Prenatal particulate air pollution exposure and sleep disruption in preschoolers: windows of susceptibility

Abstract

BACKGROUND:

The programming of sleep architecture begins in pregnancy and depends upon optimal in utero formation and maturation of the neural connectivity of the brain. Particulate air pollution exposure can disrupt fetal brain development but associations between fine particulate matter (PM_2.5) exposure during pregnancy and child sleep outcomes have not been previously explored.

METHODS:

Analyses included 397 mother-child pairs enrolled in a pregnancy cohort in Mexico City. Daily ambient prenatal PM_2.5 exposure was estimated using a validated satellite-based spatio-temporally resolved prediction model. Child sleep periods were estimated objectively using wrist-worn, continuous actigraphy over a 1-week period at age 4–5 years. Data-driven advanced statistical methods (distributed lag models (DLMs)) were employed to identify sensitive windows whereby PM_2.5 exposure during gestation was significantly associated with changes in sleep duration or efficiency. Models were adjusted for maternal education, season, child’s age, sex, and BMI z-score.

RESULTS:

Mother’s average age was 27.7 years, with 59% having at least a high school education. Children slept an average of 7.7 hours at night, with mean 80.1% efficiency. The adjusted DLM identified windows of PM_2.5 exposure between 31–35 gestation that were significantly associated with decreased sleep duration in children. In addition, increased PM_2.5 during weeks 1–8 was associated with decreased sleep efficiency. In other exposure windows (weeks 39–40), PM_2.5 was associated with increased sleep duration.

CONCLUSION:

Prenatal PM_2.5 exposure is associated with altered sleep in preschool-aged children in Mexico City. Pollutant exposure during sensitive windows of pregnancy may have critical influence upon sleep programming.

INTRODUCTION

Insufficient or poor-quality sleep in early childhood is one of the most common problems reported by caregivers ¹, with 20–40% of children experiencing sleep disturbance before reaching school age ^2,3. Abnormal sleep patterns persist into later life ^4,5, and have life-long adverse effects on health, including increased neurodevelopmental, metabolic, cardiovascular, and respiratory disorders ^6,7. Disordered sleep in young children, manifesting as altered duration, timing, maintenance, or quality of sleep ⁸, is associated with hyperactivity ^9,10, impaired attention ^11,12, cognitive function ^13,14, and behavior problems ^15,16. While treating sleep disorders in young children improves some outcomes ^17,18, complete response is the exception rather than the rule ^19,20, suggesting there may be a critical early window in which sleep disruption may have lasting health and neurocognitive impacts ^21,22. Therefore, identifying factors that disrupt the ontogeny of sleep is critical in developing prevention strategies to promote healthy sleep development.

Sleep is an active, complex neurophysiological process ^23,24. Optimal formation of the neural connectivity required to process signals involved in sleep regulation begins in utero ^25,26, as early as seven weeks of gestation, when neuronal production and migration start laying down necessary myelinated neural networks across the brain. While there is continued maturation of homeostatic, circadian, and ultradian regulation of sleep following birth, the basic components – phasic rapid eye movement (REM) and non-REM (NREM) sleep - are also established prenatally, as early as 28 weeks ²⁷, and timely progression of the formation of these neural networks in utero until birth is necessary to establish the mature sleep wake cycle observed in the full-term newborn ²⁵. Hence, developmental interruption--such as by preterm birth--may result in reduced sleep duration and efficiency ²⁸, altered sleep architecture ²⁸, and circadian rhythm abnormalities ²⁹, manifesting postnatally.

