Abstract
Objective
Single-substance exposure effects on neurodevelopmental outcomes, such as problem behavior and intelligence quotient (IQ), have been studied in children for decades. However, the long-term consequences of polysubstance exposure are poorly understood.
Study Design
Longitudinal neurodevelopmental data were gathered from cohorts across the United States through the Environmental Influences on Child Health Outcomes Program. Data on prenatal exposure to opioids, nicotine, marijuana, and alcohol were collected from children ages 6 to 11 years (N=256). Problem behavior was assessed using the Child Behavior Checklist (school-age version), and verbal IQ (VIQ) and performance IQ (PIQ) were assessed using the Weschler Intelligence Scale for Children, Fifth Edition. We first identified latent profiles in the overall sample, then evaluated differences in profile membership for children with and without prenatal substance exposure.
Results
Latent profile analysis identified two mutually exclusive categories: average VIQ and PIQ, with typical problem behavior, and below-average VIQ with average PIQ and clinically significant problem behavior. Children with prenatal nicotine and polysubstance exposures were more likely to be classified in the below-average VIQ, elevated problem behavior profile compared with children without prenatal nicotine exposure.
Conclusion
The presence of clinically significant behavior problems in children with average PIQ, but below-average VIQ, could represent a unique endophenotype related to prenatal nicotine exposure in the context of other prenatal substance exposures.
Keywords: prenatal polysubstance exposure, problem behavior, intelligence, pregnancy
Prenatal exposure to substances is a significant public health concern. A 2019 U.S. survey on drug use showed that one in five pregnant women used substances while pregnant, which ranged from legal substances, such as nicotine and alcohol, to illegal drugs, including opioids.1 Most research has examined the effect of a specific substance, controlling for the effects of other substances.2 However, polysubstance use, or use of at least two substances during pregnancy, is the norm rather than the exception.3 Despite this understanding, limited information is available on polysubstance exposure effects on neurodevelopmental outcomes.
Studies over the past several decades have examined the effects of prenatal exposure to single substances (e.g., nicotine or alcohol) on problem behavior outcomes, such as internalizing and externalizing behavior, and on cognitive outcomes, such as intelligence quotient (IQ). These studies show that the effects of prenatal exposure to alcohol, marijuana, cocaine, and nicotine appear to be more pronounced for externalizing behavior than internalizing behavior.4,5 These studies also suggest that prenatal exposure to multiple substances, including cocaine, opioids, nicotine, marijuana, and alcohol, are associated with elevations in externalizing, but not internalizing behavior, at age 116 and also with growth in problem behavior across adolescence.7
In these studies, IQ and problem behaviors are typically examined as outcomes in isolation. However, discrete groups of children may show challenges in one domain (such as problem behavior) and strengths in another (such as IQ). A recent review of 52 studies found that 23 studies observed an impact of prenatal opioid exposure on behavioral outcomes, whereas 17 studies detected cognitive effects.2 Prenatal exposure to cocaine and other substances predicted increased internalizing and externalizing problems at age 3 but no effects on verbal IQ (VIQ) or performance IQ (PIQ) in early childhood.8 Prenatal exposure to opioids and other substances may therefore have stronger effects on problem behavior outcomes than IQ outcomes.2
In contrast, clinicians typically take a multidimensional, or “whole child,” approach by examining multiple cognitive and behavioral outcomes and integrating the data in a clinical setting instead of only examining single domains. We used this approach in the current study by examining whether clinically significant scores in one domain (e.g., cognition) were found in other domains (e.g., internalizing and externalizing behavior) to find initial evidence of an endophenotype for prenatal substance exposure. As part of the Environmental Influences on Child Health Outcomes (ECHO) Program, we harmonized prenatal substance exposure, IQ, and problem behavior data from 256 pregnant women and their children ages 6 to 11 years from two cohorts in the United States. We hypothesized that children with high levels of polysubstance exposure would have the highest externalizing behavior and attention problems and the lowest IQ scores.
Materials and Methods
Environmental Influences on Child Health Outcomes Program
The ECHO Program is a consortium funded by the U.S. National Institutes of Health with 69 existing pediatric cohorts across the United States designed to investigate the effects of early life exposures on five key areas of child health: pre-, peri-, and postnatal health; obesity; respiratory conditions, including asthma; neurodevelopment; and positive health/well-being.9,10 The ECHO Program includes the ECHO Neonatal Opioid Withdrawal Syndrome special interest group for studying substance exposures in children. The data for this study come from two cohorts in the NIH ECHO Program (the Behavior and Mood of Babies, Adolescents, and Mothers study [BAMBAM] and The Infant Development and Environment Study [TIDES], described below). The study protocol was approved by the Local and/or Central ECHO Institutional Review Board. Written informed consent or parent’s/guardian’s permission was obtained along with child assent as appropriate.
Study Population
Participants were mothers with nonmissing information on opioid use during a singleton pregnancy with a biological child 6 to 11 years of age for whom data were available from the Child Behavior Checklist, school-age version (CBCL-Sch), with Verbal Comprehension Index and Perceptual Reasoning Index scores from the Weschler Intelligence Scale for Children, Fifth Edition (WISC-5), for assessing VIQ and PIQ, respectively. Problem behavior and IQ data were collected prospectively. A total of 256 mother–child pairs were included in this analysis from two ECHO cohorts. The BAMBAM (n=12311,12) is a prospective longitudinal analysis of brain development in healthy term children in Rhode Island beginning during pregnancy (2010–2017). Prospective longitudinal cognitive data were collected from age 3 months to 12 years as part of a larger study on typical brain development across childhood. The TIDES (n=133) recruited pregnant women from 2010 to 2012 from prenatal clinics in Minnesota, Washington, California, and New York to examine phthalate exposure, genital outcomes, and neurobehavior in children between 2010 and 2012.13,14
Prenatal Substance Use
Any detection of prenatal substance exposure as indicated on medical records was used for this study. We created binary variables (yes/no) for each prenatal substance of interest. Nicotine, alcohol, marijuana, and opioid use was defined as any use of the respective substance during pregnancy as ascertained by self-report, medical record abstraction, or maternal toxicology screen. Tobacco/nicotine exposure included cigarette smoking, e-cigarette smoking (vaping, electronic nicotine delivery devices/Electronic Nicotine Delivery Devices (ENDS), vape pens, or mods, etc.), and other forms of tobacco/nicotine (chewing tobacco/snuff, nicotine patch, nicotine gum/lozenges, cigar, pipe, hookah, and Bidi [Beedi]).
