | 144 | 502 (110.3) | 3 (1.9) | 11 (4.3) | |||
| Q1 (median=0.2) | 20 | 525.3 (59.3) | Reference | 3.7 (1.8) | Reference | 11.4 (3.7) | Reference |
| Q2 (median=0.3) | 19 | 548.8 (113.7) | 1.4 (−47.7,50.5) | 3.6 (2.3) | −0.3 (−1.1,0.4) | 12.4 (3.8) | −0.6 (−2.7,1.4) |
| Q3 (median=0.4) | 22 | 523.2 (131.3) | −10.8 (−58.1,36.4) | 3.2 (2.3) | −0.5 (−1.2,0.2) | 10.3 (5.3) | −1.7 (−3.6,0.3) |
| Q4 (median=0.5) | 24 | 497.6 (76.9) | −27.5 (−73.8,18.9) | 3.5 (2.1) | −0.2 (−0.9,0.5) | 11.9 (4.2) | −1 (−2.9,1) |
| Q5 (median=0.6) | 20 | 487 (129) | −50.9 (−99.2,−2.5)* | 3 (0.9) | −0.7 (−1.4,0.1) | 9.5 (3.8) | −2.8 (−4.9,−0.8)* |
| Q6 (median=1.1) | 39 | 492.2 (117.5) | −16.3 (−58.4,25.7) | 3.3 (1.9) | −0.3 (−0.9,0.4) | 12 (3.9) | −0.4 (−2.1,1.4) |
| β, log transformed5 | −9.6 (−27.3,8.2) | −0.1 (−0.3,0.2) | −0.2 (−1,0.5) | ||||
| P value for VCM6 | 0.12 | 0.70 | 0.33 | ||||
| Hg, urine | 139 | 502 (112) | 3 (2) | 11 (4.2) | |||
| Q1 (median=0.2) | 16 | 475 (87.7) | Reference | 3.1 (1.3) | Reference | 11 (3.3) | Reference |
| Q2 (median=0.3) | 23 | 499.4 (116) | 8.4 (−42.6,59.5) | 3.4 (1.2) | 0.3 (−0.5,1) | 12.3 (7.1) | 0.1 (−2.1,2.3) |
| Q3 (median=0.4) | 20 | 488.4 (116.1) | 9.8 (−42.2,61.7) | 3.4 (2.8) | 0.4 (−0.3,1.2) | 12.2 (5.2) | −0.1 (−2.3,2.1) |
| Q4 (median=0.6) | 18 | 516.3 (70.5) | 11.3 (−42.3,64.8) | 3.7 (2.1) | 0.7 (−0.1,1.5) | 10.8 (4.9) | −0.7 (−3,1.6) |
| Q5 (median=0.8) | 25 | 549 (107.6) | 50 (0,100.1)* | 3.4 (1.4) | 0.2 (−0.5,1) | 12.8 (3.9) | 0.8 (−1.4,2.9) |
| Q6 (median=1.9) | 37 | 508.6 (130.4) | 24 (−23,70.9) | 3.6 (2.3) | 0.7 (0,1.4) | 10.8 (3) | −1 (−3,1) |
| β, log transformed5 | 6 (−7.9,19.8) | 0.2 (0,0.4) | −0.3 (−0.9,0.2) | ||||
| P value for VCM6 | 0.40 | 0.028 | 0.07 | ||||
Sleep duration averaged over the weekday nights (Sunday through Thursday)
From linear regression models with sleep characteristic (separately for duration, midpoint, and fragmentation index) as the dependent variable, indicator variables for quartiles of mercury as the independent predictor; and continuous age and socioeconomic status.
Sleep midpoint (median of bedtime and wake time) averaged over the weekdays
Calculated as the percentage of one-minute periods of sleep out of total number of sleep bouts of any length, averaged over the 7-day wear time
From linear regression models with log transformed Hg exposure.
VCM= varying coefficient model; tests the significance of an overall association and allows for non-linear associations
Degrees of freedom= 6
Degrees of freedom= 1
Among boys, there was evidence of an association between higher blood mercury and later weekday midpoint, such that those with higher blood mercury had later sleep midpoints, such that each unit log Hg was associated with an 0.4 hour later sleep midpoint (95% CI 0.1, 0.8; varying coefficient P value=0.04).
In sensitivity analysis, additional adjustment for concurrent blood lead (Pb) or recent fish intake (consumption of the most common fish sources- tuna, shrimp, and schoolshark- in the previous 7 days) did not alter any of the findings.
4. DISCUSSION
In a cohort of Mexican adolescents, there was little evidence of associations between Hg levels in biomarkers during mid-childhood and adolescence with sleep during adolescence, although a few notable relationships existed. The primary analysis showed no associations between concurrent urine Hg levels and sleep characteristics. However, in secondary analysis that used mercury measurements from mid-childhood, a non-linear association existed between mercury measured in blood and weekday sleep duration among girls, with the longest sleep duration at mid-levels of Hg. In addition, higher mid-childhood Hg was associated with later adolescent sleep timing in both boys and girls, although in different biomarkers (urine for girls, and blood for boys).