Notably, the plasticity of neurodevelopmental phenomena involved in establishing sleep phenotypes during gestation ³⁰ also renders it vulnerable to the influence of environmental exposures, which may in turn result in childhood sleep disorders. For example, disruption of critical neurodevelopmental processes during pregnancy, such as by prenatal tobacco smoke and other toxins, has been linked with abnormal sleep-wake transitions, excessive arousals, and disorganized sleep patterns in infancy ^31,32 that persist into later childhood ^33,34. Overlapping animal and epidemiologic data link ambient air pollution, in particular fine particulate matter (PM_2.5), to adverse effects on central nervous system (CNS) development via oxidative stress and neuroinflammation ^35,36. Ambient PM_2.5 impacts fetal brain structure, function, and neurohormonal development ³⁷. For example, Peterson et al. demonstrated that prenatal air pollution is associated with diffusely reduced white matter on MRI by school age, with concomitant cognitive and behavioral disorders in children ³⁸. Notably, these anatomic alterations were more pronounced for prenatal rather than postnatal exposure. Traffic pollution is also linked to lower functional brain connectivity in children, presumably through its effects on white matter cell maturation and myelination ³⁹, and neuroinflammatory pathways known to impair brain development ^35,40. In addition, prenatal PM alters neurotransmitters across multiple brain regions in animal models ⁴¹, and consequent central changes in availability or sensitivity to neurotransmitters such as impaired glycine and GABA_A receptor function in animal studies results in abnormal neuro-inhibitory transmission and sleep fragmentation, mimicking human REM sleep disorders ⁴². Lastly, in humans, prenatal PM_2.5 exposure influences placental methylation in the promoter regions of circadian pathway genes (e.g. Clock), specifically during the 3^rd trimester ⁴³, highlighting another potential mechanism for the alteration of sleep regulation by particulate air pollution.

While recent epidemiologic data demonstrate associations between PM and sleep difficulties among children exposed postnatally ^44–46, to date, to our knowledge, there are no published studies examining the impact of prenatal ambient PM_2.5 on sleep in early childhood. In addition, evidence shows that neurotoxic effects of prenatal PM_2.5 exposure in children occur during specific time periods during pregnancy ⁴⁷, suggesting that associations between prenatal toxic exposures and postnatal sleep disorders may depend on timing of exposure as well as dose. In order to examine associations between prenatal PM_2.5 exposure on early childhood sleep outcomes accounting for timing of exposure, we implemented data-driven distributed lag models (DLMs) to estimate windows of vulnerability between prenatal PM_2.5 exposure and sleep duration and efficiency estimated objectively at preschool age within a prospective pregnancy cohort established in Mexico City.

METHODS

Participants

Pregnant mothers were recruited from prenatal clinics in Mexico City belonging to the Mexican Social Security System from July 2007 to February 2011. Enrollment into the Programming Research in Obesity, Growth, Environment, and Social Stressors (PROGRESS) cohort included mothers at least 18 years of age, <20 weeks pregnant, planning to stay in Mexico city for the next 3 years, with access to a telephone, and having no medical history of heart or kidney disease, daily alcohol consumption, or steroid or anti-epilepsy medication use. Pregnant women were excluded from participation if they had multiple gestation, maternal preeclampsia, psychiatric disease or use of psychotropic drugs, and for logistical reasons, those mothers living in a household outside the metropolitan area. Additional details of the cohort have been described previously ⁴⁸. Eight-hundred and fifteen mother-child dyads were enrolled and continued longitudinal follow-up postnatally. Mothers provided written consent in their primary language and the study was approved by the Institutional Review Boards at the Harvard School of Public Health, the Mexican National Institute of Public Health, and the Icahn School of Medicine at Mount Sinai. Premature births with gestation less than 37 weeks were specifically excluded from these analyses. Our analyses included n=397 mothers and full term-infants (born>37 weeks gestation) who completed the 4–5 year visit and had complete environmental air pollution and sleep outcome data. Those included in the analysis and those excluded did not differ significantly based on average PM_2.5 exposure over pregnancy (22.7 vs 22.6 μg/m³.), mean maternal age (27.7 vs 27.5 years), maternal education (59 vs. 60%), child BMI z-score (0.07 vs. 0.26), age at sleep evaluation (4.8 vs. 4.8 years), or sex (51 vs. 54% males), respectively (p-value for all >0.05).

Ambient air pollution assessment

Individual maternal daily pollutant exposure over pregnancy was estimated from home addresses obtained at enrollment and updated if the participant moved. PM_2.5 levels across Mexico City were estimated at a 1 × 1 km spatial resolution, using hybrid models incorporating day-specific calibrations of aerosol optimal depth (AOD) data with land use regression (LUR) and meteorological variables (roadway density, temperature, relative humidity, planetary boundary layer, and daily precipitation), and calibrated against ground measurements performed at 12 monitoring stations covering Mexico City, as described previously ⁴⁹. Mixed effect models with spatial and temporal predictors and day-specific random effects were used to account for temporal variations in the PM_2.5−AOD relationship. For days without AOD data, models were fit with a seasonal smooth function of latitude and longitude and time-varying average based on local monitoring. The out of sample ten-fold cross validation R² of 0.724. We similarly derived postnatal PM_2.5 over the first year of the child’s life.