Child Behavior Outcomes: Child Behavior Checklist, School-Age Version
The CBCL is used in clinical and nonclinical settings for assessing behavioral and emotional problems in children.15 It is a questionnaire administered to caregivers that asks about a target child’s emotional and behavioral problems and is used in clinical and nonclinical settings. CBCL items ask a caregiver to rate the degree to which a child has each of 100 problems on a scale from 0 (Not True) to 3 (Very True or Often True). The CBCL yields T-scores with a mean of 50 and a standard deviation (SD) of 10; scores above or equal to 60 are considered to fall above the clinical cutoff for the problem behavior domain (borderline: 60–63; clinical: greater than 64). The CBCL-Sch was administered to ECHO caregivers when children were 6 to 11 years of age. The CBCL-Sch consists of 120 items related to child behavior issues scored on a 3-point scale ranging from Not True (value of 0) to Often True or Very True (value of 2). Raw CBCL scores, calculated by summing the corresponding CBCL responses, were transformed into T-scores for analysis according to the CBCL manual.
Child Cognition Outcomes: Wechsler Verbal Comprehension Index and Perceptual Reasoning Index
The Wechsler Verbal Comprehension Index and Perceptual Reasoning Index are scores that quantify verbal and nonverbal ability, respectively, and are produced by the larger WISC-V, a well-validated set of tests (e.g., define words, complete patterns) administered to children to test their cognitive abilities. We calculated VIQ and PIQ scores from the WISC-5.16 When index scores were not available from the participating cohorts, but subtest scores were available, we calculated VIQ and PIQ scores by averaging the scores for subtests included in the VIQ and PIQ index scores separately, converting those averages to the IQ metric (mean of 100, SD of 15; scores less than 85 considered “below-average IQ”) to yield calculated VIQ and PIQ scores. The subtests in the WISC-5 are reliable and valid indicators of the measured domains;16 therefore, scores obtained from this method are comparable to scores calculated by cohorts using the test manual. Overall, 122 individuals had VIQ scores provided by cohorts, and scores were calculated for an additional 134 children using Vocabulary (n=133) and Similarities (n=1). PIQ scores were available for 121 individuals in the cohorts, and we calculated PIQ as the average of the Block Design and Matrix Reasoning for the remaining 135 individuals.
Covariates
The highest level of maternal education attained was categorized as less than high school, high school, or some college and above. Household income was categorized as <$30,000, 30,000–49,999, 50,000–74,999, 75,000–99,000, and 100,000 or more. Child and maternal race and ethnicity were categorized as non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, non-Hispanic other race, and Hispanic categories. These data were included as covariates because they are considered proxy measures for unmeasured social determinants of health, such as exposure to racial discrimination. Non-Hispanic other race included non-Hispanic Alaska Native, non-Hispanic American Indian, non-Hispanic Native Hawaiian, and non-Hispanic Pacific Islander. Prenatal marital status was defined as married or living with a partner or not married. Insurance type was categorized as follows: Medicare, Medicaid, medical assistance, Children’s Health Insurance Program, or any kind of state or government assistance plan based on the income or disability of the biological mother (yes/no); private insurance, including Tricare/military healthcare and Indian Health Service coverage (yes/no); other insurance (yes/no); and no insurance (yes/no).
Maternal history of depression was any diagnosis of depression at any point before the pregnancy through 8 weeks’ postpartum by medical records, ICD-9/10 codes (ICD-9: 296.2X, 296.3X, 300.4, 296.9X; ICD-10: F32.XX, F33.XX, F34.1, F39.XX), or self-report.
Child characteristics included year of birth, sex at birth (male, female, ambiguous), gestational age at birth in completed weeks, and birth weight (g). Child data were obtained through maternal medical records, child medical records, or parent report.
Statistical Analysis
We described maternal sociodemographic characteristics, prenatal substance use, history of depression, childbirth characteristics, and child neurodevelopment outcomes by class and for the entire sample (Table 1).
Table 1.