Only one prior study examined low-level exposures of mercury measured in blood (which reflects exposure largely to organic methylmercury derived from seafood consumption) in relation to sleep duration in a pediatric population. In this cross-sectional survey of 100 US children with average Hg levels of 0.46 μg/L, authors found that higher levels of blood Hg were linearly related to objectively-measured shorter sleep duration (Gump et al., 2014). In our primary analysis, we found no association between concurrent Hg and sleep duration, although the Hg was measured in urine. Urine Hg is typically reflective of elemental/inorganic sources of Hg exposure, though methylmercury from seafood consumption can contribute to some of the total Hg in urine among frequent fish consumers (Basu et al., 2018; Sherman et al., 2013). Interestingly, our secondary analysis revealed a non-linear association between blood Hg and sleep duration among girls, such that girls with both low levels of Hg exposure and higher levels of Hg exposures had the shortest sleep duration. It is possible that our mostly null but also non-linear findings for Hg exposure and sleep could potentially reflect opposing effects of Hg on sleep duration, including both fatigue and sleepiness (which may cause greater nighttime sleep duration) and insomnia symptoms (causing shorter sleep duration). One case report in Poland that included 15 adults and 7 children who were exposed to mercury vapors (inorganic mercury and thus consistent with higher urinary Hg) documented that the most common complaints were fatigue/weariness and excessive sleepiness, and that children were more sensitive than adults (Kasznia-Kocot et al., 2010). In contrast, although not directly comparable to our study population of youth with low-level mercury exposures, associations between high levels of mercury exposure and sleep disturbances including insomnia (which would likely be related to shorter sleep duration) have been previously documented in adult populations exposed to mercury through occupation. For example, mercury exposure through waste removal jobs (primarily elemental mercury exposure) in Thailand (Decharat, 2018) and Korea(Do et al., 2017) has been related to insomnia, fatigue, and sleep disturbances. Furthermore, a small randomized trial of Ecuadorian adults working in gold mines (exposed to elemental mercury) showed that men who took the chelator N,N’bis-(2-mercaptoethyl) isophthalamide experienced a reduction in fatigue and in self-reported sleeping problems relative to the control group (Schutzmeier et al., 2018). As an alternative explanation for our null findings, it is entirely plausible that the levels of mercury experienced within this population was not high enough for effects on sleep duration.
A more consistent finding was that there were positive associations between mid-childhood Hg levels and sleep midpoint assessed in adolescence in both boys and girls (blood Hg and urine Hg, respectively). Although the prior US study did not examine sleep timing, an association between higher Hg and delayed sleep timing has biological plausibility. In particular, mercury has been shown to accumulate in the pineal gland(Falnoga et al., 2000), where the timing of sleep onset is regulated by melatonin. Melatonin is also thought to play a role in the onset of puberty(Macchi and Bruce, 2004), a time period during which sleep timing patterns change substantially. Thus, if melatonin is involved in the Hg-sleep timing pathway, pubertal timing may be a plausible mediator. Sleep/wake disturbances have been observed in animal studies and among adult populations with high levels of exposure/intoxication (Parmalee and Aschner, 2017).
Overall, the fact that we had some mixed findings across different biomarkers could mean that effects depend on the type of mercury exposure. In general, Hg is the most commonly used biomarker that reflects elemental mercury exposure from sources such as amalgams and approximately 80% of elemental Hg vapor from dental amalgams is absorbed through inhalation (Sherman et al., 2013). Blood Hg levels are correlated with recently ingested methylmercury from fish consumption(Basu et al., 2018), and hair Hg reflects longer term exposure to methylmercury from fish. In the Mexican population, the main seafood sources of methylmercury are canned tuna and schoolshark (Cantoral et al., 2017).
Across the different measures of Hg, our non-null findings were sex-specific. Altogether, the fact that there were sex-specific findings is not surprising, as endocrine-disrupting toxicants like mercury may interact with female and male sex hormones in distinct manners, and males and females progress through puberty at different rates. Furthermore, sex steroid receptors have been found in the pineal gland(Luboshitzky and Lavie, 1999), suggesting a role of sex steroids in melatonin production. In addition, girls secrete more melatonin than boys during puberty(Crowley et al., 2012); thus, it is possible that boys may be more sensitive to a melatonin-related pathways. Future studies with mechanistic biomarkers (e.g. melatonin) are needed to explore these pathways.