Sleep outcomes

Child sleep duration and sleep efficiency were the outcomes evaluated. Child sleep duration was estimated objectively using ActiGraph GT3x+ accelerometers, measuring motion in oscillations, which was placed on the child’s non-dominant wrist and worn over a 7-day period. Sleep duration was identified by periods with low to no physical activity (defined as <1000 oscillations per minute). Primary outcomes included nocturnal sleep duration and sleep efficiency determined as follows. Using accelerometry software ActiLife v6.11.9, we used the Sadeh sleep algorithm to determine total nocturnal sleep time, defined as the number of minutes the child was asleep each night, ⁵⁰ and sleep periods were crosschecked against parent-completed sleep diaries. Nocturnal sleep efficiency, defined as the percentage of time in bed spent asleep, was calculated by dividing accelerometer-derived sleep periods by the total time spent in bed. In addition to analyzing periods coincident with nighttime, sleep periods during the day (i.e. naps) were also identified and analyzed using the same algorithm. Compliance with accelerometry was high; 96.8% of children wore the Actigraph for at least 6 nights.

Other Covariates

We considered covariates related to air pollution exposure and child sleep. Demographic and health-related data were collected during an initial prenatal study visit. Maternal age, education level as an indicator of socioeconomic status (SES), and maternal prenatal smoking status were self-reported through questionnaire administered at enrollment. Hospital delivery records were used to collect newborn characteristics such as sex and birthweight. Gestational age was based on maternal report of last menstrual period (LMP) and by a standardized physical examination at birth, with discrepancies of more than 3 weeks resolved by using the latter. We derived Fenton birth weight for gestational age z-scores, which accounts for the non-linear birth weight patterns during pregnancy ⁵¹. Child height and weight were recorded at the 4–5 year visit by research staff using a wall-mounted stadiometer; body mass index (BMI) was calculated by dividing weight by height squared (kg/m²), and z-scores were derived using World Health Organization norms. Season of visit was defined according to weather patterns in Mexico City as dry cold (January-February; November-December), dry warm (March-April) and rainy (May-October).

Statistical analyses

We restricted our sample to infants born ≥37 weeks gestation, as we considered prematurity to be a pathway variable, with well-established associations describing prenatal air pollution and prematurity ⁵², as well as between prematurity and disruption of neural networks relevant to sleep ^53,54. Among the 397 mother-child dyads included in the analyses, for those infants (n=289) born between 37 and 40 weeks of gestation, we used ambient post-natal exposure of the remaining days of a 40-week gestational period as PM_2.5 exposure estimates in our models as in prior studies ^55,56.

We first evaluated the relationship between mean PM_2.5 exposure averaged over the full gestational period, as well as individually for each of the trimesters (adjusting for the remaining trimester exposures), and the sleep outcomes using linear regression models. Next, weekly averages of daily PM_2.5 exposure estimates throughout pregnancy were used in order to determine the time-varying associations between pollutant exposures and sleep outcomes and identify sensitive windows during gestation where these associations were significant. Specifically, we estimated overall associations between prenatal PM_2.5 exposure and nocturnal sleep duration and sleep efficiency, in separate models, and identified sensitive windows by constructing distributed lag models (DLM) using the previously described approach ⁵⁷ with no interaction. When the association varies with time, time periods with greater association (sensitive windows) graphically appear as a bump where PM_2.5 is associated with sleep outcomes. In addition to determining sensitive windows, we employed similar DLMs to estimate cumulative effects over the entire pregnancy of PM_2.5. All primary models were adjusted for maternal age and education, season of visit, child’s BMI z-score, age, and sex. Secondary sensitivity analyses were conducted for sleep duration that included daytime naps, as well as for further adjustment for the following covariates: birthweight for gestational age (BWGA) Fenton z-score, maternal smoking in pregnancy, and post-natal PM_2.5 exposure averaged over the first year of life, in separate models.

RESULTS

Participant characteristics

Mothers had an average (mean, SD) age of 27.7 (5.7) years at enrollment; 59% completed at least a high school education. Median [interquartile range (IQR)]) ambient prenatal PM_2.5 concentration was 23.0 [21.1–24.3] μg/m³. Child sleep was estimated objectively at a mean [SD] age of 4.8 [0.5] years. Mean [SD] nocturnal sleep duration and efficiency were 7.7 [0.74] hours, and 80.1 [5.8]%, respectively (Table 1).