Sample characteristics overall and by class membership
| Class 1 (average IQ and problem behavior) | Class 2 (low VIQ and elevated problem behavior) | Overall | p-Value | |
|---|---|---|---|---|
| Number of women | 223 (87.1%) | 33 (12.9%) | 256 (100%) | |
| Maternal demographics | ||||
| Age at delivery, N (%) with data | 222 (99.6%) | 33 (100%) | 255 (99.6%) | |
| Mean (SD) | 31 (6) | 30 (6.1) | 31 (6) | 0.243 |
| Race/ethnicity, N (%) with data | 220 (98.7%) | 33 (100%) | 253 (98.8%) | 0.771 |
| Non-Hispanic White, N (%) | 162 (74%) | 24 (73%) | 186 (74%) | |
| Non-Hispanic Black, N (%) | 12 (5%) | 1 (3%) | 13 (5%) | |
| Non-Hispanic Asian, N (%) | 6 (3%) | 0 (0%) | 6 (2%) | |
| Non-Hispanic other race, N (%) | 16 (7%) | 3 (9%) | 19 (8%) | |
| Hispanic, N (%) | 24 (11%) | 5 (15%) | 29 (11%) | |
| Calendar year of childbirth, N (%) with data | 223 (100%) | 33 (100%) | 256 (100%) | |
| 2001–2010, N (%) | 62 (28%) | 4 (12%) | 66 (26%) | 0.055 |
| 2011–2020, N (%) | 161 (72%) | 29 (88%) | 190 (74%) | 0.055 |
| Socioeconomic Status | ||||
| Education, N (%) with data | 221 (99.1%) | 32 (97%) | 253 (98.8%) | |
| Some college and above, N (%) | 199 (90%) | 26 (81%) | 225 (89%) | 0.138 |
| Employment, N (%) with data | 112 (50.2%) | 21 (63.6%) | 133 (52%) | |
| Employed, N (%) | 80 (71%) | 12 (57%) | 92 (69%) | 0.193 |
| Income, N (%) with data | 27 (12.1%) | 10 (30.3%) | 37 (14.5%) | 0.216 |
| < $30,000, N (%) | 19 (70%) | 9 (90%) | 28 (76%) | |
| $30,000–$49,999, N (%) | 0 (0%) | 0 (0%) | 0 (0%) | |
| $50,000–$74,999, N (%) | 8 (30%) | 1 (10%) | 9 (24%) | |
| $75,000–$99,999, N (%) | 0 (0%) | 0 (0%) | 0 (0%) | |
| $100,000 or more, N (%) | 0 (0%) | 0 (0%) | 0 (0%) | |
| Marital status, N (%) with data | 112 (50.2%) | 21 (63.6%) | 133 (52%) | |
| Married or living with a partner, N (%) | 102 (91%) | 14 (67%) | 116 (87%) | 0.002 |
| Comorbidities | ||||
| History of depression through 8-wk postpartum, N (%) with data | 116 (52%) | 20 (60.6%) | 136 (53.1%) | |
| Yes, N (%) | 9 (8%) | 6 (30%) | 15 (11%) | 0.003 |
| Depression during pregnancy, N (%) with data | 116 (52%) | 20 (60.6%) | 136 (53.1%) | |
| Yes, N (%) | 8 (7%) | 6 (30%) | 14 (10%) | 0.002 |
| Substance use during pregnancy | ||||
| Any substance use, N (%) | 29 (15%) | 8 (30%) | 37 (17%) | 0.003 |
| Total number of substances used, N (%) with data | 194 (87%) | 27 (81.8%) | 221 (86.3%) | |
| Mean (SD) | 0.2 (0.5) | 0.4 (0.6) | 0.2 (0.5) | 0.061 |
| Any substance use, N (%) | 29 (15%) | 8 (30%) | 37 (17%) | 0.056 |
| Alcohol, N (%) with data | 220 (98.7%) | 31 (93.9%) | 251 (98%) | |
| Yes, N (%) | 26 (12%) | 5 (16%) | 31 (12%) | 0.495 |
| Tobacco, N (%) with data | 199 (89.2%) | 27 (81.8%) | 226 (88.3%) | |
| Yes, N (%) | 5 (3%) | 4 (15%) | 9 (4%) | 0.002 |
| Marijuana, N (%) with data | 202 (90.6%) | 30 (90.9%) | 232 (90.6%) | |
| Yes, N (%) | 4 (2%) | 1 (3%) | 5 (2%) | 0.634 |
| Any illicit drugs*, N (%) with data | 221 (99.1%) | 33 (100%) | 254 (99.2%) | |
| Yes, N (%) | 0 (0%) | 0 (0%) | 0 (0%) | |
| Methamphetamine, N (%) with data | 221 (99.1%) | 33 (100%) | 254 (99.2%) | |
| Yes, N (%) | 0 (0%) | 0 (0%) | 0 (0%) | |
| Cocaine, N (%) with data | 221 (99.1%) | 33 (100%) | 254 (99.2%) | |
| Yes, N (%) | 0 (0%) | 0 (0%) | 0 (0%) | |
| Opioid use | ||||
| Use during pregnancy, N (%) with data | 223 (100%) | 33 (100%) | 256 (100%) | |
| Yes, N (%) | 5 (2%) | 2 (6%) | 7 (3%) | 0.209 |
| Use by trimester, N (%) with data | 222 (99.6%) | 33 (100%) | 255 (99.6%) | |
| First trimester, N (%) | <5 | <5 | <5 | 0.117 |
| Second trimester, N (%) | <5 | <5 | <5 | 0.29 |
| Third trimester, N (%) | 3 (1%) | 2 (6%) | 5 (2%) | 0.069 |
| Prescription use, N (%) with data | 116 (52%) | 21 (63.6%) | 137 (53.5%) | |
| Yes, N (%) | 5 (4%) | 2 (10%) | 7 (5%) | 0.318 |
| Heroin, N (%) with data | 222 (99.6%) | 33 (100%) | 255 (99.6%) | |
| Child characteristics | ||||
| Sex, N (%) with data | 223 (100%) | 33 (100%) | 256 (100%) | |