Our study had both strengths and limitations to consider. One of the primary strengths is a much larger sample size than previous studies. Second is the fact that we had multiple measurements of Hg exposure- in hair, urine, and blood- and repeated measures over time. Finally, we measured sleep with a wearable device over a 7-day period rather than relying on self-reported sleep behaviors, which is prone to measurement error. One of the limitations is that sleep was not assessed during the ELEMENT 2008 visit in order to examine concurrent 2008 Hg levels and sleep, or to examine change in sleep over time. Second, funding did not allow us to examine Hg in blood or hair during the adolescent visit. Generalizability may be limited to children of low-middle socioeconomic status from Mexico City. Finally, there may be unmeasured or residual confounding from factors that are independent predictors of sleep behaviors as well as associated with mercury exposure, including other substances found in seafood. Nonetheless, controlling for fish intake and docosahexaenoic acid, an omega 3 fatty acid found in fish, did not alter the results.
In summary, in a population with low-moderate levels of mercury exposure, we found little evidence of associations between Hg and sleep characteristics among adolescents, except for some indication of a potential association between Hg levels and delayed sleep timing in both boys and girls. The findings may be relevant for children (especially girls) with dental amalgams and/for those with consumption of fish with moderate to high levels of mercury (e.g., tuna). Ultimately, replication of these findings in other populations with relatively low levels of intake is needed before conclusions can be made.
Supplementary Material
Table 1.
Mercury levels of 372 youth from Mexico City, according to sociodemographic and lifestyle characteristics
| Sociodemographic and lifestyle characteristics | N | Hg, urine, 2015 |
|---|---|---|
| Sex | ||
| Male | 180 | 0.66 (0.78) |
| Female | 192 | 0.59 (0.58) |
| P value1 | 0.85 | |
| Maternal education, years (y) | ||
| 8 y or less (secondary or primary) | 42 | 0.73 (0.7) |
| 9 to 11 y (some high school) | 149 | 0.55 (0.57) |
| 12 y (completed high school) | 131 | 0.66 (0.84) |
| >12 y | 50 | 0.67 (0.51) |
| P, trend | 0.65 | |
| Age group, years (y) | ||
| 9 to <12 y | 85 | 0.73 (0.78) |
| 12 to <14 y | 141 | 0.58 (0.51) |
| 14 to <16 y | 78 | 0.59 (0.83) |
| 16 to 17 y | 68 | 0.63 (0.68) |
| P, trend | 0.41 | |
| BMI for age Z score | ||
| <0 | 130 | 0.66 (0.73) |
| 0 to <1 | 102 | 0.51 (0.44) |
| 1 to <2 | 87 | 0.72 (0.83) |
| ≥2 | 53 | 0.6 (0.68) |
| P, trend | 0.99 | |
| Puberty status | ||
| Has not reached puberty | 77 | 0.69 (0.79) |
| Has reached puberty | 287 | 0.6 (0.65) |
| P value | 0.35 | |
| Socioeconomic status (AMAI) | ||
| Quartile 1- “E” (lowest) | 40 | 0.57 (0.45) |
| Quartile 2- “D” or “D+” | 131 | 0.54 (0.52) |
| Quartile 3- “C or C+” | 176 | 0.68 (0.81) |
| Quartile 4- “A/B” (highest) | 25 | 0.79 (0.75) |
| P, trend | 0.05 | |
| Plasma DHA, % of total fatty acids | ||
| Quartile 1 | 80 | 0.59 (0.65) |
| Quartile 2 | 79 | 0.57 (0.6) |
| Quartile 3 | 79 | 0.59 (0.49) |
| Quartile 4 | 80 | 0.85 (0.97) |
| P, trend | 0.03 | |
For dichotomous characteristics, P values are from Wilcoxon tests. For ordinal characteristics, P for trends are from linear regression models with plasma fatty acid as the dependent variable and a continuous variable representing ordinal categories of the sociodemographic or lifestyle predictor as the independent variable
Table 2 b).
Cross-sectional associations between mercury levels in urine at P01 and actigraphy-assessed sleep characteristics in a sample of 191 Mexican girls
| N | Weekday Duration (minutes)1, Median (IQR) | Adjusted Difference2 | Weekday Midpoint (decimal hour)3, Median (IQR) | Adjusted Difference2 | Fragmentation Index4, Median (IQR) | Adjusted Difference2 | |
|---|---|---|---|---|---|---|---|
| Hg, urine in P01 | 191 | 510.2 (95.9) | 3.4 (1.9) | 11.4 (4.3) | |||
| Q1 (median=0.16) | 34 | 497.2 (68.2) | Reference | 3.4 (2.1) | Reference | 11.4 (5.4) | Reference |
| Q2 (median=0.25) | 23 | 516.6 (104.6) | 16.3 (−23.1,55.6) | 4.4 (2.3) | 0.1 (−0.5,0.8) | 12 (4.2) | −0.5 (−2.3,1.2) |
| Q3 (median=0.34) | 26 | 488.5 (96.4) | −4.9 (−42.9,33.1) | 3.4 (1.1) | −0.2 (−0.8,0.4) | 12 (4.3) | −0.6 (−2.3,1.1) |
| Q4 (median=0.45) | 31 | 495.8 (110.6) | −1.3 (−37.6,35) | 3.4 (1.2) | 0.1 (−0.5,0.7) | 9.8 (4.7) | −1.9 (−3.5,−0.3)* |
| Q5 (median=0.58) | 27 | 541.4 (80.9) | 38.2 (0.1,76.2)* | 3.4 (1.7) | 0 (−0.6,0.6) | 11 (3.5) | −1.3 (−3,0.4) |
| Q6 (median=0.95) | 50 | 517.2 (98.9) | 24.9 (−7.5,57.4) | 3.3 (2.7) | 0.2 (−0.3,0.8) | 11.5 (3.9) | −0.7 (−2.1,0.8) |
| β, log transformed5 | 19.6 (−7.7,46.8) | 0.1 (−0.3,0.6) | −0.2 (−1.4,1) | ||||
| P value for VCM6 | 0.24 | 0.94 | 0.18 | ||||
Sleep duration averaged over the weekday nights (Sunday through Thursday)
From linear regression models with sleep characteristic (separately for duration, midpoint, and fragmentation index) as the dependent variable, indicator variables for quartiles of mercury as the independent predictor; and continuous age, and socioeconomic status.
