Association Between Prenatal Exposure to Alcohol and Tobacco and Neonatal Brain Activity

This cohort study uses electroencephalography to examine the association of prenatal exposure to alcohol and tobacco smoking with brain activity in newborns.


Introduction
Negative long-term effects of excessive prenatal alcohol exposure (PAE) and prenatal tobacco exposure (PTE) on risk for multiple adverse outcomes have been well established. PAE is the leading cause of preventable intellectual disability, and smoking during pregnancy is one of the most modifiable causes of perinatal morbidity and mortality. 1,2 Understanding the associations of quantity, timing, and various combinations of in utero alcohol and smoking exposures with early brain function could help identify mechanisms that underlie adverse long-term neurobehavioral outcomes.
Assessing neonatal brain activity through electroencephalography (EEG) provides a means of examining potential associations of PAE and PTE with brain activity in the immediate postnatal period. EEG is a noninvasive, sensitive measure of brain activity reflecting electrical activity generated by spatially aligned postsynaptic potentials in cortical pyramidal cells owing to their orientation in relation to the cortex. [3][4][5][6] EEG spectral power is a measure of the amplitude of the EEG signal from peak to peak across a specified length of time. [3][4][5][6] As children age, there is a developmental decrease in low-frequency brain oscillations, delta (δ) and theta (θ), starting around 4 years of age and a developmental increase in high-frequency brain activity beta (β) and gamma (γ). [3][4][5][6] Longitudinal studies of alpha (α) power during infancy suggest a developmental increase in 6to 9-Hz activity with central alpha peaking around 2 years of age. 7 Although less is known regarding the development of EEG power during the neonatal period, developmental changes in oscillatory activity are postulated to reflect decreases in synaptic density that underlie neural pruning to increase functional specialization. 7,8 EEG power in newborns has been shown to predict developmental outcomes at later ages when controlling for gestational age at birth and sleep state. [9][10][11][12][13] Although there is significant heterogeneity in prior studies, relative to neurotypical populations, infants at risk for developmental disorders often exhibit atypical developmental trajectories in neural oscillations, such as increased delta and theta or decreased beta or gamma power. [14][15][16][17][18][19][20][21][22] To our knowledge, only 3 prior studies have examined associations between PAE or PTE and infant brain function by using EEG. [23][24][25] Although these studies were restricted by small sample size and only examined high levels of PAE, each reported increased EEG power in infants with PAE, described as hypersynchronous EEG. [23][24][25] Increased EEG power in infants of alcoholic mothers was highest in rapid eye movement (REM) sleep, where power was approximately 200% greater than controls. 23 This finding was further supported by preclinical studies demonstrating increased hippocampal theta rhythm activity in rats with PAE. 26 Increased EEG power in infants with PAE is likely not attributable to alcohol withdrawal syndrome given that neurophysiologic differences persist 4 to 6 weeks postnatally. 25 More recently, a study has identified hypersynchrony in magnetoencephalography spectral power in awake 6-month-old infants with low to moderate PAE compared with controls. 27 The association was persistent across all frequency bands and was most prominent in left anterior and posterior temporal regions. 27 Although 1 study found no differences in EEG power in infants with PTE compared with controls with minimal or no exposure, 24 more recent data suggest neurophysiologic sensitivity to prenatal nicotine exposure (PNE). An EEG/event-related potential study examined the auditory K-complex in infants 3 to 5 months old with PNE and found reduced delta power compared with unexposed infants in non-REM sleep. 28 However, a study of PNE on sleep/wake ontogenesis in neonatal rats demonstrated increased delta and theta Hz activity in REM sleep. 29 Recent evidence also suggests that PAE and PNE combined induce oxidative stress and increase monoamine oxidase activity and caspase expression in the cerebellum. 30 These biochemical aberrations from dual PAE and PNE suggest potential compounding neurotoxic effects on the developing brain.
The current prospective study of 1739 newborns examines associations of PAE and PTE with neonatal brain activity measured via EEG spectral power. To examine several different patterns of PAE and PTE, we have used cluster analysis to carefully characterize drinking and smoking patterns.
Based on prior clinical and preclinical studies, we hypothesized that dual prenatal exposure to alcohol and smoking would be associated with increased low-frequency EEG power and decreased highfrequency EEG power in neonates.

Participants
Participants were a subset of neonates with neonatal EEG data enrolled in the Safe Passage Study conducted by the Prenatal Alcohol and SIDS and Stillbirth (PASS) Network, a multicenter study investigating the role of prenatal exposure to alcohol and smoking in risk for multiple adverse outcomes. 31 Mother-newborn dyads were enrolled from December 2011 through August 2015. Data were analyzed between June 2018 through June 2019. Participants were excluded from the present analysis on the basis of multiple birth, birth before 37 weeks' gestation or after 41 weeks' gestation, or prenatal exposure to any psychiatric medications at any point during pregnancy (selective serotonin reuptake inhibitors, antidepressants, classic antipsychotics, atypical antipsychotics, mood stabilizers, stimulants, anxiolytics, or anticonvulsants). Written informed consent to record neonate brain activity using EEG was obtained as part of the consent for the main study. Ethical approval was obtained from Stellenbosch University, Sanford Health, the Indian Health Service, and New York State Psychiatric Institute.