| Males, N (%) | 115 (52%) | 22 (67%) | 137 (54%) | 0.105 |
| Females, N (%) | 108 (48%) | 11 (33%) | 119 (46%) | |
| Child race/ethnicity, N(%) with data | 221 (99.1%) | 33 (100%) | 254 (99.2%) | 0.601 |
| Non-Hispanic White, N (%) | 146 (66%) | 21 (64%) | 167 (66%) | |
| Non-Hispanic Black, N (%) | 14 (6%) | 2 (6%) | 16 (6%) | |
| Non-Hispanic Asian, N (%) | 9 (4%) | 0 (0%) | 9 (4%) | |
| Non-Hispanic other race, N (%) | 23 (10%) | 3 (9%) | 26 (10%) | |
| Hispanic, N (%) | 29 (13%) | 7 (21%) | 36 (14%) | |
| Gestational age at birth, N (%) with data | 221 (99.1%) | 33 (100%) | 254 (99.2%) | |
| Mean (SD) | 39 (1.8) | 39.3 (2) | 39 (1.8) | 0.315 |
| Birthweight (kg), N (%) with data | 220 (98.7%) | 32 (97%) | 252 (98.4%) | |
| Mean (SD) | 3.3 (0.5) | 3.4 (0.5) | 3.4 (0.5) | 0.419 |
| Large or small for gestational age at birth, N (%) with data | 218 (97.8%) | 32 (97%) | 250 (97.7%) | |
| Large for gestational age, N (%) | 39 (18%) | 5 (16%) | 44 (18%) | 0.753 |
| Small for gestational age, N (%) | 18 (8%) | 1 (3%) | 19 (8%) | 0.306 |
| Cognition summary | ||||
| Verbal comprehension IQ (VIQ) score, mean (SD) | 11 2(15) | 106 (19.1) | 111 (15.7) | 0.051 |
| 1 SD < mean, VIQ, N (%) | 15 (7%) | 7 (21%) | 22 (9%) | 0.006 |
| Perceptual reasoning IQ (PIQ) score, mean (SD) | 107 (14) | 103 (14.8) | 106 (14.1) | 0.183 |
| 1 SD < mean, PIQ, N (%) | 21 (9%) | 5 (15%) | 26 (10%) | 0.309 |
| Child age (y) at IQ assessment, mean (SD) | 7.4 (1.7) | 6.9 (1.4) | 7.3 (1.7) | 0.151 |
| Behavior summary | ||||
| Anxious/Depressed syndrome T-score, Mean (SD) | 52 (4.5) | 65 (7.4) | 54 (6.5) | <0.001 |
| Withdrawn/Depressed syndrome T-score, Mean (SD) | 52 (3.8) | 63 (6.8) | 54 (5.6) | <0.001 |
| Somatic complaints T-score, mean (SD) | 54 (5) | 62 (7.7) | 55 (6) | <0.001 |
| Social problems syndrome T-score, mean (SD) | 52 (2.7) | 64 (6.5) | 54 (5.4) | <0.001 |
| Thought problems T-score, mean (SD) | 52 (3.6) | 63 (7.6) | 54 (5.6) | <0.001 |
| Attention problems T-score, mean (SD) | 54 (5.1) | 64 (8.8) | 55 (6.7) | <0.001 |
| Rule-Breaking behavior T-score, Mean (SD) | 52 (3.5) | 62 (8.1) | 54 (5.4) | <0.001 |
| Aggressive behavior T-score, mean (SD) | 52 (4.4) | 65 (8.7) | 54 (6.7) | <0.001 |
| Internalizing problems clinical ranges, N (%) with data | 223 (100%) | 33 (100%) | 256 (100%) | <0.001 |
| Normal range, N (%) | 203 (91%) | 5 (15%) | 208 (81%) | |
| Borderline clinical range, N (%) | 13 (6%) | 4 (12%) | 17 (7%) | |
| Clinical range, N (%) | 7 (3%) | 24 (73%) | 31 (12%) | |
| Externalizing problems clinical ranges, N (%) with data | 223 (100%) | 33 (100%) | 256 (100%) | <0.001 |
| Normal range, N (%) | 211 (95%) | 9 (27%) | 220 (86%) | |
| Borderline clinical range, N (%) | 9 (4%) | 7 (21%) | 16 (6%) | |
| Clinical range, N (%) | 3 (1%) | 17 (52%) | 20 (8%) | |
| Total problems clinical ranges, N (%) with data | 223 (100%) | 33 (100%) | 256 (100%) | <0.001 |
| Normal range, N (%) | 216 (97%) | 1 (3%) | 217 (85%) | |
| Borderline clinical range, N (%) | 7 (3%) | 9 (27%) | 16 (6%) | |
| Clinical range, N (%) | 0 (0%) | 23 (70%) | 23 (9%) | |
| Child age (y) at behavior assessment, mean (SD) | 8 (1.5) | 7.4 (1.4) | 7.9 (1.5) | 0.032 |
Abbreviations: PIQ, performance intelligence quotient; SD, standard deviation; VIQ, verbal intelligence quotient.
To impute missing data, we used multiple imputation by fully conditional specification (FCS) with a discriminant function17 for the categorical and binary variables, including race/ethnicity, maternal education, and prenatal alcohol, nicotine, and marijuana use. FCS predictive mean matching methods were used for imputing the continuous variables, such as maternal age at delivery, gestational age of child, and birth weight of child. Imputation models included prenatal substance use (opioid, alcohol, nicotine, and marijuana use), outcome (VIQ, PIQ, or CBCL scores), all covariates in the main analysis model, and Cohort-ID as a classification variable.