Sleep midpoint (median of bedtime and wake time) averaged over the weekdays
Calculated as the percentage of one-minute periods of sleep out of total number of sleep bouts of any length, averaged over the 7-day wear time
From linear regression models with log transformed Hg exposure.
VCM= varying coefficient model; tests the significance of an overall association and allows for non-linear associations
Table 3 a).
Associations between mercury levels and actigraphy-assessed sleep characteristics in a sample of Mexican youth
| N | Weekday Duration (minutes)1, Median (IQR) | Adjusted Difference2 | Weekday Midpoint (decimal hour)3, Median (IQR) | Adjusted Difference2 | Fragmentation Index4, Median (IQR) | Adjusted Difference2 | |
|---|---|---|---|---|---|---|---|
| Hg, blood | 222 | 504.3 (98) | 3.6 (2.1) | 11.6 (4.8) | |||
| Q1 (median=0.6) | 34 | 511.3 (83.7) | Reference | 3.2 (1.7) | Reference | 11.4 (4) | Reference |
| Q2 (median=0.9) | 33 | 508.6 (116.6) | 12.3 (−22.5,47) | 4.2 (2.5) | 0.5 (−0.1,1) | 11.7 (4.8) | 0 (−1.7,1.8) |
| Q3 (median=1.3) | 33 | 489.2 (97.8) | −6.2 (−40.8,28.5) | 4.2 (2.1) | 0.6 (0,1.2)* | 11.6 (4.3) | 0.1 (−1.6,1.9) |
| Q4 (median=1.7) | 33 | 496.6 (83.6) | −9.3 (−44.1,25.5) | 3.2 (2.4) | 0.3 (−0.3,0.9) | 11.8 (5.4) | −0.1 (−1.8,1.7) |
| Q5 (median=2.1) | 33 | 501.6 (70.4) | 1.7 (−33,36.4) | 3.8 (1.7) | 0.5 (−0.1,1.1) | 11.8 (5.4) | −0.4 (−2.2,1.3) |
| Q6 (median=3.4) | 56 | 510 (107.8) | −3.8 (−34.6,27) | 3.7 (1.9) | 0.4 (−0.1,1) | 11.8 (4.8) | 0.2 (−1.3,1.8) |
| β, log transformed5 | −4 (−17.9,9.9) | 0.1 (−0.1,0.4) | 0 (−0.7,0.7) | ||||
| P value for VCM6 | 0.88 | 0.08 | 0.96 | ||||
| Hg, hair | 274 | 504 (94.6) | 3 (2.2) | 12 (5) | |||
| Q1 (median=0.2) | 41 | 510.2 (79.6) | Reference | 3.4 (2) | Reference | 11.5 (6.9) | Reference |
| Q2 (median=0.3) | 41 | 526.6 (89.4) | 17.4 (−13,47.8) | 3.6 (2.5) | 0 (−0.6,0.5) | 12.3 (4.1) | 0.3 (−1.3,1.9) |
| Q3 (median=0.4) | 41 | 484.6 (96.2) | −8.1 (−38.5,22.3) | 4.3 (2.4) | 0.2 (−0.3,0.8) | 11.2 (4.9) | −0.8 (−2.3,0.8) |
| Q4 (median=0.5) | 41 | 493.8 (81.4) | −24.4 (−55,6.2) | 3.4 (2.3) | 0 (−0.5,0.6) | 12.1 (4.8) | −0.3 (−1.8,1.3) |
| Q5 (median=0.6) | 41 | 510.4 (89.7) | −21.4 (−52,9.1) | 3.4 (2) | −0.1 (−0.6,0.5) | 11.1 (4.7) | −1.4 (−3,0.1) |
| Q6 (median=1.1) | 69 | 508.4 (93) | −1.6 (−28.8,25.5) | 3.4 (1.9) | 0 (−0.4,0.5) | 12 (5.3) | −0.1 (−1.5,1.3) |
| β, log transformed5 | −5.9 (−17.5,5.7) | 0 (−0.2,0.2) | −0.3 (−0.9,0.3) | ||||
| P value for VCM6 | 0.23 | 0.60 | 0.78 | ||||
| Hg, urine | 268 | 508 (97) | 3 (2.1) | 12 (5) | |||
| Q1 (median=0.2) | 41 | 491.6 (100.3) | Reference | 3.4 (2.3) | Reference | 11.4 (4.2) | Reference |
| Q2 (median=0.3) | 40 | 509.5 (103.5) | −2.6 (−33.8,28.6) | 3.4 (1.7) | −0.3 (−0.8,0.3) | 13.1 (6.3) | 0.3 (−1.3,1.9) |
| Q3 (median=0.4) | 40 | 491.2 (112.9) | −8.4 (−39.4,22.7) | 3.2 (2.4) | −0.2 (−0.7,0.4) | 11.3 (4) | −0.9 (−2.5,0.7) |
| Q4 (median=0.6) | 40 | 494.1 (80.8) | −13.4 (−44.4,17.6) | 3.2 (2.6) | −0.1 (−0.7,0.4) | 12 (5.3) | −0.5 (−2.1,1.1) |
| Q5 (median=0.8) | 40 | 512.1 (92.4) | 20 (−11.4,51.4) | 3.5 (1.6) | −0.2 (−0.8,0.3) | 12.8 (4.7) | 0.5 (−1.1,2.2) |