Table 1.

PROGRESS participant characteristics (N=397)

Variables	All Children
Maternal age (year; mean, SD)	27.7	5.7
Maternal education (n, %)
< High school	163	41.1
≥High school	234	59.0
Maternal smoking (n, %)	3	0.8
Child sex (n, %)
Male	203	51.1
Female	194	48.9
Child age (year; mean, SD)	4.8	0.5
Child’s BMI z-score (z-score; median, IQR)	0.07	−0.47 – 0.65
Season of visit (n, %)
Dry / Cold	127	32.0
Dry / Warm	70	17.6
Rain	200	50.4
Prenatal averaged PM2.5 level (μg/m3; median, IQR)	23.0	21.1 – 24.3
Postnatal 1st year averaged PM2.5 level (μg/m3; median, IQR)	22.8	20.8 – 24.0
Nocturnal sleep duration (hours; mean, SD)	7.7	0.74
Total sleep duration per 24 hour period (hours; mean, SD)	7.8	0.72
Nocturnal sleep efficiency (%; mean, SD)	80.1	5.8
Birthweight for gestational age (Fenton z-score; mean, SD)	−0.49	0.85

Association between prenatal PM_2.5 and child nocturnal sleep duration

We evaluated the relationship between mean PM_2.5 exposure averaged over the course of pregnancy, as well as individually for each of the trimesters, and nocturnal sleep duration. After adjusting for confounders, linear regression models did not identify significant associations between any of the average exposures (over gestation or within each trimester) and nocturnal sleep duration.

Adjusted DLMs were used to estimate the overall associations between weekly PM_2.5 during gestation and nocturnal sleep duration. We identified a late sensitive window at 31–35 weeks whereby higher PM exposure during pregnancy was significantly associated with decreased total sleep time at preschool age (Figure 1). In addition, there was also a second significant critical window at 39–40 weeks that was positively associated with total sleep time. There was a suggestive effect of an IQR change in PM exposure being associated with increased nocturnal sleep duration in weeks 21–27, but this did not reach statistical significance. There was no significant cumulative effect of PM_2.5 exposure over the course of gestation upon child nocturnal sleep time.

Figure 1. Association between weekly PM_2.5 over gestation and child sleep: nocturnal sleep duration.

Distributed lag models (DLMs) were used to estimate week-specific effects for preschool children. Models were adjusted for maternal age and education, season of visit, child BMI z-score, age, and sex. The x-axis indicates the gestational age in weeks. The y-axis represents the change in nocturnal sleep time corresponding to each interquartile range (IQR) of PM_2.5 across the entire pregnancy. Solid lines represent the predicted time-varying change in sleep duration and gray areas indicate the 95% confidence intervals (CI). A sensitive window is identified when the estimated point-wise 95% CI does not include zero.

In addition to nocturnal sleep, we performed sensitivity analyses between PM_2.5 exposure and total 24 hour sleep duration to account for daytime nap periods, which had a similar shape and where the association between PM and sleep duration was significant between 21–27 weeks. (Figure S1). The cumulative effect of PM_2.5 exposure over the course of gestation was not significant in either linear regression or DLM models (data not shown).

Association between prenatal PM_2.5 and child nocturnal sleep efficiency

Adjusted linear regression models demonstrated a statistically significant association between mean PM_2.5 exposure averaged over pregnancy and nocturnal sleep efficiency (β= −0.247, p=0.037, for each 1 μg/m³ change in PM_2.5). Furthermore, when average exposures were examined by trimester, statistically significant associations were noted for the first trimester only (β = −0.204, p= 0.004 per 1 μg/m³ change in PM_2.5). Consistent with these findings, best-fit DLMs examining the time-varying association between prenatal PM_2.5 exposure and child sleep efficiency also identified a distinct sensitive window at 1–8 weeks of gestation, whereby increased pollutant exposure was associated with decreased nocturnal sleep efficiency in preschool children (Figure 2). In addition, the estimated cumulative effect of pollutant exposure over the entire gestational period demonstrated significantly reduced sleep efficiency (β = −0.78, 95%CI= −1.52 to −0.13 for each IQR change in PM_2.5).

Figure 2. Association between weekly PM_2.5 over gestation and child sleep: nocturnal sleep efficiency.