We modelled the heterogeneity of child behavior problems and IQ using latent profile analysis. This approach groups children into distinct profiles based on subscales from the CBCL, VIQ, and PIQ. To determine the number of latent profiles, we compared goodness of fit indices using standard fit statistics (Bayesian information criterion, BIC), where a lower number is indicative of better fit, and the Vuong–Lo–Mendell–Rubin adjusted likelihood ratio test (LRT18,19). Once the number of profiles was chosen, the model was run without covariates to obtain the most likely class membership and associated classification probabilities, which were then used in additional models with outcomes and covariates. This method avoids class switching and accounts for measurement error that exists when using this type of latent approach.20 Standard errors were adjusted for clustering at the cohort level using sandwich estimators. Odds ratios (ORs) of being in Class 1 compared with Class 2 were adjusted for prenatal substance use, race/ethnicity, child’s sex, gestational age, birth weight, and maternal education, and 95% confidence intervals (CIs) were obtained. Analyses were conducted in Mplus 8.16.21
Results
Sample Characteristics
Sample characteristics for the 256 children in this study are summarized in Table 1. Approximately half of the participants were male (53.7%) and primarily non-Hispanic White (80.2%). Overall, 12.6% of the children were exposed to alcohol, 4.2% exposed to nicotine, 2.2% exposed to marijuana, and 2.7% exposed to opioids in utero. The mean age at CBCL assessment was 7.3 years, and the mean age at WISC-5 administration was 7.9 years.
Outcome: Latent Profile Analysis
Fit statistics for class enumeration (Table 2) showed that a two-class solution fit the data well (Fig. 1) based on the BIC values and the LRT. The largest profile, Class 1 (C1; n=223, 87%), was characterized by low behavioral problems across all CBCL subscales and average scores for IQ (a mean PIQ value of 107 [SD=14] and a mean VIQ value of 112 [SD=15]). The second profile, Class 2 (C2; n=33, 13%), was characterized by elevated scores for all CBCL subscales and significant elevations above the clinical cutoff for internalizing, externalizing, and total behavior problems (Table 1). The C2 group was characterized by average IQ scores with a mean PIQ value of 103 (SD=14.8) and a mean VIQ value of 106 (SD=19.1; Table 1) and significantly more children in the C2 group had VIQ scores 1 SD below the mean compared with children in the C1 group (p=0.006). All CBCL subscales differed significantly between the two profiles, with the C2 group having higher scores than the C1 group (p<0.001). In the C1 group, 15% of children had exposure to any substance compared with 30% of children in the C2 group (chi-squared, p=0.03; Table 1).
Table 2.
Model fit indices
| Class | Log likelihood | No. of free par | BIC | ssaBIC | Entropy | LMR | LMR–LRT p- value | Smallest class |
|---|---|---|---|---|---|---|---|---|
| 1 | −8,799.314 | 20 | 17,709.76 | 17,646.356 | ||||
| 2 | −8,365.392 | 31 | 16,903.05 | 16,804.764 | 0.986 | 853.874 | 0.0031 | 13.10% |
| 3 | −8,241.36 | 42 | 16,716.11 | 16,582.952 | 0.971 | 244.07 | 0.22 | 4.60% |
| 4 | −8,177.74 | 53 | 16,649.99 | 16,481.962 | 0.982 | 125.194 | 0.3659 | 4.60% |
Abbreviations: BIC, Bayesian information criterion; LMR, likelihood ratio test; LMR-LRT, Lo–Mendell–Rubin adjusted likelihood ratio test; ssaBIC, sample size–adjusted Bayesian information criterion.
Fig. 1.
Latent profiles of IQ and problem behavior outcomes. IQ and problem behavior variables are plotted for each latent profile. IQ, intelligence quotient. PIQ, performance intelligence quotient; VIQ, verbal intelligence quotient; summary scores from the Child Behavior Checklist include Anx/Dep, anxiety and depression; With/Dep, withdrawn/depressed; Somatic, somatic problems; Social Prb, social problems; Thought Prb, thought problems; Attention, attention problems; Rule breaking, rule-breaking behavior; Aggression, aggression problems.
Several covariates were used to predict class membership (Table 3). Nicotine use during the prenatal period was a significant predictor of class membership (OR=0.16; 95% CI: 0.04–0.60) such that children with exposure to nicotine were more likely to be in the C2 group. Prenatal opioid, alcohol, and marijuana use were not significant predictors of profile membership. Overall exposure to substances during the prenatal period was higher in the C2 group. Compared with non-Hispanic White children, non-Hispanic Black children were significantly more likely to be in the C1 group (OR=2.46; 95% CI: 1.85–3.26).
Table 3.
Adjusted odds ratio for the association of prenatal substance exposure with class membership
| Full sample (n = 256) | Class 1 (n = 223, 87%) | Class 2 (n = 33, 13%) | Adjusted OR (95% CI) | p-Value | |
|---|---|---|---|---|---|
| Prenatal opioid use | 2.7% | 2.2% | 6.0% | 0.52 (0.07–3.70) | 0.58 |
| Prenatal alcohol use | 12.6% | 11.7% | 18.7% | 0.76 (0.41–1.43) | 0.48 |
| Prenatal tobacco use | 4.2% | 2.9% | 13.0% | 0.16 (0.04–0.60) | 0.02 |
| Prenatal marijuana use | 2.2% | 2.0% | 3.1% | 1.54 (0.01–293.5) | 0.89 |
| Race/ethnicity | |||||
| Non-Hispanic Black | 6.2% | 6.2% | 5.9% | 2.46 (1.85–3.26) | <0.001 |
| Non-Hispanic other race | 27.9% | 27.7% | 29.4% | 1.42 (0.54–3.69) | 0.55 |
| Hispanic | 14.3% | 13.4% | 20.7% | 0.59 (0.06–6.01) | 0.71 |
| Sex (male) | 53.7% | 51.5% | 68.0% | 0.37 (0.14–1.02) | 0.11 |
| Gestational age, mean completed weeks | 39.1 | 39.0 | 39.4 | 0.85 (0.80–0.90) | <0.001 |
| Birth weight (g) | 3357.1 | 3341.8 | 3460.2 | 1.00 (1.00–1.00) | 0.78 |
| Maternal education | 89.2% | 90.2% | 82.2% | 2.01 (0.54–3.69) | 0.31 |
Abbreviations: CI, confidence interval; OR, odds ratio.