| Q6 (median=1.8) | 67 | 514.8 (80.3) | 10.1 (−17.7,38) | 3.8 (2.1) | 0.1 (−0.4,0.6) | 11.4 (4.9) | −0.3 (−1.7,1.2) |
| β, log transformed5 | 2.9 (−6.4,12.2) | 0.1 (−0.1,0.2) | −0.1 (−0.5,0.4) | ||||
| P value for VCM6 | 0.42 | 0.11 | 0.037 | ||||
Sleep duration averaged over the weekday nights (Sunday through Thursday)
From linear regression models with sleep characteristic (separately for duration, midpoint, and fragmentation index) as the dependent variable, indicator variables for quartiles of mercury as the independent predictor; and sex, continuous age, and socioeconomic status.
Sleep midpoint (median of bedtime and wake time) averaged over the weekdays
Calculated as the percentage of one-minute periods of sleep out of total number of sleep bouts of any length, averaged over the 7-day wear time
From linear regression models with log transformed Hg exposure.
VCM= varying coefficient model; tests the significance of an overall association and allows for non-linear associations
Degrees of freedom= 4
Table 3 c).
Associations between mercury levels and actigraphy-assessed sleep characteristics in a sample of Mexican boys
| N | Weekday Duration (minutes)1, Median (IQR) | Adjusted Difference2 | Weekday Midpoint (decimal hour)3, Median (IQR) | Adjusted Difference2 | Fragmentation Index4, Median (IQR) | Adjusted Difference2 | |
|---|---|---|---|---|---|---|---|
| Hg, blood | 109 | 499.4 (73.8) | 3.9 (2.3) | 12.3 (5) | |||
| Q1 (median=0.6) | 15 | 526.6 (80.9) | Reference | 3.1 (1.7) | Reference | 11.5 (4) | Reference |
| Q2 (median=0.9) | 17 | 467.8 (45) | −48.7 (−91.4,−6)* | 3.1 (2.5) | 0.3 (−0.6,1.2) | 11 (5.2) | 0 (−2.7,2.7) |
| Q3 (median=1.3) | 14 | 480.9 (95.6) | −25.6 (−70.5,19.3) | 4.6 (2.3) | 1 (0.1,2)* | 12.7 (3.8) | 1.1 (−1.8,3.9) |
| Q4 (median=1.7) | 19 | 496.6 (64.1) | −24 (−66.4,18.5) | 3.3 (3.2) | 1 (0.1,1.8)* | 12.1 (4.3) | 0.3 (−2.4,3) |
| Q5 (median=2.1) | 17 | 510.4 (59) | −17.2 (−60.3,26) | 4.2 (1.4) | 1.1 (0.2,1.9)* | 11.8 (5.3) | 0.4 (−2.3,3.1) |
| Q6 (median=3.2) | 27 | 515.8 (73.8) | −7.8 (−46.5,30.9) | 4.3 (1.9) | 0.8 (0,1.6) | 13.4 (5) | 0.8 (−1.6,3.3) |
| β, log transformed5 | 3 (−15.6,21.6) | 0.4 (0.1,0.8)* | 0.2 (−0.9,1.4) | ||||
| P value for VCM6 | 0.36 | 0.047 | 0.85 | ||||
| Hg, hair | 130 | 509 (79.7) | 4 (2.3) | 12 (5.1) | |||
| Q1 (median=0.2) | 21 | 487.8 (82.5) | Reference | 3.1 (1.7) | Reference | 11.7 (8.5) | Reference |
| Q2 (median=0.3) | 22 | 523.2 (78.5) | 35.5 (−0.7,71.6) | 3.5 (2.5) | 0.3 (−0.5,1.1) | 11.7 (4) | 1.3 (−1.1,3.7) |
| Q3 (median=0.4) | 19 | 478 (25.5) | −8.9 (−46.4,28.6) | 4.9 (2.3) | 1 (0.2,1.8)* | 13 (3.9) | 0 (−2.5,2.5) |
| Q4 (median=0.4) | 17 | 464.4 (99) | −20 (−59.5,19.5) | 2.9 (1.5) | 0.1 (−0.7,1) | 13.6 (5) | 0.7 (−1.9,3.4) |
| Q5 (median=0.6) | 21 | 514.8 (64.9) | 8.1 (−29.1,45.3) | 4.4 (2.2) | 0.6 (−0.2,1.4) | 12.7 (3.5) | −0.1 (−2.6,2.3) |
| Q6 (median=1.1) | 30 | 514.2 (63.4) | 15.2 (−18.6,49.1) | 3.6 (1.8) | 0.3 (−0.4,1.1) | 12.4 (5.8) | 0 (−2.2,2.3) |
| β, log transformed5 | −1.4 (−16.2,13.3) | 0.1 (−0.2,0.4) | −0.4 (−1.3,0.6) | ||||