Distributed lag models (DLMs) were used to estimate week-specific effects for preschool children. Models were adjusted for maternal age and education, season of visit, child BMI, age, and sex. The x-axis indicates the gestational age in weeks. The y-axis represents the change in sleep efficiency corresponding to each interquartile range (IQR) of PM_2.5 across the entire pregnancy. Solid lines represent the predicted time-varying change in sleep efficiency and gray areas indicate the 95% confidence intervals (CI). A sensitive window is identified when the estimated point-wise 95% CI does not include zero.

Sensitivity analyses were also performed to account, individually, for BWGA Fenton z-score (Figures S2 and S3), maternal smoking (Figures S4 and S5), and post-natal PM_2.5 exposure over the first year of life (Figures S6 and S7), each of which revealed similar findings for both sleep duration and efficiency.

DISCUSSION

To our knowledge, this is the first prospective study to examine associations between prenatal PM_2.5 and sleep outcomes in childhood. In addition to highlighting significant associations between average PM exposure over pregnancy and sleep, we leveraged advanced statistical modeling approaches with highly temporally- and spatially-resolved ambient PM_2.5 exposure estimates to identify sensitive windows during which exposure to ambient fine particulate matter was significantly associated with alterations in nocturnal sleep duration and sleep efficiency in preschool children. Notably, while current American Association of Sleep Medicine (AASM) guidelines recommend that children 3–5 years of age receive 10–13 hours of sleep per 24 hours ⁵⁸, children in our cohort slept on average 7.8 hours over a 24 hour period, consistent with other data demonstrating similarly short sleep durations among young children, especially within minority populations in the United States ^59–61. These results highlight the need to identify early risk factors for sleep deficiencies in childhood, particularly in higher risk, minority populations.

There is a growing literature demonstrating the neurotoxic effects of prenatal exposure to fine particulate air pollution and early life neurodevelopmental processes. These effects may be mediated by oxidative stress and neuroinflammatory pathways, as PM has been shown to induce local and systemic inflammation. We believe this also provides biologic plausibility for our findings related to sleep outcomes. Basic sleep programming occurs before birth ²⁷, and is highly dependent upon staged progression of neuronal growth and maturation to optimize both structural and functional connectivity of the central nervous system ^25,30. Disruptions to the development of this connectivity, such as by prematurity, hypoxic insult, and neuro-inflammation may impair neural network development that in turn alter sleep states in children ²⁵. Indeed, data from both animal and human studies demonstrate that growth restriction and impaired neural architecture are associated with reduced active sleep and increased indeterminate sleep (IS) in the fetus, as well as altered sleep staging in the postnatal period ²⁵. In addition to perinatal changes, studies of young children raise concern for long-term effects on sleep quality, including poor sleep, lower sleep efficiency, and increased awakenings tied to prenatal insults. Of these, prenatal environmental pollutants such as maternal tobacco smoke exposure-- known to have adverse effects on fetal neural development ^62-64-- has been shown to affect downstream sleep outcomes after birth, resulting in disruptions in sleep structure and continuity that are reflective of disturbed sleep patterns and altered arousability ³². In this report, we present findings that for the first time extend this literature to demonstrate that prenatal ambient PM_2.5 exposure during critical periods of pregnancy is also associated with altered sleep outcomes in young children. Notably, overall estimates of prenatal PM_2.5 in this Mexico City cohort of 23 μg/m³ were found to be relatively higher than third trimester exposures in other United States-based pregnancy cohorts, for example, in Boston (10.75 μg/m³) or New York (8.2μg/m³) ⁶⁵.

Another novel aspect of our work lies in the identification of sensitive windows during gestation. As neurodevelopment in utero is a highly orchestrated, timed process, using data-driven approaches to identify sensitive windows of gestation allows us to overlay these time periods with corresponding knowledge of embryologic events (instead of arbitrary intervals of pregnancy), providing insight into the underlying mechanisms being perturbed. For example, the early window identified to be associated with decreased sleep efficiency (1–8 weeks) corresponds to the time period during which neuron production and migration allow for myelin-producing cells to facilitate white matter maturation required for functional connectivity of the brain ²⁵. One may conceptualize that early disruptions in laying down this foundation of neural architecture may reduce the ultimate potential of the “connectome ⁶⁶” and alter sleep development by limiting prerequisite neural arborization. In contrast, the later window identified (30–32 weeks) for PM effects on sleep duration occurs just after the emergence of identifiable sleep-wake states (active and quiet sleep) but coincides with subsequent key adjustments that occur in the proportions of time the fetus spends in IS and QS (achieved by 35–36 weeks gestation to approximate the sleep states of the full term infant) ²⁷. As a result, disruptions during this period and later may interfere with optimal sleep patterns and lead to the disorganized sleep states noted above.