Notes: Non-Hispanic other race includes non-Hispanic Asian and non-Hispanic Alaska Native, non-Hispanic American Indian, non-Hispanic Native Hawaiian, and non-Hispanic Pacific Islander. Maternal education categorized as some college and above versus less than college.
Discussion
A critical gap in the literature concerns the long-term neurodevelopmental outcomes of polysubstance exposure in children. We sought to identify which subgroups of children with prenatal exposure to alcohol, nicotine, marijuana, and opioids may be most in need of neurodevelopmental support (clinically significant behavior problems and low IQ). Two latent profiles of problem behavior and IQ outcomes emerged in this study that were predicted by prenatal maternal substance use. The majority of children (C1 group) had average scores on the problem behavior assessment and slightly above-average IQ. Prenatal exposure to any substance was low. In contrast, 73% of children in the C2 group had internalizing behaviors above the clinical cutoff, and 52% had clinically elevated externalizing problems. More of the children in the C2 group had low (<85) VIQ scores. In addition, twice as many children in the C2 group had polysubstance exposure compared with children in the C1 group (30 vs. 15%, respectively).
Children in the high problem behavior, average IQ class (C2 group) were significantly more likely to have prenatal nicotine exposure (15% of children in the C2 group had prenatal nicotine exposure compared with 3% in the C1 group). Children in the C2 group were polysubstance exposed and also had prenatal exposure to alcohol, nicotine, marijuana, and illicit substances. In previous studies using magnetic resonance imaging, prenatal nicotine exposure throughout pregnancy predicted cortical thinning in the precentral and superior frontal cortices, which in turn predicted increased affective problems in 6 year olds.22 Prenatal nicotine exposure could also disrupt placental blood flow23 and increase risk for iron deficiency24 that may explain the behavioral and verbal reasoning challenges observed in the C2 group.
Prenatal nicotine exposure also predicts increased externalizing behavior in middle childhood.5 These findings suggest that pediatricians should monitor prenatal nicotine exposure effects closely as they may screen for prenatal exposure to other substances, such as opioids. Furthermore, for children with prenatal nicotine exposure, a typical IQ score may not generalize to average problem behavior. Clinicians should apply a multidimensional approach to screening that includes cognitive functioning in addition to behavior problems. The discrepancy we observed between VIQ and PIQ in this class may suggest that a child with prenatal nicotine and other substance exposure could have challenges with verbal expression, which may be frustrating and increase the risk for aggression or withdrawn behavior. For example, VIQ, but not PIQ, predicted elevated externalizing problems in boys with attention-deficit/hyperactivity disorder (ADHD25). It is possible that the hypothesized endophenotype is the result of nicotine-driven polysubstance exposure and its effect on behavior problems explained, in part, challenges in verbal reasoning.
The effects of nicotine on problem behavior and IQ were stronger than expected. Studies of prenatal substance exposure usually employ one of two methods: identifying cohorts with prenatal exposure to the drug of interest compared with a substance with no prenatal substance exposure26 or studying cohorts with polysubstance exposure and control for (or covary) other substances.7 Both of these methods are treated as identifying “substance effects.” However, finding a substance effect in the context of other substances may be different than finding a substance effect without the presence of other substances. The latent profile analysis in our study generated two groups of children with prenatal substance exposure. Thus, the nicotine effects that we observed were in the context of other substance exposures. It is also noteworthy that this is not a population with a high sociodemographic risk. Most participants had some college education, were employed, and were married, or had a partner. Many children were not exposed to any substances. Yet, not only were there prenatal substance use effects, but also the effects were large enough to reach clinical significance. These findings indicate how pervasive the effects of substance use are on children and that even low levels of exposure to commonly used substances reach all segments of society.
We attempted to consider the effects of a number of relevant confounding variables, such as prenatal maternal depression exposure and socioeconomic status, given that women who use substances prenatally also report high levels of stress, psychopathology, and financial stress (Conradt; others). Symptoms of maternal depression were elevated in Class 2 relative to Class 1, which could also explain the elevations in problem behavior and verbal reasoning observed in Class 2. Some mothers with depression may struggle with providing an enriching verbal environment for their children, which may have impacted verbal development.27 There is also evidence that some mothers who report elevated symptoms of depression also have infants who show more irritability, making it challenging to parent the infant, which could over time and via coercive parent–child interactions lead to elevated behavioral difficulties in the child.28 Prospective studies that include observations of parenting and child behavior may help to clarify potential mechanisms implicated in the development of behavior dysregulation.
Study Strength and Limitations
These findings have important clinical and policy applications. The majority of studies that investigate the neurodevelopmental consequences of prenatal substance exposure examine outcomes in isolation. However, clinicians often assess children’s functioning in one domain (such as problem behavior) while neglecting others (such as IQ). Examining the combination of problem behavior and IQ may help to identify which specific groups of children are most impacted by prenatal substance exposure. Both groups of children showed average VIQ and PIQ, although significantly more children in the C2 group had VIQ scores that fell at least 1 SD below the mean, the cutoff for early intervention services. Problem behavior scores and the number of children with clinically significant elevations in problem behavior also distinguished the groups. Children in the C2 group had elevated problem behavior scores, with most children falling in the clinical range for internalizing and externalizing behavior (73 and 52%, respectively). This pattern was different from other studies wherein children with prenatal substance exposure typically show elevations in externalizing and attention problems, but not always in internalizing behavior.6 It may be that polysubstance exposure has a more pervasive impact on behavior, affecting all problem behaviors measured. If replicated, these results may also begin to describe a polysubstance exposure endophenotype characterized, in part, by clinically significant elevations in internalizing, externalizing, and attention problems, below-average VIQ, and average nonverbal IQ.