| P value for VCM6 | 0.74 | 0.64 | 0.48 | ||||
| Hg, urine | 129 | 509 (80.6) | 4 (2.3) | 13 (5.1) | |||
| Q1 (median=0.2) | 25 | 519.8 (98.6) | Reference | 4.1 (2.3) | Reference | 11.9 (5.8) | Reference |
| Q2 (median=0.3) | 17 | 514.4 (63.4) | −3.4 (−41.4,34.7) | 3.1 (1.8) | −0.6 (−1.4,0.2) | 14 (7) | 0.7 (−1.7,3.1) |
| Q3 (median=0.4) | 20 | 495.8 (95.1) | −21.5 (−58,14.9) | 3.1 (2.3) | −0.6 (−1.4,0.1) | 10.8 (2) | −1.8 (−4.1,0.6) |
| Q4 (median=0.6) | 22 | 477.7 (79.6) | −29.2 (−64.7,6.2) | 2.9 (2.5) | −0.7 (−1.4,0.1) | 13.3 (4.1) | −0.3 (−2.5,2) |
| Q5 (median=0.8) | 15 | 508.4 (38.1) | −15.2 (−55.2,24.8) | 4.2 (1.5) | −0.4 (−1.3,0.4) | 12.9 (4.4) | −0.1 (−2.7,2.4) |
| Q6 (median=1.7) | 30 | 530.4 (62.4) | 1.8 (−31,34.6) | 3.9 (2.1) | −0.3 (−1,0.4) | 13.1 (6.9) | 0.5 (−1.6,2.6) |
| β, log transformed5 | −1.7 (−13.9,10.5) | −0.1 (−0.4,0.1) | 0.3 (−0.5,1) | ||||
| P value for VCM6 | 0.82 | 0.52 | 0.15 | ||||
Sleep duration averaged over the weekday nights (Sunday through Thursday)
From linear regression models with sleep characteristic (separately for duration, midpoint, and fragmentation index) as the dependent variable, indicator variables for quartiles of mercury as the independent predictor; and continuous age and socioeconomic status.
Sleep midpoint (median of bedtime and wake time) averaged over the weekdays
Calculated as the percentage of one-minute periods of sleep out of total number of sleep bouts of any length, averaged over the 7-day wear time
From linear regression models with log transformed Hg exposure.
VCM= varying coefficient model; tests the significance of an overall association and allows for non-linear associations
Degrees of freedom=2
HIGHLIGHTS.
Whether low-level mercury exposure is related to poor sleep is an underexplored research area
In cross-sectional analyses, urine Hg was not related to sleep in adolescents
Yet, prospective analysis provided some evidence that Hg may be related to sleep
Prospective positive associations between Hg and sleep duration among girls were found
Prospective associations between Hg and later sleep timing for both boys and girls were noted
Acknowledgements:
We gratefully acknowledge the research staff and the British Cowdray Medical Center (ABC) for use of their research facilities.
Funding Source: This work was supported by the US Environmental Protection Agency (US EPA) grant RD83543601 and National Institute for Environmental Health Sciences grants P01 ES02284401, P30 ES017885, and R24ES02850. This study was also supported and partially funded by the National Institute of Public Health/Ministry of Health of Mexico.
AUTHOR CONTRIBUTIONS
EJ conceptualized the research question and wrote the original draft of the manuscript. EH conducted formal analysis and reviewed and edited the manuscript. JG, AC, MM, and NB aided in investigation and in review and editing. NTO provided project administration and reviewed and edited the manuscript PXKS provided supervision of analysis and reviewed and edited the manuscript. KEP was responsible for funding acquisition and review and editing.