With regards to particular sleep outcomes, our findings were particularly strong for relationships between prenatal PM_2.5 exposure and nocturnal sleep efficiency during early gestation. The specific timing of the sensitive window of 1–8 weeks gestation and the association during that window was consistently estimated with a number of different degrees of freedom for the DLM. The result is further supported by both the significant cumulative effect from the DLM and the association with mean PM_2.5 exposure averaged over pregnancy, and in particular, the first trimester. For sleep duration, we found a negative association during in weeks 30–32 weeks. However, we also identified a late gestational window at 39–40 weeks whereby increased PM_2.5 exposure was associated with increased sleep duration, though models estimating associations with sleep duration were less robust to the parameterization. Nonetheless, we speculate that while decreased sleep time may be the pathologic result of a reduced inhibitory state and increased degree of arousals, increased sleep time may reflect the need to increase sleep time to overcome toxin-induced poor sleep efficiency. Further studies are warranted to replicate these findings and investigate pathologic mechanisms underlying these observed relationships.

Our study has several strengths to highlight. To our knowledge, this is the first study to examine the effects of prenatal ambient particulate pollutant exposures on sleep outcomes in children. While a few recent studies have considered post-natal air pollutant exposures and sleep quality, no study, to our knowledge, has investigated the effect of exposures that occur during the prenatal period, when the foundation of sleep architecture is established. The relationships we report between PM_2.5 exposures during pregnancy and sleep outcomes in childhood emphasize the importance of critical periods during gestation for which early alteration of sleep programming can have lasting results. In addition, we have leveraged highly spatially (1 km x 1 km) and temporally resolved (daily) ambient pollution data reflecting individual maternal environmental exposure estimates during the complete period of pregnancy to conduct data-driven analytic approaches that identify sensitive windows most susceptible to pollutant effects. These approaches overcome the potential bias that is introduced by considering exposure timing by arbitrary cut-offs (e.g. trimesters) ⁶⁷, and place critical windows within the context of embryologic processes relevant to sleep development.

Certain limitations worth noting include the generalizability of our study to other populations, given that this is a relatively homogenous Mexican cohort in an urban area with higher levels of air pollution. In addition, as the literature on prenatal environmental factors known to influence sleep development is relatively sparse, we may not have accounted for unidentified confounders in our models. Notably, we considered a number of key covariates known to be significant from prior work regarding particulate air pollution and other neurodevelopmental outcomes. Furthermore, while the associations we report have biological plausibility, further work is needed to strengthen potential causal relationships between prenatal pollutant exposure and child sleep outcomes. Lastly, we used accelerometry to capture sleep-wake patterns over time, which is limited in that it estimates sleep duration by movement and does not identify EEG-defined staging of sleep. Nevertheless, actigraphy allowed us to measure sleep duration and efficiency in a large cohort of children within their home environment over several nights to generate a robust dataset reflective of sleep patterns. Future studies using polysomnography to assess for sleep disordered breathing, movement disorders, and analysis of sleep stage cycling, continuity, and spectral analysis will help further characterize the adverse effects of prenatal PM_2.5 exposure and the development of healthy sleep in children.

In conclusion, increased prenatal exposure to PM_2.5 during multiple critical windows over gestation was associated with altered sleep duration and decreased sleep efficiency in preschool-aged children. Given the high prevalence of sleep disorders in early childhood, as well as the tendency of these sleep abnormalities to persist later in life, these findings are not only significant for young children, but may have implications for sleep health across the lifespan. Furthermore, in light of the known adverse effects of inadequate sleep upon multiple other trajectories of health outcomes (e.g. metabolic, cognitive, behavioral), identification of early modifiable risk factors influencing suboptimal sleep programming is critical for designing targeted interventions to minimize potentially irreversible, compounding developmental consequences.

Abbreviations:

(DLM)

distributive lag models

(PM)

particulate matter

(CNS)

central nervous system

(BMI)

body mass index

(REM)

rapid eye movement

(AS)

active sleep

(QS)

quiet sleep

(IS)

indeterminate sleep

(AOD)

aerosol optimal depth

(LUR)

land use regression