With respect to limitations, the sample size of the C2 group was small, as was the number of women reporting nicotine use during pregnancy. Similar to other cohort studies (e.g,29 we had little information about the frequency, quantity, or trimester-specific effects of substance use on behavioral and cognitive outcomes, and we were unable to test for trimester-specific effects due to a lack of data, a priority for future studies. In addition, maternal education and socioeconomic status were used as covariates, which may serve as proxies for unmeasured direct effects of poverty on child problem behavior and IQ outcomes, such as quality of language exposure. We were also not able to measure additional postnatal risk factors that may contribute to the development of problem behavior, such as parenting, parental history of learning problems or ADHD, and postnatal substance use, which are also associated with prenatal nicotine use.30 Additional limitations of this study include the assessment of prenatal substance exposure, which relied on medical records or maternal report rather than toxicology analysis.
Conclusion
A key contribution of this research to clinical practice is the identification of two subgroups of children with different profiles of IQ and behavioral data depending on the type of substance to which they were exposed. The group of children with elevated internalizing, externalizing, and attention problems, average PIQ, and lower VIQ had significantly more prenatal exposure to nicotine. This approach to modelling data using multiple outcomes simultaneously could be replicated in other high-risk samples, for example, infants born premature. Evidence from this study and other longitudinal studies indicates that pediatricians and policy makers should continue to screen for psychopathology well into middle childhood, particularly in children with known prenatal nicotine exposure. Attempts should also be made to replicate our problem behavior and IQ profiles in children with other known early life risks, such as those reared in poverty, to determine if our multidimensional approach to examining developmental outcomes is consistent across diverse groups of children.
Key Points.
The neurodevelopmental consequences of prenatal polysubstance exposure are poorly understood.
Children with prenatal polysubstance exposure exhibited reduced IQ and elevated problem behavior.
We found significant behavior problems in children with average PIQ and below-average VIQ.
This may represent a unique endophenotype related to prenatal nicotine exposure.
Acknowledgments
The authors wish to thank our ECHO colleagues, the medical, nursing, and program staff, as well as the children and families participating in the ECHO cohorts. We also acknowledge the contribution of the following ECHO program collaborators:
Environmental Influences on Child Health Outcomes Components
Coordinating Center: Duke Clinical Research Institute, Durham, North Carolina: Smith PB, Newby KL; and Data Analysis Center: Johns Hopkins University Bloomberg School of Public Health, Baltimore, Maryland: Jacobson LP; Research Triangle Institute, Durham, North Carolina: Parker CB.
Environmental Influences on Child Health Outcomes Awardees and Cohorts
Memorial Hospital of Rhode Island, Pawtucket: Deoni S; University of Washington, Seattle: Karr C, Sathyanarayana S.
Funding
This work was supported by the Environmental Influences on Child Health Outcomes (ECHO) program, Office of The Director, National Institutes of Health, under Award Numbers U2COD023375 (Coordinating Center), U24OD023382 (Data Analysis Center), U24OD023319 (PRO Core), and UH3OD023347 (Lester), UH3OD023347 (McEvoy), 4UH3OD023271-03 (Karr, Sathyanarayana), UH3 OD023285 (Paneth), 4UH3OD023282-03 (Gern). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Ethical Approval and Consent to Participate
The study protocol was approved by the local and/or central ECHO Institutional Review Board.
Written informed consent or parent’s/guardian’s permission was obtained along with child assent as appropriate.
Conflict of Interest
None declared.
Data Availability Statement
Deidentified data from the ECHO Program are available through NICHD’s Data and Specimen Hub (DASH). DASH is a centralized resource that allows researchers to access data from various studies via a controlled-access mechanism. Researchers can now request access to these data by creating a DASH account and submitting a Data Request Form. The NICHD DASH Data Access Committee will review the request and provide a response in approximately 2 to 3 weeks. Once granted access, researchers will be able to use the data for 3 years. See the DASH Tutorial for more detailed information on the process.