Self-reported Exposure Measures
The procedures used to obtain detailed information about quantity and timing of PAE and PTE have previously been described by the PASS Network. 31,32 In brief, information regarding PAE was acquired using a modified 30-day Timeline Followback interview 31,32 in which women self-reported their daily alcohol consumption for their last drinking day and 30 days prior up to 4 times during pregnancy. Detailed information was acquired regarding drink sharing, the type and brand of alcohol, container size, and duration of drinking to estimate the amount of alcohol consumed as accurately as possible. 31,32 This information was used to calculate an estimate of total grams of alcohol consumed per day for each day during pregnancy. Agreement between the Safe Passage Study Timeline Followback interview and neonate meconium alcohol marker ethyl glucuronide demonstrated 82% sensitivity (95% CI, 71.6%-92.0%) and 75% specificity (95% CI, 63.2%-86.8%) between PAE and ethyl glucuronide. 33 PTE information was also obtained up to 4 times during pregnancy in which women reported their smoking habits for their last reported smoking day and 30 days prior. Women indicated how often and the quantity they smoked tobacco cigarettes on an average smoking day. 31,34 These estimates were used to calculate average cigarettes smoked per week for each week of pregnancy.

Missing Data Imputation
We imputed missing daily alcohol and weekly smoking data by a nonparametric machine learning algorithm called the k-nearest neighbor approach. 36-38 More detail is available in the eMethods in the Supplement.

Alcohol and Smoking Cluster Analysis
To discern associations between different patterns of PAE and PTE during pregnancy on newborn brain activity, we implemented cluster analysis to characterize multiple patterns of maternal drinking and smoking behaviors (eMethods in the Supplement). 38 In the present analysis, we collapsed the PASS alcohol and smoking cluster groups 38 to create a 4-level PAE variable (no alcohol, low continuous alcohol, quit early alcohol, moderate or high continuous alcohol) and a 3-level PTE variable (no smoking, low continuous or quit early smoking, moderate or high continuous smoking) ( Table 1). More detail is available in the eMethods and eTables 2 and 3 in the Supplement.

Computing EEG Power
Absolute EEG power, representing the square of EEG magnitude, was calculated for 12 scalp regions hours; P < .001), the standardized residual of EEG power, after adjusting for postnatal age assessment within clinical site, was used for all subsequent analyses.

Statistical Analyses of EEG Power
Analyses for examining the association of PAE and PTE with EEG power controlled for sex, gestational age at birth, clinical site, and recreational drug exposure. Analyses of covariance were run by sleep state for each frequency bin and scalp region to examine the main effect of alcohol, the main effect of smoking, and an interaction term between alcohol and smoking, resulting in 180 statistical comparisons. Hypothesis tests were 2-sided. A 10% false discovery rate (FDR) correction was implemented to correct for multiple comparisons using the Benjamini-Hochberg procedure. 39 In the presence of a significant main effect, all pairwise comparisons were run where we reported 95% CIs for the difference and pairwise P values. Statistical analyses were performed with R, version 3.6.1 (R Studio) and SPSS, version 26 (IBM Corp). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Summary Demographic Information
The final sample consisted of a subset of 1739 term-age neonates from the Safe Passage Study with available maternal prenatal exposure information and neonatal EEG data (886 [50.9%] were female)

Main Effect of Alcohol on EEG Power in Active Sleep
After FDR correction for multiple comparisons, there was a significant main effect of alcohol on righttemporal 4-to 6-Hz, left-temporal 4-to 6-Hz, and left-temporal 7-to 9-Hz theta and (infant) alpha EEG power (Table 3; Figure 1) (Figure 1).

Associations of Alcohol and Smoking With EEG Power in Active and Quiet Sleep
There were no statistically significant associations between alcohol and smoking and EEG power in AS. Sample size in QS was reduced from 1739 in AS to 1201. After FDR correction, there were no statistically significant main effects or significant associations between PAE and PTE and EEG power in QS.

Discussion
To our knowledge, the present report is the largest study to date to investigate associations between PAE and PTE and brain activity in newborns. Our PAE findings are in partial agreement with prior reports that demonstrated increased EEG power in infants of alcoholic mothers, described as hypersynchronous EEG, [23][24][25] and with the recent report of increased magnetoencephalography power in 6-month-old infants with low to moderate PAE. 27 However, we are the first to report associations between PAE and brain activity even in infants with low continuous PAE and in infants whose mothers quit drinking before the second trimester.
This finding has significant public health relevance in the context of media reports on the lack of perinatal effects from light drinking during pregnancy. The consistent finding of increased EEG power from PAE may reflect an imbalance in the excitatory glutamate to inhibitory γ-aminobutyric acid ratio resulting in weakened neural inhibition, increased neural excitation, or aberrant neuronal differentiation. 49 Evidence from in vitro studies has demonstrated that PAE results in increased amplitude and duration of excitatory hippocampal pyramidal cellular activity. 50 Although we have not yet determined the association between changes in brain activity at birth from PAE and subsequent cognitive or behavioral outcomes, our present findings from a diverse, multinational cohort are especially important in the context of recent reports suggesting either no effect or a protective effect from low to moderate PAE on birth, academic, cognitive, and attentional outcomes in women from high socioeconomic households. 51-53 Because women with low levels of education or advanced maternal age who consume alcohol during pregnancy are at the greatest risk for having a child with fetal alcohol spectrum disorders, 54 it is important to assess brain function at birth independent of potential modifying factors in an enriched postnatal environment. 54 Taken together, our findings suggest that any level of PAE or PTE has robust associations with newborn brain activity, reaffirming the public health message that research has not yet determined a safe level of alcohol or tobacco use during pregnancy.

Limitations
Although we attempted to accurately estimate drinking and smoking behaviors, it is possible there could be underreporting or overreporting of PAE or PTE. The present study only reports the associations of PAE and PTE in term-age infants, which may vary in preterm birth, and we measured EEG at only 1 time point. In future reports, we plan to examine additional exposures and long-term neurodevelopmental outcomes.

Conclusions
Examining neonatal EEG may prove to be a reliable proximal marker of the potential downstream associations of PAE and PTE with neurodevelopment. We hope these data can elucidate potential mechanisms underlying risk for adverse outcomes.