Footnotes
Competing Interests: The authors declare no competing interests.
Declaration of interests
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
5. REFERENCES
- Basu N, Horvat M, Evers DC, Zastenskaya I, Weihe P, Tempowski J, 2018. A state-of-the-science review of mercury biomarkers in human populations worldwide between 2000 and 2018. Environ. Health Perspect. 126 10.1289/EHP3904 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Basu N, Tutino R, Zhang Z, Cantonwine DE, Goodrich JM, Somers EC, Rodriguez L, Schnaas L, Solano M, Mercado A, Peterson K, Sánchez BN, Hernández-Avila M, Hu H, Maria Téllez-Rojo M, 2014. Mercury levels in pregnant women, children, and seafood from Mexico City. Environ. Res. 135, 63–69. 10.1016/j.envres.2014.08.029 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bose-O’Reilly S, Schierl R, Nowak D, Siebert U, William JF, Owi FT, Ir YI, 2016. A preliminary study on health effects in villagers exposed to mercury in a small-scale artisanal gold mining area in Indonesia. Environ. Res. 149, 274–281. 10.1016/j.envres.2016.04.007 [DOI] [PubMed] [Google Scholar]
- Cantoral A, Batis C, Basu N, 2017. National estimation of seafood consumption in Mexico: Implications for exposure to methylmercury and polyunsaturated fatty acids. Chemosphere 174, 289–296. 10.1016/j.chemosphere.2017.01.109 [DOI] [PubMed] [Google Scholar]
- Chambers JM, 2017. Linear models, in: Statistical Models in S. pp. 95–144. 10.1201/9780203738535 [DOI] [Google Scholar]
- Chavarro JE, Watkins DJ, Afeiche MC, Zhang Z, Sánchez BN, Cantonwine D, Mercado-García A, Blank-Goldenberg C, Meeker JD, Téllez-Rojo MM, Peterson KE, 2017. Validity of Self-Assessed Sexual Maturation Against Physician Assessments and Hormone Levels. J. Pediatr. 186, 172–178.e3. 10.1016/j.jpeds.2017.03.050 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chung S, Youn S, Lee C, Jo M-W, Park J, Jo SW, Lee J, Sung JH, Sim CS, 2016. Environmental Noise and Sleep Disturbance: Night-to-Night Variability of Sleep/Wake Pattern. Sleep Med. Res. 7, 78–81. 10.17241/smr.2016.00122 [DOI] [Google Scholar]
- Crowley SJ, Acebo C, Carskadon MA, 2012. Human puberty: Salivary melatonin profiles in constant conditions. Dev. Psychobiol. 54, 468–473. 10.1002/dev.20605 [DOI] [PMC free article] [PubMed] [Google Scholar]
- de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J, 2007. Development of a WHO growth reference for school-aged children and adolescents. Bull. World Health Organ. 85, 660–667. https://doi.org/S0042-96862007000900010 [pii] [DOI] [PMC free article] [PubMed] [Google Scholar]
- Decharat S, 2018. Urinary mercury levels among workers in e-waste shops in Nakhon Si [DOI] [PMC free article] [PubMed] [Google Scholar]
- Thammarat Province, Thailand J Prev. Med. Public Heal. 51, 196–204. 10.3961/jpmph.18.04 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Do SY, Lee CG, Kim JY, Moon YH, Kim MS, Bae IH, Song HS, 2017. Cases of acute mercury poisoning by mercury vapor exposure during the demolition of a fluorescent lamp factory. Ann. Occup. Environ. Med. 29 10.1186/s40557-017-0184-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- Falnoga I, Tušek-Žnidarič M, Horvat M, Stegnar P, 2000. Mercury, selenium, and cadmium in human autopsy samples from Idrija residents and mercury mine workers, in: Environmental Research. pp. 211–218. 10.1006/enrs.2000.4116 [DOI] [PubMed] [Google Scholar]
- Gradisar M, Gardner G, Dohnt H, 2011. Recent worldwide sleep patterns and problems during adolescence: A review and meta-analysis of age, region, and sleep. Sleep Med. 12, 110–118. 10.1016/j.sleep.2010.11.008 [DOI] [PubMed] [Google Scholar]