References
- 1.Substance Abuse and Mental Health Services Administration, U.S. Department of Health and Human Services. 2019 National Survey on Drug Use and Health: Women. Accessed May 23, 2023 at: https://www.samhsa.gov/data/sites/default/files/reports/rpt31102/2019NSDUH-Women/Women%202019%20NSDUH.pdf
- 2.Conradt E, Flannery T, Aschner JL, et al. Prenatal opioid exposure: neurodevelopmental consequences and future research priorities. Pediatrics 2019;144(03):e20190128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Liu Y, Williamson V, Setlow B, Cottler LB, Knackstedt LA. The importance of considering polysubstance use: lessons from cocaine research. Drug Alcohol Depend 2018;192:16–28 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Cornelius MD, De Genna NM, Leech SL, Willford JA, Goldschmidt L, Day NL. Effects of prenatal cigarette smoke exposure on neurobehavioral outcomes in 10-year-old children of adolescent mothers. Neurotoxicol Teratol 2011;33(01):137–144 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Min MO, Minnes S, Park H, et al. Developmental trajectories of externalizing behavior from ages 4 to 12: prenatal cocaine exposure and adolescent correlates. Drug Alcohol Depend 2018; 192:223–232 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Conradt E, Abar B, Lester BM, et al. Cortisol reactivity to social stress as a mediator of early adversity on risk and adaptive outcomes. Child Dev 2014;85(06):2279–2298 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Fisher PA, Lester BM, DeGarmo DS, et al. The combined effects of prenatal drug exposure and early adversity on neurobehavioral disinhibition in childhood and adolescence. Dev Psychopathol 2011;23(03):777–788 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Liu J, Bann C, Lester B, et al. Neonatal neurobehavior predicts medical and behavioral outcome. Pediatrics 2010;125(01):e90–e98 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Romano ME, Buckley JP, Elliott AJ, Johnson CC, Paneth N. on behalf of program collaborators for Environmental Influences on Child Health Outcomes. SPR perspectives: scientific opportunities in the environmental influences on child health outcomes program. Pediatr Res 2022;92(05):1255–1261 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Paneth N, Monk C. The importance of cohort research starting early in life to understanding child health. Curr Opin Pediatr 2018; 30(02):292–296 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Deoni SCL, O’Muircheartaigh J, Elison JT, et al. White matter maturation profiles through early childhood predict general cognitive ability. Brain Struct Funct 2016;221(02):1189–1203 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Deoni SCL, Dean DC III, O’Muircheartaigh J, Dirks H, Jerskey BA. Investigating white matter development in infancy and early childhood using myelin water faction and relaxation time mapping. Neuroimage 2012;63(03):1038–1053 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Sathyanarayana S, Grady R, Redmon JB, et al. ; TIDES Study Team. Anogenital distance and penile width measurements in The Infant Development and the Environment Study (TIDES): methods and predictors. J Pediatr Urol 2015;11(02):76.e1–76.e6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Day DB, Collett BR, Barrett ES, et al. Phthalate mixtures in pregnancy, autistic traits, and adverse childhood behavioral outcomes. Environ Int 2021;147:106330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Achenbach TM, Ruffle TM. The Child Behavior Checklist and related forms for assessing behavioral/emotional problems and competencies. Pediatr Rev 2000;21(08):265–271 [DOI] [PubMed] [Google Scholar]
- 16.Wechsler D Wechsler Intelligence Scale for Children–Fifth Edition Technical and Interpretive Manual. NCS Pearson; 2014 [Google Scholar]
- 17.Liu Y, De A. Multiple imputation by fully conditional specification for dealing with missing data in a large epidemiologic study. Int J Stat Med Res 2015;4(03):287–295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Nylund KL, Asparouhov T, Muthén BO. Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study. Struct Equ Model Multidiscip J. 2007;14(04):535–569 [Google Scholar]
- 19.Nylund-Gibson K, Choi AY. Ten frequently asked questions about latent class analysis. Transl Issues Psychol Sci 2018;4(04): 440–461 [Google Scholar]
- 20.Nylund-Gibson K, Grimm RP, Masyn KE. Prediction from latent classes: a demonstration of different approaches to include distal outcomes in mixture models. Struct Equ Model Multidiscip J. 2019;26(06):967–985 [Google Scholar]
- 21.Muthén LK, Muthén BO. Mplus User’s Guide. 5th ed. Muthén & Muthén; 2017 [Google Scholar]
- 22.El Marroun H, Schmidt MN, Franken IHA, et al. Prenatal tobacco exposure and brain morphology: a prospective study in young children. Neuropsychopharmacology 2014;39(04):792–800 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Machado JdeB, Plínio Filho VM, Petersen GO, Chatkin JM. Quantitative effects of tobacco smoking exposure on the maternal-fetal circulation. BMC Pregnancy Childbirth 2011;11(01):24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Carter RC, Jacobson SW, Molteno CD, Jacobson JL. Fetal alcohol exposure, iron-deficiency anemia, and infant growth. Pediatrics 2007;120(03):559–567 [DOI] [PubMed] [Google Scholar]
- 25.Kebir O, Grizenko N, Sengupta S, Joober R. Verbal but not performance IQ is highly correlated to externalizing behavior in boys with ADHD carrying both DRD4 and DAT1 risk genotypes. Prog Neuropsychopharmacol Biol Psychiatry 2009;33 (06):939–944 [DOI] [PubMed] [Google Scholar]
- 26.Himes SK, Stroud LR, Scheidweiler KB, Niaura RS, Huestis MA. Prenatal tobacco exposure, biomarkers for tobacco in meconium, and neonatal growth outcomes. J Pediatr 2013;162(05):970–975 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Ahun MN, Geoffroy MC, Herba CM, et al. Timing and chronicity of maternal depression symptoms and children’s verbal abilities. J Pediatr 2017;190:251–257 [DOI] [PubMed] [Google Scholar]
- 28.Hails KA, Reuben JD, Shaw DS, Dishion TJ, Wilson MN. Transactional associations among maternal depression, parent-child coercion, and child conduct problems during early childhood. J Clin Child Adolesc Psychol 2018;47(Suppl 1): S291–S305 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Baranger DAA, Paul SE, Colbert SMC, et al. Association of mental health burden with prenatal cannabis exposure from childhood to early adolescence: longitudinal findings from the adolescent brain cognitive development (ABCD) study. JAMA Pediatr 2022; 176(12):1261–1265 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Palmer RHC, Bidwell LC, Heath AC, Brick LA, Madden PAF, Knopik VS. Effects of maternal smoking during pregnancy on offspring externalizing problems: contextual effects in a sample of female twins. Behav Genet 2016;46(03):403–415 [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.
Data Availability Statement
Deidentified data from the ECHO Program are available through NICHD’s Data and Specimen Hub (DASH). DASH is a centralized resource that allows researchers to access data from various studies via a controlled-access mechanism. Researchers can now request access to these data by creating a DASH account and submitting a Data Request Form. The NICHD DASH Data Access Committee will review the request and provide a response in approximately 2 to 3 weeks. Once granted access, researchers will be able to use the data for 3 years. See the DASH Tutorial for more detailed information on the process.