- Gump BB, Gabrikova E, Bendinskas K, Dumas AK, Palmer CD, Parsons PJ, MacKenzie JA, 2014. Low-level mercury in children: Associations with sleep duration and cytokines TNF-α and IL-6. Environ. Res. 134, 228–232. 10.1016/j.envres.2014.07.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jansen EC, Conroy DA, Burgess HJ, O’Brien LM, Cantoral A, Téllez-Rojo MM, Peterson KE, Baylin A, 2020. Plasma DHA Is Related to Sleep Timing and Duration in a Cohort of Mexican Adolescents. J. Nutr. 150, 592–598. 10.1093/jn/nxz286 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jansen EC, Dunietz GL, Dababneh A, Peterson KE, Chervin RD, Baek J, O’Brien L, Song PXK, Cantoral A, Hu H, Téllez-Rojo MM, 2019. Cumulative Childhood Lead Levels in Relation to Sleep During Adolescence. J. Clin. Sleep Med. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kasznia-Kocot J, Dabkowska B, Muszyńska-Graca M, Zlotkowska R, Brewczyński PZ, 2010. Diagnostic and environmental procedures in accidental mercury vapor intoxication - experiences from an outpatient clinic. Polish J. Environ. Stud. 19, 579–585. [Google Scholar]
- Kothari S, Kruse D, Karimi R, Silbernagel S, Gursoy N, Jaber R, Roppelt H, Awan R, Gold A, Meliker JR, 2015. High mercury seafood consumption associated with fatigue at specialty medical clinics on Long Island, NY. Prev. Med. Reports 2, 798–802. 10.1016/j.pmedr.2015.09.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Luboshitzky R, Lavie P, 1999. Melatonin and sex hormone interrelationships - A review. J. Pediatr. Endocrinol. Metab. 12, 355–362. 10.1515/JPEM.1999.12.3.355 [DOI] [PubMed] [Google Scholar]
- Macchi MM, Bruce JN, 2004. Human pineal physiology and functional significance of melatonin. Front. Neuroendocrinol. 25, 177–195. 10.1016/j.yfrne.2004.08.001 [DOI] [PubMed] [Google Scholar]
- Owens J, Adolescent Sleep Working Group, A.S.W., Committee on Adolescence, C.O., 2014. Insufficient sleep in adolescents and young adults: an update on causes and consequences. Pediatrics; 134, e921–32. 10.1542/peds.2014-1696 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Parmalee NL, Aschner M, 2017. Metals and Circadian Rhythms. pp. 119–130. 10.1016/bs.ant.2017.07.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Paruchuri Y, Siuniak A, Johnson N, Levin E, Mitchell K, Goodrich JM, Renne EP, Basu N, 2010. Occupational and environmental mercury exposure among small-scale gold miners in the Talensi-Nabdam District of Ghana’s Upper East region. Sci. Total Environ. 408, 6079–6085. 10.1016/j.scitotenv.2010.08.022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Perng W, Tamayo-Ortiz M, Tang L, Sánchez BN, Cantoral A, Meeker JD, Dolinoy DC, Roberts EF, Martinez-Mier EA, Lamadrid-Figueroa H, Song PXK, Ettinger AS, Wright R, Arora M, Schnaas L, Watkins DJ, Goodrich JM, Garcia RC, Solano-Gonzalez M, Bautista-Arredondo LF, Mercado-Garcia A, Hu H, Hernandez-Avila M, Tellez-Rojo MM, Peterson KE, 2019. Early Life Exposure in Mexico to ENvironmental Toxicants (ELEMENT) Project. BMJ Open 9 10.1136/bmjopen-2019-030427 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sadeh A, Sharkey KM, Carskadon MA, 1994. Activity-based sleep-wake identification: An empirical test of methodological issues. Sleep 17, 201–207. 10.1093/sleep/17.3.201 [DOI] [PubMed] [Google Scholar]
- Sauvé JF, Lévesque M, Huard M, Drolet D, Lavoué J, Tardif R, Truchon G, 2015. Creatinine and specific gravity normalization in biological monitoring of occupational exposures. J. Occup. Environ. Hyg. 12, 123–129. 10.1080/15459624.2014.955179 [DOI] [PubMed] [Google Scholar]
- Schutzmeier P, Focil Baquerizo A, Castillo-Tandazo W, Focil N, Bose-O’Reilly S, 2018. Efficacy of N,N’bis-(2-mercaptoethyl) isophthalamide on mercury intoxication: A randomized controlled trial. Environ. Heal. A Glob. Access Sci. Source 17 10.1186/s12940-018-0358-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sherman LS, Blum JD, Franzblau A, Basu N, 2013. New insight into biomarkers of human mercury exposure using naturally occurring mercury stable isotopes. Environ. Sci. Technol. 47, 3403–3409. 10.1021/es305250z [DOI] [PubMed] [Google Scholar]
- Shiue I, 2017. Urinary arsenic, pesticides, heavy metals, phthalates, polyaromatic hydrocarbons, and polyfluoroalkyl compounds are associated with sleep troubles in adults: USA NHANES, 2005–2006. Environ. Sci. Pollut. Res. 24, 3108–3116. 10.1007/s11356-016-8054-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wheaton AG, Jones SE, Cooper AC, Croft JB, 2018. Short Sleep Duration Among Middle School and High School Students — United States, 2015. MMWR. Morb. Mortal. Wkly. Rep. 67, 85–90. 10.15585/mmwr.mm6703a1 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.